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