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/export.h> 17 #include <linux/extable.h> 18 #include <linux/module.h> 19 #include <linux/proc_fs.h> 20 #include <linux/binfmts.h> 21 #include <linux/kernel.h> 22 #include <linux/syscalls.h> 23 #include <linux/stackprotector.h> 24 #include <linux/string.h> 25 #include <linux/ctype.h> 26 #include <linux/delay.h> 27 #include <linux/ioport.h> 28 #include <linux/init.h> 29 #include <linux/initrd.h> 30 #include <linux/memblock.h> 31 #include <linux/acpi.h> 32 #include <linux/bootconfig.h> 33 #include <linux/console.h> 34 #include <linux/nmi.h> 35 #include <linux/percpu.h> 36 #include <linux/kmod.h> 37 #include <linux/kprobes.h> 38 #include <linux/kmsan.h> 39 #include <linux/ksysfs.h> 40 #include <linux/vmalloc.h> 41 #include <linux/kernel_stat.h> 42 #include <linux/start_kernel.h> 43 #include <linux/security.h> 44 #include <linux/smp.h> 45 #include <linux/profile.h> 46 #include <linux/kfence.h> 47 #include <linux/rcupdate.h> 48 #include <linux/srcu.h> 49 #include <linux/moduleparam.h> 50 #include <linux/kallsyms.h> 51 #include <linux/buildid.h> 52 #include <linux/writeback.h> 53 #include <linux/cpu.h> 54 #include <linux/cpuset.h> 55 #include <linux/memcontrol.h> 56 #include <linux/cgroup.h> 57 #include <linux/tick.h> 58 #include <linux/sched/isolation.h> 59 #include <linux/interrupt.h> 60 #include <linux/taskstats_kern.h> 61 #include <linux/delayacct.h> 62 #include <linux/unistd.h> 63 #include <linux/utsname.h> 64 #include <linux/rmap.h> 65 #include <linux/mempolicy.h> 66 #include <linux/key.h> 67 #include <linux/debug_locks.h> 68 #include <linux/debugobjects.h> 69 #include <linux/lockdep.h> 70 #include <linux/kmemleak.h> 71 #include <linux/padata.h> 72 #include <linux/pid_namespace.h> 73 #include <linux/device/driver.h> 74 #include <linux/kthread.h> 75 #include <linux/sched.h> 76 #include <linux/sched/init.h> 77 #include <linux/signal.h> 78 #include <linux/idr.h> 79 #include <linux/kgdb.h> 80 #include <linux/ftrace.h> 81 #include <linux/async.h> 82 #include <linux/shmem_fs.h> 83 #include <linux/slab.h> 84 #include <linux/perf_event.h> 85 #include <linux/ptrace.h> 86 #include <linux/pti.h> 87 #include <linux/blkdev.h> 88 #include <linux/sched/clock.h> 89 #include <linux/sched/task.h> 90 #include <linux/sched/task_stack.h> 91 #include <linux/context_tracking.h> 92 #include <linux/random.h> 93 #include <linux/moduleloader.h> 94 #include <linux/list.h> 95 #include <linux/integrity.h> 96 #include <linux/proc_ns.h> 97 #include <linux/io.h> 98 #include <linux/cache.h> 99 #include <linux/rodata_test.h> 100 #include <linux/jump_label.h> 101 #include <linux/kcsan.h> 102 #include <linux/init_syscalls.h> 103 #include <linux/stackdepot.h> 104 #include <linux/randomize_kstack.h> 105 #include <linux/pidfs.h> 106 #include <linux/ptdump.h> 107 #include <linux/time_namespace.h> 108 #include <linux/unaligned.h> 109 #include <linux/vdso_datastore.h> 110 #include <net/net_namespace.h> 111 112 #include <asm/io.h> 113 #include <asm/setup.h> 114 #include <asm/sections.h> 115 #include <asm/cacheflush.h> 116 117 #define CREATE_TRACE_POINTS 118 #include <trace/events/initcall.h> 119 120 #include <kunit/test.h> 121 122 static int kernel_init(void *); 123 124 /* 125 * Debug helper: via this flag we know that we are in 'early bootup code' 126 * where only the boot processor is running with IRQ disabled. This means 127 * two things - IRQ must not be enabled before the flag is cleared and some 128 * operations which are not allowed with IRQ disabled are allowed while the 129 * flag is set. 130 */ 131 bool early_boot_irqs_disabled __read_mostly; 132 133 enum system_states system_state __read_mostly; 134 EXPORT_SYMBOL(system_state); 135 136 /* 137 * Boot command-line arguments 138 */ 139 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT 140 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT 141 142 /* 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 __ro_after_init; 149 unsigned int saved_command_line_len __ro_after_init; 150 /* Command line for parameter parsing */ 151 static char *static_command_line; 152 /* Untouched extra command line */ 153 static char *extra_command_line; 154 /* Extra init arguments */ 155 static char *extra_init_args; 156 157 #ifdef CONFIG_BOOT_CONFIG 158 /* Is bootconfig on command line? */ 159 static bool bootconfig_found; 160 static size_t initargs_offs; 161 #else 162 # define bootconfig_found false 163 # define initargs_offs 0 164 #endif 165 166 static char *execute_command; 167 static char *ramdisk_execute_command = "/init"; 168 static bool __initdata ramdisk_execute_command_set; 169 170 /* 171 * Used to generate warnings if static_key manipulation functions are used 172 * before jump_label_init is called. 173 */ 174 bool static_key_initialized __read_mostly; 175 EXPORT_SYMBOL_GPL(static_key_initialized); 176 177 /* 178 * If set, this is an indication to the drivers that reset the underlying 179 * device before going ahead with the initialization otherwise driver might 180 * rely on the BIOS and skip the reset operation. 181 * 182 * This is useful if kernel is booting in an unreliable environment. 183 * For ex. kdump situation where previous kernel has crashed, BIOS has been 184 * skipped and devices will be in unknown state. 185 */ 186 unsigned int reset_devices; 187 EXPORT_SYMBOL(reset_devices); 188 189 static int __init set_reset_devices(char *str) 190 { 191 reset_devices = 1; 192 return 1; 193 } 194 195 __setup("reset_devices", set_reset_devices); 196 197 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 198 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 199 static const char *panic_later, *panic_param; 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 u8 *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 = (u8 *)(data - 8); 296 size = get_unaligned_le32(hdr); 297 csum = get_unaligned_le32(hdr + 4); 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 /* Make an extra command line under given key word */ 328 static char * __init xbc_make_cmdline(const char *key) 329 { 330 struct xbc_node *root; 331 char *new_cmdline; 332 int ret, len = 0; 333 334 root = xbc_find_node(key); 335 if (!root) 336 return NULL; 337 338 /* Count required buffer size */ 339 len = xbc_snprint_cmdline(NULL, 0, root); 340 if (len <= 0) 341 return NULL; 342 343 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES); 344 if (!new_cmdline) { 345 pr_err("Failed to allocate memory for extra kernel cmdline.\n"); 346 return NULL; 347 } 348 349 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root); 350 if (ret < 0 || ret > len) { 351 pr_err("Failed to print extra kernel cmdline.\n"); 352 memblock_free(new_cmdline, len + 1); 353 return NULL; 354 } 355 356 return new_cmdline; 357 } 358 359 static int __init bootconfig_params(char *param, char *val, 360 const char *unused, void *arg) 361 { 362 if (strcmp(param, "bootconfig") == 0) { 363 bootconfig_found = true; 364 } 365 return 0; 366 } 367 368 static int __init warn_bootconfig(char *str) 369 { 370 /* The 'bootconfig' has been handled by bootconfig_params(). */ 371 return 0; 372 } 373 374 static void __init setup_boot_config(void) 375 { 376 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 377 const char *msg, *data; 378 int pos, ret; 379 size_t size; 380 char *err; 381 382 /* Cut out the bootconfig data even if we have no bootconfig option */ 383 data = get_boot_config_from_initrd(&size); 384 /* If there is no bootconfig in initrd, try embedded one. */ 385 if (!data) 386 data = xbc_get_embedded_bootconfig(&size); 387 388 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 389 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL, 390 bootconfig_params); 391 392 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE))) 393 return; 394 395 /* parse_args() stops at the next param of '--' and returns an address */ 396 if (err) 397 initargs_offs = err - tmp_cmdline; 398 399 if (!data) { 400 /* If user intended to use bootconfig, show an error level message */ 401 if (bootconfig_found) 402 pr_err("'bootconfig' found on command line, but no bootconfig found\n"); 403 else 404 pr_info("No bootconfig data provided, so skipping bootconfig"); 405 return; 406 } 407 408 if (size >= XBC_DATA_MAX) { 409 pr_err("bootconfig size %ld greater than max size %d\n", 410 (long)size, XBC_DATA_MAX); 411 return; 412 } 413 414 ret = xbc_init(data, size, &msg, &pos); 415 if (ret < 0) { 416 if (pos < 0) 417 pr_err("Failed to init bootconfig: %s.\n", msg); 418 else 419 pr_err("Failed to parse bootconfig: %s at %d.\n", 420 msg, pos); 421 } else { 422 xbc_get_info(&ret, NULL); 423 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret); 424 /* keys starting with "kernel." are passed via cmdline */ 425 extra_command_line = xbc_make_cmdline("kernel"); 426 /* Also, "init." keys are init arguments */ 427 extra_init_args = xbc_make_cmdline("init"); 428 } 429 return; 430 } 431 432 static void __init exit_boot_config(void) 433 { 434 xbc_exit(); 435 } 436 437 #else /* !CONFIG_BOOT_CONFIG */ 438 439 static void __init setup_boot_config(void) 440 { 441 /* Remove bootconfig data from initrd */ 442 get_boot_config_from_initrd(NULL); 443 } 444 445 static int __init warn_bootconfig(char *str) 446 { 447 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n"); 448 return 0; 449 } 450 451 #define exit_boot_config() do {} while (0) 452 453 #endif /* CONFIG_BOOT_CONFIG */ 454 455 early_param("bootconfig", warn_bootconfig); 456 457 bool __init cmdline_has_extra_options(void) 458 { 459 return extra_command_line || extra_init_args; 460 } 461 462 /* Change NUL term back to "=", to make "param" the whole string. */ 463 static void __init repair_env_string(char *param, char *val) 464 { 465 if (val) { 466 /* param=val or param="val"? */ 467 if (val == param+strlen(param)+1) 468 val[-1] = '='; 469 else if (val == param+strlen(param)+2) { 470 val[-2] = '='; 471 memmove(val-1, val, strlen(val)+1); 472 } else 473 BUG(); 474 } 475 } 476 477 /* Anything after -- gets handed straight to init. */ 478 static int __init set_init_arg(char *param, char *val, 479 const char *unused, void *arg) 480 { 481 unsigned int i; 482 483 if (panic_later) 484 return 0; 485 486 repair_env_string(param, val); 487 488 for (i = 0; argv_init[i]; i++) { 489 if (i == MAX_INIT_ARGS) { 490 panic_later = "init"; 491 panic_param = param; 492 return 0; 493 } 494 } 495 argv_init[i] = param; 496 return 0; 497 } 498 499 /* 500 * Unknown boot options get handed to init, unless they look like 501 * unused parameters (modprobe will find them in /proc/cmdline). 502 */ 503 static int __init unknown_bootoption(char *param, char *val, 504 const char *unused, void *arg) 505 { 506 size_t len = strlen(param); 507 /* 508 * Well-known bootloader identifiers: 509 * 1. LILO/Grub pass "BOOT_IMAGE=..."; 510 * 2. kexec/kdump (kexec-tools) pass "kexec". 511 */ 512 const char *bootloader[] = { "BOOT_IMAGE=", "kexec", NULL }; 513 514 /* Handle params aliased to sysctls */ 515 if (sysctl_is_alias(param)) 516 return 0; 517 518 repair_env_string(param, val); 519 520 /* Handle bootloader identifier */ 521 for (int i = 0; bootloader[i]; i++) { 522 if (strstarts(param, bootloader[i])) 523 return 0; 524 } 525 526 /* Handle obsolete-style parameters */ 527 if (obsolete_checksetup(param)) 528 return 0; 529 530 /* Unused module parameter. */ 531 if (strnchr(param, len, '.')) 532 return 0; 533 534 if (panic_later) 535 return 0; 536 537 if (val) { 538 /* Environment option */ 539 unsigned int i; 540 for (i = 0; envp_init[i]; i++) { 541 if (i == MAX_INIT_ENVS) { 542 panic_later = "env"; 543 panic_param = param; 544 } 545 if (!strncmp(param, envp_init[i], len+1)) 546 break; 547 } 548 envp_init[i] = param; 549 } else { 550 /* Command line option */ 551 unsigned int i; 552 for (i = 0; argv_init[i]; i++) { 553 if (i == MAX_INIT_ARGS) { 554 panic_later = "init"; 555 panic_param = param; 556 } 557 } 558 argv_init[i] = param; 559 } 560 return 0; 561 } 562 563 static int __init init_setup(char *str) 564 { 565 unsigned int i; 566 567 execute_command = str; 568 /* 569 * In case LILO is going to boot us with default command line, 570 * it prepends "auto" before the whole cmdline which makes 571 * the shell think it should execute a script with such name. 572 * So we ignore all arguments entered _before_ init=... [MJ] 573 */ 574 for (i = 1; i < MAX_INIT_ARGS; i++) 575 argv_init[i] = NULL; 576 return 1; 577 } 578 __setup("init=", init_setup); 579 580 static int __init rdinit_setup(char *str) 581 { 582 unsigned int i; 583 584 ramdisk_execute_command = str; 585 ramdisk_execute_command_set = true; 586 /* See "auto" comment in init_setup */ 587 for (i = 1; i < MAX_INIT_ARGS; i++) 588 argv_init[i] = NULL; 589 return 1; 590 } 591 __setup("rdinit=", rdinit_setup); 592 593 #ifndef CONFIG_SMP 594 static inline void setup_nr_cpu_ids(void) { } 595 static inline void smp_prepare_cpus(unsigned int maxcpus) { } 596 #endif 597 598 /* 599 * We need to store the untouched command line for future reference. 600 * We also need to store the touched command line since the parameter 601 * parsing is performed in place, and we should allow a component to 602 * store reference of name/value for future reference. 603 */ 604 static void __init setup_command_line(char *command_line) 605 { 606 size_t len, xlen = 0, ilen = 0; 607 608 if (extra_command_line) 609 xlen = strlen(extra_command_line); 610 if (extra_init_args) { 611 extra_init_args = strim(extra_init_args); /* remove trailing space */ 612 ilen = strlen(extra_init_args) + 4; /* for " -- " */ 613 } 614 615 len = xlen + strlen(boot_command_line) + ilen + 1; 616 617 saved_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES); 618 619 len = xlen + strlen(command_line) + 1; 620 621 static_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES); 622 623 if (xlen) { 624 /* 625 * We have to put extra_command_line before boot command 626 * lines because there could be dashes (separator of init 627 * command line) in the command lines. 628 */ 629 strcpy(saved_command_line, extra_command_line); 630 strcpy(static_command_line, extra_command_line); 631 } 632 strcpy(saved_command_line + xlen, boot_command_line); 633 strcpy(static_command_line + xlen, command_line); 634 635 if (ilen) { 636 /* 637 * Append supplemental init boot args to saved_command_line 638 * so that user can check what command line options passed 639 * to init. 640 * The order should always be 641 * " -- "[bootconfig init-param][cmdline init-param] 642 */ 643 if (initargs_offs) { 644 len = xlen + initargs_offs; 645 strcpy(saved_command_line + len, extra_init_args); 646 len += ilen - 4; /* strlen(extra_init_args) */ 647 strcpy(saved_command_line + len, 648 boot_command_line + initargs_offs - 1); 649 } else { 650 len = strlen(saved_command_line); 651 strcpy(saved_command_line + len, " -- "); 652 len += 4; 653 strcpy(saved_command_line + len, extra_init_args); 654 } 655 } 656 657 saved_command_line_len = strlen(saved_command_line); 658 } 659 660 /* 661 * We need to finalize in a non-__init function or else race conditions 662 * between the root thread and the init thread may cause start_kernel to 663 * be reaped by free_initmem before the root thread has proceeded to 664 * cpu_idle. 665 * 666 * gcc-3.4 accidentally inlines this function, so use noinline. 667 */ 668 669 static __initdata DECLARE_COMPLETION(kthreadd_done); 670 671 static noinline void __ref __noreturn rest_init(void) 672 { 673 struct task_struct *tsk; 674 int pid; 675 676 rcu_scheduler_starting(); 677 /* 678 * We need to spawn init first so that it obtains pid 1, however 679 * the init task will end up wanting to create kthreads, which, if 680 * we schedule it before we create kthreadd, will OOPS. 681 */ 682 pid = user_mode_thread(kernel_init, NULL, CLONE_FS); 683 /* 684 * Pin init on the boot CPU. Task migration is not properly working 685 * until sched_init_smp() has been run. It will set the allowed 686 * CPUs for init to the non isolated CPUs. 687 */ 688 rcu_read_lock(); 689 tsk = find_task_by_pid_ns(pid, &init_pid_ns); 690 tsk->flags |= PF_NO_SETAFFINITY; 691 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); 692 rcu_read_unlock(); 693 694 numa_default_policy(); 695 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES); 696 rcu_read_lock(); 697 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 698 rcu_read_unlock(); 699 700 /* 701 * Enable might_sleep() and smp_processor_id() checks. 702 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y 703 * kernel_thread() would trigger might_sleep() splats. With 704 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled 705 * already, but it's stuck on the kthreadd_done completion. 706 */ 707 system_state = SYSTEM_SCHEDULING; 708 709 complete(&kthreadd_done); 710 711 /* 712 * The boot idle thread must execute schedule() 713 * at least once to get things moving: 714 */ 715 schedule_preempt_disabled(); 716 /* Call into cpu_idle with preempt disabled */ 717 cpu_startup_entry(CPUHP_ONLINE); 718 } 719 720 /* Check for early params. */ 721 static int __init do_early_param(char *param, char *val, 722 const char *unused, void *arg) 723 { 724 const struct obs_kernel_param *p; 725 726 for (p = __setup_start; p < __setup_end; p++) { 727 if (p->early && parameq(param, p->str)) { 728 if (p->setup_func(val) != 0) 729 pr_warn("Malformed early option '%s'\n", param); 730 } 731 } 732 /* We accept everything at this stage. */ 733 return 0; 734 } 735 736 void __init parse_early_options(char *cmdline) 737 { 738 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, 739 do_early_param); 740 } 741 742 /* Arch code calls this early on, or if not, just before other parsing. */ 743 void __init parse_early_param(void) 744 { 745 static int done __initdata; 746 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 747 748 if (done) 749 return; 750 751 /* All fall through to do_early_param. */ 752 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 753 parse_early_options(tmp_cmdline); 754 done = 1; 755 } 756 757 void __init __weak arch_post_acpi_subsys_init(void) { } 758 759 void __init __weak smp_setup_processor_id(void) 760 { 761 } 762 763 void __init __weak smp_prepare_boot_cpu(void) 764 { 765 } 766 767 # if THREAD_SIZE >= PAGE_SIZE 768 void __init __weak thread_stack_cache_init(void) 769 { 770 } 771 #endif 772 773 void __init __weak poking_init(void) { } 774 775 void __init __weak pgtable_cache_init(void) { } 776 777 void __init __weak trap_init(void) { } 778 779 bool initcall_debug; 780 core_param(initcall_debug, initcall_debug, bool, 0644); 781 782 #ifdef TRACEPOINTS_ENABLED 783 static void __init initcall_debug_enable(void); 784 #else 785 static inline void initcall_debug_enable(void) 786 { 787 } 788 #endif 789 790 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET 791 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT, 792 randomize_kstack_offset); 793 DEFINE_PER_CPU(struct rnd_state, kstack_rnd_state); 794 795 static int __init random_kstack_init(void) 796 { 797 prandom_seed_full_state(&kstack_rnd_state); 798 return 0; 799 } 800 late_initcall(random_kstack_init); 801 802 static int __init early_randomize_kstack_offset(char *buf) 803 { 804 int ret; 805 bool bool_result; 806 807 ret = kstrtobool(buf, &bool_result); 808 if (ret) 809 return ret; 810 811 if (bool_result) 812 static_branch_enable(&randomize_kstack_offset); 813 else 814 static_branch_disable(&randomize_kstack_offset); 815 return 0; 816 } 817 early_param("randomize_kstack_offset", early_randomize_kstack_offset); 818 #endif 819 820 static void __init print_unknown_bootoptions(void) 821 { 822 char *unknown_options; 823 char *end; 824 const char *const *p; 825 size_t len; 826 827 if (panic_later || (!argv_init[1] && !envp_init[2])) 828 return; 829 830 /* 831 * Determine how many options we have to print out, plus a space 832 * before each 833 */ 834 len = 1; /* null terminator */ 835 for (p = &argv_init[1]; *p; p++) { 836 len++; 837 len += strlen(*p); 838 } 839 for (p = &envp_init[2]; *p; p++) { 840 len++; 841 len += strlen(*p); 842 } 843 844 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES); 845 if (!unknown_options) { 846 pr_err("%s: Failed to allocate %zu bytes\n", 847 __func__, len); 848 return; 849 } 850 end = unknown_options; 851 852 for (p = &argv_init[1]; *p; p++) 853 end += sprintf(end, " %s", *p); 854 for (p = &envp_init[2]; *p; p++) 855 end += sprintf(end, " %s", *p); 856 857 /* Start at unknown_options[1] to skip the initial space */ 858 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n", 859 &unknown_options[1]); 860 memblock_free(unknown_options, len); 861 } 862 863 static void __init early_numa_node_init(void) 864 { 865 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID 866 #ifndef cpu_to_node 867 int cpu; 868 869 /* The early_cpu_to_node() should be ready here. */ 870 for_each_possible_cpu(cpu) 871 set_cpu_numa_node(cpu, early_cpu_to_node(cpu)); 872 #endif 873 #endif 874 } 875 876 #define KERNEL_CMDLINE_PREFIX "Kernel command line: " 877 #define KERNEL_CMDLINE_PREFIX_LEN (sizeof(KERNEL_CMDLINE_PREFIX) - 1) 878 #define KERNEL_CMDLINE_CONTINUATION " \\" 879 #define KERNEL_CMDLINE_CONTINUATION_LEN (sizeof(KERNEL_CMDLINE_CONTINUATION) - 1) 880 881 #define MIN_CMDLINE_LOG_WRAP_IDEAL_LEN (KERNEL_CMDLINE_PREFIX_LEN + \ 882 KERNEL_CMDLINE_CONTINUATION_LEN) 883 #define CMDLINE_LOG_WRAP_IDEAL_LEN (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN > \ 884 MIN_CMDLINE_LOG_WRAP_IDEAL_LEN ? \ 885 CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN : \ 886 MIN_CMDLINE_LOG_WRAP_IDEAL_LEN) 887 888 #define IDEAL_CMDLINE_LEN (CMDLINE_LOG_WRAP_IDEAL_LEN - KERNEL_CMDLINE_PREFIX_LEN) 889 #define IDEAL_CMDLINE_SPLIT_LEN (IDEAL_CMDLINE_LEN - KERNEL_CMDLINE_CONTINUATION_LEN) 890 891 /** 892 * print_kernel_cmdline() - Print the kernel cmdline with wrapping. 893 * @cmdline: The cmdline to print. 894 * 895 * Print the kernel command line, trying to wrap based on the Kconfig knob 896 * CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN. 897 * 898 * Wrapping is based on spaces, ignoring quotes. All lines are prefixed 899 * with "Kernel command line: " and lines that are not the last line have 900 * a " \" suffix added to them. The prefix and suffix count towards the 901 * line length for wrapping purposes. The ideal length will be exceeded 902 * if no appropriate place to wrap is found. 903 * 904 * Example output if CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN is 40: 905 * Kernel command line: loglevel=7 \ 906 * Kernel command line: init=/sbin/init \ 907 * Kernel command line: root=PARTUUID=8c3efc1a-768b-6642-8d0c-89eb782f19f0/PARTNROFF=1 \ 908 * Kernel command line: rootwait ro \ 909 * Kernel command line: my_quoted_arg="The \ 910 * Kernel command line: quick brown fox \ 911 * Kernel command line: jumps over the \ 912 * Kernel command line: lazy dog." 913 */ 914 static void __init print_kernel_cmdline(const char *cmdline) 915 { 916 size_t len; 917 918 /* Config option of 0 or anything longer than the max disables wrapping */ 919 if (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN == 0 || 920 IDEAL_CMDLINE_LEN >= COMMAND_LINE_SIZE - 1) { 921 pr_notice("%s%s\n", KERNEL_CMDLINE_PREFIX, cmdline); 922 return; 923 } 924 925 len = strlen(cmdline); 926 while (len > IDEAL_CMDLINE_LEN) { 927 const char *first_space; 928 const char *prev_cutoff; 929 const char *cutoff; 930 int to_print; 931 size_t used; 932 933 /* Find the last ' ' that wouldn't make the line too long */ 934 prev_cutoff = NULL; 935 cutoff = cmdline; 936 while (true) { 937 cutoff = strchr(cutoff + 1, ' '); 938 if (!cutoff || cutoff - cmdline > IDEAL_CMDLINE_SPLIT_LEN) 939 break; 940 prev_cutoff = cutoff; 941 } 942 if (prev_cutoff) 943 cutoff = prev_cutoff; 944 else if (!cutoff) 945 break; 946 947 /* Find the beginning and end of the string of spaces */ 948 first_space = cutoff; 949 while (first_space > cmdline && first_space[-1] == ' ') 950 first_space--; 951 to_print = first_space - cmdline; 952 while (*cutoff == ' ') 953 cutoff++; 954 used = cutoff - cmdline; 955 956 /* If the whole string is used, break and do the final printout */ 957 if (len == used) 958 break; 959 960 if (to_print) 961 pr_notice("%s%.*s%s\n", KERNEL_CMDLINE_PREFIX, 962 to_print, cmdline, KERNEL_CMDLINE_CONTINUATION); 963 964 len -= used; 965 cmdline += used; 966 } 967 if (len) 968 pr_notice("%s%s\n", KERNEL_CMDLINE_PREFIX, cmdline); 969 } 970 971 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector 972 void start_kernel(void) 973 { 974 char *command_line; 975 char *after_dashes; 976 977 set_task_stack_end_magic(&init_task); 978 smp_setup_processor_id(); 979 debug_objects_early_init(); 980 init_vmlinux_build_id(); 981 982 cgroup_init_early(); 983 984 local_irq_disable(); 985 early_boot_irqs_disabled = true; 986 987 /* 988 * Interrupts are still disabled. Do necessary setups, then 989 * enable them. 990 */ 991 boot_cpu_init(); 992 page_address_init(); 993 pr_notice("%s", linux_banner); 994 setup_arch(&command_line); 995 mm_core_init_early(); 996 /* Static keys and static calls are needed by LSMs */ 997 jump_label_init(); 998 static_call_init(); 999 early_security_init(); 1000 setup_boot_config(); 1001 setup_command_line(command_line); 1002 setup_nr_cpu_ids(); 1003 setup_per_cpu_areas(); 1004 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 1005 early_numa_node_init(); 1006 boot_cpu_hotplug_init(); 1007 1008 print_kernel_cmdline(saved_command_line); 1009 /* parameters may set static keys */ 1010 parse_early_param(); 1011 after_dashes = parse_args("Booting kernel", 1012 static_command_line, __start___param, 1013 __stop___param - __start___param, 1014 -1, -1, NULL, &unknown_bootoption); 1015 print_unknown_bootoptions(); 1016 if (!IS_ERR_OR_NULL(after_dashes)) 1017 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 1018 NULL, set_init_arg); 1019 if (extra_init_args) 1020 parse_args("Setting extra init args", extra_init_args, 1021 NULL, 0, -1, -1, NULL, set_init_arg); 1022 1023 /* Architectural and non-timekeeping rng init, before allocator init */ 1024 random_init_early(command_line); 1025 1026 /* 1027 * These use large bootmem allocations and must precede 1028 * initalization of page allocator 1029 */ 1030 setup_log_buf(0); 1031 vfs_caches_init_early(); 1032 sort_main_extable(); 1033 trap_init(); 1034 mm_core_init(); 1035 maple_tree_init(); 1036 poking_init(); 1037 ftrace_init(); 1038 1039 /* trace_printk can be enabled here */ 1040 early_trace_init(); 1041 1042 /* 1043 * Set up the scheduler prior starting any interrupts (such as the 1044 * timer interrupt). Full topology setup happens at smp_init() 1045 * time - but meanwhile we still have a functioning scheduler. 1046 */ 1047 sched_init(); 1048 1049 if (WARN(!irqs_disabled(), 1050 "Interrupts were enabled *very* early, fixing it\n")) 1051 local_irq_disable(); 1052 radix_tree_init(); 1053 1054 /* 1055 * Set up housekeeping before setting up workqueues to allow the unbound 1056 * workqueue to take non-housekeeping into account. 1057 */ 1058 housekeeping_init(); 1059 1060 /* 1061 * Allow workqueue creation and work item queueing/cancelling 1062 * early. Work item execution depends on kthreads and starts after 1063 * workqueue_init(). 1064 */ 1065 workqueue_init_early(); 1066 1067 rcu_init(); 1068 kvfree_rcu_init(); 1069 1070 /* Trace events are available after this */ 1071 trace_init(); 1072 1073 if (initcall_debug) 1074 initcall_debug_enable(); 1075 1076 context_tracking_init(); 1077 /* init some links before init_ISA_irqs() */ 1078 early_irq_init(); 1079 init_IRQ(); 1080 tick_init(); 1081 rcu_init_nohz(); 1082 timers_init(); 1083 srcu_init(); 1084 hrtimers_init(); 1085 softirq_init(); 1086 vdso_setup_data_pages(); 1087 timekeeping_init(); 1088 time_init(); 1089 1090 /* This must be after timekeeping is initialized */ 1091 random_init(); 1092 1093 /* These make use of the fully initialized rng */ 1094 kfence_init(); 1095 boot_init_stack_canary(); 1096 1097 perf_event_init(); 1098 profile_init(); 1099 call_function_init(); 1100 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 1101 1102 early_boot_irqs_disabled = false; 1103 local_irq_enable(); 1104 1105 kmem_cache_init_late(); 1106 1107 /* 1108 * HACK ALERT! This is early. We're enabling the console before 1109 * we've done PCI setups etc, and console_init() must be aware of 1110 * this. But we do want output early, in case something goes wrong. 1111 */ 1112 console_init(); 1113 if (panic_later) 1114 panic("Too many boot %s vars at `%s'", panic_later, 1115 panic_param); 1116 1117 lockdep_init(); 1118 1119 /* 1120 * Need to run this when irqs are enabled, because it wants 1121 * to self-test [hard/soft]-irqs on/off lock inversion bugs 1122 * too: 1123 */ 1124 locking_selftest(); 1125 1126 #ifdef CONFIG_BLK_DEV_INITRD 1127 if (initrd_start && !initrd_below_start_ok && 1128 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 1129 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 1130 page_to_pfn(virt_to_page((void *)initrd_start)), 1131 min_low_pfn); 1132 initrd_start = 0; 1133 } 1134 #endif 1135 setup_per_cpu_pageset(); 1136 numa_policy_init(); 1137 acpi_early_init(); 1138 if (late_time_init) 1139 late_time_init(); 1140 sched_clock_init(); 1141 calibrate_delay(); 1142 1143 arch_cpu_finalize_init(); 1144 1145 pid_idr_init(); 1146 anon_vma_init(); 1147 thread_stack_cache_init(); 1148 cred_init(); 1149 fork_init(); 1150 proc_caches_init(); 1151 uts_ns_init(); 1152 time_ns_init(); 1153 key_init(); 1154 security_init(); 1155 dbg_late_init(); 1156 net_ns_init(); 1157 vfs_caches_init(); 1158 pagecache_init(); 1159 signals_init(); 1160 seq_file_init(); 1161 proc_root_init(); 1162 nsfs_init(); 1163 pidfs_init(); 1164 cpuset_init(); 1165 mem_cgroup_init(); 1166 cgroup_init(); 1167 taskstats_init_early(); 1168 delayacct_init(); 1169 1170 acpi_subsystem_init(); 1171 arch_post_acpi_subsys_init(); 1172 kcsan_init(); 1173 1174 /* Do the rest non-__init'ed, we're now alive */ 1175 rest_init(); 1176 1177 /* 1178 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10 1179 * and older. 1180 */ 1181 #if !__has_attribute(__no_stack_protector__) 1182 prevent_tail_call_optimization(); 1183 #endif 1184 } 1185 1186 /* Call all constructor functions linked into the kernel. */ 1187 static void __init do_ctors(void) 1188 { 1189 /* 1190 * For UML, the constructors have already been called by the 1191 * normal setup code as it's just a normal ELF binary, so we 1192 * cannot do it again - but we do need CONFIG_CONSTRUCTORS 1193 * even on UML for modules. 1194 */ 1195 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML) 1196 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 1197 1198 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 1199 (*fn)(); 1200 #endif 1201 } 1202 1203 #ifdef CONFIG_KALLSYMS 1204 struct blacklist_entry { 1205 struct list_head next; 1206 char *buf; 1207 }; 1208 1209 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 1210 1211 static int __init initcall_blacklist(char *str) 1212 { 1213 char *str_entry; 1214 struct blacklist_entry *entry; 1215 1216 /* str argument is a comma-separated list of functions */ 1217 do { 1218 str_entry = strsep(&str, ","); 1219 if (str_entry) { 1220 pr_debug("blacklisting initcall %s\n", str_entry); 1221 entry = memblock_alloc_or_panic(sizeof(*entry), 1222 SMP_CACHE_BYTES); 1223 entry->buf = memblock_alloc_or_panic(strlen(str_entry) + 1, 1224 SMP_CACHE_BYTES); 1225 strcpy(entry->buf, str_entry); 1226 list_add(&entry->next, &blacklisted_initcalls); 1227 } 1228 } while (str_entry); 1229 1230 return 1; 1231 } 1232 1233 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1234 { 1235 struct blacklist_entry *entry; 1236 char fn_name[KSYM_SYMBOL_LEN]; 1237 unsigned long addr; 1238 1239 if (list_empty(&blacklisted_initcalls)) 1240 return false; 1241 1242 addr = (unsigned long) dereference_function_descriptor(fn); 1243 sprint_symbol_no_offset(fn_name, addr); 1244 1245 /* 1246 * fn will be "function_name [module_name]" where [module_name] is not 1247 * displayed for built-in init functions. Strip off the [module_name]. 1248 */ 1249 strreplace(fn_name, ' ', '\0'); 1250 1251 list_for_each_entry(entry, &blacklisted_initcalls, next) { 1252 if (!strcmp(fn_name, entry->buf)) { 1253 pr_debug("initcall %s blacklisted\n", fn_name); 1254 return true; 1255 } 1256 } 1257 1258 return false; 1259 } 1260 #else 1261 static int __init initcall_blacklist(char *str) 1262 { 1263 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 1264 return 0; 1265 } 1266 1267 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1268 { 1269 return false; 1270 } 1271 #endif 1272 __setup("initcall_blacklist=", initcall_blacklist); 1273 1274 static __init_or_module void 1275 trace_initcall_start_cb(void *data, initcall_t fn) 1276 { 1277 ktime_t *calltime = data; 1278 1279 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current)); 1280 *calltime = ktime_get(); 1281 } 1282 1283 static __init_or_module void 1284 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 1285 { 1286 ktime_t rettime, *calltime = data; 1287 1288 rettime = ktime_get(); 1289 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n", 1290 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime)); 1291 } 1292 1293 static __init_or_module void 1294 trace_initcall_level_cb(void *data, const char *level) 1295 { 1296 printk(KERN_DEBUG "entering initcall level: %s\n", level); 1297 } 1298 1299 static ktime_t initcall_calltime; 1300 1301 #ifdef TRACEPOINTS_ENABLED 1302 static void __init initcall_debug_enable(void) 1303 { 1304 int ret; 1305 1306 ret = register_trace_initcall_start(trace_initcall_start_cb, 1307 &initcall_calltime); 1308 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 1309 &initcall_calltime); 1310 ret |= register_trace_initcall_level(trace_initcall_level_cb, NULL); 1311 WARN(ret, "Failed to register initcall tracepoints\n"); 1312 } 1313 # define do_trace_initcall_start trace_initcall_start 1314 # define do_trace_initcall_finish trace_initcall_finish 1315 # define do_trace_initcall_level trace_initcall_level 1316 #else 1317 static inline void do_trace_initcall_start(initcall_t fn) 1318 { 1319 if (!initcall_debug) 1320 return; 1321 trace_initcall_start_cb(&initcall_calltime, fn); 1322 } 1323 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 1324 { 1325 if (!initcall_debug) 1326 return; 1327 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 1328 } 1329 static inline void do_trace_initcall_level(const char *level) 1330 { 1331 if (!initcall_debug) 1332 return; 1333 trace_initcall_level_cb(NULL, level); 1334 } 1335 #endif /* !TRACEPOINTS_ENABLED */ 1336 1337 int __init_or_module do_one_initcall(initcall_t fn) 1338 { 1339 int count = preempt_count(); 1340 char msgbuf[64]; 1341 int ret; 1342 1343 if (initcall_blacklisted(fn)) 1344 return -EPERM; 1345 1346 do_trace_initcall_start(fn); 1347 ret = fn(); 1348 do_trace_initcall_finish(fn, ret); 1349 1350 msgbuf[0] = 0; 1351 1352 if (preempt_count() != count) { 1353 sprintf(msgbuf, "preemption imbalance "); 1354 preempt_count_set(count); 1355 } 1356 if (irqs_disabled()) { 1357 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 1358 local_irq_enable(); 1359 } 1360 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 1361 1362 add_latent_entropy(); 1363 return ret; 1364 } 1365 1366 1367 static initcall_entry_t *initcall_levels[] __initdata = { 1368 __initcall0_start, 1369 __initcall1_start, 1370 __initcall2_start, 1371 __initcall3_start, 1372 __initcall4_start, 1373 __initcall5_start, 1374 __initcall6_start, 1375 __initcall7_start, 1376 __initcall_end, 1377 }; 1378 1379 /* Keep these in sync with initcalls in include/linux/init.h */ 1380 static const char *initcall_level_names[] __initdata = { 1381 "pure", 1382 "core", 1383 "postcore", 1384 "arch", 1385 "subsys", 1386 "fs", 1387 "device", 1388 "late", 1389 }; 1390 1391 static int __init ignore_unknown_bootoption(char *param, char *val, 1392 const char *unused, void *arg) 1393 { 1394 return 0; 1395 } 1396 1397 static void __init do_initcall_level(int level, char *command_line) 1398 { 1399 initcall_entry_t *fn; 1400 1401 parse_args(initcall_level_names[level], 1402 command_line, __start___param, 1403 __stop___param - __start___param, 1404 level, level, 1405 NULL, ignore_unknown_bootoption); 1406 1407 do_trace_initcall_level(initcall_level_names[level]); 1408 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1409 do_one_initcall(initcall_from_entry(fn)); 1410 } 1411 1412 static void __init do_initcalls(void) 1413 { 1414 int level; 1415 size_t len = saved_command_line_len + 1; 1416 char *command_line; 1417 1418 command_line = kzalloc(len, GFP_KERNEL); 1419 if (!command_line) 1420 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 1421 1422 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { 1423 /* Parser modifies command_line, restore it each time */ 1424 strcpy(command_line, saved_command_line); 1425 do_initcall_level(level, command_line); 1426 } 1427 1428 kfree(command_line); 1429 } 1430 1431 /* 1432 * Ok, the machine is now initialized. None of the devices 1433 * have been touched yet, but the CPU subsystem is up and 1434 * running, and memory and process management works. 1435 * 1436 * Now we can finally start doing some real work.. 1437 */ 1438 static void __init do_basic_setup(void) 1439 { 1440 cpuset_init_smp(); 1441 ksysfs_init(); 1442 driver_init(); 1443 init_irq_proc(); 1444 do_ctors(); 1445 do_initcalls(); 1446 } 1447 1448 static void __init do_pre_smp_initcalls(void) 1449 { 1450 initcall_entry_t *fn; 1451 1452 do_trace_initcall_level("early"); 1453 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1454 do_one_initcall(initcall_from_entry(fn)); 1455 } 1456 1457 static int run_init_process(const char *init_filename) 1458 { 1459 const char *const *p; 1460 1461 argv_init[0] = init_filename; 1462 pr_info("Run %s as init process\n", init_filename); 1463 pr_debug(" with arguments:\n"); 1464 for (p = argv_init; *p; p++) 1465 pr_debug(" %s\n", *p); 1466 pr_debug(" with environment:\n"); 1467 for (p = envp_init; *p; p++) 1468 pr_debug(" %s\n", *p); 1469 return kernel_execve(init_filename, argv_init, envp_init); 1470 } 1471 1472 static int try_to_run_init_process(const char *init_filename) 1473 { 1474 int ret; 1475 1476 ret = run_init_process(init_filename); 1477 1478 if (ret && ret != -ENOENT) { 1479 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1480 init_filename, ret); 1481 } 1482 1483 return ret; 1484 } 1485 1486 static noinline void __init kernel_init_freeable(void); 1487 1488 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1489 bool rodata_enabled __ro_after_init = true; 1490 1491 #ifndef arch_parse_debug_rodata 1492 static inline bool arch_parse_debug_rodata(char *str) { return false; } 1493 #endif 1494 1495 static int __init set_debug_rodata(char *str) 1496 { 1497 if (arch_parse_debug_rodata(str)) 1498 return 0; 1499 1500 if (str && !strcmp(str, "on")) 1501 rodata_enabled = true; 1502 else if (str && !strcmp(str, "off")) 1503 rodata_enabled = false; 1504 else 1505 pr_warn("Invalid option string for rodata: '%s'\n", str); 1506 return 0; 1507 } 1508 early_param("rodata", set_debug_rodata); 1509 #endif 1510 1511 static void mark_readonly(void) 1512 { 1513 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) { 1514 /* 1515 * load_module() results in W+X mappings, which are cleaned 1516 * up with init_free_wq. Let's make sure that queued work is 1517 * flushed so that we don't hit false positives looking for 1518 * insecure pages which are W+X. 1519 */ 1520 flush_module_init_free_work(); 1521 jump_label_init_ro(); 1522 mark_rodata_ro(); 1523 debug_checkwx(); 1524 rodata_test(); 1525 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) { 1526 pr_info("Kernel memory protection disabled.\n"); 1527 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) { 1528 pr_warn("Kernel memory protection not selected by kernel config.\n"); 1529 } else { 1530 pr_warn("This architecture does not have kernel memory protection.\n"); 1531 } 1532 } 1533 1534 void __weak free_initmem(void) 1535 { 1536 free_initmem_default(POISON_FREE_INITMEM); 1537 } 1538 1539 static int __ref kernel_init(void *unused) 1540 { 1541 int ret; 1542 1543 /* 1544 * Wait until kthreadd is all set-up. 1545 */ 1546 wait_for_completion(&kthreadd_done); 1547 1548 kernel_init_freeable(); 1549 /* need to finish all async __init code before freeing the memory */ 1550 async_synchronize_full(); 1551 1552 system_state = SYSTEM_FREEING_INITMEM; 1553 kprobe_free_init_mem(); 1554 ftrace_free_init_mem(); 1555 kgdb_free_init_mem(); 1556 exit_boot_config(); 1557 free_initmem(); 1558 mark_readonly(); 1559 1560 /* 1561 * Kernel mappings are now finalized - update the userspace page-table 1562 * to finalize PTI. 1563 */ 1564 pti_finalize(); 1565 1566 system_state = SYSTEM_RUNNING; 1567 numa_default_policy(); 1568 1569 rcu_end_inkernel_boot(); 1570 1571 do_sysctl_args(); 1572 1573 if (ramdisk_execute_command) { 1574 ret = run_init_process(ramdisk_execute_command); 1575 if (!ret) 1576 return 0; 1577 pr_err("Failed to execute %s (error %d)\n", 1578 ramdisk_execute_command, ret); 1579 } 1580 1581 /* 1582 * We try each of these until one succeeds. 1583 * 1584 * The Bourne shell can be used instead of init if we are 1585 * trying to recover a really broken machine. 1586 */ 1587 if (execute_command) { 1588 ret = run_init_process(execute_command); 1589 if (!ret) 1590 return 0; 1591 panic("Requested init %s failed (error %d).", 1592 execute_command, ret); 1593 } 1594 1595 if (CONFIG_DEFAULT_INIT[0] != '\0') { 1596 ret = run_init_process(CONFIG_DEFAULT_INIT); 1597 if (ret) 1598 pr_err("Default init %s failed (error %d)\n", 1599 CONFIG_DEFAULT_INIT, ret); 1600 else 1601 return 0; 1602 } 1603 1604 if (!try_to_run_init_process("/sbin/init") || 1605 !try_to_run_init_process("/etc/init") || 1606 !try_to_run_init_process("/bin/init") || 1607 !try_to_run_init_process("/bin/sh")) 1608 return 0; 1609 1610 panic("No working init found. Try passing init= option to kernel. " 1611 "See Linux Documentation/admin-guide/init.rst for guidance."); 1612 } 1613 1614 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */ 1615 void __init console_on_rootfs(void) 1616 { 1617 struct file *file = filp_open("/dev/console", O_RDWR, 0); 1618 1619 if (IS_ERR(file)) { 1620 pr_err("Warning: unable to open an initial console.\n"); 1621 return; 1622 } 1623 init_dup(file); 1624 init_dup(file); 1625 init_dup(file); 1626 fput(file); 1627 } 1628 1629 static noinline void __init kernel_init_freeable(void) 1630 { 1631 /* Now the scheduler is fully set up and can do blocking allocations */ 1632 gfp_allowed_mask = __GFP_BITS_MASK; 1633 1634 /* 1635 * init can allocate pages on any node 1636 */ 1637 set_mems_allowed(node_states[N_MEMORY]); 1638 1639 cad_pid = get_pid(task_pid(current)); 1640 1641 smp_prepare_cpus(setup_max_cpus); 1642 1643 workqueue_init(); 1644 1645 init_mm_internals(); 1646 1647 do_pre_smp_initcalls(); 1648 lockup_detector_init(); 1649 1650 smp_init(); 1651 sched_init_smp(); 1652 1653 workqueue_init_topology(); 1654 async_init(); 1655 padata_init(); 1656 page_alloc_init_late(); 1657 1658 do_basic_setup(); 1659 1660 kunit_run_all_tests(); 1661 1662 wait_for_initramfs(); 1663 console_on_rootfs(); 1664 1665 /* 1666 * check if there is an early userspace init. If yes, let it do all 1667 * the work 1668 */ 1669 int ramdisk_command_access; 1670 ramdisk_command_access = init_eaccess(ramdisk_execute_command); 1671 if (ramdisk_command_access != 0) { 1672 if (ramdisk_execute_command_set) 1673 pr_warn("check access for rdinit=%s failed: %i, ignoring\n", 1674 ramdisk_execute_command, ramdisk_command_access); 1675 ramdisk_execute_command = NULL; 1676 prepare_namespace(); 1677 } 1678 1679 /* 1680 * Ok, we have completed the initial bootup, and 1681 * we're essentially up and running. Get rid of the 1682 * initmem segments and start the user-mode stuff.. 1683 * 1684 * rootfs is available now, try loading the public keys 1685 * and default modules 1686 */ 1687 1688 integrity_load_keys(); 1689 } 1690