1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * linux/init/main.c 4 * 5 * Copyright (C) 1991, 1992 Linus Torvalds 6 * 7 * GK 2/5/95 - Changed to support mounting root fs via NFS 8 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96 9 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96 10 * Simplified starting of init: Michael A. Griffith <grif@acm.org> 11 */ 12 13 #define DEBUG /* Enable initcall_debug */ 14 15 #include <linux/types.h> 16 #include <linux/extable.h> 17 #include <linux/module.h> 18 #include <linux/proc_fs.h> 19 #include <linux/binfmts.h> 20 #include <linux/kernel.h> 21 #include <linux/syscalls.h> 22 #include <linux/stackprotector.h> 23 #include <linux/string.h> 24 #include <linux/ctype.h> 25 #include <linux/delay.h> 26 #include <linux/ioport.h> 27 #include <linux/init.h> 28 #include <linux/initrd.h> 29 #include <linux/memblock.h> 30 #include <linux/acpi.h> 31 #include <linux/bootconfig.h> 32 #include <linux/console.h> 33 #include <linux/nmi.h> 34 #include <linux/percpu.h> 35 #include <linux/kmod.h> 36 #include <linux/kprobes.h> 37 #include <linux/kmsan.h> 38 #include <linux/vmalloc.h> 39 #include <linux/kernel_stat.h> 40 #include <linux/start_kernel.h> 41 #include <linux/security.h> 42 #include <linux/smp.h> 43 #include <linux/profile.h> 44 #include <linux/kfence.h> 45 #include <linux/rcupdate.h> 46 #include <linux/srcu.h> 47 #include <linux/moduleparam.h> 48 #include <linux/kallsyms.h> 49 #include <linux/buildid.h> 50 #include <linux/writeback.h> 51 #include <linux/cpu.h> 52 #include <linux/cpuset.h> 53 #include <linux/cgroup.h> 54 #include <linux/efi.h> 55 #include <linux/tick.h> 56 #include <linux/sched/isolation.h> 57 #include <linux/interrupt.h> 58 #include <linux/taskstats_kern.h> 59 #include <linux/delayacct.h> 60 #include <linux/unistd.h> 61 #include <linux/utsname.h> 62 #include <linux/rmap.h> 63 #include <linux/mempolicy.h> 64 #include <linux/key.h> 65 #include <linux/page_ext.h> 66 #include <linux/debug_locks.h> 67 #include <linux/debugobjects.h> 68 #include <linux/lockdep.h> 69 #include <linux/kmemleak.h> 70 #include <linux/padata.h> 71 #include <linux/pid_namespace.h> 72 #include <linux/device/driver.h> 73 #include <linux/kthread.h> 74 #include <linux/sched.h> 75 #include <linux/sched/init.h> 76 #include <linux/signal.h> 77 #include <linux/idr.h> 78 #include <linux/kgdb.h> 79 #include <linux/ftrace.h> 80 #include <linux/async.h> 81 #include <linux/shmem_fs.h> 82 #include <linux/slab.h> 83 #include <linux/perf_event.h> 84 #include <linux/ptrace.h> 85 #include <linux/pti.h> 86 #include <linux/blkdev.h> 87 #include <linux/sched/clock.h> 88 #include <linux/sched/task.h> 89 #include <linux/sched/task_stack.h> 90 #include <linux/context_tracking.h> 91 #include <linux/random.h> 92 #include <linux/list.h> 93 #include <linux/integrity.h> 94 #include <linux/proc_ns.h> 95 #include <linux/io.h> 96 #include <linux/cache.h> 97 #include <linux/rodata_test.h> 98 #include <linux/jump_label.h> 99 #include <linux/mem_encrypt.h> 100 #include <linux/kcsan.h> 101 #include <linux/init_syscalls.h> 102 #include <linux/stackdepot.h> 103 #include <linux/randomize_kstack.h> 104 #include <net/net_namespace.h> 105 106 #include <asm/io.h> 107 #include <asm/bugs.h> 108 #include <asm/setup.h> 109 #include <asm/sections.h> 110 #include <asm/cacheflush.h> 111 112 #define CREATE_TRACE_POINTS 113 #include <trace/events/initcall.h> 114 115 #include <kunit/test.h> 116 117 static int kernel_init(void *); 118 119 extern void init_IRQ(void); 120 extern void radix_tree_init(void); 121 extern void maple_tree_init(void); 122 123 /* 124 * Debug helper: via this flag we know that we are in 'early bootup code' 125 * where only the boot processor is running with IRQ disabled. This means 126 * two things - IRQ must not be enabled before the flag is cleared and some 127 * operations which are not allowed with IRQ disabled are allowed while the 128 * flag is set. 129 */ 130 bool early_boot_irqs_disabled __read_mostly; 131 132 enum system_states system_state __read_mostly; 133 EXPORT_SYMBOL(system_state); 134 135 /* 136 * Boot command-line arguments 137 */ 138 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT 139 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT 140 141 extern void time_init(void); 142 /* Default late time init is NULL. archs can override this later. */ 143 void (*__initdata late_time_init)(void); 144 145 /* Untouched command line saved by arch-specific code. */ 146 char __initdata boot_command_line[COMMAND_LINE_SIZE]; 147 /* Untouched saved command line (eg. for /proc) */ 148 char *saved_command_line; 149 /* Command line for parameter parsing */ 150 static char *static_command_line; 151 /* Untouched extra command line */ 152 static char *extra_command_line; 153 /* Extra init arguments */ 154 static char *extra_init_args; 155 156 #ifdef CONFIG_BOOT_CONFIG 157 /* Is bootconfig on command line? */ 158 static bool bootconfig_found; 159 static size_t initargs_offs; 160 #else 161 # define bootconfig_found false 162 # define initargs_offs 0 163 #endif 164 165 static char *execute_command; 166 static char *ramdisk_execute_command = "/init"; 167 168 /* 169 * Used to generate warnings if static_key manipulation functions are used 170 * before jump_label_init is called. 171 */ 172 bool static_key_initialized __read_mostly; 173 EXPORT_SYMBOL_GPL(static_key_initialized); 174 175 /* 176 * If set, this is an indication to the drivers that reset the underlying 177 * device before going ahead with the initialization otherwise driver might 178 * rely on the BIOS and skip the reset operation. 179 * 180 * This is useful if kernel is booting in an unreliable environment. 181 * For ex. kdump situation where previous kernel has crashed, BIOS has been 182 * skipped and devices will be in unknown state. 183 */ 184 unsigned int reset_devices; 185 EXPORT_SYMBOL(reset_devices); 186 187 static int __init set_reset_devices(char *str) 188 { 189 reset_devices = 1; 190 return 1; 191 } 192 193 __setup("reset_devices", set_reset_devices); 194 195 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 196 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 197 static const char *panic_later, *panic_param; 198 199 extern const struct obs_kernel_param __setup_start[], __setup_end[]; 200 201 static bool __init obsolete_checksetup(char *line) 202 { 203 const struct obs_kernel_param *p; 204 bool had_early_param = false; 205 206 p = __setup_start; 207 do { 208 int n = strlen(p->str); 209 if (parameqn(line, p->str, n)) { 210 if (p->early) { 211 /* Already done in parse_early_param? 212 * (Needs exact match on param part). 213 * Keep iterating, as we can have early 214 * params and __setups of same names 8( */ 215 if (line[n] == '\0' || line[n] == '=') 216 had_early_param = true; 217 } else if (!p->setup_func) { 218 pr_warn("Parameter %s is obsolete, ignored\n", 219 p->str); 220 return true; 221 } else if (p->setup_func(line + n)) 222 return true; 223 } 224 p++; 225 } while (p < __setup_end); 226 227 return had_early_param; 228 } 229 230 /* 231 * This should be approx 2 Bo*oMips to start (note initial shift), and will 232 * still work even if initially too large, it will just take slightly longer 233 */ 234 unsigned long loops_per_jiffy = (1<<12); 235 EXPORT_SYMBOL(loops_per_jiffy); 236 237 static int __init debug_kernel(char *str) 238 { 239 console_loglevel = CONSOLE_LOGLEVEL_DEBUG; 240 return 0; 241 } 242 243 static int __init quiet_kernel(char *str) 244 { 245 console_loglevel = CONSOLE_LOGLEVEL_QUIET; 246 return 0; 247 } 248 249 early_param("debug", debug_kernel); 250 early_param("quiet", quiet_kernel); 251 252 static int __init loglevel(char *str) 253 { 254 int newlevel; 255 256 /* 257 * Only update loglevel value when a correct setting was passed, 258 * to prevent blind crashes (when loglevel being set to 0) that 259 * are quite hard to debug 260 */ 261 if (get_option(&str, &newlevel)) { 262 console_loglevel = newlevel; 263 return 0; 264 } 265 266 return -EINVAL; 267 } 268 269 early_param("loglevel", loglevel); 270 271 #ifdef CONFIG_BLK_DEV_INITRD 272 static void * __init get_boot_config_from_initrd(size_t *_size) 273 { 274 u32 size, csum; 275 char *data; 276 u32 *hdr; 277 int i; 278 279 if (!initrd_end) 280 return NULL; 281 282 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN; 283 /* 284 * Since Grub may align the size of initrd to 4, we must 285 * check the preceding 3 bytes as well. 286 */ 287 for (i = 0; i < 4; i++) { 288 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN)) 289 goto found; 290 data--; 291 } 292 return NULL; 293 294 found: 295 hdr = (u32 *)(data - 8); 296 size = le32_to_cpu(hdr[0]); 297 csum = le32_to_cpu(hdr[1]); 298 299 data = ((void *)hdr) - size; 300 if ((unsigned long)data < initrd_start) { 301 pr_err("bootconfig size %d is greater than initrd size %ld\n", 302 size, initrd_end - initrd_start); 303 return NULL; 304 } 305 306 if (xbc_calc_checksum(data, size) != csum) { 307 pr_err("bootconfig checksum failed\n"); 308 return NULL; 309 } 310 311 /* Remove bootconfig from initramfs/initrd */ 312 initrd_end = (unsigned long)data; 313 if (_size) 314 *_size = size; 315 316 return data; 317 } 318 #else 319 static void * __init get_boot_config_from_initrd(size_t *_size) 320 { 321 return NULL; 322 } 323 #endif 324 325 #ifdef CONFIG_BOOT_CONFIG 326 327 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata; 328 329 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0) 330 331 static int __init xbc_snprint_cmdline(char *buf, size_t size, 332 struct xbc_node *root) 333 { 334 struct xbc_node *knode, *vnode; 335 char *end = buf + size; 336 const char *val; 337 int ret; 338 339 xbc_node_for_each_key_value(root, knode, val) { 340 ret = xbc_node_compose_key_after(root, knode, 341 xbc_namebuf, XBC_KEYLEN_MAX); 342 if (ret < 0) 343 return ret; 344 345 vnode = xbc_node_get_child(knode); 346 if (!vnode) { 347 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf); 348 if (ret < 0) 349 return ret; 350 buf += ret; 351 continue; 352 } 353 xbc_array_for_each_value(vnode, val) { 354 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ", 355 xbc_namebuf, val); 356 if (ret < 0) 357 return ret; 358 buf += ret; 359 } 360 } 361 362 return buf - (end - size); 363 } 364 #undef rest 365 366 /* Make an extra command line under given key word */ 367 static char * __init xbc_make_cmdline(const char *key) 368 { 369 struct xbc_node *root; 370 char *new_cmdline; 371 int ret, len = 0; 372 373 root = xbc_find_node(key); 374 if (!root) 375 return NULL; 376 377 /* Count required buffer size */ 378 len = xbc_snprint_cmdline(NULL, 0, root); 379 if (len <= 0) 380 return NULL; 381 382 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES); 383 if (!new_cmdline) { 384 pr_err("Failed to allocate memory for extra kernel cmdline.\n"); 385 return NULL; 386 } 387 388 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root); 389 if (ret < 0 || ret > len) { 390 pr_err("Failed to print extra kernel cmdline.\n"); 391 memblock_free(new_cmdline, len + 1); 392 return NULL; 393 } 394 395 return new_cmdline; 396 } 397 398 static int __init bootconfig_params(char *param, char *val, 399 const char *unused, void *arg) 400 { 401 if (strcmp(param, "bootconfig") == 0) { 402 bootconfig_found = true; 403 } 404 return 0; 405 } 406 407 static int __init warn_bootconfig(char *str) 408 { 409 /* The 'bootconfig' has been handled by bootconfig_params(). */ 410 return 0; 411 } 412 413 static void __init setup_boot_config(void) 414 { 415 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 416 const char *msg, *data; 417 int pos, ret; 418 size_t size; 419 char *err; 420 421 /* Cut out the bootconfig data even if we have no bootconfig option */ 422 data = get_boot_config_from_initrd(&size); 423 /* If there is no bootconfig in initrd, try embedded one. */ 424 if (!data) 425 data = xbc_get_embedded_bootconfig(&size); 426 427 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 428 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL, 429 bootconfig_params); 430 431 if (IS_ERR(err) || !bootconfig_found) 432 return; 433 434 /* parse_args() stops at the next param of '--' and returns an address */ 435 if (err) 436 initargs_offs = err - tmp_cmdline; 437 438 if (!data) { 439 pr_err("'bootconfig' found on command line, but no bootconfig found\n"); 440 return; 441 } 442 443 if (size >= XBC_DATA_MAX) { 444 pr_err("bootconfig size %ld greater than max size %d\n", 445 (long)size, XBC_DATA_MAX); 446 return; 447 } 448 449 ret = xbc_init(data, size, &msg, &pos); 450 if (ret < 0) { 451 if (pos < 0) 452 pr_err("Failed to init bootconfig: %s.\n", msg); 453 else 454 pr_err("Failed to parse bootconfig: %s at %d.\n", 455 msg, pos); 456 } else { 457 xbc_get_info(&ret, NULL); 458 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret); 459 /* keys starting with "kernel." are passed via cmdline */ 460 extra_command_line = xbc_make_cmdline("kernel"); 461 /* Also, "init." keys are init arguments */ 462 extra_init_args = xbc_make_cmdline("init"); 463 } 464 return; 465 } 466 467 static void __init exit_boot_config(void) 468 { 469 xbc_exit(); 470 } 471 472 #else /* !CONFIG_BOOT_CONFIG */ 473 474 static void __init setup_boot_config(void) 475 { 476 /* Remove bootconfig data from initrd */ 477 get_boot_config_from_initrd(NULL); 478 } 479 480 static int __init warn_bootconfig(char *str) 481 { 482 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n"); 483 return 0; 484 } 485 486 #define exit_boot_config() do {} while (0) 487 488 #endif /* CONFIG_BOOT_CONFIG */ 489 490 early_param("bootconfig", warn_bootconfig); 491 492 /* Change NUL term back to "=", to make "param" the whole string. */ 493 static void __init repair_env_string(char *param, char *val) 494 { 495 if (val) { 496 /* param=val or param="val"? */ 497 if (val == param+strlen(param)+1) 498 val[-1] = '='; 499 else if (val == param+strlen(param)+2) { 500 val[-2] = '='; 501 memmove(val-1, val, strlen(val)+1); 502 } else 503 BUG(); 504 } 505 } 506 507 /* Anything after -- gets handed straight to init. */ 508 static int __init set_init_arg(char *param, char *val, 509 const char *unused, void *arg) 510 { 511 unsigned int i; 512 513 if (panic_later) 514 return 0; 515 516 repair_env_string(param, val); 517 518 for (i = 0; argv_init[i]; i++) { 519 if (i == MAX_INIT_ARGS) { 520 panic_later = "init"; 521 panic_param = param; 522 return 0; 523 } 524 } 525 argv_init[i] = param; 526 return 0; 527 } 528 529 /* 530 * Unknown boot options get handed to init, unless they look like 531 * unused parameters (modprobe will find them in /proc/cmdline). 532 */ 533 static int __init unknown_bootoption(char *param, char *val, 534 const char *unused, void *arg) 535 { 536 size_t len = strlen(param); 537 538 repair_env_string(param, val); 539 540 /* Handle obsolete-style parameters */ 541 if (obsolete_checksetup(param)) 542 return 0; 543 544 /* Unused module parameter. */ 545 if (strnchr(param, len, '.')) 546 return 0; 547 548 if (panic_later) 549 return 0; 550 551 if (val) { 552 /* Environment option */ 553 unsigned int i; 554 for (i = 0; envp_init[i]; i++) { 555 if (i == MAX_INIT_ENVS) { 556 panic_later = "env"; 557 panic_param = param; 558 } 559 if (!strncmp(param, envp_init[i], len+1)) 560 break; 561 } 562 envp_init[i] = param; 563 } else { 564 /* Command line option */ 565 unsigned int i; 566 for (i = 0; argv_init[i]; i++) { 567 if (i == MAX_INIT_ARGS) { 568 panic_later = "init"; 569 panic_param = param; 570 } 571 } 572 argv_init[i] = param; 573 } 574 return 0; 575 } 576 577 static int __init init_setup(char *str) 578 { 579 unsigned int i; 580 581 execute_command = str; 582 /* 583 * In case LILO is going to boot us with default command line, 584 * it prepends "auto" before the whole cmdline which makes 585 * the shell think it should execute a script with such name. 586 * So we ignore all arguments entered _before_ init=... [MJ] 587 */ 588 for (i = 1; i < MAX_INIT_ARGS; i++) 589 argv_init[i] = NULL; 590 return 1; 591 } 592 __setup("init=", init_setup); 593 594 static int __init rdinit_setup(char *str) 595 { 596 unsigned int i; 597 598 ramdisk_execute_command = str; 599 /* See "auto" comment in init_setup */ 600 for (i = 1; i < MAX_INIT_ARGS; i++) 601 argv_init[i] = NULL; 602 return 1; 603 } 604 __setup("rdinit=", rdinit_setup); 605 606 #ifndef CONFIG_SMP 607 static const unsigned int setup_max_cpus = NR_CPUS; 608 static inline void setup_nr_cpu_ids(void) { } 609 static inline void smp_prepare_cpus(unsigned int maxcpus) { } 610 #endif 611 612 /* 613 * We need to store the untouched command line for future reference. 614 * We also need to store the touched command line since the parameter 615 * parsing is performed in place, and we should allow a component to 616 * store reference of name/value for future reference. 617 */ 618 static void __init setup_command_line(char *command_line) 619 { 620 size_t len, xlen = 0, ilen = 0; 621 622 if (extra_command_line) 623 xlen = strlen(extra_command_line); 624 if (extra_init_args) 625 ilen = strlen(extra_init_args) + 4; /* for " -- " */ 626 627 len = xlen + strlen(boot_command_line) + 1; 628 629 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES); 630 if (!saved_command_line) 631 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen); 632 633 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES); 634 if (!static_command_line) 635 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 636 637 if (xlen) { 638 /* 639 * We have to put extra_command_line before boot command 640 * lines because there could be dashes (separator of init 641 * command line) in the command lines. 642 */ 643 strcpy(saved_command_line, extra_command_line); 644 strcpy(static_command_line, extra_command_line); 645 } 646 strcpy(saved_command_line + xlen, boot_command_line); 647 strcpy(static_command_line + xlen, command_line); 648 649 if (ilen) { 650 /* 651 * Append supplemental init boot args to saved_command_line 652 * so that user can check what command line options passed 653 * to init. 654 * The order should always be 655 * " -- "[bootconfig init-param][cmdline init-param] 656 */ 657 if (initargs_offs) { 658 len = xlen + initargs_offs; 659 strcpy(saved_command_line + len, extra_init_args); 660 len += ilen - 4; /* strlen(extra_init_args) */ 661 strcpy(saved_command_line + len, 662 boot_command_line + initargs_offs - 1); 663 } else { 664 len = strlen(saved_command_line); 665 strcpy(saved_command_line + len, " -- "); 666 len += 4; 667 strcpy(saved_command_line + len, extra_init_args); 668 } 669 } 670 } 671 672 /* 673 * We need to finalize in a non-__init function or else race conditions 674 * between the root thread and the init thread may cause start_kernel to 675 * be reaped by free_initmem before the root thread has proceeded to 676 * cpu_idle. 677 * 678 * gcc-3.4 accidentally inlines this function, so use noinline. 679 */ 680 681 static __initdata DECLARE_COMPLETION(kthreadd_done); 682 683 noinline void __ref rest_init(void) 684 { 685 struct task_struct *tsk; 686 int pid; 687 688 rcu_scheduler_starting(); 689 /* 690 * We need to spawn init first so that it obtains pid 1, however 691 * the init task will end up wanting to create kthreads, which, if 692 * we schedule it before we create kthreadd, will OOPS. 693 */ 694 pid = user_mode_thread(kernel_init, NULL, CLONE_FS); 695 /* 696 * Pin init on the boot CPU. Task migration is not properly working 697 * until sched_init_smp() has been run. It will set the allowed 698 * CPUs for init to the non isolated CPUs. 699 */ 700 rcu_read_lock(); 701 tsk = find_task_by_pid_ns(pid, &init_pid_ns); 702 tsk->flags |= PF_NO_SETAFFINITY; 703 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); 704 rcu_read_unlock(); 705 706 numa_default_policy(); 707 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); 708 rcu_read_lock(); 709 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 710 rcu_read_unlock(); 711 712 /* 713 * Enable might_sleep() and smp_processor_id() checks. 714 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y 715 * kernel_thread() would trigger might_sleep() splats. With 716 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled 717 * already, but it's stuck on the kthreadd_done completion. 718 */ 719 system_state = SYSTEM_SCHEDULING; 720 721 complete(&kthreadd_done); 722 723 /* 724 * The boot idle thread must execute schedule() 725 * at least once to get things moving: 726 */ 727 schedule_preempt_disabled(); 728 /* Call into cpu_idle with preempt disabled */ 729 cpu_startup_entry(CPUHP_ONLINE); 730 } 731 732 /* Check for early params. */ 733 static int __init do_early_param(char *param, char *val, 734 const char *unused, void *arg) 735 { 736 const struct obs_kernel_param *p; 737 738 for (p = __setup_start; p < __setup_end; p++) { 739 if ((p->early && parameq(param, p->str)) || 740 (strcmp(param, "console") == 0 && 741 strcmp(p->str, "earlycon") == 0) 742 ) { 743 if (p->setup_func(val) != 0) 744 pr_warn("Malformed early option '%s'\n", param); 745 } 746 } 747 /* We accept everything at this stage. */ 748 return 0; 749 } 750 751 void __init parse_early_options(char *cmdline) 752 { 753 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, 754 do_early_param); 755 } 756 757 /* Arch code calls this early on, or if not, just before other parsing. */ 758 void __init parse_early_param(void) 759 { 760 static int done __initdata; 761 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 762 763 if (done) 764 return; 765 766 /* All fall through to do_early_param. */ 767 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 768 parse_early_options(tmp_cmdline); 769 done = 1; 770 } 771 772 void __init __weak arch_post_acpi_subsys_init(void) { } 773 774 void __init __weak smp_setup_processor_id(void) 775 { 776 } 777 778 # if THREAD_SIZE >= PAGE_SIZE 779 void __init __weak thread_stack_cache_init(void) 780 { 781 } 782 #endif 783 784 void __init __weak mem_encrypt_init(void) { } 785 786 void __init __weak poking_init(void) { } 787 788 void __init __weak pgtable_cache_init(void) { } 789 790 void __init __weak trap_init(void) { } 791 792 bool initcall_debug; 793 core_param(initcall_debug, initcall_debug, bool, 0644); 794 795 #ifdef TRACEPOINTS_ENABLED 796 static void __init initcall_debug_enable(void); 797 #else 798 static inline void initcall_debug_enable(void) 799 { 800 } 801 #endif 802 803 /* Report memory auto-initialization states for this boot. */ 804 static void __init report_meminit(void) 805 { 806 const char *stack; 807 808 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN)) 809 stack = "all(pattern)"; 810 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO)) 811 stack = "all(zero)"; 812 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL)) 813 stack = "byref_all(zero)"; 814 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF)) 815 stack = "byref(zero)"; 816 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER)) 817 stack = "__user(zero)"; 818 else 819 stack = "off"; 820 821 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n", 822 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off", 823 want_init_on_free() ? "on" : "off"); 824 if (want_init_on_free()) 825 pr_info("mem auto-init: clearing system memory may take some time...\n"); 826 } 827 828 /* 829 * Set up kernel memory allocators 830 */ 831 static void __init mm_init(void) 832 { 833 /* 834 * page_ext requires contiguous pages, 835 * bigger than MAX_ORDER unless SPARSEMEM. 836 */ 837 page_ext_init_flatmem(); 838 init_mem_debugging_and_hardening(); 839 kfence_alloc_pool(); 840 report_meminit(); 841 kmsan_init_shadow(); 842 stack_depot_early_init(); 843 mem_init(); 844 mem_init_print_info(); 845 kmem_cache_init(); 846 /* 847 * page_owner must be initialized after buddy is ready, and also after 848 * slab is ready so that stack_depot_init() works properly 849 */ 850 page_ext_init_flatmem_late(); 851 kmemleak_init(); 852 pgtable_init(); 853 debug_objects_mem_init(); 854 vmalloc_init(); 855 /* Should be run after vmap initialization */ 856 if (early_page_ext_enabled()) 857 page_ext_init(); 858 /* Should be run before the first non-init thread is created */ 859 init_espfix_bsp(); 860 /* Should be run after espfix64 is set up. */ 861 pti_init(); 862 kmsan_init_runtime(); 863 } 864 865 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET 866 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT, 867 randomize_kstack_offset); 868 DEFINE_PER_CPU(u32, kstack_offset); 869 870 static int __init early_randomize_kstack_offset(char *buf) 871 { 872 int ret; 873 bool bool_result; 874 875 ret = kstrtobool(buf, &bool_result); 876 if (ret) 877 return ret; 878 879 if (bool_result) 880 static_branch_enable(&randomize_kstack_offset); 881 else 882 static_branch_disable(&randomize_kstack_offset); 883 return 0; 884 } 885 early_param("randomize_kstack_offset", early_randomize_kstack_offset); 886 #endif 887 888 void __init __weak arch_call_rest_init(void) 889 { 890 rest_init(); 891 } 892 893 static void __init print_unknown_bootoptions(void) 894 { 895 char *unknown_options; 896 char *end; 897 const char *const *p; 898 size_t len; 899 900 if (panic_later || (!argv_init[1] && !envp_init[2])) 901 return; 902 903 /* 904 * Determine how many options we have to print out, plus a space 905 * before each 906 */ 907 len = 1; /* null terminator */ 908 for (p = &argv_init[1]; *p; p++) { 909 len++; 910 len += strlen(*p); 911 } 912 for (p = &envp_init[2]; *p; p++) { 913 len++; 914 len += strlen(*p); 915 } 916 917 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES); 918 if (!unknown_options) { 919 pr_err("%s: Failed to allocate %zu bytes\n", 920 __func__, len); 921 return; 922 } 923 end = unknown_options; 924 925 for (p = &argv_init[1]; *p; p++) 926 end += sprintf(end, " %s", *p); 927 for (p = &envp_init[2]; *p; p++) 928 end += sprintf(end, " %s", *p); 929 930 /* Start at unknown_options[1] to skip the initial space */ 931 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n", 932 &unknown_options[1]); 933 memblock_free(unknown_options, len); 934 } 935 936 asmlinkage __visible void __init __no_sanitize_address start_kernel(void) 937 { 938 char *command_line; 939 char *after_dashes; 940 941 set_task_stack_end_magic(&init_task); 942 smp_setup_processor_id(); 943 debug_objects_early_init(); 944 init_vmlinux_build_id(); 945 946 cgroup_init_early(); 947 948 local_irq_disable(); 949 early_boot_irqs_disabled = true; 950 951 /* 952 * Interrupts are still disabled. Do necessary setups, then 953 * enable them. 954 */ 955 boot_cpu_init(); 956 page_address_init(); 957 pr_notice("%s", linux_banner); 958 early_security_init(); 959 setup_arch(&command_line); 960 setup_boot_config(); 961 setup_command_line(command_line); 962 setup_nr_cpu_ids(); 963 setup_per_cpu_areas(); 964 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 965 boot_cpu_hotplug_init(); 966 967 build_all_zonelists(NULL); 968 page_alloc_init(); 969 970 pr_notice("Kernel command line: %s\n", saved_command_line); 971 /* parameters may set static keys */ 972 jump_label_init(); 973 parse_early_param(); 974 after_dashes = parse_args("Booting kernel", 975 static_command_line, __start___param, 976 __stop___param - __start___param, 977 -1, -1, NULL, &unknown_bootoption); 978 print_unknown_bootoptions(); 979 if (!IS_ERR_OR_NULL(after_dashes)) 980 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 981 NULL, set_init_arg); 982 if (extra_init_args) 983 parse_args("Setting extra init args", extra_init_args, 984 NULL, 0, -1, -1, NULL, set_init_arg); 985 986 /* 987 * These use large bootmem allocations and must precede 988 * kmem_cache_init() 989 */ 990 setup_log_buf(0); 991 vfs_caches_init_early(); 992 sort_main_extable(); 993 trap_init(); 994 mm_init(); 995 996 ftrace_init(); 997 998 /* trace_printk can be enabled here */ 999 early_trace_init(); 1000 1001 /* 1002 * Set up the scheduler prior starting any interrupts (such as the 1003 * timer interrupt). Full topology setup happens at smp_init() 1004 * time - but meanwhile we still have a functioning scheduler. 1005 */ 1006 sched_init(); 1007 1008 if (WARN(!irqs_disabled(), 1009 "Interrupts were enabled *very* early, fixing it\n")) 1010 local_irq_disable(); 1011 radix_tree_init(); 1012 maple_tree_init(); 1013 1014 /* 1015 * Set up housekeeping before setting up workqueues to allow the unbound 1016 * workqueue to take non-housekeeping into account. 1017 */ 1018 housekeeping_init(); 1019 1020 /* 1021 * Allow workqueue creation and work item queueing/cancelling 1022 * early. Work item execution depends on kthreads and starts after 1023 * workqueue_init(). 1024 */ 1025 workqueue_init_early(); 1026 1027 rcu_init(); 1028 1029 /* Trace events are available after this */ 1030 trace_init(); 1031 1032 if (initcall_debug) 1033 initcall_debug_enable(); 1034 1035 context_tracking_init(); 1036 /* init some links before init_ISA_irqs() */ 1037 early_irq_init(); 1038 init_IRQ(); 1039 tick_init(); 1040 rcu_init_nohz(); 1041 init_timers(); 1042 srcu_init(); 1043 hrtimers_init(); 1044 softirq_init(); 1045 timekeeping_init(); 1046 kfence_init(); 1047 time_init(); 1048 1049 /* 1050 * For best initial stack canary entropy, prepare it after: 1051 * - setup_arch() for any UEFI RNG entropy and boot cmdline access 1052 * - timekeeping_init() for ktime entropy used in random_init() 1053 * - time_init() for making random_get_entropy() work on some platforms 1054 * - random_init() to initialize the RNG from from early entropy sources 1055 */ 1056 random_init(command_line); 1057 boot_init_stack_canary(); 1058 1059 perf_event_init(); 1060 profile_init(); 1061 call_function_init(); 1062 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 1063 1064 early_boot_irqs_disabled = false; 1065 local_irq_enable(); 1066 1067 kmem_cache_init_late(); 1068 1069 /* 1070 * HACK ALERT! This is early. We're enabling the console before 1071 * we've done PCI setups etc, and console_init() must be aware of 1072 * this. But we do want output early, in case something goes wrong. 1073 */ 1074 console_init(); 1075 if (panic_later) 1076 panic("Too many boot %s vars at `%s'", panic_later, 1077 panic_param); 1078 1079 lockdep_init(); 1080 1081 /* 1082 * Need to run this when irqs are enabled, because it wants 1083 * to self-test [hard/soft]-irqs on/off lock inversion bugs 1084 * too: 1085 */ 1086 locking_selftest(); 1087 1088 /* 1089 * This needs to be called before any devices perform DMA 1090 * operations that might use the SWIOTLB bounce buffers. It will 1091 * mark the bounce buffers as decrypted so that their usage will 1092 * not cause "plain-text" data to be decrypted when accessed. 1093 */ 1094 mem_encrypt_init(); 1095 1096 #ifdef CONFIG_BLK_DEV_INITRD 1097 if (initrd_start && !initrd_below_start_ok && 1098 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 1099 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 1100 page_to_pfn(virt_to_page((void *)initrd_start)), 1101 min_low_pfn); 1102 initrd_start = 0; 1103 } 1104 #endif 1105 setup_per_cpu_pageset(); 1106 numa_policy_init(); 1107 acpi_early_init(); 1108 if (late_time_init) 1109 late_time_init(); 1110 sched_clock_init(); 1111 calibrate_delay(); 1112 pid_idr_init(); 1113 anon_vma_init(); 1114 #ifdef CONFIG_X86 1115 if (efi_enabled(EFI_RUNTIME_SERVICES)) 1116 efi_enter_virtual_mode(); 1117 #endif 1118 thread_stack_cache_init(); 1119 cred_init(); 1120 fork_init(); 1121 proc_caches_init(); 1122 uts_ns_init(); 1123 key_init(); 1124 security_init(); 1125 dbg_late_init(); 1126 net_ns_init(); 1127 vfs_caches_init(); 1128 pagecache_init(); 1129 signals_init(); 1130 seq_file_init(); 1131 proc_root_init(); 1132 nsfs_init(); 1133 cpuset_init(); 1134 cgroup_init(); 1135 taskstats_init_early(); 1136 delayacct_init(); 1137 1138 poking_init(); 1139 check_bugs(); 1140 1141 acpi_subsystem_init(); 1142 arch_post_acpi_subsys_init(); 1143 kcsan_init(); 1144 1145 /* Do the rest non-__init'ed, we're now alive */ 1146 arch_call_rest_init(); 1147 1148 prevent_tail_call_optimization(); 1149 } 1150 1151 /* Call all constructor functions linked into the kernel. */ 1152 static void __init do_ctors(void) 1153 { 1154 /* 1155 * For UML, the constructors have already been called by the 1156 * normal setup code as it's just a normal ELF binary, so we 1157 * cannot do it again - but we do need CONFIG_CONSTRUCTORS 1158 * even on UML for modules. 1159 */ 1160 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML) 1161 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 1162 1163 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 1164 (*fn)(); 1165 #endif 1166 } 1167 1168 #ifdef CONFIG_KALLSYMS 1169 struct blacklist_entry { 1170 struct list_head next; 1171 char *buf; 1172 }; 1173 1174 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 1175 1176 static int __init initcall_blacklist(char *str) 1177 { 1178 char *str_entry; 1179 struct blacklist_entry *entry; 1180 1181 /* str argument is a comma-separated list of functions */ 1182 do { 1183 str_entry = strsep(&str, ","); 1184 if (str_entry) { 1185 pr_debug("blacklisting initcall %s\n", str_entry); 1186 entry = memblock_alloc(sizeof(*entry), 1187 SMP_CACHE_BYTES); 1188 if (!entry) 1189 panic("%s: Failed to allocate %zu bytes\n", 1190 __func__, sizeof(*entry)); 1191 entry->buf = memblock_alloc(strlen(str_entry) + 1, 1192 SMP_CACHE_BYTES); 1193 if (!entry->buf) 1194 panic("%s: Failed to allocate %zu bytes\n", 1195 __func__, strlen(str_entry) + 1); 1196 strcpy(entry->buf, str_entry); 1197 list_add(&entry->next, &blacklisted_initcalls); 1198 } 1199 } while (str_entry); 1200 1201 return 1; 1202 } 1203 1204 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1205 { 1206 struct blacklist_entry *entry; 1207 char fn_name[KSYM_SYMBOL_LEN]; 1208 unsigned long addr; 1209 1210 if (list_empty(&blacklisted_initcalls)) 1211 return false; 1212 1213 addr = (unsigned long) dereference_function_descriptor(fn); 1214 sprint_symbol_no_offset(fn_name, addr); 1215 1216 /* 1217 * fn will be "function_name [module_name]" where [module_name] is not 1218 * displayed for built-in init functions. Strip off the [module_name]. 1219 */ 1220 strreplace(fn_name, ' ', '\0'); 1221 1222 list_for_each_entry(entry, &blacklisted_initcalls, next) { 1223 if (!strcmp(fn_name, entry->buf)) { 1224 pr_debug("initcall %s blacklisted\n", fn_name); 1225 return true; 1226 } 1227 } 1228 1229 return false; 1230 } 1231 #else 1232 static int __init initcall_blacklist(char *str) 1233 { 1234 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 1235 return 0; 1236 } 1237 1238 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1239 { 1240 return false; 1241 } 1242 #endif 1243 __setup("initcall_blacklist=", initcall_blacklist); 1244 1245 static __init_or_module void 1246 trace_initcall_start_cb(void *data, initcall_t fn) 1247 { 1248 ktime_t *calltime = (ktime_t *)data; 1249 1250 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current)); 1251 *calltime = ktime_get(); 1252 } 1253 1254 static __init_or_module void 1255 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 1256 { 1257 ktime_t rettime, *calltime = (ktime_t *)data; 1258 1259 rettime = ktime_get(); 1260 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n", 1261 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime)); 1262 } 1263 1264 static ktime_t initcall_calltime; 1265 1266 #ifdef TRACEPOINTS_ENABLED 1267 static void __init initcall_debug_enable(void) 1268 { 1269 int ret; 1270 1271 ret = register_trace_initcall_start(trace_initcall_start_cb, 1272 &initcall_calltime); 1273 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 1274 &initcall_calltime); 1275 WARN(ret, "Failed to register initcall tracepoints\n"); 1276 } 1277 # define do_trace_initcall_start trace_initcall_start 1278 # define do_trace_initcall_finish trace_initcall_finish 1279 #else 1280 static inline void do_trace_initcall_start(initcall_t fn) 1281 { 1282 if (!initcall_debug) 1283 return; 1284 trace_initcall_start_cb(&initcall_calltime, fn); 1285 } 1286 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 1287 { 1288 if (!initcall_debug) 1289 return; 1290 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 1291 } 1292 #endif /* !TRACEPOINTS_ENABLED */ 1293 1294 int __init_or_module do_one_initcall(initcall_t fn) 1295 { 1296 int count = preempt_count(); 1297 char msgbuf[64]; 1298 int ret; 1299 1300 if (initcall_blacklisted(fn)) 1301 return -EPERM; 1302 1303 do_trace_initcall_start(fn); 1304 ret = fn(); 1305 do_trace_initcall_finish(fn, ret); 1306 1307 msgbuf[0] = 0; 1308 1309 if (preempt_count() != count) { 1310 sprintf(msgbuf, "preemption imbalance "); 1311 preempt_count_set(count); 1312 } 1313 if (irqs_disabled()) { 1314 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 1315 local_irq_enable(); 1316 } 1317 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 1318 1319 add_latent_entropy(); 1320 return ret; 1321 } 1322 1323 1324 extern initcall_entry_t __initcall_start[]; 1325 extern initcall_entry_t __initcall0_start[]; 1326 extern initcall_entry_t __initcall1_start[]; 1327 extern initcall_entry_t __initcall2_start[]; 1328 extern initcall_entry_t __initcall3_start[]; 1329 extern initcall_entry_t __initcall4_start[]; 1330 extern initcall_entry_t __initcall5_start[]; 1331 extern initcall_entry_t __initcall6_start[]; 1332 extern initcall_entry_t __initcall7_start[]; 1333 extern initcall_entry_t __initcall_end[]; 1334 1335 static initcall_entry_t *initcall_levels[] __initdata = { 1336 __initcall0_start, 1337 __initcall1_start, 1338 __initcall2_start, 1339 __initcall3_start, 1340 __initcall4_start, 1341 __initcall5_start, 1342 __initcall6_start, 1343 __initcall7_start, 1344 __initcall_end, 1345 }; 1346 1347 /* Keep these in sync with initcalls in include/linux/init.h */ 1348 static const char *initcall_level_names[] __initdata = { 1349 "pure", 1350 "core", 1351 "postcore", 1352 "arch", 1353 "subsys", 1354 "fs", 1355 "device", 1356 "late", 1357 }; 1358 1359 static int __init ignore_unknown_bootoption(char *param, char *val, 1360 const char *unused, void *arg) 1361 { 1362 return 0; 1363 } 1364 1365 static void __init do_initcall_level(int level, char *command_line) 1366 { 1367 initcall_entry_t *fn; 1368 1369 parse_args(initcall_level_names[level], 1370 command_line, __start___param, 1371 __stop___param - __start___param, 1372 level, level, 1373 NULL, ignore_unknown_bootoption); 1374 1375 trace_initcall_level(initcall_level_names[level]); 1376 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1377 do_one_initcall(initcall_from_entry(fn)); 1378 } 1379 1380 static void __init do_initcalls(void) 1381 { 1382 int level; 1383 size_t len = strlen(saved_command_line) + 1; 1384 char *command_line; 1385 1386 command_line = kzalloc(len, GFP_KERNEL); 1387 if (!command_line) 1388 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 1389 1390 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { 1391 /* Parser modifies command_line, restore it each time */ 1392 strcpy(command_line, saved_command_line); 1393 do_initcall_level(level, command_line); 1394 } 1395 1396 kfree(command_line); 1397 } 1398 1399 /* 1400 * Ok, the machine is now initialized. None of the devices 1401 * have been touched yet, but the CPU subsystem is up and 1402 * running, and memory and process management works. 1403 * 1404 * Now we can finally start doing some real work.. 1405 */ 1406 static void __init do_basic_setup(void) 1407 { 1408 cpuset_init_smp(); 1409 driver_init(); 1410 init_irq_proc(); 1411 do_ctors(); 1412 do_initcalls(); 1413 } 1414 1415 static void __init do_pre_smp_initcalls(void) 1416 { 1417 initcall_entry_t *fn; 1418 1419 trace_initcall_level("early"); 1420 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1421 do_one_initcall(initcall_from_entry(fn)); 1422 } 1423 1424 static int run_init_process(const char *init_filename) 1425 { 1426 const char *const *p; 1427 1428 argv_init[0] = init_filename; 1429 pr_info("Run %s as init process\n", init_filename); 1430 pr_debug(" with arguments:\n"); 1431 for (p = argv_init; *p; p++) 1432 pr_debug(" %s\n", *p); 1433 pr_debug(" with environment:\n"); 1434 for (p = envp_init; *p; p++) 1435 pr_debug(" %s\n", *p); 1436 return kernel_execve(init_filename, argv_init, envp_init); 1437 } 1438 1439 static int try_to_run_init_process(const char *init_filename) 1440 { 1441 int ret; 1442 1443 ret = run_init_process(init_filename); 1444 1445 if (ret && ret != -ENOENT) { 1446 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1447 init_filename, ret); 1448 } 1449 1450 return ret; 1451 } 1452 1453 static noinline void __init kernel_init_freeable(void); 1454 1455 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1456 bool rodata_enabled __ro_after_init = true; 1457 1458 #ifndef arch_parse_debug_rodata 1459 static inline bool arch_parse_debug_rodata(char *str) { return false; } 1460 #endif 1461 1462 static int __init set_debug_rodata(char *str) 1463 { 1464 if (arch_parse_debug_rodata(str)) 1465 return 0; 1466 1467 if (str && !strcmp(str, "on")) 1468 rodata_enabled = true; 1469 else if (str && !strcmp(str, "off")) 1470 rodata_enabled = false; 1471 else 1472 pr_warn("Invalid option string for rodata: '%s'\n", str); 1473 return 0; 1474 } 1475 early_param("rodata", set_debug_rodata); 1476 #endif 1477 1478 #ifdef CONFIG_STRICT_KERNEL_RWX 1479 static void mark_readonly(void) 1480 { 1481 if (rodata_enabled) { 1482 /* 1483 * load_module() results in W+X mappings, which are cleaned 1484 * up with call_rcu(). Let's make sure that queued work is 1485 * flushed so that we don't hit false positives looking for 1486 * insecure pages which are W+X. 1487 */ 1488 rcu_barrier(); 1489 mark_rodata_ro(); 1490 rodata_test(); 1491 } else 1492 pr_info("Kernel memory protection disabled.\n"); 1493 } 1494 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX) 1495 static inline void mark_readonly(void) 1496 { 1497 pr_warn("Kernel memory protection not selected by kernel config.\n"); 1498 } 1499 #else 1500 static inline void mark_readonly(void) 1501 { 1502 pr_warn("This architecture does not have kernel memory protection.\n"); 1503 } 1504 #endif 1505 1506 void __weak free_initmem(void) 1507 { 1508 free_initmem_default(POISON_FREE_INITMEM); 1509 } 1510 1511 static int __ref kernel_init(void *unused) 1512 { 1513 int ret; 1514 1515 /* 1516 * Wait until kthreadd is all set-up. 1517 */ 1518 wait_for_completion(&kthreadd_done); 1519 1520 kernel_init_freeable(); 1521 /* need to finish all async __init code before freeing the memory */ 1522 async_synchronize_full(); 1523 1524 system_state = SYSTEM_FREEING_INITMEM; 1525 kprobe_free_init_mem(); 1526 ftrace_free_init_mem(); 1527 kgdb_free_init_mem(); 1528 exit_boot_config(); 1529 free_initmem(); 1530 mark_readonly(); 1531 1532 /* 1533 * Kernel mappings are now finalized - update the userspace page-table 1534 * to finalize PTI. 1535 */ 1536 pti_finalize(); 1537 1538 system_state = SYSTEM_RUNNING; 1539 numa_default_policy(); 1540 1541 rcu_end_inkernel_boot(); 1542 1543 do_sysctl_args(); 1544 1545 if (ramdisk_execute_command) { 1546 ret = run_init_process(ramdisk_execute_command); 1547 if (!ret) 1548 return 0; 1549 pr_err("Failed to execute %s (error %d)\n", 1550 ramdisk_execute_command, ret); 1551 } 1552 1553 /* 1554 * We try each of these until one succeeds. 1555 * 1556 * The Bourne shell can be used instead of init if we are 1557 * trying to recover a really broken machine. 1558 */ 1559 if (execute_command) { 1560 ret = run_init_process(execute_command); 1561 if (!ret) 1562 return 0; 1563 panic("Requested init %s failed (error %d).", 1564 execute_command, ret); 1565 } 1566 1567 if (CONFIG_DEFAULT_INIT[0] != '\0') { 1568 ret = run_init_process(CONFIG_DEFAULT_INIT); 1569 if (ret) 1570 pr_err("Default init %s failed (error %d)\n", 1571 CONFIG_DEFAULT_INIT, ret); 1572 else 1573 return 0; 1574 } 1575 1576 if (!try_to_run_init_process("/sbin/init") || 1577 !try_to_run_init_process("/etc/init") || 1578 !try_to_run_init_process("/bin/init") || 1579 !try_to_run_init_process("/bin/sh")) 1580 return 0; 1581 1582 panic("No working init found. Try passing init= option to kernel. " 1583 "See Linux Documentation/admin-guide/init.rst for guidance."); 1584 } 1585 1586 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */ 1587 void __init console_on_rootfs(void) 1588 { 1589 struct file *file = filp_open("/dev/console", O_RDWR, 0); 1590 1591 if (IS_ERR(file)) { 1592 pr_err("Warning: unable to open an initial console.\n"); 1593 return; 1594 } 1595 init_dup(file); 1596 init_dup(file); 1597 init_dup(file); 1598 fput(file); 1599 } 1600 1601 static noinline void __init kernel_init_freeable(void) 1602 { 1603 /* Now the scheduler is fully set up and can do blocking allocations */ 1604 gfp_allowed_mask = __GFP_BITS_MASK; 1605 1606 /* 1607 * init can allocate pages on any node 1608 */ 1609 set_mems_allowed(node_states[N_MEMORY]); 1610 1611 cad_pid = get_pid(task_pid(current)); 1612 1613 smp_prepare_cpus(setup_max_cpus); 1614 1615 workqueue_init(); 1616 1617 init_mm_internals(); 1618 1619 rcu_init_tasks_generic(); 1620 do_pre_smp_initcalls(); 1621 lockup_detector_init(); 1622 1623 smp_init(); 1624 sched_init_smp(); 1625 1626 padata_init(); 1627 page_alloc_init_late(); 1628 /* Initialize page ext after all struct pages are initialized. */ 1629 if (!early_page_ext_enabled()) 1630 page_ext_init(); 1631 1632 do_basic_setup(); 1633 1634 kunit_run_all_tests(); 1635 1636 wait_for_initramfs(); 1637 console_on_rootfs(); 1638 1639 /* 1640 * check if there is an early userspace init. If yes, let it do all 1641 * the work 1642 */ 1643 if (init_eaccess(ramdisk_execute_command) != 0) { 1644 ramdisk_execute_command = NULL; 1645 prepare_namespace(); 1646 } 1647 1648 /* 1649 * Ok, we have completed the initial bootup, and 1650 * we're essentially up and running. Get rid of the 1651 * initmem segments and start the user-mode stuff.. 1652 * 1653 * rootfs is available now, try loading the public keys 1654 * and default modules 1655 */ 1656 1657 integrity_load_keys(); 1658 } 1659