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setup_func) { 198 pr_warn("Parameter %s is obsolete, ignored\n", 199 p-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