at v3.9 23 kB view raw
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#include <linux/types.h> 13#include <linux/module.h> 14#include <linux/proc_fs.h> 15#include <linux/kernel.h> 16#include <linux/syscalls.h> 17#include <linux/stackprotector.h> 18#include <linux/string.h> 19#include <linux/ctype.h> 20#include <linux/delay.h> 21#include <linux/ioport.h> 22#include <linux/init.h> 23#include <linux/initrd.h> 24#include <linux/bootmem.h> 25#include <linux/acpi.h> 26#include <linux/tty.h> 27#include <linux/percpu.h> 28#include <linux/kmod.h> 29#include <linux/vmalloc.h> 30#include <linux/kernel_stat.h> 31#include <linux/start_kernel.h> 32#include <linux/security.h> 33#include <linux/smp.h> 34#include <linux/profile.h> 35#include <linux/rcupdate.h> 36#include <linux/moduleparam.h> 37#include <linux/kallsyms.h> 38#include <linux/writeback.h> 39#include <linux/cpu.h> 40#include <linux/cpuset.h> 41#include <linux/cgroup.h> 42#include <linux/efi.h> 43#include <linux/tick.h> 44#include <linux/interrupt.h> 45#include <linux/taskstats_kern.h> 46#include <linux/delayacct.h> 47#include <linux/unistd.h> 48#include <linux/rmap.h> 49#include <linux/mempolicy.h> 50#include <linux/key.h> 51#include <linux/buffer_head.h> 52#include <linux/page_cgroup.h> 53#include <linux/debug_locks.h> 54#include <linux/debugobjects.h> 55#include <linux/lockdep.h> 56#include <linux/kmemleak.h> 57#include <linux/pid_namespace.h> 58#include <linux/device.h> 59#include <linux/kthread.h> 60#include <linux/sched.h> 61#include <linux/signal.h> 62#include <linux/idr.h> 63#include <linux/kgdb.h> 64#include <linux/ftrace.h> 65#include <linux/async.h> 66#include <linux/kmemcheck.h> 67#include <linux/sfi.h> 68#include <linux/shmem_fs.h> 69#include <linux/slab.h> 70#include <linux/perf_event.h> 71#include <linux/file.h> 72#include <linux/ptrace.h> 73#include <linux/blkdev.h> 74#include <linux/elevator.h> 75 76#include <asm/io.h> 77#include <asm/bugs.h> 78#include <asm/setup.h> 79#include <asm/sections.h> 80#include <asm/cacheflush.h> 81 82#ifdef CONFIG_X86_LOCAL_APIC 83#include <asm/smp.h> 84#endif 85 86static int kernel_init(void *); 87 88extern void init_IRQ(void); 89extern void fork_init(unsigned long); 90extern void mca_init(void); 91extern void sbus_init(void); 92extern void radix_tree_init(void); 93#ifndef CONFIG_DEBUG_RODATA 94static inline void mark_rodata_ro(void) { } 95#endif 96 97#ifdef CONFIG_TC 98extern void tc_init(void); 99#endif 100 101/* 102 * Debug helper: via this flag we know that we are in 'early bootup code' 103 * where only the boot processor is running with IRQ disabled. This means 104 * two things - IRQ must not be enabled before the flag is cleared and some 105 * operations which are not allowed with IRQ disabled are allowed while the 106 * flag is set. 107 */ 108bool early_boot_irqs_disabled __read_mostly; 109 110enum system_states system_state __read_mostly; 111EXPORT_SYMBOL(system_state); 112 113/* 114 * Boot command-line arguments 115 */ 116#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT 117#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT 118 119extern void time_init(void); 120/* Default late time init is NULL. archs can override this later. */ 121void (*__initdata late_time_init)(void); 122extern void softirq_init(void); 123 124/* Untouched command line saved by arch-specific code. */ 125char __initdata boot_command_line[COMMAND_LINE_SIZE]; 126/* Untouched saved command line (eg. for /proc) */ 127char *saved_command_line; 128/* Command line for parameter parsing */ 129static char *static_command_line; 130 131static char *execute_command; 132static char *ramdisk_execute_command; 133 134/* 135 * If set, this is an indication to the drivers that reset the underlying 136 * device before going ahead with the initialization otherwise driver might 137 * rely on the BIOS and skip the reset operation. 138 * 139 * This is useful if kernel is booting in an unreliable environment. 140 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been 141 * skipped and devices will be in unknown state. 142 */ 143unsigned int reset_devices; 144EXPORT_SYMBOL(reset_devices); 145 146static int __init set_reset_devices(char *str) 147{ 148 reset_devices = 1; 149 return 1; 150} 151 152__setup("reset_devices", set_reset_devices); 153 154static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 155const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 156static const char *panic_later, *panic_param; 157 158extern const struct obs_kernel_param __setup_start[], __setup_end[]; 159 160static int __init obsolete_checksetup(char *line) 161{ 162 const struct obs_kernel_param *p; 163 int had_early_param = 0; 164 165 p = __setup_start; 166 do { 167 int n = strlen(p->str); 168 if (parameqn(line, p->str, n)) { 169 if (p->early) { 170 /* Already done in parse_early_param? 171 * (Needs exact match on param part). 172 * Keep iterating, as we can have early 173 * params and __setups of same names 8( */ 174 if (line[n] == '\0' || line[n] == '=') 175 had_early_param = 1; 176 } else if (!p->setup_func) { 177 printk(KERN_WARNING "Parameter %s is obsolete," 178 " ignored\n", p->str); 179 return 1; 180 } else if (p->setup_func(line + n)) 181 return 1; 182 } 183 p++; 184 } while (p < __setup_end); 185 186 return had_early_param; 187} 188 189/* 190 * This should be approx 2 Bo*oMips to start (note initial shift), and will 191 * still work even if initially too large, it will just take slightly longer 192 */ 193unsigned long loops_per_jiffy = (1<<12); 194 195EXPORT_SYMBOL(loops_per_jiffy); 196 197static int __init debug_kernel(char *str) 198{ 199 console_loglevel = 10; 200 return 0; 201} 202 203static int __init quiet_kernel(char *str) 204{ 205 console_loglevel = 4; 206 return 0; 207} 208 209early_param("debug", debug_kernel); 210early_param("quiet", quiet_kernel); 211 212static int __init loglevel(char *str) 213{ 214 int newlevel; 215 216 /* 217 * Only update loglevel value when a correct setting was passed, 218 * to prevent blind crashes (when loglevel being set to 0) that 219 * are quite hard to debug 220 */ 221 if (get_option(&str, &newlevel)) { 222 console_loglevel = newlevel; 223 return 0; 224 } 225 226 return -EINVAL; 227} 228 229early_param("loglevel", loglevel); 230 231/* Change NUL term back to "=", to make "param" the whole string. */ 232static int __init repair_env_string(char *param, char *val, const char *unused) 233{ 234 if (val) { 235 /* param=val or param="val"? */ 236 if (val == param+strlen(param)+1) 237 val[-1] = '='; 238 else if (val == param+strlen(param)+2) { 239 val[-2] = '='; 240 memmove(val-1, val, strlen(val)+1); 241 val--; 242 } else 243 BUG(); 244 } 245 return 0; 246} 247 248/* 249 * Unknown boot options get handed to init, unless they look like 250 * unused parameters (modprobe will find them in /proc/cmdline). 251 */ 252static int __init unknown_bootoption(char *param, char *val, const char *unused) 253{ 254 repair_env_string(param, val, unused); 255 256 /* Handle obsolete-style parameters */ 257 if (obsolete_checksetup(param)) 258 return 0; 259 260 /* Unused module parameter. */ 261 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) 262 return 0; 263 264 if (panic_later) 265 return 0; 266 267 if (val) { 268 /* Environment option */ 269 unsigned int i; 270 for (i = 0; envp_init[i]; i++) { 271 if (i == MAX_INIT_ENVS) { 272 panic_later = "Too many boot env vars at `%s'"; 273 panic_param = param; 274 } 275 if (!strncmp(param, envp_init[i], val - param)) 276 break; 277 } 278 envp_init[i] = param; 279 } else { 280 /* Command line option */ 281 unsigned int i; 282 for (i = 0; argv_init[i]; i++) { 283 if (i == MAX_INIT_ARGS) { 284 panic_later = "Too many boot init vars at `%s'"; 285 panic_param = param; 286 } 287 } 288 argv_init[i] = param; 289 } 290 return 0; 291} 292 293static int __init init_setup(char *str) 294{ 295 unsigned int i; 296 297 execute_command = str; 298 /* 299 * In case LILO is going to boot us with default command line, 300 * it prepends "auto" before the whole cmdline which makes 301 * the shell think it should execute a script with such name. 302 * So we ignore all arguments entered _before_ init=... [MJ] 303 */ 304 for (i = 1; i < MAX_INIT_ARGS; i++) 305 argv_init[i] = NULL; 306 return 1; 307} 308__setup("init=", init_setup); 309 310static int __init rdinit_setup(char *str) 311{ 312 unsigned int i; 313 314 ramdisk_execute_command = str; 315 /* See "auto" comment in init_setup */ 316 for (i = 1; i < MAX_INIT_ARGS; i++) 317 argv_init[i] = NULL; 318 return 1; 319} 320__setup("rdinit=", rdinit_setup); 321 322#ifndef CONFIG_SMP 323static const unsigned int setup_max_cpus = NR_CPUS; 324#ifdef CONFIG_X86_LOCAL_APIC 325static void __init smp_init(void) 326{ 327 APIC_init_uniprocessor(); 328} 329#else 330#define smp_init() do { } while (0) 331#endif 332 333static inline void setup_nr_cpu_ids(void) { } 334static inline void smp_prepare_cpus(unsigned int maxcpus) { } 335#endif 336 337/* 338 * We need to store the untouched command line for future reference. 339 * We also need to store the touched command line since the parameter 340 * parsing is performed in place, and we should allow a component to 341 * store reference of name/value for future reference. 342 */ 343static void __init setup_command_line(char *command_line) 344{ 345 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1); 346 static_command_line = alloc_bootmem(strlen (command_line)+1); 347 strcpy (saved_command_line, boot_command_line); 348 strcpy (static_command_line, command_line); 349} 350 351/* 352 * We need to finalize in a non-__init function or else race conditions 353 * between the root thread and the init thread may cause start_kernel to 354 * be reaped by free_initmem before the root thread has proceeded to 355 * cpu_idle. 356 * 357 * gcc-3.4 accidentally inlines this function, so use noinline. 358 */ 359 360static __initdata DECLARE_COMPLETION(kthreadd_done); 361 362static noinline void __init_refok rest_init(void) 363{ 364 int pid; 365 366 rcu_scheduler_starting(); 367 /* 368 * We need to spawn init first so that it obtains pid 1, however 369 * the init task will end up wanting to create kthreads, which, if 370 * we schedule it before we create kthreadd, will OOPS. 371 */ 372 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND); 373 numa_default_policy(); 374 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); 375 rcu_read_lock(); 376 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 377 rcu_read_unlock(); 378 complete(&kthreadd_done); 379 380 /* 381 * The boot idle thread must execute schedule() 382 * at least once to get things moving: 383 */ 384 init_idle_bootup_task(current); 385 schedule_preempt_disabled(); 386 /* Call into cpu_idle with preempt disabled */ 387 cpu_idle(); 388} 389 390/* Check for early params. */ 391static int __init do_early_param(char *param, char *val, const char *unused) 392{ 393 const struct obs_kernel_param *p; 394 395 for (p = __setup_start; p < __setup_end; p++) { 396 if ((p->early && parameq(param, p->str)) || 397 (strcmp(param, "console") == 0 && 398 strcmp(p->str, "earlycon") == 0) 399 ) { 400 if (p->setup_func(val) != 0) 401 printk(KERN_WARNING 402 "Malformed early option '%s'\n", param); 403 } 404 } 405 /* We accept everything at this stage. */ 406 return 0; 407} 408 409void __init parse_early_options(char *cmdline) 410{ 411 parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param); 412} 413 414/* Arch code calls this early on, or if not, just before other parsing. */ 415void __init parse_early_param(void) 416{ 417 static __initdata int done = 0; 418 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE]; 419 420 if (done) 421 return; 422 423 /* All fall through to do_early_param. */ 424 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 425 parse_early_options(tmp_cmdline); 426 done = 1; 427} 428 429/* 430 * Activate the first processor. 431 */ 432 433static void __init boot_cpu_init(void) 434{ 435 int cpu = smp_processor_id(); 436 /* Mark the boot cpu "present", "online" etc for SMP and UP case */ 437 set_cpu_online(cpu, true); 438 set_cpu_active(cpu, true); 439 set_cpu_present(cpu, true); 440 set_cpu_possible(cpu, true); 441} 442 443void __init __weak smp_setup_processor_id(void) 444{ 445} 446 447# if THREAD_SIZE >= PAGE_SIZE 448void __init __weak thread_info_cache_init(void) 449{ 450} 451#endif 452 453/* 454 * Set up kernel memory allocators 455 */ 456static void __init mm_init(void) 457{ 458 /* 459 * page_cgroup requires contiguous pages, 460 * bigger than MAX_ORDER unless SPARSEMEM. 461 */ 462 page_cgroup_init_flatmem(); 463 mem_init(); 464 kmem_cache_init(); 465 percpu_init_late(); 466 pgtable_cache_init(); 467 vmalloc_init(); 468} 469 470asmlinkage void __init start_kernel(void) 471{ 472 char * command_line; 473 extern const struct kernel_param __start___param[], __stop___param[]; 474 475 /* 476 * Need to run as early as possible, to initialize the 477 * lockdep hash: 478 */ 479 lockdep_init(); 480 smp_setup_processor_id(); 481 debug_objects_early_init(); 482 483 /* 484 * Set up the the initial canary ASAP: 485 */ 486 boot_init_stack_canary(); 487 488 cgroup_init_early(); 489 490 local_irq_disable(); 491 early_boot_irqs_disabled = true; 492 493/* 494 * Interrupts are still disabled. Do necessary setups, then 495 * enable them 496 */ 497 tick_init(); 498 boot_cpu_init(); 499 page_address_init(); 500 printk(KERN_NOTICE "%s", linux_banner); 501 setup_arch(&command_line); 502 mm_init_owner(&init_mm, &init_task); 503 mm_init_cpumask(&init_mm); 504 setup_command_line(command_line); 505 setup_nr_cpu_ids(); 506 setup_per_cpu_areas(); 507 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 508 509 build_all_zonelists(NULL, NULL); 510 page_alloc_init(); 511 512 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line); 513 parse_early_param(); 514 parse_args("Booting kernel", static_command_line, __start___param, 515 __stop___param - __start___param, 516 -1, -1, &unknown_bootoption); 517 518 jump_label_init(); 519 520 /* 521 * These use large bootmem allocations and must precede 522 * kmem_cache_init() 523 */ 524 setup_log_buf(0); 525 pidhash_init(); 526 vfs_caches_init_early(); 527 sort_main_extable(); 528 trap_init(); 529 mm_init(); 530 531 /* 532 * Set up the scheduler prior starting any interrupts (such as the 533 * timer interrupt). Full topology setup happens at smp_init() 534 * time - but meanwhile we still have a functioning scheduler. 535 */ 536 sched_init(); 537 /* 538 * Disable preemption - early bootup scheduling is extremely 539 * fragile until we cpu_idle() for the first time. 540 */ 541 preempt_disable(); 542 if (!irqs_disabled()) { 543 printk(KERN_WARNING "start_kernel(): bug: interrupts were " 544 "enabled *very* early, fixing it\n"); 545 local_irq_disable(); 546 } 547 idr_init_cache(); 548 perf_event_init(); 549 rcu_init(); 550 radix_tree_init(); 551 /* init some links before init_ISA_irqs() */ 552 early_irq_init(); 553 init_IRQ(); 554 init_timers(); 555 hrtimers_init(); 556 softirq_init(); 557 timekeeping_init(); 558 time_init(); 559 profile_init(); 560 call_function_init(); 561 if (!irqs_disabled()) 562 printk(KERN_CRIT "start_kernel(): bug: interrupts were " 563 "enabled early\n"); 564 early_boot_irqs_disabled = false; 565 local_irq_enable(); 566 567 kmem_cache_init_late(); 568 569 /* 570 * HACK ALERT! This is early. We're enabling the console before 571 * we've done PCI setups etc, and console_init() must be aware of 572 * this. But we do want output early, in case something goes wrong. 573 */ 574 console_init(); 575 if (panic_later) 576 panic(panic_later, panic_param); 577 578 lockdep_info(); 579 580 /* 581 * Need to run this when irqs are enabled, because it wants 582 * to self-test [hard/soft]-irqs on/off lock inversion bugs 583 * too: 584 */ 585 locking_selftest(); 586 587#ifdef CONFIG_BLK_DEV_INITRD 588 if (initrd_start && !initrd_below_start_ok && 589 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 590 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - " 591 "disabling it.\n", 592 page_to_pfn(virt_to_page((void *)initrd_start)), 593 min_low_pfn); 594 initrd_start = 0; 595 } 596#endif 597 page_cgroup_init(); 598 debug_objects_mem_init(); 599 kmemleak_init(); 600 setup_per_cpu_pageset(); 601 numa_policy_init(); 602 if (late_time_init) 603 late_time_init(); 604 sched_clock_init(); 605 calibrate_delay(); 606 pidmap_init(); 607 anon_vma_init(); 608#ifdef CONFIG_X86 609 if (efi_enabled(EFI_RUNTIME_SERVICES)) 610 efi_enter_virtual_mode(); 611#endif 612 thread_info_cache_init(); 613 cred_init(); 614 fork_init(totalram_pages); 615 proc_caches_init(); 616 buffer_init(); 617 key_init(); 618 security_init(); 619 dbg_late_init(); 620 vfs_caches_init(totalram_pages); 621 signals_init(); 622 /* rootfs populating might need page-writeback */ 623 page_writeback_init(); 624#ifdef CONFIG_PROC_FS 625 proc_root_init(); 626#endif 627 cgroup_init(); 628 cpuset_init(); 629 taskstats_init_early(); 630 delayacct_init(); 631 632 check_bugs(); 633 634 acpi_early_init(); /* before LAPIC and SMP init */ 635 sfi_init_late(); 636 637 if (efi_enabled(EFI_RUNTIME_SERVICES)) { 638 efi_late_init(); 639 efi_free_boot_services(); 640 } 641 642 ftrace_init(); 643 644 /* Do the rest non-__init'ed, we're now alive */ 645 rest_init(); 646} 647 648/* Call all constructor functions linked into the kernel. */ 649static void __init do_ctors(void) 650{ 651#ifdef CONFIG_CONSTRUCTORS 652 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 653 654 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 655 (*fn)(); 656#endif 657} 658 659bool initcall_debug; 660core_param(initcall_debug, initcall_debug, bool, 0644); 661 662static char msgbuf[64]; 663 664static int __init_or_module do_one_initcall_debug(initcall_t fn) 665{ 666 ktime_t calltime, delta, rettime; 667 unsigned long long duration; 668 int ret; 669 670 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current)); 671 calltime = ktime_get(); 672 ret = fn(); 673 rettime = ktime_get(); 674 delta = ktime_sub(rettime, calltime); 675 duration = (unsigned long long) ktime_to_ns(delta) >> 10; 676 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn, 677 ret, duration); 678 679 return ret; 680} 681 682int __init_or_module do_one_initcall(initcall_t fn) 683{ 684 int count = preempt_count(); 685 int ret; 686 687 if (initcall_debug) 688 ret = do_one_initcall_debug(fn); 689 else 690 ret = fn(); 691 692 msgbuf[0] = 0; 693 694 if (ret && ret != -ENODEV && initcall_debug) 695 sprintf(msgbuf, "error code %d ", ret); 696 697 if (preempt_count() != count) { 698 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf)); 699 preempt_count() = count; 700 } 701 if (irqs_disabled()) { 702 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 703 local_irq_enable(); 704 } 705 if (msgbuf[0]) { 706 printk("initcall %pF returned with %s\n", fn, msgbuf); 707 } 708 709 return ret; 710} 711 712 713extern initcall_t __initcall_start[]; 714extern initcall_t __initcall0_start[]; 715extern initcall_t __initcall1_start[]; 716extern initcall_t __initcall2_start[]; 717extern initcall_t __initcall3_start[]; 718extern initcall_t __initcall4_start[]; 719extern initcall_t __initcall5_start[]; 720extern initcall_t __initcall6_start[]; 721extern initcall_t __initcall7_start[]; 722extern initcall_t __initcall_end[]; 723 724static initcall_t *initcall_levels[] __initdata = { 725 __initcall0_start, 726 __initcall1_start, 727 __initcall2_start, 728 __initcall3_start, 729 __initcall4_start, 730 __initcall5_start, 731 __initcall6_start, 732 __initcall7_start, 733 __initcall_end, 734}; 735 736/* Keep these in sync with initcalls in include/linux/init.h */ 737static char *initcall_level_names[] __initdata = { 738 "early", 739 "core", 740 "postcore", 741 "arch", 742 "subsys", 743 "fs", 744 "device", 745 "late", 746}; 747 748static void __init do_initcall_level(int level) 749{ 750 extern const struct kernel_param __start___param[], __stop___param[]; 751 initcall_t *fn; 752 753 strcpy(static_command_line, saved_command_line); 754 parse_args(initcall_level_names[level], 755 static_command_line, __start___param, 756 __stop___param - __start___param, 757 level, level, 758 &repair_env_string); 759 760 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 761 do_one_initcall(*fn); 762} 763 764static void __init do_initcalls(void) 765{ 766 int level; 767 768 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) 769 do_initcall_level(level); 770} 771 772/* 773 * Ok, the machine is now initialized. None of the devices 774 * have been touched yet, but the CPU subsystem is up and 775 * running, and memory and process management works. 776 * 777 * Now we can finally start doing some real work.. 778 */ 779static void __init do_basic_setup(void) 780{ 781 cpuset_init_smp(); 782 usermodehelper_init(); 783 shmem_init(); 784 driver_init(); 785 init_irq_proc(); 786 do_ctors(); 787 usermodehelper_enable(); 788 do_initcalls(); 789} 790 791static void __init do_pre_smp_initcalls(void) 792{ 793 initcall_t *fn; 794 795 for (fn = __initcall_start; fn < __initcall0_start; fn++) 796 do_one_initcall(*fn); 797} 798 799/* 800 * This function requests modules which should be loaded by default and is 801 * called twice right after initrd is mounted and right before init is 802 * exec'd. If such modules are on either initrd or rootfs, they will be 803 * loaded before control is passed to userland. 804 */ 805void __init load_default_modules(void) 806{ 807 load_default_elevator_module(); 808} 809 810static int run_init_process(const char *init_filename) 811{ 812 argv_init[0] = init_filename; 813 return do_execve(init_filename, 814 (const char __user *const __user *)argv_init, 815 (const char __user *const __user *)envp_init); 816} 817 818static noinline void __init kernel_init_freeable(void); 819 820static int __ref kernel_init(void *unused) 821{ 822 kernel_init_freeable(); 823 /* need to finish all async __init code before freeing the memory */ 824 async_synchronize_full(); 825 free_initmem(); 826 mark_rodata_ro(); 827 system_state = SYSTEM_RUNNING; 828 numa_default_policy(); 829 830 flush_delayed_fput(); 831 832 if (ramdisk_execute_command) { 833 if (!run_init_process(ramdisk_execute_command)) 834 return 0; 835 printk(KERN_WARNING "Failed to execute %s\n", 836 ramdisk_execute_command); 837 } 838 839 /* 840 * We try each of these until one succeeds. 841 * 842 * The Bourne shell can be used instead of init if we are 843 * trying to recover a really broken machine. 844 */ 845 if (execute_command) { 846 if (!run_init_process(execute_command)) 847 return 0; 848 printk(KERN_WARNING "Failed to execute %s. Attempting " 849 "defaults...\n", execute_command); 850 } 851 if (!run_init_process("/sbin/init") || 852 !run_init_process("/etc/init") || 853 !run_init_process("/bin/init") || 854 !run_init_process("/bin/sh")) 855 return 0; 856 857 panic("No init found. Try passing init= option to kernel. " 858 "See Linux Documentation/init.txt for guidance."); 859} 860 861static noinline void __init kernel_init_freeable(void) 862{ 863 /* 864 * Wait until kthreadd is all set-up. 865 */ 866 wait_for_completion(&kthreadd_done); 867 868 /* Now the scheduler is fully set up and can do blocking allocations */ 869 gfp_allowed_mask = __GFP_BITS_MASK; 870 871 /* 872 * init can allocate pages on any node 873 */ 874 set_mems_allowed(node_states[N_MEMORY]); 875 /* 876 * init can run on any cpu. 877 */ 878 set_cpus_allowed_ptr(current, cpu_all_mask); 879 880 cad_pid = task_pid(current); 881 882 smp_prepare_cpus(setup_max_cpus); 883 884 do_pre_smp_initcalls(); 885 lockup_detector_init(); 886 887 smp_init(); 888 sched_init_smp(); 889 890 do_basic_setup(); 891 892 /* Open the /dev/console on the rootfs, this should never fail */ 893 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) 894 printk(KERN_WARNING "Warning: unable to open an initial console.\n"); 895 896 (void) sys_dup(0); 897 (void) sys_dup(0); 898 /* 899 * check if there is an early userspace init. If yes, let it do all 900 * the work 901 */ 902 903 if (!ramdisk_execute_command) 904 ramdisk_execute_command = "/init"; 905 906 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { 907 ramdisk_execute_command = NULL; 908 prepare_namespace(); 909 } 910 911 /* 912 * Ok, we have completed the initial bootup, and 913 * we're essentially up and running. Get rid of the 914 * initmem segments and start the user-mode stuff.. 915 */ 916 917 /* rootfs is available now, try loading default modules */ 918 load_default_modules(); 919}