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/fs_struct.h> 107 #include <linux/ptdump.h> 108 #include <linux/time_namespace.h> 109 #include <linux/unaligned.h> 110 #include <linux/vdso_datastore.h> 111 #include <net/net_namespace.h> 112 113 #include <asm/io.h> 114 #include <asm/setup.h> 115 #include <asm/sections.h> 116 #include <asm/cacheflush.h> 117 118 #define CREATE_TRACE_POINTS 119 #include <trace/events/initcall.h> 120 121 #include <kunit/test.h> 122 123 static int kernel_init(void *); 124 125 /* 126 * Debug helper: via this flag we know that we are in 'early bootup code' 127 * where only the boot processor is running with IRQ disabled. This means 128 * two things - IRQ must not be enabled before the flag is cleared and some 129 * operations which are not allowed with IRQ disabled are allowed while the 130 * flag is set. 131 */ 132 bool early_boot_irqs_disabled __read_mostly; 133 134 enum system_states system_state __read_mostly; 135 EXPORT_SYMBOL(system_state); 136 137 /* 138 * Boot command-line arguments 139 */ 140 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT 141 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT 142 143 /* Default late time init is NULL. archs can override this later. */ 144 void (*__initdata late_time_init)(void); 145 146 /* Untouched command line saved by arch-specific code. */ 147 char __initdata boot_command_line[COMMAND_LINE_SIZE]; 148 /* Untouched saved command line (eg. for /proc) */ 149 char *saved_command_line __ro_after_init; 150 unsigned int saved_command_line_len __ro_after_init; 151 /* Command line for parameter parsing */ 152 static char *static_command_line; 153 /* Untouched extra command line */ 154 static char *extra_command_line; 155 /* Extra init arguments */ 156 static char *extra_init_args; 157 158 #ifdef CONFIG_BOOT_CONFIG 159 /* Is bootconfig on command line? */ 160 static bool bootconfig_found; 161 static size_t initargs_offs; 162 #else 163 # define bootconfig_found false 164 # define initargs_offs 0 165 #endif 166 167 static char *execute_command; 168 static char *ramdisk_execute_command = "/init"; 169 static bool __initdata ramdisk_execute_command_set; 170 171 /* 172 * Used to generate warnings if static_key manipulation functions are used 173 * before jump_label_init is called. 174 */ 175 bool static_key_initialized __read_mostly; 176 EXPORT_SYMBOL_GPL(static_key_initialized); 177 178 /* 179 * If set, this is an indication to the drivers that reset the underlying 180 * device before going ahead with the initialization otherwise driver might 181 * rely on the BIOS and skip the reset operation. 182 * 183 * This is useful if kernel is booting in an unreliable environment. 184 * For ex. kdump situation where previous kernel has crashed, BIOS has been 185 * skipped and devices will be in unknown state. 186 */ 187 unsigned int reset_devices; 188 EXPORT_SYMBOL(reset_devices); 189 190 static int __init set_reset_devices(char *str) 191 { 192 reset_devices = 1; 193 return 1; 194 } 195 196 __setup("reset_devices", set_reset_devices); 197 198 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 199 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 200 static const char *panic_later, *panic_param; 201 202 static bool __init obsolete_checksetup(char *line) 203 { 204 const struct obs_kernel_param *p; 205 bool had_early_param = false; 206 207 p = __setup_start; 208 do { 209 int n = strlen(p->str); 210 if (parameqn(line, p->str, n)) { 211 if (p->early) { 212 /* Already done in parse_early_param? 213 * (Needs exact match on param part). 214 * Keep iterating, as we can have early 215 * params and __setups of same names 8( */ 216 if (line[n] == '\0' || line[n] == '=') 217 had_early_param = true; 218 } else if (!p->setup_func) { 219 pr_warn("Parameter %s is obsolete, ignored\n", 220 p->str); 221 return true; 222 } else if (p->setup_func(line + n)) 223 return true; 224 } 225 p++; 226 } while (p < __setup_end); 227 228 return had_early_param; 229 } 230 231 /* 232 * This should be approx 2 Bo*oMips to start (note initial shift), and will 233 * still work even if initially too large, it will just take slightly longer 234 */ 235 unsigned long loops_per_jiffy = (1<<12); 236 EXPORT_SYMBOL(loops_per_jiffy); 237 238 static int __init debug_kernel(char *str) 239 { 240 console_loglevel = CONSOLE_LOGLEVEL_DEBUG; 241 return 0; 242 } 243 244 static int __init quiet_kernel(char *str) 245 { 246 console_loglevel = CONSOLE_LOGLEVEL_QUIET; 247 return 0; 248 } 249 250 early_param("debug", debug_kernel); 251 early_param("quiet", quiet_kernel); 252 253 static int __init loglevel(char *str) 254 { 255 int newlevel; 256 257 /* 258 * Only update loglevel value when a correct setting was passed, 259 * to prevent blind crashes (when loglevel being set to 0) that 260 * are quite hard to debug 261 */ 262 if (get_option(&str, &newlevel)) { 263 console_loglevel = newlevel; 264 return 0; 265 } 266 267 return -EINVAL; 268 } 269 270 early_param("loglevel", loglevel); 271 272 #ifdef CONFIG_BLK_DEV_INITRD 273 static void * __init get_boot_config_from_initrd(size_t *_size) 274 { 275 u32 size, csum; 276 char *data; 277 u8 *hdr; 278 int i; 279 280 if (!initrd_end) 281 return NULL; 282 283 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN; 284 /* 285 * Since Grub may align the size of initrd to 4, we must 286 * check the preceding 3 bytes as well. 287 */ 288 for (i = 0; i < 4; i++) { 289 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN)) 290 goto found; 291 data--; 292 } 293 return NULL; 294 295 found: 296 hdr = (u8 *)(data - 8); 297 size = get_unaligned_le32(hdr); 298 csum = get_unaligned_le32(hdr + 4); 299 300 data = ((void *)hdr) - size; 301 if ((unsigned long)data < initrd_start) { 302 pr_err("bootconfig size %d is greater than initrd size %ld\n", 303 size, initrd_end - initrd_start); 304 return NULL; 305 } 306 307 if (xbc_calc_checksum(data, size) != csum) { 308 pr_err("bootconfig checksum failed\n"); 309 return NULL; 310 } 311 312 /* Remove bootconfig from initramfs/initrd */ 313 initrd_end = (unsigned long)data; 314 if (_size) 315 *_size = size; 316 317 return data; 318 } 319 #else 320 static void * __init get_boot_config_from_initrd(size_t *_size) 321 { 322 return NULL; 323 } 324 #endif 325 326 #ifdef CONFIG_BOOT_CONFIG 327 328 /* Make an extra command line under given key word */ 329 static char * __init xbc_make_cmdline(const char *key) 330 { 331 struct xbc_node *root; 332 char *new_cmdline; 333 int ret, len = 0; 334 335 root = xbc_find_node(key); 336 if (!root) 337 return NULL; 338 339 /* Count required buffer size */ 340 len = xbc_snprint_cmdline(NULL, 0, root); 341 if (len <= 0) 342 return NULL; 343 344 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES); 345 if (!new_cmdline) { 346 pr_err("Failed to allocate memory for extra kernel cmdline.\n"); 347 return NULL; 348 } 349 350 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root); 351 if (ret < 0 || ret > len) { 352 pr_err("Failed to print extra kernel cmdline.\n"); 353 memblock_free(new_cmdline, len + 1); 354 return NULL; 355 } 356 357 return new_cmdline; 358 } 359 360 static int __init bootconfig_params(char *param, char *val, 361 const char *unused, void *arg) 362 { 363 if (strcmp(param, "bootconfig") == 0) { 364 bootconfig_found = true; 365 } 366 return 0; 367 } 368 369 static int __init warn_bootconfig(char *str) 370 { 371 /* The 'bootconfig' has been handled by bootconfig_params(). */ 372 return 0; 373 } 374 375 static void __init setup_boot_config(void) 376 { 377 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 378 const char *msg, *data; 379 int pos, ret; 380 size_t size; 381 char *err; 382 383 /* Cut out the bootconfig data even if we have no bootconfig option */ 384 data = get_boot_config_from_initrd(&size); 385 /* If there is no bootconfig in initrd, try embedded one. */ 386 if (!data) 387 data = xbc_get_embedded_bootconfig(&size); 388 389 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 390 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL, 391 bootconfig_params); 392 393 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE))) 394 return; 395 396 /* parse_args() stops at the next param of '--' and returns an address */ 397 if (err) 398 initargs_offs = err - tmp_cmdline; 399 400 if (!data) { 401 /* If user intended to use bootconfig, show an error level message */ 402 if (bootconfig_found) 403 pr_err("'bootconfig' found on command line, but no bootconfig found\n"); 404 else 405 pr_info("No bootconfig data provided, so skipping bootconfig"); 406 return; 407 } 408 409 if (size >= XBC_DATA_MAX) { 410 pr_err("bootconfig size %ld greater than max size %d\n", 411 (long)size, XBC_DATA_MAX); 412 return; 413 } 414 415 ret = xbc_init(data, size, &msg, &pos); 416 if (ret < 0) { 417 if (pos < 0) 418 pr_err("Failed to init bootconfig: %s.\n", msg); 419 else 420 pr_err("Failed to parse bootconfig: %s at %d.\n", 421 msg, pos); 422 } else { 423 xbc_get_info(&ret, NULL); 424 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret); 425 /* keys starting with "kernel." are passed via cmdline */ 426 extra_command_line = xbc_make_cmdline("kernel"); 427 /* Also, "init." keys are init arguments */ 428 extra_init_args = xbc_make_cmdline("init"); 429 } 430 return; 431 } 432 433 static void __init exit_boot_config(void) 434 { 435 xbc_exit(); 436 } 437 438 #else /* !CONFIG_BOOT_CONFIG */ 439 440 static void __init setup_boot_config(void) 441 { 442 /* Remove bootconfig data from initrd */ 443 get_boot_config_from_initrd(NULL); 444 } 445 446 static int __init warn_bootconfig(char *str) 447 { 448 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n"); 449 return 0; 450 } 451 452 #define exit_boot_config() do {} while (0) 453 454 #endif /* CONFIG_BOOT_CONFIG */ 455 456 early_param("bootconfig", warn_bootconfig); 457 458 bool __init cmdline_has_extra_options(void) 459 { 460 return extra_command_line || extra_init_args; 461 } 462 463 /* Change NUL term back to "=", to make "param" the whole string. */ 464 static void __init repair_env_string(char *param, char *val) 465 { 466 if (val) { 467 /* param=val or param="val"? */ 468 if (val == param+strlen(param)+1) 469 val[-1] = '='; 470 else if (val == param+strlen(param)+2) { 471 val[-2] = '='; 472 memmove(val-1, val, strlen(val)+1); 473 } else 474 BUG(); 475 } 476 } 477 478 /* Anything after -- gets handed straight to init. */ 479 static int __init set_init_arg(char *param, char *val, 480 const char *unused, void *arg) 481 { 482 unsigned int i; 483 484 if (panic_later) 485 return 0; 486 487 repair_env_string(param, val); 488 489 for (i = 0; argv_init[i]; i++) { 490 if (i == MAX_INIT_ARGS) { 491 panic_later = "init"; 492 panic_param = param; 493 return 0; 494 } 495 } 496 argv_init[i] = param; 497 return 0; 498 } 499 500 /* 501 * Unknown boot options get handed to init, unless they look like 502 * unused parameters (modprobe will find them in /proc/cmdline). 503 */ 504 static int __init unknown_bootoption(char *param, char *val, 505 const char *unused, void *arg) 506 { 507 size_t len = strlen(param); 508 /* 509 * Well-known bootloader identifiers: 510 * 1. LILO/Grub pass "BOOT_IMAGE=..."; 511 * 2. kexec/kdump (kexec-tools) pass "kexec". 512 */ 513 const char *bootloader[] = { "BOOT_IMAGE=", "kexec", NULL }; 514 515 /* Handle params aliased to sysctls */ 516 if (sysctl_is_alias(param)) 517 return 0; 518 519 repair_env_string(param, val); 520 521 /* Handle bootloader identifier */ 522 for (int i = 0; bootloader[i]; i++) { 523 if (strstarts(param, bootloader[i])) 524 return 0; 525 } 526 527 /* Handle obsolete-style parameters */ 528 if (obsolete_checksetup(param)) 529 return 0; 530 531 /* Unused module parameter. */ 532 if (strnchr(param, len, '.')) 533 return 0; 534 535 if (panic_later) 536 return 0; 537 538 if (val) { 539 /* Environment option */ 540 unsigned int i; 541 for (i = 0; envp_init[i]; i++) { 542 if (i == MAX_INIT_ENVS) { 543 panic_later = "env"; 544 panic_param = param; 545 } 546 if (!strncmp(param, envp_init[i], len+1)) 547 break; 548 } 549 envp_init[i] = param; 550 } else { 551 /* Command line option */ 552 unsigned int i; 553 for (i = 0; argv_init[i]; i++) { 554 if (i == MAX_INIT_ARGS) { 555 panic_later = "init"; 556 panic_param = param; 557 } 558 } 559 argv_init[i] = param; 560 } 561 return 0; 562 } 563 564 static int __init init_setup(char *str) 565 { 566 unsigned int i; 567 568 execute_command = str; 569 /* 570 * In case LILO is going to boot us with default command line, 571 * it prepends "auto" before the whole cmdline which makes 572 * the shell think it should execute a script with such name. 573 * So we ignore all arguments entered _before_ init=... [MJ] 574 */ 575 for (i = 1; i < MAX_INIT_ARGS; i++) 576 argv_init[i] = NULL; 577 return 1; 578 } 579 __setup("init=", init_setup); 580 581 static int __init rdinit_setup(char *str) 582 { 583 unsigned int i; 584 585 ramdisk_execute_command = str; 586 ramdisk_execute_command_set = true; 587 /* See "auto" comment in init_setup */ 588 for (i = 1; i < MAX_INIT_ARGS; i++) 589 argv_init[i] = NULL; 590 return 1; 591 } 592 __setup("rdinit=", rdinit_setup); 593 594 #ifndef CONFIG_SMP 595 static inline void setup_nr_cpu_ids(void) { } 596 static inline void smp_prepare_cpus(unsigned int maxcpus) { } 597 #endif 598 599 /* 600 * We need to store the untouched command line for future reference. 601 * We also need to store the touched command line since the parameter 602 * parsing is performed in place, and we should allow a component to 603 * store reference of name/value for future reference. 604 */ 605 static void __init setup_command_line(char *command_line) 606 { 607 size_t len, xlen = 0, ilen = 0; 608 609 if (extra_command_line) 610 xlen = strlen(extra_command_line); 611 if (extra_init_args) { 612 extra_init_args = strim(extra_init_args); /* remove trailing space */ 613 ilen = strlen(extra_init_args) + 4; /* for " -- " */ 614 } 615 616 len = xlen + strlen(boot_command_line) + ilen + 1; 617 618 saved_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES); 619 620 len = xlen + strlen(command_line) + 1; 621 622 static_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES); 623 624 if (xlen) { 625 /* 626 * We have to put extra_command_line before boot command 627 * lines because there could be dashes (separator of init 628 * command line) in the command lines. 629 */ 630 strcpy(saved_command_line, extra_command_line); 631 strcpy(static_command_line, extra_command_line); 632 } 633 strcpy(saved_command_line + xlen, boot_command_line); 634 strcpy(static_command_line + xlen, command_line); 635 636 if (ilen) { 637 /* 638 * Append supplemental init boot args to saved_command_line 639 * so that user can check what command line options passed 640 * to init. 641 * The order should always be 642 * " -- "[bootconfig init-param][cmdline init-param] 643 */ 644 if (initargs_offs) { 645 len = xlen + initargs_offs; 646 strcpy(saved_command_line + len, extra_init_args); 647 len += ilen - 4; /* strlen(extra_init_args) */ 648 strcpy(saved_command_line + len, 649 boot_command_line + initargs_offs - 1); 650 } else { 651 len = strlen(saved_command_line); 652 strcpy(saved_command_line + len, " -- "); 653 len += 4; 654 strcpy(saved_command_line + len, extra_init_args); 655 } 656 } 657 658 saved_command_line_len = strlen(saved_command_line); 659 } 660 661 /* 662 * We need to finalize in a non-__init function or else race conditions 663 * between the root thread and the init thread may cause start_kernel to 664 * be reaped by free_initmem before the root thread has proceeded to 665 * cpu_idle. 666 * 667 * gcc-3.4 accidentally inlines this function, so use noinline. 668 */ 669 670 static __initdata DECLARE_COMPLETION(kthreadd_done); 671 672 static noinline void __ref __noreturn rest_init(void) 673 { 674 struct kernel_clone_args init_args = { 675 .flags = (CLONE_VM | CLONE_UNTRACED), 676 .fn = kernel_init, 677 .fn_arg = NULL, 678 }; 679 struct task_struct *tsk; 680 int pid; 681 682 rcu_scheduler_starting(); 683 /* 684 * We need to spawn init first so that it obtains pid 1, however 685 * the init task will end up wanting to create kthreads, which, if 686 * we schedule it before we create kthreadd, will OOPS. 687 */ 688 pid = kernel_clone(&init_args); 689 /* 690 * Pin init on the boot CPU. Task migration is not properly working 691 * until sched_init_smp() has been run. It will set the allowed 692 * CPUs for init to the non isolated CPUs. 693 */ 694 rcu_read_lock(); 695 tsk = find_task_by_pid_ns(pid, &init_pid_ns); 696 tsk->flags |= PF_NO_SETAFFINITY; 697 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); 698 rcu_read_unlock(); 699 700 numa_default_policy(); 701 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES); 702 rcu_read_lock(); 703 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 704 rcu_read_unlock(); 705 706 /* 707 * Enable might_sleep() and smp_processor_id() checks. 708 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y 709 * kernel_thread() would trigger might_sleep() splats. With 710 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled 711 * already, but it's stuck on the kthreadd_done completion. 712 */ 713 system_state = SYSTEM_SCHEDULING; 714 715 complete(&kthreadd_done); 716 717 /* 718 * The boot idle thread must execute schedule() 719 * at least once to get things moving: 720 */ 721 schedule_preempt_disabled(); 722 /* Call into cpu_idle with preempt disabled */ 723 cpu_startup_entry(CPUHP_ONLINE); 724 } 725 726 /* Check for early params. */ 727 static int __init do_early_param(char *param, char *val, 728 const char *unused, void *arg) 729 { 730 const struct obs_kernel_param *p; 731 732 for (p = __setup_start; p < __setup_end; p++) { 733 if (p->early && parameq(param, p->str)) { 734 if (p->setup_func(val) != 0) 735 pr_warn("Malformed early option '%s'\n", param); 736 } 737 } 738 /* We accept everything at this stage. */ 739 return 0; 740 } 741 742 void __init parse_early_options(char *cmdline) 743 { 744 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, 745 do_early_param); 746 } 747 748 /* Arch code calls this early on, or if not, just before other parsing. */ 749 void __init parse_early_param(void) 750 { 751 static int done __initdata; 752 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 753 754 if (done) 755 return; 756 757 /* All fall through to do_early_param. */ 758 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 759 parse_early_options(tmp_cmdline); 760 done = 1; 761 } 762 763 void __init __weak arch_post_acpi_subsys_init(void) { } 764 765 void __init __weak smp_setup_processor_id(void) 766 { 767 } 768 769 void __init __weak smp_prepare_boot_cpu(void) 770 { 771 } 772 773 # if THREAD_SIZE >= PAGE_SIZE 774 void __init __weak thread_stack_cache_init(void) 775 { 776 } 777 #endif 778 779 void __init __weak poking_init(void) { } 780 781 void __init __weak pgtable_cache_init(void) { } 782 783 void __init __weak trap_init(void) { } 784 785 bool initcall_debug; 786 core_param(initcall_debug, initcall_debug, bool, 0644); 787 788 #ifdef TRACEPOINTS_ENABLED 789 static void __init initcall_debug_enable(void); 790 #else 791 static inline void initcall_debug_enable(void) 792 { 793 } 794 #endif 795 796 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET 797 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT, 798 randomize_kstack_offset); 799 DEFINE_PER_CPU(struct rnd_state, kstack_rnd_state); 800 801 static int __init random_kstack_init(void) 802 { 803 prandom_seed_full_state(&kstack_rnd_state); 804 return 0; 805 } 806 late_initcall(random_kstack_init); 807 808 static int __init early_randomize_kstack_offset(char *buf) 809 { 810 int ret; 811 bool bool_result; 812 813 ret = kstrtobool(buf, &bool_result); 814 if (ret) 815 return ret; 816 817 if (bool_result) 818 static_branch_enable(&randomize_kstack_offset); 819 else 820 static_branch_disable(&randomize_kstack_offset); 821 return 0; 822 } 823 early_param("randomize_kstack_offset", early_randomize_kstack_offset); 824 #endif 825 826 static void __init print_unknown_bootoptions(void) 827 { 828 char *unknown_options; 829 char *end; 830 const char *const *p; 831 size_t len; 832 833 if (panic_later || (!argv_init[1] && !envp_init[2])) 834 return; 835 836 /* 837 * Determine how many options we have to print out, plus a space 838 * before each 839 */ 840 len = 1; /* null terminator */ 841 for (p = &argv_init[1]; *p; p++) { 842 len++; 843 len += strlen(*p); 844 } 845 for (p = &envp_init[2]; *p; p++) { 846 len++; 847 len += strlen(*p); 848 } 849 850 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES); 851 if (!unknown_options) { 852 pr_err("%s: Failed to allocate %zu bytes\n", 853 __func__, len); 854 return; 855 } 856 end = unknown_options; 857 858 for (p = &argv_init[1]; *p; p++) 859 end += sprintf(end, " %s", *p); 860 for (p = &envp_init[2]; *p; p++) 861 end += sprintf(end, " %s", *p); 862 863 /* Start at unknown_options[1] to skip the initial space */ 864 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n", 865 &unknown_options[1]); 866 memblock_free(unknown_options, len); 867 } 868 869 static void __init early_numa_node_init(void) 870 { 871 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID 872 #ifndef cpu_to_node 873 int cpu; 874 875 /* The early_cpu_to_node() should be ready here. */ 876 for_each_possible_cpu(cpu) 877 set_cpu_numa_node(cpu, early_cpu_to_node(cpu)); 878 #endif 879 #endif 880 } 881 882 #define KERNEL_CMDLINE_PREFIX "Kernel command line: " 883 #define KERNEL_CMDLINE_PREFIX_LEN (sizeof(KERNEL_CMDLINE_PREFIX) - 1) 884 #define KERNEL_CMDLINE_CONTINUATION " \\" 885 #define KERNEL_CMDLINE_CONTINUATION_LEN (sizeof(KERNEL_CMDLINE_CONTINUATION) - 1) 886 887 #define MIN_CMDLINE_LOG_WRAP_IDEAL_LEN (KERNEL_CMDLINE_PREFIX_LEN + \ 888 KERNEL_CMDLINE_CONTINUATION_LEN) 889 #define CMDLINE_LOG_WRAP_IDEAL_LEN (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN > \ 890 MIN_CMDLINE_LOG_WRAP_IDEAL_LEN ? \ 891 CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN : \ 892 MIN_CMDLINE_LOG_WRAP_IDEAL_LEN) 893 894 #define IDEAL_CMDLINE_LEN (CMDLINE_LOG_WRAP_IDEAL_LEN - KERNEL_CMDLINE_PREFIX_LEN) 895 #define IDEAL_CMDLINE_SPLIT_LEN (IDEAL_CMDLINE_LEN - KERNEL_CMDLINE_CONTINUATION_LEN) 896 897 /** 898 * print_kernel_cmdline() - Print the kernel cmdline with wrapping. 899 * @cmdline: The cmdline to print. 900 * 901 * Print the kernel command line, trying to wrap based on the Kconfig knob 902 * CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN. 903 * 904 * Wrapping is based on spaces, ignoring quotes. All lines are prefixed 905 * with "Kernel command line: " and lines that are not the last line have 906 * a " \" suffix added to them. The prefix and suffix count towards the 907 * line length for wrapping purposes. The ideal length will be exceeded 908 * if no appropriate place to wrap is found. 909 * 910 * Example output if CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN is 40: 911 * Kernel command line: loglevel=7 \ 912 * Kernel command line: init=/sbin/init \ 913 * Kernel command line: root=PARTUUID=8c3efc1a-768b-6642-8d0c-89eb782f19f0/PARTNROFF=1 \ 914 * Kernel command line: rootwait ro \ 915 * Kernel command line: my_quoted_arg="The \ 916 * Kernel command line: quick brown fox \ 917 * Kernel command line: jumps over the \ 918 * Kernel command line: lazy dog." 919 */ 920 static void __init print_kernel_cmdline(const char *cmdline) 921 { 922 size_t len; 923 924 /* Config option of 0 or anything longer than the max disables wrapping */ 925 if (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN == 0 || 926 IDEAL_CMDLINE_LEN >= COMMAND_LINE_SIZE - 1) { 927 pr_notice("%s%s\n", KERNEL_CMDLINE_PREFIX, cmdline); 928 return; 929 } 930 931 len = strlen(cmdline); 932 while (len > IDEAL_CMDLINE_LEN) { 933 const char *first_space; 934 const char *prev_cutoff; 935 const char *cutoff; 936 int to_print; 937 size_t used; 938 939 /* Find the last ' ' that wouldn't make the line too long */ 940 prev_cutoff = NULL; 941 cutoff = cmdline; 942 while (true) { 943 cutoff = strchr(cutoff + 1, ' '); 944 if (!cutoff || cutoff - cmdline > IDEAL_CMDLINE_SPLIT_LEN) 945 break; 946 prev_cutoff = cutoff; 947 } 948 if (prev_cutoff) 949 cutoff = prev_cutoff; 950 else if (!cutoff) 951 break; 952 953 /* Find the beginning and end of the string of spaces */ 954 first_space = cutoff; 955 while (first_space > cmdline && first_space[-1] == ' ') 956 first_space--; 957 to_print = first_space - cmdline; 958 while (*cutoff == ' ') 959 cutoff++; 960 used = cutoff - cmdline; 961 962 /* If the whole string is used, break and do the final printout */ 963 if (len == used) 964 break; 965 966 if (to_print) 967 pr_notice("%s%.*s%s\n", KERNEL_CMDLINE_PREFIX, 968 to_print, cmdline, KERNEL_CMDLINE_CONTINUATION); 969 970 len -= used; 971 cmdline += used; 972 } 973 if (len) 974 pr_notice("%s%s\n", KERNEL_CMDLINE_PREFIX, cmdline); 975 } 976 977 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector 978 void start_kernel(void) 979 { 980 char *command_line; 981 char *after_dashes; 982 983 set_task_stack_end_magic(&init_task); 984 smp_setup_processor_id(); 985 debug_objects_early_init(); 986 init_vmlinux_build_id(); 987 988 cgroup_init_early(); 989 990 local_irq_disable(); 991 early_boot_irqs_disabled = true; 992 993 /* 994 * Interrupts are still disabled. Do necessary setups, then 995 * enable them. 996 */ 997 boot_cpu_init(); 998 page_address_init(); 999 pr_notice("%s", linux_banner); 1000 setup_arch(&command_line); 1001 mm_core_init_early(); 1002 /* Static keys and static calls are needed by LSMs */ 1003 jump_label_init(); 1004 static_call_init(); 1005 early_security_init(); 1006 setup_boot_config(); 1007 setup_command_line(command_line); 1008 setup_nr_cpu_ids(); 1009 setup_per_cpu_areas(); 1010 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 1011 early_numa_node_init(); 1012 boot_cpu_hotplug_init(); 1013 1014 print_kernel_cmdline(saved_command_line); 1015 /* parameters may set static keys */ 1016 parse_early_param(); 1017 after_dashes = parse_args("Booting kernel", 1018 static_command_line, __start___param, 1019 __stop___param - __start___param, 1020 -1, -1, NULL, &unknown_bootoption); 1021 print_unknown_bootoptions(); 1022 if (!IS_ERR_OR_NULL(after_dashes)) 1023 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 1024 NULL, set_init_arg); 1025 if (extra_init_args) 1026 parse_args("Setting extra init args", extra_init_args, 1027 NULL, 0, -1, -1, NULL, set_init_arg); 1028 1029 /* Architectural and non-timekeeping rng init, before allocator init */ 1030 random_init_early(command_line); 1031 1032 /* 1033 * These use large bootmem allocations and must precede 1034 * initalization of page allocator 1035 */ 1036 setup_log_buf(0); 1037 vfs_caches_init_early(); 1038 sort_main_extable(); 1039 trap_init(); 1040 mm_core_init(); 1041 maple_tree_init(); 1042 poking_init(); 1043 ftrace_init(); 1044 1045 /* trace_printk can be enabled here */ 1046 early_trace_init(); 1047 1048 /* 1049 * Set up the scheduler prior starting any interrupts (such as the 1050 * timer interrupt). Full topology setup happens at smp_init() 1051 * time - but meanwhile we still have a functioning scheduler. 1052 */ 1053 sched_init(); 1054 1055 if (WARN(!irqs_disabled(), 1056 "Interrupts were enabled *very* early, fixing it\n")) 1057 local_irq_disable(); 1058 radix_tree_init(); 1059 1060 /* 1061 * Set up housekeeping before setting up workqueues to allow the unbound 1062 * workqueue to take non-housekeeping into account. 1063 */ 1064 housekeeping_init(); 1065 1066 /* 1067 * Allow workqueue creation and work item queueing/cancelling 1068 * early. Work item execution depends on kthreads and starts after 1069 * workqueue_init(). 1070 */ 1071 workqueue_init_early(); 1072 1073 rcu_init(); 1074 kvfree_rcu_init(); 1075 1076 /* Trace events are available after this */ 1077 trace_init(); 1078 1079 if (initcall_debug) 1080 initcall_debug_enable(); 1081 1082 context_tracking_init(); 1083 /* init some links before init_ISA_irqs() */ 1084 early_irq_init(); 1085 init_IRQ(); 1086 tick_init(); 1087 rcu_init_nohz(); 1088 timers_init(); 1089 srcu_init(); 1090 hrtimers_init(); 1091 softirq_init(); 1092 vdso_setup_data_pages(); 1093 timekeeping_init(); 1094 time_init(); 1095 1096 /* This must be after timekeeping is initialized */ 1097 random_init(); 1098 1099 /* These make use of the fully initialized rng */ 1100 kfence_init(); 1101 boot_init_stack_canary(); 1102 1103 perf_event_init(); 1104 profile_init(); 1105 call_function_init(); 1106 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 1107 1108 early_boot_irqs_disabled = false; 1109 local_irq_enable(); 1110 1111 kmem_cache_init_late(); 1112 1113 /* 1114 * HACK ALERT! This is early. We're enabling the console before 1115 * we've done PCI setups etc, and console_init() must be aware of 1116 * this. But we do want output early, in case something goes wrong. 1117 */ 1118 console_init(); 1119 if (panic_later) 1120 panic("Too many boot %s vars at `%s'", panic_later, 1121 panic_param); 1122 1123 lockdep_init(); 1124 1125 /* 1126 * Need to run this when irqs are enabled, because it wants 1127 * to self-test [hard/soft]-irqs on/off lock inversion bugs 1128 * too: 1129 */ 1130 locking_selftest(); 1131 1132 #ifdef CONFIG_BLK_DEV_INITRD 1133 if (initrd_start && !initrd_below_start_ok && 1134 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 1135 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 1136 page_to_pfn(virt_to_page((void *)initrd_start)), 1137 min_low_pfn); 1138 initrd_start = 0; 1139 } 1140 #endif 1141 setup_per_cpu_pageset(); 1142 numa_policy_init(); 1143 acpi_early_init(); 1144 if (late_time_init) 1145 late_time_init(); 1146 sched_clock_init(); 1147 calibrate_delay(); 1148 1149 arch_cpu_finalize_init(); 1150 1151 pid_idr_init(); 1152 anon_vma_init(); 1153 thread_stack_cache_init(); 1154 cred_init(); 1155 fork_init(); 1156 proc_caches_init(); 1157 uts_ns_init(); 1158 time_ns_init(); 1159 key_init(); 1160 security_init(); 1161 dbg_late_init(); 1162 net_ns_init(); 1163 vfs_caches_init(); 1164 pagecache_init(); 1165 signals_init(); 1166 seq_file_init(); 1167 proc_root_init(); 1168 nsfs_init(); 1169 pidfs_init(); 1170 cpuset_init(); 1171 mem_cgroup_init(); 1172 cgroup_init(); 1173 taskstats_init_early(); 1174 delayacct_init(); 1175 1176 acpi_subsystem_init(); 1177 arch_post_acpi_subsys_init(); 1178 kcsan_init(); 1179 1180 /* Do the rest non-__init'ed, we're now alive */ 1181 rest_init(); 1182 1183 /* 1184 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10 1185 * and older. 1186 */ 1187 #if !__has_attribute(__no_stack_protector__) 1188 prevent_tail_call_optimization(); 1189 #endif 1190 } 1191 1192 /* Call all constructor functions linked into the kernel. */ 1193 static void __init do_ctors(void) 1194 { 1195 /* 1196 * For UML, the constructors have already been called by the 1197 * normal setup code as it's just a normal ELF binary, so we 1198 * cannot do it again - but we do need CONFIG_CONSTRUCTORS 1199 * even on UML for modules. 1200 */ 1201 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML) 1202 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 1203 1204 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 1205 (*fn)(); 1206 #endif 1207 } 1208 1209 #ifdef CONFIG_KALLSYMS 1210 struct blacklist_entry { 1211 struct list_head next; 1212 char *buf; 1213 }; 1214 1215 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 1216 1217 static int __init initcall_blacklist(char *str) 1218 { 1219 char *str_entry; 1220 struct blacklist_entry *entry; 1221 1222 /* str argument is a comma-separated list of functions */ 1223 do { 1224 str_entry = strsep(&str, ","); 1225 if (str_entry) { 1226 pr_debug("blacklisting initcall %s\n", str_entry); 1227 entry = memblock_alloc_or_panic(sizeof(*entry), 1228 SMP_CACHE_BYTES); 1229 entry->buf = memblock_alloc_or_panic(strlen(str_entry) + 1, 1230 SMP_CACHE_BYTES); 1231 strcpy(entry->buf, str_entry); 1232 list_add(&entry->next, &blacklisted_initcalls); 1233 } 1234 } while (str_entry); 1235 1236 return 1; 1237 } 1238 1239 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1240 { 1241 struct blacklist_entry *entry; 1242 char fn_name[KSYM_SYMBOL_LEN]; 1243 unsigned long addr; 1244 1245 if (list_empty(&blacklisted_initcalls)) 1246 return false; 1247 1248 addr = (unsigned long) dereference_function_descriptor(fn); 1249 sprint_symbol_no_offset(fn_name, addr); 1250 1251 /* 1252 * fn will be "function_name [module_name]" where [module_name] is not 1253 * displayed for built-in init functions. Strip off the [module_name]. 1254 */ 1255 strreplace(fn_name, ' ', '\0'); 1256 1257 list_for_each_entry(entry, &blacklisted_initcalls, next) { 1258 if (!strcmp(fn_name, entry->buf)) { 1259 pr_debug("initcall %s blacklisted\n", fn_name); 1260 return true; 1261 } 1262 } 1263 1264 return false; 1265 } 1266 #else 1267 static int __init initcall_blacklist(char *str) 1268 { 1269 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 1270 return 0; 1271 } 1272 1273 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1274 { 1275 return false; 1276 } 1277 #endif 1278 __setup("initcall_blacklist=", initcall_blacklist); 1279 1280 static __init_or_module void 1281 trace_initcall_start_cb(void *data, initcall_t fn) 1282 { 1283 ktime_t *calltime = data; 1284 1285 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current)); 1286 *calltime = ktime_get(); 1287 } 1288 1289 static __init_or_module void 1290 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 1291 { 1292 ktime_t rettime, *calltime = data; 1293 1294 rettime = ktime_get(); 1295 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n", 1296 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime)); 1297 } 1298 1299 static __init_or_module void 1300 trace_initcall_level_cb(void *data, const char *level) 1301 { 1302 printk(KERN_DEBUG "entering initcall level: %s\n", level); 1303 } 1304 1305 static ktime_t initcall_calltime; 1306 1307 #ifdef TRACEPOINTS_ENABLED 1308 static void __init initcall_debug_enable(void) 1309 { 1310 int ret; 1311 1312 ret = register_trace_initcall_start(trace_initcall_start_cb, 1313 &initcall_calltime); 1314 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 1315 &initcall_calltime); 1316 ret |= register_trace_initcall_level(trace_initcall_level_cb, NULL); 1317 WARN(ret, "Failed to register initcall tracepoints\n"); 1318 } 1319 # define do_trace_initcall_start trace_initcall_start 1320 # define do_trace_initcall_finish trace_initcall_finish 1321 # define do_trace_initcall_level trace_initcall_level 1322 #else 1323 static inline void do_trace_initcall_start(initcall_t fn) 1324 { 1325 if (!initcall_debug) 1326 return; 1327 trace_initcall_start_cb(&initcall_calltime, fn); 1328 } 1329 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 1330 { 1331 if (!initcall_debug) 1332 return; 1333 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 1334 } 1335 static inline void do_trace_initcall_level(const char *level) 1336 { 1337 if (!initcall_debug) 1338 return; 1339 trace_initcall_level_cb(NULL, level); 1340 } 1341 #endif /* !TRACEPOINTS_ENABLED */ 1342 1343 int __init_or_module do_one_initcall(initcall_t fn) 1344 { 1345 int count = preempt_count(); 1346 char msgbuf[64]; 1347 int ret; 1348 1349 if (initcall_blacklisted(fn)) 1350 return -EPERM; 1351 1352 do_trace_initcall_start(fn); 1353 ret = fn(); 1354 do_trace_initcall_finish(fn, ret); 1355 1356 msgbuf[0] = 0; 1357 1358 if (preempt_count() != count) { 1359 sprintf(msgbuf, "preemption imbalance "); 1360 preempt_count_set(count); 1361 } 1362 if (irqs_disabled()) { 1363 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 1364 local_irq_enable(); 1365 } 1366 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 1367 1368 add_latent_entropy(); 1369 return ret; 1370 } 1371 1372 1373 static initcall_entry_t *initcall_levels[] __initdata = { 1374 __initcall0_start, 1375 __initcall1_start, 1376 __initcall2_start, 1377 __initcall3_start, 1378 __initcall4_start, 1379 __initcall5_start, 1380 __initcall6_start, 1381 __initcall7_start, 1382 __initcall_end, 1383 }; 1384 1385 /* Keep these in sync with initcalls in include/linux/init.h */ 1386 static const char *initcall_level_names[] __initdata = { 1387 "pure", 1388 "core", 1389 "postcore", 1390 "arch", 1391 "subsys", 1392 "fs", 1393 "device", 1394 "late", 1395 }; 1396 1397 static int __init ignore_unknown_bootoption(char *param, char *val, 1398 const char *unused, void *arg) 1399 { 1400 return 0; 1401 } 1402 1403 static void __init do_initcall_level(int level, char *command_line) 1404 { 1405 initcall_entry_t *fn; 1406 1407 parse_args(initcall_level_names[level], 1408 command_line, __start___param, 1409 __stop___param - __start___param, 1410 level, level, 1411 NULL, ignore_unknown_bootoption); 1412 1413 do_trace_initcall_level(initcall_level_names[level]); 1414 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1415 do_one_initcall(initcall_from_entry(fn)); 1416 } 1417 1418 static void __init do_initcalls(void) 1419 { 1420 int level; 1421 size_t len = saved_command_line_len + 1; 1422 char *command_line; 1423 1424 command_line = kzalloc(len, GFP_KERNEL); 1425 if (!command_line) 1426 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 1427 1428 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { 1429 /* Parser modifies command_line, restore it each time */ 1430 strcpy(command_line, saved_command_line); 1431 do_initcall_level(level, command_line); 1432 } 1433 1434 kfree(command_line); 1435 } 1436 1437 /* 1438 * Ok, the machine is now initialized. None of the devices 1439 * have been touched yet, but the CPU subsystem is up and 1440 * running, and memory and process management works. 1441 * 1442 * Now we can finally start doing some real work.. 1443 */ 1444 static void __init do_basic_setup(void) 1445 { 1446 cpuset_init_smp(); 1447 ksysfs_init(); 1448 driver_init(); 1449 init_irq_proc(); 1450 do_ctors(); 1451 do_initcalls(); 1452 } 1453 1454 static void __init do_pre_smp_initcalls(void) 1455 { 1456 initcall_entry_t *fn; 1457 1458 do_trace_initcall_level("early"); 1459 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1460 do_one_initcall(initcall_from_entry(fn)); 1461 } 1462 1463 static int run_init_process(const char *init_filename) 1464 { 1465 const char *const *p; 1466 1467 argv_init[0] = init_filename; 1468 pr_info("Run %s as init process\n", init_filename); 1469 pr_debug(" with arguments:\n"); 1470 for (p = argv_init; *p; p++) 1471 pr_debug(" %s\n", *p); 1472 pr_debug(" with environment:\n"); 1473 for (p = envp_init; *p; p++) 1474 pr_debug(" %s\n", *p); 1475 return kernel_execve(init_filename, argv_init, envp_init); 1476 } 1477 1478 static int try_to_run_init_process(const char *init_filename) 1479 { 1480 int ret; 1481 1482 ret = run_init_process(init_filename); 1483 1484 if (ret && ret != -ENOENT) { 1485 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1486 init_filename, ret); 1487 } 1488 1489 return ret; 1490 } 1491 1492 static noinline void __init kernel_init_freeable(void); 1493 1494 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1495 bool rodata_enabled __ro_after_init = true; 1496 1497 #ifndef arch_parse_debug_rodata 1498 static inline bool arch_parse_debug_rodata(char *str) { return false; } 1499 #endif 1500 1501 static int __init set_debug_rodata(char *str) 1502 { 1503 if (arch_parse_debug_rodata(str)) 1504 return 0; 1505 1506 if (str && !strcmp(str, "on")) 1507 rodata_enabled = true; 1508 else if (str && !strcmp(str, "off")) 1509 rodata_enabled = false; 1510 else 1511 pr_warn("Invalid option string for rodata: '%s'\n", str); 1512 return 0; 1513 } 1514 early_param("rodata", set_debug_rodata); 1515 #endif 1516 1517 static void mark_readonly(void) 1518 { 1519 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) { 1520 /* 1521 * load_module() results in W+X mappings, which are cleaned 1522 * up with init_free_wq. Let's make sure that queued work is 1523 * flushed so that we don't hit false positives looking for 1524 * insecure pages which are W+X. 1525 */ 1526 flush_module_init_free_work(); 1527 jump_label_init_ro(); 1528 mark_rodata_ro(); 1529 debug_checkwx(); 1530 rodata_test(); 1531 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) { 1532 pr_info("Kernel memory protection disabled.\n"); 1533 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) { 1534 pr_warn("Kernel memory protection not selected by kernel config.\n"); 1535 } else { 1536 pr_warn("This architecture does not have kernel memory protection.\n"); 1537 } 1538 } 1539 1540 void __weak free_initmem(void) 1541 { 1542 free_initmem_default(POISON_FREE_INITMEM); 1543 } 1544 1545 static int __ref kernel_init(void *unused) 1546 { 1547 int ret; 1548 1549 init_userspace_fs(); 1550 1551 /* 1552 * Wait until kthreadd is all set-up. 1553 */ 1554 wait_for_completion(&kthreadd_done); 1555 1556 kernel_init_freeable(); 1557 /* need to finish all async __init code before freeing the memory */ 1558 async_synchronize_full(); 1559 1560 system_state = SYSTEM_FREEING_INITMEM; 1561 kprobe_free_init_mem(); 1562 ftrace_free_init_mem(); 1563 kgdb_free_init_mem(); 1564 exit_boot_config(); 1565 free_initmem(); 1566 mark_readonly(); 1567 1568 /* 1569 * Kernel mappings are now finalized - update the userspace page-table 1570 * to finalize PTI. 1571 */ 1572 pti_finalize(); 1573 1574 system_state = SYSTEM_RUNNING; 1575 numa_default_policy(); 1576 1577 rcu_end_inkernel_boot(); 1578 1579 do_sysctl_args(); 1580 1581 if (ramdisk_execute_command) { 1582 ret = run_init_process(ramdisk_execute_command); 1583 if (!ret) 1584 return 0; 1585 pr_err("Failed to execute %s (error %d)\n", 1586 ramdisk_execute_command, ret); 1587 } 1588 1589 /* 1590 * We try each of these until one succeeds. 1591 * 1592 * The Bourne shell can be used instead of init if we are 1593 * trying to recover a really broken machine. 1594 */ 1595 if (execute_command) { 1596 ret = run_init_process(execute_command); 1597 if (!ret) 1598 return 0; 1599 panic("Requested init %s failed (error %d).", 1600 execute_command, ret); 1601 } 1602 1603 if (CONFIG_DEFAULT_INIT[0] != '\0') { 1604 ret = run_init_process(CONFIG_DEFAULT_INIT); 1605 if (ret) 1606 pr_err("Default init %s failed (error %d)\n", 1607 CONFIG_DEFAULT_INIT, ret); 1608 else 1609 return 0; 1610 } 1611 1612 if (!try_to_run_init_process("/sbin/init") || 1613 !try_to_run_init_process("/etc/init") || 1614 !try_to_run_init_process("/bin/init") || 1615 !try_to_run_init_process("/bin/sh")) 1616 return 0; 1617 1618 panic("No working init found. Try passing init= option to kernel. " 1619 "See Linux Documentation/admin-guide/init.rst for guidance."); 1620 } 1621 1622 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */ 1623 void __init console_on_rootfs(void) 1624 { 1625 struct file *file = filp_open("/dev/console", O_RDWR, 0); 1626 1627 if (IS_ERR(file)) { 1628 pr_err("Warning: unable to open an initial console.\n"); 1629 return; 1630 } 1631 init_dup(file); 1632 init_dup(file); 1633 init_dup(file); 1634 fput(file); 1635 } 1636 1637 static noinline void __init kernel_init_freeable(void) 1638 { 1639 /* Now the scheduler is fully set up and can do blocking allocations */ 1640 gfp_allowed_mask = __GFP_BITS_MASK; 1641 1642 /* 1643 * init can allocate pages on any node 1644 */ 1645 set_mems_allowed(node_states[N_MEMORY]); 1646 1647 cad_pid = get_pid(task_pid(current)); 1648 1649 smp_prepare_cpus(setup_max_cpus); 1650 1651 workqueue_init(); 1652 1653 init_mm_internals(); 1654 1655 do_pre_smp_initcalls(); 1656 lockup_detector_init(); 1657 1658 smp_init(); 1659 sched_init_smp(); 1660 1661 workqueue_init_topology(); 1662 async_init(); 1663 padata_init(); 1664 page_alloc_init_late(); 1665 1666 do_basic_setup(); 1667 1668 kunit_run_all_tests(); 1669 1670 wait_for_initramfs(); 1671 console_on_rootfs(); 1672 1673 /* 1674 * check if there is an early userspace init. If yes, let it do all 1675 * the work 1676 */ 1677 int ramdisk_command_access; 1678 ramdisk_command_access = init_eaccess(ramdisk_execute_command); 1679 if (ramdisk_command_access != 0) { 1680 if (ramdisk_execute_command_set) 1681 pr_warn("check access for rdinit=%s failed: %i, ignoring\n", 1682 ramdisk_execute_command, ramdisk_command_access); 1683 ramdisk_execute_command = NULL; 1684 prepare_namespace(); 1685 } 1686 1687 /* 1688 * Ok, we have completed the initial bootup, and 1689 * we're essentially up and running. Get rid of the 1690 * initmem segments and start the user-mode stuff.. 1691 * 1692 * rootfs is available now, try loading the public keys 1693 * and default modules 1694 */ 1695 1696 integrity_load_keys(); 1697 } 1698