1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * linux/init/main.c 4 * 5 * Copyright (C) 1991, 1992 Linus Torvalds 6 * 7 * GK 2/5/95 - Changed to support mounting root fs via NFS 8 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96 9 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96 10 * Simplified starting of init: Michael A. Griffith <grif@acm.org> 11 */ 12 13 #define DEBUG /* Enable initcall_debug */ 14 15 #include <linux/types.h> 16 #include <linux/extable.h> 17 #include <linux/module.h> 18 #include <linux/proc_fs.h> 19 #include <linux/binfmts.h> 20 #include <linux/kernel.h> 21 #include <linux/syscalls.h> 22 #include <linux/stackprotector.h> 23 #include <linux/string.h> 24 #include <linux/ctype.h> 25 #include <linux/delay.h> 26 #include <linux/ioport.h> 27 #include <linux/init.h> 28 #include <linux/initrd.h> 29 #include <linux/memblock.h> 30 #include <linux/acpi.h> 31 #include <linux/console.h> 32 #include <linux/nmi.h> 33 #include <linux/percpu.h> 34 #include <linux/kmod.h> 35 #include <linux/vmalloc.h> 36 #include <linux/kernel_stat.h> 37 #include <linux/start_kernel.h> 38 #include <linux/security.h> 39 #include <linux/smp.h> 40 #include <linux/profile.h> 41 #include <linux/rcupdate.h> 42 #include <linux/moduleparam.h> 43 #include <linux/kallsyms.h> 44 #include <linux/writeback.h> 45 #include <linux/cpu.h> 46 #include <linux/cpuset.h> 47 #include <linux/cgroup.h> 48 #include <linux/efi.h> 49 #include <linux/tick.h> 50 #include <linux/sched/isolation.h> 51 #include <linux/interrupt.h> 52 #include <linux/taskstats_kern.h> 53 #include <linux/delayacct.h> 54 #include <linux/unistd.h> 55 #include <linux/utsname.h> 56 #include <linux/rmap.h> 57 #include <linux/mempolicy.h> 58 #include <linux/key.h> 59 #include <linux/buffer_head.h> 60 #include <linux/page_ext.h> 61 #include <linux/debug_locks.h> 62 #include <linux/debugobjects.h> 63 #include <linux/lockdep.h> 64 #include <linux/kmemleak.h> 65 #include <linux/pid_namespace.h> 66 #include <linux/device.h> 67 #include <linux/kthread.h> 68 #include <linux/sched.h> 69 #include <linux/sched/init.h> 70 #include <linux/signal.h> 71 #include <linux/idr.h> 72 #include <linux/kgdb.h> 73 #include <linux/ftrace.h> 74 #include <linux/async.h> 75 #include <linux/sfi.h> 76 #include <linux/shmem_fs.h> 77 #include <linux/slab.h> 78 #include <linux/perf_event.h> 79 #include <linux/ptrace.h> 80 #include <linux/pti.h> 81 #include <linux/blkdev.h> 82 #include <linux/elevator.h> 83 #include <linux/sched/clock.h> 84 #include <linux/sched/task.h> 85 #include <linux/sched/task_stack.h> 86 #include <linux/context_tracking.h> 87 #include <linux/random.h> 88 #include <linux/list.h> 89 #include <linux/integrity.h> 90 #include <linux/proc_ns.h> 91 #include <linux/io.h> 92 #include <linux/cache.h> 93 #include <linux/rodata_test.h> 94 #include <linux/jump_label.h> 95 #include <linux/mem_encrypt.h> 96 97 #include <asm/io.h> 98 #include <asm/bugs.h> 99 #include <asm/setup.h> 100 #include <asm/sections.h> 101 #include <asm/cacheflush.h> 102 103 #define CREATE_TRACE_POINTS 104 #include <trace/events/initcall.h> 105 106 static int kernel_init(void *); 107 108 extern void init_IRQ(void); 109 extern void radix_tree_init(void); 110 111 /* 112 * Debug helper: via this flag we know that we are in 'early bootup code' 113 * where only the boot processor is running with IRQ disabled. This means 114 * two things - IRQ must not be enabled before the flag is cleared and some 115 * operations which are not allowed with IRQ disabled are allowed while the 116 * flag is set. 117 */ 118 bool early_boot_irqs_disabled __read_mostly; 119 120 enum system_states system_state __read_mostly; 121 EXPORT_SYMBOL(system_state); 122 123 /* 124 * Boot command-line arguments 125 */ 126 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT 127 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT 128 129 extern void time_init(void); 130 /* Default late time init is NULL. archs can override this later. */ 131 void (*__initdata late_time_init)(void); 132 133 /* Untouched command line saved by arch-specific code. */ 134 char __initdata boot_command_line[COMMAND_LINE_SIZE]; 135 /* Untouched saved command line (eg. for /proc) */ 136 char *saved_command_line; 137 /* Command line for parameter parsing */ 138 static char *static_command_line; 139 /* Command line for per-initcall parameter parsing */ 140 static char *initcall_command_line; 141 142 static char *execute_command; 143 static char *ramdisk_execute_command; 144 145 /* 146 * Used to generate warnings if static_key manipulation functions are used 147 * before jump_label_init is called. 148 */ 149 bool static_key_initialized __read_mostly; 150 EXPORT_SYMBOL_GPL(static_key_initialized); 151 152 /* 153 * If set, this is an indication to the drivers that reset the underlying 154 * device before going ahead with the initialization otherwise driver might 155 * rely on the BIOS and skip the reset operation. 156 * 157 * This is useful if kernel is booting in an unreliable environment. 158 * For ex. kdump situation where previous kernel has crashed, BIOS has been 159 * skipped and devices will be in unknown state. 160 */ 161 unsigned int reset_devices; 162 EXPORT_SYMBOL(reset_devices); 163 164 static int __init set_reset_devices(char *str) 165 { 166 reset_devices = 1; 167 return 1; 168 } 169 170 __setup("reset_devices", set_reset_devices); 171 172 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 173 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 174 static const char *panic_later, *panic_param; 175 176 extern const struct obs_kernel_param __setup_start[], __setup_end[]; 177 178 static bool __init obsolete_checksetup(char *line) 179 { 180 const struct obs_kernel_param *p; 181 bool had_early_param = false; 182 183 p = __setup_start; 184 do { 185 int n = strlen(p->str); 186 if (parameqn(line, p->str, n)) { 187 if (p->early) { 188 /* Already done in parse_early_param? 189 * (Needs exact match on param part). 190 * Keep iterating, as we can have early 191 * params and __setups of same names 8( */ 192 if (line[n] == '\0' || line[n] == '=') 193 had_early_param = true; 194 } else if (!p->setup_func) { 195 pr_warn("Parameter %s is obsolete, ignored\n", 196 p->str); 197 return true; 198 } else if (p->setup_func(line + n)) 199 return true; 200 } 201 p++; 202 } while (p < __setup_end); 203 204 return had_early_param; 205 } 206 207 /* 208 * This should be approx 2 Bo*oMips to start (note initial shift), and will 209 * still work even if initially too large, it will just take slightly longer 210 */ 211 unsigned long loops_per_jiffy = (1<<12); 212 EXPORT_SYMBOL(loops_per_jiffy); 213 214 static int __init debug_kernel(char *str) 215 { 216 console_loglevel = CONSOLE_LOGLEVEL_DEBUG; 217 return 0; 218 } 219 220 static int __init quiet_kernel(char *str) 221 { 222 console_loglevel = CONSOLE_LOGLEVEL_QUIET; 223 return 0; 224 } 225 226 early_param("debug", debug_kernel); 227 early_param("quiet", quiet_kernel); 228 229 static int __init loglevel(char *str) 230 { 231 int newlevel; 232 233 /* 234 * Only update loglevel value when a correct setting was passed, 235 * to prevent blind crashes (when loglevel being set to 0) that 236 * are quite hard to debug 237 */ 238 if (get_option(&str, &newlevel)) { 239 console_loglevel = newlevel; 240 return 0; 241 } 242 243 return -EINVAL; 244 } 245 246 early_param("loglevel", loglevel); 247 248 /* Change NUL term back to "=", to make "param" the whole string. */ 249 static int __init repair_env_string(char *param, char *val, 250 const char *unused, void *arg) 251 { 252 if (val) { 253 /* param=val or param="val"? */ 254 if (val == param+strlen(param)+1) 255 val[-1] = '='; 256 else if (val == param+strlen(param)+2) { 257 val[-2] = '='; 258 memmove(val-1, val, strlen(val)+1); 259 val--; 260 } else 261 BUG(); 262 } 263 return 0; 264 } 265 266 /* Anything after -- gets handed straight to init. */ 267 static int __init set_init_arg(char *param, char *val, 268 const char *unused, void *arg) 269 { 270 unsigned int i; 271 272 if (panic_later) 273 return 0; 274 275 repair_env_string(param, val, unused, NULL); 276 277 for (i = 0; argv_init[i]; i++) { 278 if (i == MAX_INIT_ARGS) { 279 panic_later = "init"; 280 panic_param = param; 281 return 0; 282 } 283 } 284 argv_init[i] = param; 285 return 0; 286 } 287 288 /* 289 * Unknown boot options get handed to init, unless they look like 290 * unused parameters (modprobe will find them in /proc/cmdline). 291 */ 292 static int __init unknown_bootoption(char *param, char *val, 293 const char *unused, void *arg) 294 { 295 repair_env_string(param, val, unused, NULL); 296 297 /* Handle obsolete-style parameters */ 298 if (obsolete_checksetup(param)) 299 return 0; 300 301 /* Unused module parameter. */ 302 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) 303 return 0; 304 305 if (panic_later) 306 return 0; 307 308 if (val) { 309 /* Environment option */ 310 unsigned int i; 311 for (i = 0; envp_init[i]; i++) { 312 if (i == MAX_INIT_ENVS) { 313 panic_later = "env"; 314 panic_param = param; 315 } 316 if (!strncmp(param, envp_init[i], val - param)) 317 break; 318 } 319 envp_init[i] = param; 320 } else { 321 /* Command line option */ 322 unsigned int i; 323 for (i = 0; argv_init[i]; i++) { 324 if (i == MAX_INIT_ARGS) { 325 panic_later = "init"; 326 panic_param = param; 327 } 328 } 329 argv_init[i] = param; 330 } 331 return 0; 332 } 333 334 static int __init init_setup(char *str) 335 { 336 unsigned int i; 337 338 execute_command = str; 339 /* 340 * In case LILO is going to boot us with default command line, 341 * it prepends "auto" before the whole cmdline which makes 342 * the shell think it should execute a script with such name. 343 * So we ignore all arguments entered _before_ init=... [MJ] 344 */ 345 for (i = 1; i < MAX_INIT_ARGS; i++) 346 argv_init[i] = NULL; 347 return 1; 348 } 349 __setup("init=", init_setup); 350 351 static int __init rdinit_setup(char *str) 352 { 353 unsigned int i; 354 355 ramdisk_execute_command = str; 356 /* See "auto" comment in init_setup */ 357 for (i = 1; i < MAX_INIT_ARGS; i++) 358 argv_init[i] = NULL; 359 return 1; 360 } 361 __setup("rdinit=", rdinit_setup); 362 363 #ifndef CONFIG_SMP 364 static const unsigned int setup_max_cpus = NR_CPUS; 365 static inline void setup_nr_cpu_ids(void) { } 366 static inline void smp_prepare_cpus(unsigned int maxcpus) { } 367 #endif 368 369 /* 370 * We need to store the untouched command line for future reference. 371 * We also need to store the touched command line since the parameter 372 * parsing is performed in place, and we should allow a component to 373 * store reference of name/value for future reference. 374 */ 375 static void __init setup_command_line(char *command_line) 376 { 377 size_t len = strlen(boot_command_line) + 1; 378 379 saved_command_line = memblock_alloc(len, SMP_CACHE_BYTES); 380 if (!saved_command_line) 381 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 382 383 initcall_command_line = memblock_alloc(len, SMP_CACHE_BYTES); 384 if (!initcall_command_line) 385 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 386 387 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES); 388 if (!static_command_line) 389 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 390 391 strcpy(saved_command_line, boot_command_line); 392 strcpy(static_command_line, command_line); 393 } 394 395 /* 396 * We need to finalize in a non-__init function or else race conditions 397 * between the root thread and the init thread may cause start_kernel to 398 * be reaped by free_initmem before the root thread has proceeded to 399 * cpu_idle. 400 * 401 * gcc-3.4 accidentally inlines this function, so use noinline. 402 */ 403 404 static __initdata DECLARE_COMPLETION(kthreadd_done); 405 406 noinline void __ref rest_init(void) 407 { 408 struct task_struct *tsk; 409 int pid; 410 411 rcu_scheduler_starting(); 412 /* 413 * We need to spawn init first so that it obtains pid 1, however 414 * the init task will end up wanting to create kthreads, which, if 415 * we schedule it before we create kthreadd, will OOPS. 416 */ 417 pid = kernel_thread(kernel_init, NULL, CLONE_FS); 418 /* 419 * Pin init on the boot CPU. Task migration is not properly working 420 * until sched_init_smp() has been run. It will set the allowed 421 * CPUs for init to the non isolated CPUs. 422 */ 423 rcu_read_lock(); 424 tsk = find_task_by_pid_ns(pid, &init_pid_ns); 425 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); 426 rcu_read_unlock(); 427 428 numa_default_policy(); 429 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); 430 rcu_read_lock(); 431 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 432 rcu_read_unlock(); 433 434 /* 435 * Enable might_sleep() and smp_processor_id() checks. 436 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y 437 * kernel_thread() would trigger might_sleep() splats. With 438 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled 439 * already, but it's stuck on the kthreadd_done completion. 440 */ 441 system_state = SYSTEM_SCHEDULING; 442 443 complete(&kthreadd_done); 444 445 /* 446 * The boot idle thread must execute schedule() 447 * at least once to get things moving: 448 */ 449 schedule_preempt_disabled(); 450 /* Call into cpu_idle with preempt disabled */ 451 cpu_startup_entry(CPUHP_ONLINE); 452 } 453 454 /* Check for early params. */ 455 static int __init do_early_param(char *param, char *val, 456 const char *unused, void *arg) 457 { 458 const struct obs_kernel_param *p; 459 460 for (p = __setup_start; p < __setup_end; p++) { 461 if ((p->early && parameq(param, p->str)) || 462 (strcmp(param, "console") == 0 && 463 strcmp(p->str, "earlycon") == 0) 464 ) { 465 if (p->setup_func(val) != 0) 466 pr_warn("Malformed early option '%s'\n", param); 467 } 468 } 469 /* We accept everything at this stage. */ 470 return 0; 471 } 472 473 void __init parse_early_options(char *cmdline) 474 { 475 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, 476 do_early_param); 477 } 478 479 /* Arch code calls this early on, or if not, just before other parsing. */ 480 void __init parse_early_param(void) 481 { 482 static int done __initdata; 483 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 484 485 if (done) 486 return; 487 488 /* All fall through to do_early_param. */ 489 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 490 parse_early_options(tmp_cmdline); 491 done = 1; 492 } 493 494 void __init __weak arch_post_acpi_subsys_init(void) { } 495 496 void __init __weak smp_setup_processor_id(void) 497 { 498 } 499 500 # if THREAD_SIZE >= PAGE_SIZE 501 void __init __weak thread_stack_cache_init(void) 502 { 503 } 504 #endif 505 506 void __init __weak mem_encrypt_init(void) { } 507 508 void __init __weak poking_init(void) { } 509 510 void __init __weak pgtable_cache_init(void) { } 511 512 bool initcall_debug; 513 core_param(initcall_debug, initcall_debug, bool, 0644); 514 515 #ifdef TRACEPOINTS_ENABLED 516 static void __init initcall_debug_enable(void); 517 #else 518 static inline void initcall_debug_enable(void) 519 { 520 } 521 #endif 522 523 /* Report memory auto-initialization states for this boot. */ 524 static void __init report_meminit(void) 525 { 526 const char *stack; 527 528 if (IS_ENABLED(CONFIG_INIT_STACK_ALL)) 529 stack = "all"; 530 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL)) 531 stack = "byref_all"; 532 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF)) 533 stack = "byref"; 534 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER)) 535 stack = "__user"; 536 else 537 stack = "off"; 538 539 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n", 540 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off", 541 want_init_on_free() ? "on" : "off"); 542 if (want_init_on_free()) 543 pr_info("mem auto-init: clearing system memory may take some time...\n"); 544 } 545 546 /* 547 * Set up kernel memory allocators 548 */ 549 static void __init mm_init(void) 550 { 551 /* 552 * page_ext requires contiguous pages, 553 * bigger than MAX_ORDER unless SPARSEMEM. 554 */ 555 page_ext_init_flatmem(); 556 report_meminit(); 557 mem_init(); 558 kmem_cache_init(); 559 kmemleak_init(); 560 pgtable_init(); 561 debug_objects_mem_init(); 562 vmalloc_init(); 563 ioremap_huge_init(); 564 /* Should be run before the first non-init thread is created */ 565 init_espfix_bsp(); 566 /* Should be run after espfix64 is set up. */ 567 pti_init(); 568 } 569 570 void __init __weak arch_call_rest_init(void) 571 { 572 rest_init(); 573 } 574 575 asmlinkage __visible void __init start_kernel(void) 576 { 577 char *command_line; 578 char *after_dashes; 579 580 set_task_stack_end_magic(&init_task); 581 smp_setup_processor_id(); 582 debug_objects_early_init(); 583 584 cgroup_init_early(); 585 586 local_irq_disable(); 587 early_boot_irqs_disabled = true; 588 589 /* 590 * Interrupts are still disabled. Do necessary setups, then 591 * enable them. 592 */ 593 boot_cpu_init(); 594 page_address_init(); 595 pr_notice("%s", linux_banner); 596 setup_arch(&command_line); 597 setup_command_line(command_line); 598 setup_nr_cpu_ids(); 599 setup_per_cpu_areas(); 600 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 601 boot_cpu_hotplug_init(); 602 603 build_all_zonelists(NULL); 604 page_alloc_init(); 605 606 pr_notice("Kernel command line: %s\n", boot_command_line); 607 /* parameters may set static keys */ 608 jump_label_init(); 609 parse_early_param(); 610 after_dashes = parse_args("Booting kernel", 611 static_command_line, __start___param, 612 __stop___param - __start___param, 613 -1, -1, NULL, &unknown_bootoption); 614 if (!IS_ERR_OR_NULL(after_dashes)) 615 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 616 NULL, set_init_arg); 617 618 /* 619 * These use large bootmem allocations and must precede 620 * kmem_cache_init() 621 */ 622 setup_log_buf(0); 623 vfs_caches_init_early(); 624 sort_main_extable(); 625 trap_init(); 626 mm_init(); 627 628 ftrace_init(); 629 630 /* trace_printk can be enabled here */ 631 early_trace_init(); 632 633 /* 634 * Set up the scheduler prior starting any interrupts (such as the 635 * timer interrupt). Full topology setup happens at smp_init() 636 * time - but meanwhile we still have a functioning scheduler. 637 */ 638 sched_init(); 639 /* 640 * Disable preemption - early bootup scheduling is extremely 641 * fragile until we cpu_idle() for the first time. 642 */ 643 preempt_disable(); 644 if (WARN(!irqs_disabled(), 645 "Interrupts were enabled *very* early, fixing it\n")) 646 local_irq_disable(); 647 radix_tree_init(); 648 649 /* 650 * Set up housekeeping before setting up workqueues to allow the unbound 651 * workqueue to take non-housekeeping into account. 652 */ 653 housekeeping_init(); 654 655 /* 656 * Allow workqueue creation and work item queueing/cancelling 657 * early. Work item execution depends on kthreads and starts after 658 * workqueue_init(). 659 */ 660 workqueue_init_early(); 661 662 rcu_init(); 663 664 /* Trace events are available after this */ 665 trace_init(); 666 667 if (initcall_debug) 668 initcall_debug_enable(); 669 670 context_tracking_init(); 671 /* init some links before init_ISA_irqs() */ 672 early_irq_init(); 673 init_IRQ(); 674 tick_init(); 675 rcu_init_nohz(); 676 init_timers(); 677 hrtimers_init(); 678 softirq_init(); 679 timekeeping_init(); 680 681 /* 682 * For best initial stack canary entropy, prepare it after: 683 * - setup_arch() for any UEFI RNG entropy and boot cmdline access 684 * - timekeeping_init() for ktime entropy used in rand_initialize() 685 * - rand_initialize() to get any arch-specific entropy like RDRAND 686 * - add_latent_entropy() to get any latent entropy 687 * - adding command line entropy 688 */ 689 rand_initialize(); 690 add_latent_entropy(); 691 add_device_randomness(command_line, strlen(command_line)); 692 boot_init_stack_canary(); 693 694 time_init(); 695 printk_safe_init(); 696 perf_event_init(); 697 profile_init(); 698 call_function_init(); 699 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 700 701 early_boot_irqs_disabled = false; 702 local_irq_enable(); 703 704 kmem_cache_init_late(); 705 706 /* 707 * HACK ALERT! This is early. We're enabling the console before 708 * we've done PCI setups etc, and console_init() must be aware of 709 * this. But we do want output early, in case something goes wrong. 710 */ 711 console_init(); 712 if (panic_later) 713 panic("Too many boot %s vars at `%s'", panic_later, 714 panic_param); 715 716 lockdep_init(); 717 718 /* 719 * Need to run this when irqs are enabled, because it wants 720 * to self-test [hard/soft]-irqs on/off lock inversion bugs 721 * too: 722 */ 723 locking_selftest(); 724 725 /* 726 * This needs to be called before any devices perform DMA 727 * operations that might use the SWIOTLB bounce buffers. It will 728 * mark the bounce buffers as decrypted so that their usage will 729 * not cause "plain-text" data to be decrypted when accessed. 730 */ 731 mem_encrypt_init(); 732 733 #ifdef CONFIG_BLK_DEV_INITRD 734 if (initrd_start && !initrd_below_start_ok && 735 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 736 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 737 page_to_pfn(virt_to_page((void *)initrd_start)), 738 min_low_pfn); 739 initrd_start = 0; 740 } 741 #endif 742 setup_per_cpu_pageset(); 743 numa_policy_init(); 744 acpi_early_init(); 745 if (late_time_init) 746 late_time_init(); 747 sched_clock_init(); 748 calibrate_delay(); 749 pid_idr_init(); 750 anon_vma_init(); 751 #ifdef CONFIG_X86 752 if (efi_enabled(EFI_RUNTIME_SERVICES)) 753 efi_enter_virtual_mode(); 754 #endif 755 thread_stack_cache_init(); 756 cred_init(); 757 fork_init(); 758 proc_caches_init(); 759 uts_ns_init(); 760 buffer_init(); 761 key_init(); 762 security_init(); 763 dbg_late_init(); 764 vfs_caches_init(); 765 pagecache_init(); 766 signals_init(); 767 seq_file_init(); 768 proc_root_init(); 769 nsfs_init(); 770 cpuset_init(); 771 cgroup_init(); 772 taskstats_init_early(); 773 delayacct_init(); 774 775 poking_init(); 776 check_bugs(); 777 778 acpi_subsystem_init(); 779 arch_post_acpi_subsys_init(); 780 sfi_init_late(); 781 782 /* Do the rest non-__init'ed, we're now alive */ 783 arch_call_rest_init(); 784 } 785 786 /* Call all constructor functions linked into the kernel. */ 787 static void __init do_ctors(void) 788 { 789 #ifdef CONFIG_CONSTRUCTORS 790 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 791 792 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 793 (*fn)(); 794 #endif 795 } 796 797 #ifdef CONFIG_KALLSYMS 798 struct blacklist_entry { 799 struct list_head next; 800 char *buf; 801 }; 802 803 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 804 805 static int __init initcall_blacklist(char *str) 806 { 807 char *str_entry; 808 struct blacklist_entry *entry; 809 810 /* str argument is a comma-separated list of functions */ 811 do { 812 str_entry = strsep(&str, ","); 813 if (str_entry) { 814 pr_debug("blacklisting initcall %s\n", str_entry); 815 entry = memblock_alloc(sizeof(*entry), 816 SMP_CACHE_BYTES); 817 if (!entry) 818 panic("%s: Failed to allocate %zu bytes\n", 819 __func__, sizeof(*entry)); 820 entry->buf = memblock_alloc(strlen(str_entry) + 1, 821 SMP_CACHE_BYTES); 822 if (!entry->buf) 823 panic("%s: Failed to allocate %zu bytes\n", 824 __func__, strlen(str_entry) + 1); 825 strcpy(entry->buf, str_entry); 826 list_add(&entry->next, &blacklisted_initcalls); 827 } 828 } while (str_entry); 829 830 return 0; 831 } 832 833 static bool __init_or_module initcall_blacklisted(initcall_t fn) 834 { 835 struct blacklist_entry *entry; 836 char fn_name[KSYM_SYMBOL_LEN]; 837 unsigned long addr; 838 839 if (list_empty(&blacklisted_initcalls)) 840 return false; 841 842 addr = (unsigned long) dereference_function_descriptor(fn); 843 sprint_symbol_no_offset(fn_name, addr); 844 845 /* 846 * fn will be "function_name [module_name]" where [module_name] is not 847 * displayed for built-in init functions. Strip off the [module_name]. 848 */ 849 strreplace(fn_name, ' ', '\0'); 850 851 list_for_each_entry(entry, &blacklisted_initcalls, next) { 852 if (!strcmp(fn_name, entry->buf)) { 853 pr_debug("initcall %s blacklisted\n", fn_name); 854 return true; 855 } 856 } 857 858 return false; 859 } 860 #else 861 static int __init initcall_blacklist(char *str) 862 { 863 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 864 return 0; 865 } 866 867 static bool __init_or_module initcall_blacklisted(initcall_t fn) 868 { 869 return false; 870 } 871 #endif 872 __setup("initcall_blacklist=", initcall_blacklist); 873 874 static __init_or_module void 875 trace_initcall_start_cb(void *data, initcall_t fn) 876 { 877 ktime_t *calltime = (ktime_t *)data; 878 879 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current)); 880 *calltime = ktime_get(); 881 } 882 883 static __init_or_module void 884 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 885 { 886 ktime_t *calltime = (ktime_t *)data; 887 ktime_t delta, rettime; 888 unsigned long long duration; 889 890 rettime = ktime_get(); 891 delta = ktime_sub(rettime, *calltime); 892 duration = (unsigned long long) ktime_to_ns(delta) >> 10; 893 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n", 894 fn, ret, duration); 895 } 896 897 static ktime_t initcall_calltime; 898 899 #ifdef TRACEPOINTS_ENABLED 900 static void __init initcall_debug_enable(void) 901 { 902 int ret; 903 904 ret = register_trace_initcall_start(trace_initcall_start_cb, 905 &initcall_calltime); 906 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 907 &initcall_calltime); 908 WARN(ret, "Failed to register initcall tracepoints\n"); 909 } 910 # define do_trace_initcall_start trace_initcall_start 911 # define do_trace_initcall_finish trace_initcall_finish 912 #else 913 static inline void do_trace_initcall_start(initcall_t fn) 914 { 915 if (!initcall_debug) 916 return; 917 trace_initcall_start_cb(&initcall_calltime, fn); 918 } 919 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 920 { 921 if (!initcall_debug) 922 return; 923 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 924 } 925 #endif /* !TRACEPOINTS_ENABLED */ 926 927 int __init_or_module do_one_initcall(initcall_t fn) 928 { 929 int count = preempt_count(); 930 char msgbuf[64]; 931 int ret; 932 933 if (initcall_blacklisted(fn)) 934 return -EPERM; 935 936 do_trace_initcall_start(fn); 937 ret = fn(); 938 do_trace_initcall_finish(fn, ret); 939 940 msgbuf[0] = 0; 941 942 if (preempt_count() != count) { 943 sprintf(msgbuf, "preemption imbalance "); 944 preempt_count_set(count); 945 } 946 if (irqs_disabled()) { 947 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 948 local_irq_enable(); 949 } 950 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 951 952 add_latent_entropy(); 953 return ret; 954 } 955 956 957 extern initcall_entry_t __initcall_start[]; 958 extern initcall_entry_t __initcall0_start[]; 959 extern initcall_entry_t __initcall1_start[]; 960 extern initcall_entry_t __initcall2_start[]; 961 extern initcall_entry_t __initcall3_start[]; 962 extern initcall_entry_t __initcall4_start[]; 963 extern initcall_entry_t __initcall5_start[]; 964 extern initcall_entry_t __initcall6_start[]; 965 extern initcall_entry_t __initcall7_start[]; 966 extern initcall_entry_t __initcall_end[]; 967 968 static initcall_entry_t *initcall_levels[] __initdata = { 969 __initcall0_start, 970 __initcall1_start, 971 __initcall2_start, 972 __initcall3_start, 973 __initcall4_start, 974 __initcall5_start, 975 __initcall6_start, 976 __initcall7_start, 977 __initcall_end, 978 }; 979 980 /* Keep these in sync with initcalls in include/linux/init.h */ 981 static const char *initcall_level_names[] __initdata = { 982 "pure", 983 "core", 984 "postcore", 985 "arch", 986 "subsys", 987 "fs", 988 "device", 989 "late", 990 }; 991 992 static void __init do_initcall_level(int level) 993 { 994 initcall_entry_t *fn; 995 996 strcpy(initcall_command_line, saved_command_line); 997 parse_args(initcall_level_names[level], 998 initcall_command_line, __start___param, 999 __stop___param - __start___param, 1000 level, level, 1001 NULL, &repair_env_string); 1002 1003 trace_initcall_level(initcall_level_names[level]); 1004 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1005 do_one_initcall(initcall_from_entry(fn)); 1006 } 1007 1008 static void __init do_initcalls(void) 1009 { 1010 int level; 1011 1012 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) 1013 do_initcall_level(level); 1014 } 1015 1016 /* 1017 * Ok, the machine is now initialized. None of the devices 1018 * have been touched yet, but the CPU subsystem is up and 1019 * running, and memory and process management works. 1020 * 1021 * Now we can finally start doing some real work.. 1022 */ 1023 static void __init do_basic_setup(void) 1024 { 1025 cpuset_init_smp(); 1026 driver_init(); 1027 init_irq_proc(); 1028 do_ctors(); 1029 usermodehelper_enable(); 1030 do_initcalls(); 1031 } 1032 1033 static void __init do_pre_smp_initcalls(void) 1034 { 1035 initcall_entry_t *fn; 1036 1037 trace_initcall_level("early"); 1038 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1039 do_one_initcall(initcall_from_entry(fn)); 1040 } 1041 1042 static int run_init_process(const char *init_filename) 1043 { 1044 argv_init[0] = init_filename; 1045 pr_info("Run %s as init process\n", init_filename); 1046 return do_execve(getname_kernel(init_filename), 1047 (const char __user *const __user *)argv_init, 1048 (const char __user *const __user *)envp_init); 1049 } 1050 1051 static int try_to_run_init_process(const char *init_filename) 1052 { 1053 int ret; 1054 1055 ret = run_init_process(init_filename); 1056 1057 if (ret && ret != -ENOENT) { 1058 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1059 init_filename, ret); 1060 } 1061 1062 return ret; 1063 } 1064 1065 static noinline void __init kernel_init_freeable(void); 1066 1067 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1068 bool rodata_enabled __ro_after_init = true; 1069 static int __init set_debug_rodata(char *str) 1070 { 1071 return strtobool(str, &rodata_enabled); 1072 } 1073 __setup("rodata=", set_debug_rodata); 1074 #endif 1075 1076 #ifdef CONFIG_STRICT_KERNEL_RWX 1077 static void mark_readonly(void) 1078 { 1079 if (rodata_enabled) { 1080 /* 1081 * load_module() results in W+X mappings, which are cleaned 1082 * up with call_rcu(). Let's make sure that queued work is 1083 * flushed so that we don't hit false positives looking for 1084 * insecure pages which are W+X. 1085 */ 1086 rcu_barrier(); 1087 mark_rodata_ro(); 1088 rodata_test(); 1089 } else 1090 pr_info("Kernel memory protection disabled.\n"); 1091 } 1092 #else 1093 static inline void mark_readonly(void) 1094 { 1095 pr_warn("This architecture does not have kernel memory protection.\n"); 1096 } 1097 #endif 1098 1099 void __weak free_initmem(void) 1100 { 1101 free_initmem_default(POISON_FREE_INITMEM); 1102 } 1103 1104 static int __ref kernel_init(void *unused) 1105 { 1106 int ret; 1107 1108 kernel_init_freeable(); 1109 /* need to finish all async __init code before freeing the memory */ 1110 async_synchronize_full(); 1111 ftrace_free_init_mem(); 1112 free_initmem(); 1113 mark_readonly(); 1114 1115 /* 1116 * Kernel mappings are now finalized - update the userspace page-table 1117 * to finalize PTI. 1118 */ 1119 pti_finalize(); 1120 1121 system_state = SYSTEM_RUNNING; 1122 numa_default_policy(); 1123 1124 rcu_end_inkernel_boot(); 1125 1126 if (ramdisk_execute_command) { 1127 ret = run_init_process(ramdisk_execute_command); 1128 if (!ret) 1129 return 0; 1130 pr_err("Failed to execute %s (error %d)\n", 1131 ramdisk_execute_command, ret); 1132 } 1133 1134 /* 1135 * We try each of these until one succeeds. 1136 * 1137 * The Bourne shell can be used instead of init if we are 1138 * trying to recover a really broken machine. 1139 */ 1140 if (execute_command) { 1141 ret = run_init_process(execute_command); 1142 if (!ret) 1143 return 0; 1144 panic("Requested init %s failed (error %d).", 1145 execute_command, ret); 1146 } 1147 if (!try_to_run_init_process("/sbin/init") || 1148 !try_to_run_init_process("/etc/init") || 1149 !try_to_run_init_process("/bin/init") || 1150 !try_to_run_init_process("/bin/sh")) 1151 return 0; 1152 1153 panic("No working init found. Try passing init= option to kernel. " 1154 "See Linux Documentation/admin-guide/init.rst for guidance."); 1155 } 1156 1157 static noinline void __init kernel_init_freeable(void) 1158 { 1159 /* 1160 * Wait until kthreadd is all set-up. 1161 */ 1162 wait_for_completion(&kthreadd_done); 1163 1164 /* Now the scheduler is fully set up and can do blocking allocations */ 1165 gfp_allowed_mask = __GFP_BITS_MASK; 1166 1167 /* 1168 * init can allocate pages on any node 1169 */ 1170 set_mems_allowed(node_states[N_MEMORY]); 1171 1172 cad_pid = task_pid(current); 1173 1174 smp_prepare_cpus(setup_max_cpus); 1175 1176 workqueue_init(); 1177 1178 init_mm_internals(); 1179 1180 do_pre_smp_initcalls(); 1181 lockup_detector_init(); 1182 1183 smp_init(); 1184 sched_init_smp(); 1185 1186 page_alloc_init_late(); 1187 /* Initialize page ext after all struct pages are initialized. */ 1188 page_ext_init(); 1189 1190 do_basic_setup(); 1191 1192 /* Open the /dev/console on the rootfs, this should never fail */ 1193 if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) 1194 pr_err("Warning: unable to open an initial console.\n"); 1195 1196 (void) ksys_dup(0); 1197 (void) ksys_dup(0); 1198 /* 1199 * check if there is an early userspace init. If yes, let it do all 1200 * the work 1201 */ 1202 1203 if (!ramdisk_execute_command) 1204 ramdisk_execute_command = "/init"; 1205 1206 if (ksys_access((const char __user *) 1207 ramdisk_execute_command, 0) != 0) { 1208 ramdisk_execute_command = NULL; 1209 prepare_namespace(); 1210 } 1211 1212 /* 1213 * Ok, we have completed the initial bootup, and 1214 * we're essentially up and running. Get rid of the 1215 * initmem segments and start the user-mode stuff.. 1216 * 1217 * rootfs is available now, try loading the public keys 1218 * and default modules 1219 */ 1220 1221 integrity_load_keys(); 1222 } 1223