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