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#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/utsname.h>
  22#include <linux/ioport.h>
  23#include <linux/init.h>
  24#include <linux/smp_lock.h>
  25#include <linux/initrd.h>
  26#include <linux/bootmem.h>
  27#include <linux/acpi.h>
  28#include <linux/tty.h>
  29#include <linux/gfp.h>
  30#include <linux/percpu.h>
  31#include <linux/kmod.h>
  32#include <linux/vmalloc.h>
  33#include <linux/kernel_stat.h>
  34#include <linux/start_kernel.h>
  35#include <linux/security.h>
  36#include <linux/smp.h>
  37#include <linux/workqueue.h>
  38#include <linux/profile.h>
  39#include <linux/rcupdate.h>
  40#include <linux/moduleparam.h>
  41#include <linux/kallsyms.h>
  42#include <linux/writeback.h>
  43#include <linux/cpu.h>
  44#include <linux/cpuset.h>
  45#include <linux/cgroup.h>
  46#include <linux/efi.h>
  47#include <linux/tick.h>
  48#include <linux/interrupt.h>
  49#include <linux/taskstats_kern.h>
  50#include <linux/delayacct.h>
  51#include <linux/unistd.h>
  52#include <linux/rmap.h>
  53#include <linux/mempolicy.h>
  54#include <linux/key.h>
  55#include <linux/buffer_head.h>
  56#include <linux/page_cgroup.h>
  57#include <linux/debug_locks.h>
  58#include <linux/debugobjects.h>
  59#include <linux/lockdep.h>
  60#include <linux/kmemleak.h>
  61#include <linux/pid_namespace.h>
  62#include <linux/device.h>
  63#include <linux/kthread.h>
  64#include <linux/sched.h>
  65#include <linux/signal.h>
  66#include <linux/idr.h>
  67#include <linux/ftrace.h>
  68#include <linux/async.h>
  69#include <linux/kmemcheck.h>
  70#include <linux/kmemtrace.h>
  71#include <trace/boot.h>
  72
  73#include <asm/io.h>
  74#include <asm/bugs.h>
  75#include <asm/setup.h>
  76#include <asm/sections.h>
  77#include <asm/cacheflush.h>
  78
  79#ifdef CONFIG_X86_LOCAL_APIC
  80#include <asm/smp.h>
  81#endif
  82
  83static int kernel_init(void *);
  84
  85extern void init_IRQ(void);
  86extern void fork_init(unsigned long);
  87extern void mca_init(void);
  88extern void sbus_init(void);
  89extern void prio_tree_init(void);
  90extern void radix_tree_init(void);
  91extern void free_initmem(void);
  92#ifndef CONFIG_DEBUG_RODATA
  93static inline void mark_rodata_ro(void) { }
  94#endif
  95
  96#ifdef CONFIG_TC
  97extern void tc_init(void);
  98#endif
  99
 100enum system_states system_state __read_mostly;
 101EXPORT_SYMBOL(system_state);
 102
 103/*
 104 * Boot command-line arguments
 105 */
 106#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 107#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 108
 109extern void time_init(void);
 110/* Default late time init is NULL. archs can override this later. */
 111void (*__initdata late_time_init)(void);
 112extern void softirq_init(void);
 113
 114/* Untouched command line saved by arch-specific code. */
 115char __initdata boot_command_line[COMMAND_LINE_SIZE];
 116/* Untouched saved command line (eg. for /proc) */
 117char *saved_command_line;
 118/* Command line for parameter parsing */
 119static char *static_command_line;
 120
 121static char *execute_command;
 122static char *ramdisk_execute_command;
 123
 124#ifdef CONFIG_SMP
 125/* Setup configured maximum number of CPUs to activate */
 126unsigned int __initdata setup_max_cpus = NR_CPUS;
 127
 128/*
 129 * Setup routine for controlling SMP activation
 130 *
 131 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 132 * activation entirely (the MPS table probe still happens, though).
 133 *
 134 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 135 * greater than 0, limits the maximum number of CPUs activated in
 136 * SMP mode to <NUM>.
 137 */
 138
 139void __weak arch_disable_smp_support(void) { }
 140
 141static int __init nosmp(char *str)
 142{
 143        setup_max_cpus = 0;
 144        arch_disable_smp_support();
 145
 146        return 0;
 147}
 148
 149early_param("nosmp", nosmp);
 150
 151static int __init maxcpus(char *str)
 152{
 153        get_option(&str, &setup_max_cpus);
 154        if (setup_max_cpus == 0)
 155                arch_disable_smp_support();
 156
 157        return 0;
 158}
 159
 160early_param("maxcpus", maxcpus);
 161#else
 162const unsigned int setup_max_cpus = NR_CPUS;
 163#endif
 164
 165/*
 166 * If set, this is an indication to the drivers that reset the underlying
 167 * device before going ahead with the initialization otherwise driver might
 168 * rely on the BIOS and skip the reset operation.
 169 *
 170 * This is useful if kernel is booting in an unreliable environment.
 171 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 172 * skipped and devices will be in unknown state.
 173 */
 174unsigned int reset_devices;
 175EXPORT_SYMBOL(reset_devices);
 176
 177static int __init set_reset_devices(char *str)
 178{
 179        reset_devices = 1;
 180        return 1;
 181}
 182
 183__setup("reset_devices", set_reset_devices);
 184
 185static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 186char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 187static const char *panic_later, *panic_param;
 188
 189extern struct obs_kernel_param __setup_start[], __setup_end[];
 190
 191static int __init obsolete_checksetup(char *line)
 192{
 193        struct obs_kernel_param *p;
 194        int had_early_param = 0;
 195
 196        p = __setup_start;
 197        do {
 198                int n = strlen(p->str);
 199                if (!strncmp(line, p->str, n)) {
 200                        if (p->early) {
 201                                /* Already done in parse_early_param?
 202                                 * (Needs exact match on param part).
 203                                 * Keep iterating, as we can have early
 204                                 * params and __setups of same names 8( */
 205                                if (line[n] == '\0' || line[n] == '=')
 206                                        had_early_param = 1;
 207                        } else if (!p->setup_func) {
 208                                printk(KERN_WARNING "Parameter %s is obsolete,"
 209                                       " ignored\n", p->str);
 210                                return 1;
 211                        } else if (p->setup_func(line + n))
 212                                return 1;
 213                }
 214                p++;
 215        } while (p < __setup_end);
 216
 217        return had_early_param;
 218}
 219
 220/*
 221 * This should be approx 2 Bo*oMips to start (note initial shift), and will
 222 * still work even if initially too large, it will just take slightly longer
 223 */
 224unsigned long loops_per_jiffy = (1<<12);
 225
 226EXPORT_SYMBOL(loops_per_jiffy);
 227
 228static int __init debug_kernel(char *str)
 229{
 230        console_loglevel = 10;
 231        return 0;
 232}
 233
 234static int __init quiet_kernel(char *str)
 235{
 236        console_loglevel = 4;
 237        return 0;
 238}
 239
 240early_param("debug", debug_kernel);
 241early_param("quiet", quiet_kernel);
 242
 243static int __init loglevel(char *str)
 244{
 245        get_option(&str, &console_loglevel);
 246        return 0;
 247}
 248
 249early_param("loglevel", loglevel);
 250
 251/*
 252 * Unknown boot options get handed to init, unless they look like
 253 * failed parameters
 254 */
 255static int __init unknown_bootoption(char *param, char *val)
 256{
 257        /* Change NUL term back to "=", to make "param" the whole string. */
 258        if (val) {
 259                /* param=val or param="val"? */
 260                if (val == param+strlen(param)+1)
 261                        val[-1] = '=';
 262                else if (val == param+strlen(param)+2) {
 263                        val[-2] = '=';
 264                        memmove(val-1, val, strlen(val)+1);
 265                        val--;
 266                } else
 267                        BUG();
 268        }
 269
 270        /* Handle obsolete-style parameters */
 271        if (obsolete_checksetup(param))
 272                return 0;
 273
 274        /*
 275         * Preemptive maintenance for "why didn't my misspelled command
 276         * line work?"
 277         */
 278        if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
 279                printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
 280                return 0;
 281        }
 282
 283        if (panic_later)
 284                return 0;
 285
 286        if (val) {
 287                /* Environment option */
 288                unsigned int i;
 289                for (i = 0; envp_init[i]; i++) {
 290                        if (i == MAX_INIT_ENVS) {
 291                                panic_later = "Too many boot env vars at `%s'";
 292                                panic_param = param;
 293                        }
 294                        if (!strncmp(param, envp_init[i], val - param))
 295                                break;
 296                }
 297                envp_init[i] = param;
 298        } else {
 299                /* Command line option */
 300                unsigned int i;
 301                for (i = 0; argv_init[i]; i++) {
 302                        if (i == MAX_INIT_ARGS) {
 303                                panic_later = "Too many boot init vars at `%s'";
 304                                panic_param = param;
 305                        }
 306                }
 307                argv_init[i] = param;
 308        }
 309        return 0;
 310}
 311
 312#ifdef CONFIG_DEBUG_PAGEALLOC
 313int __read_mostly debug_pagealloc_enabled = 0;
 314#endif
 315
 316static int __init init_setup(char *str)
 317{
 318        unsigned int i;
 319
 320        execute_command = str;
 321        /*
 322         * In case LILO is going to boot us with default command line,
 323         * it prepends "auto" before the whole cmdline which makes
 324         * the shell think it should execute a script with such name.
 325         * So we ignore all arguments entered _before_ init=... [MJ]
 326         */
 327        for (i = 1; i < MAX_INIT_ARGS; i++)
 328                argv_init[i] = NULL;
 329        return 1;
 330}
 331__setup("init=", init_setup);
 332
 333static int __init rdinit_setup(char *str)
 334{
 335        unsigned int i;
 336
 337        ramdisk_execute_command = str;
 338        /* See "auto" comment in init_setup */
 339        for (i = 1; i < MAX_INIT_ARGS; i++)
 340                argv_init[i] = NULL;
 341        return 1;
 342}
 343__setup("rdinit=", rdinit_setup);
 344
 345#ifndef CONFIG_SMP
 346
 347#ifdef CONFIG_X86_LOCAL_APIC
 348static void __init smp_init(void)
 349{
 350        APIC_init_uniprocessor();
 351}
 352#else
 353#define smp_init()      do { } while (0)
 354#endif
 355
 356static inline void setup_per_cpu_areas(void) { }
 357static inline void setup_nr_cpu_ids(void) { }
 358static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 359
 360#else
 361
 362#if NR_CPUS > BITS_PER_LONG
 363cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
 364EXPORT_SYMBOL(cpu_mask_all);
 365#endif
 366
 367/* Setup number of possible processor ids */
 368int nr_cpu_ids __read_mostly = NR_CPUS;
 369EXPORT_SYMBOL(nr_cpu_ids);
 370
 371/* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
 372static void __init setup_nr_cpu_ids(void)
 373{
 374        nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
 375}
 376
 377#ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
 378unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
 379
 380EXPORT_SYMBOL(__per_cpu_offset);
 381
 382static void __init setup_per_cpu_areas(void)
 383{
 384        unsigned long size, i;
 385        char *ptr;
 386        unsigned long nr_possible_cpus = num_possible_cpus();
 387
 388        /* Copy section for each CPU (we discard the original) */
 389        size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
 390        ptr = alloc_bootmem_pages(size * nr_possible_cpus);
 391
 392        for_each_possible_cpu(i) {
 393                __per_cpu_offset[i] = ptr - __per_cpu_start;
 394                memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 395                ptr += size;
 396        }
 397}
 398#endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
 399
 400/* Called by boot processor to activate the rest. */
 401static void __init smp_init(void)
 402{
 403        unsigned int cpu;
 404
 405        /*
 406         * Set up the current CPU as possible to migrate to.
 407         * The other ones will be done by cpu_up/cpu_down()
 408         */
 409        set_cpu_active(smp_processor_id(), true);
 410
 411        /* FIXME: This should be done in userspace --RR */
 412        for_each_present_cpu(cpu) {
 413                if (num_online_cpus() >= setup_max_cpus)
 414                        break;
 415                if (!cpu_online(cpu))
 416                        cpu_up(cpu);
 417        }
 418
 419        /* Any cleanup work */
 420        printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 421        smp_cpus_done(setup_max_cpus);
 422}
 423
 424#endif
 425
 426/*
 427 * We need to store the untouched command line for future reference.
 428 * We also need to store the touched command line since the parameter
 429 * parsing is performed in place, and we should allow a component to
 430 * store reference of name/value for future reference.
 431 */
 432static void __init setup_command_line(char *command_line)
 433{
 434        saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 435        static_command_line = alloc_bootmem(strlen (command_line)+1);
 436        strcpy (saved_command_line, boot_command_line);
 437        strcpy (static_command_line, command_line);
 438}
 439
 440/*
 441 * We need to finalize in a non-__init function or else race conditions
 442 * between the root thread and the init thread may cause start_kernel to
 443 * be reaped by free_initmem before the root thread has proceeded to
 444 * cpu_idle.
 445 *
 446 * gcc-3.4 accidentally inlines this function, so use noinline.
 447 */
 448
 449static noinline void __init_refok rest_init(void)
 450        __releases(kernel_lock)
 451{
 452        int pid;
 453
 454        kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 455        numa_default_policy();
 456        pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 457        kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 458        unlock_kernel();
 459
 460        /*
 461         * The boot idle thread must execute schedule()
 462         * at least once to get things moving:
 463         */
 464        init_idle_bootup_task(current);
 465        rcu_scheduler_starting();
 466        preempt_enable_no_resched();
 467        schedule();
 468        preempt_disable();
 469
 470        /* Call into cpu_idle with preempt disabled */
 471        cpu_idle();
 472}
 473
 474/* Check for early params. */
 475static int __init do_early_param(char *param, char *val)
 476{
 477        struct obs_kernel_param *p;
 478
 479        for (p = __setup_start; p < __setup_end; p++) {
 480                if ((p->early && strcmp(param, p->str) == 0) ||
 481                    (strcmp(param, "console") == 0 &&
 482                     strcmp(p->str, "earlycon") == 0)
 483                ) {
 484                        if (p->setup_func(val) != 0)
 485                                printk(KERN_WARNING
 486                                       "Malformed early option '%s'\n", param);
 487                }
 488        }
 489        /* We accept everything at this stage. */
 490        return 0;
 491}
 492
 493void __init parse_early_options(char *cmdline)
 494{
 495        parse_args("early options", cmdline, NULL, 0, do_early_param);
 496}
 497
 498/* Arch code calls this early on, or if not, just before other parsing. */
 499void __init parse_early_param(void)
 500{
 501        static __initdata int done = 0;
 502        static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 503
 504        if (done)
 505                return;
 506
 507        /* All fall through to do_early_param. */
 508        strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 509        parse_early_options(tmp_cmdline);
 510        done = 1;
 511}
 512
 513/*
 514 *      Activate the first processor.
 515 */
 516
 517static void __init boot_cpu_init(void)
 518{
 519        int cpu = smp_processor_id();
 520        /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 521        set_cpu_online(cpu, true);
 522        set_cpu_present(cpu, true);
 523        set_cpu_possible(cpu, true);
 524}
 525
 526void __init __weak smp_setup_processor_id(void)
 527{
 528}
 529
 530void __init __weak thread_info_cache_init(void)
 531{
 532}
 533
 534/*
 535 * Set up kernel memory allocators
 536 */
 537static void __init mm_init(void)
 538{
 539        /*
 540         * page_cgroup requires countinous pages as memmap
 541         * and it's bigger than MAX_ORDER unless SPARSEMEM.
 542         */
 543        page_cgroup_init_flatmem();
 544        mem_init();
 545        kmem_cache_init();
 546        pgtable_cache_init();
 547        vmalloc_init();
 548}
 549
 550asmlinkage void __init start_kernel(void)
 551{
 552        char * command_line;
 553        extern struct kernel_param __start___param[], __stop___param[];
 554
 555        smp_setup_processor_id();
 556
 557        /*
 558         * Need to run as early as possible, to initialize the
 559         * lockdep hash:
 560         */
 561        lockdep_init();
 562        debug_objects_early_init();
 563
 564        /*
 565         * Set up the the initial canary ASAP:
 566         */
 567        boot_init_stack_canary();
 568
 569        cgroup_init_early();
 570
 571        local_irq_disable();
 572        early_boot_irqs_off();
 573        early_init_irq_lock_class();
 574
 575/*
 576 * Interrupts are still disabled. Do necessary setups, then
 577 * enable them
 578 */
 579        lock_kernel();
 580        tick_init();
 581        boot_cpu_init();
 582        page_address_init();
 583        printk(KERN_NOTICE "%s", linux_banner);
 584        setup_arch(&command_line);
 585        mm_init_owner(&init_mm, &init_task);
 586        setup_command_line(command_line);
 587        setup_nr_cpu_ids();
 588        setup_per_cpu_areas();
 589        smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 590
 591        build_all_zonelists();
 592        page_alloc_init();
 593
 594        printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 595        parse_early_param();
 596        parse_args("Booting kernel", static_command_line, __start___param,
 597                   __stop___param - __start___param,
 598                   &unknown_bootoption);
 599        /*
 600         * These use large bootmem allocations and must precede
 601         * kmem_cache_init()
 602         */
 603        pidhash_init();
 604        vfs_caches_init_early();
 605        sort_main_extable();
 606        trap_init();
 607        mm_init();
 608        /*
 609         * Set up the scheduler prior starting any interrupts (such as the
 610         * timer interrupt). Full topology setup happens at smp_init()
 611         * time - but meanwhile we still have a functioning scheduler.
 612         */
 613        sched_init();
 614        /*
 615         * Disable preemption - early bootup scheduling is extremely
 616         * fragile until we cpu_idle() for the first time.
 617         */
 618        preempt_disable();
 619        if (!irqs_disabled()) {
 620                printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 621                                "enabled *very* early, fixing it\n");
 622                local_irq_disable();
 623        }
 624        rcu_init();
 625        /* init some links before init_ISA_irqs() */
 626        early_irq_init();
 627        init_IRQ();
 628        prio_tree_init();
 629        init_timers();
 630        hrtimers_init();
 631        softirq_init();
 632        timekeeping_init();
 633        time_init();
 634        sched_clock_init();
 635        profile_init();
 636        if (!irqs_disabled())
 637                printk(KERN_CRIT "start_kernel(): bug: interrupts were "
 638                                 "enabled early\n");
 639        early_boot_irqs_on();
 640        local_irq_enable();
 641
 642        /* Interrupts are enabled now so all GFP allocations are safe. */
 643        set_gfp_allowed_mask(__GFP_BITS_MASK);
 644
 645        kmem_cache_init_late();
 646
 647        /*
 648         * HACK ALERT! This is early. We're enabling the console before
 649         * we've done PCI setups etc, and console_init() must be aware of
 650         * this. But we do want output early, in case something goes wrong.
 651         */
 652        console_init();
 653        if (panic_later)
 654                panic(panic_later, panic_param);
 655
 656        lockdep_info();
 657
 658        /*
 659         * Need to run this when irqs are enabled, because it wants
 660         * to self-test [hard/soft]-irqs on/off lock inversion bugs
 661         * too:
 662         */
 663        locking_selftest();
 664
 665#ifdef CONFIG_BLK_DEV_INITRD
 666        if (initrd_start && !initrd_below_start_ok &&
 667            page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 668                printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 669                    "disabling it.\n",
 670                    page_to_pfn(virt_to_page((void *)initrd_start)),
 671                    min_low_pfn);
 672                initrd_start = 0;
 673        }
 674#endif
 675        page_cgroup_init();
 676        enable_debug_pagealloc();
 677        kmemtrace_init();
 678        kmemleak_init();
 679        debug_objects_mem_init();
 680        idr_init_cache();
 681        setup_per_cpu_pageset();
 682        numa_policy_init();
 683        if (late_time_init)
 684                late_time_init();
 685        calibrate_delay();
 686        pidmap_init();
 687        anon_vma_init();
 688#ifdef CONFIG_X86
 689        if (efi_enabled)
 690                efi_enter_virtual_mode();
 691#endif
 692        thread_info_cache_init();
 693        cred_init();
 694        fork_init(num_physpages);
 695        proc_caches_init();
 696        buffer_init();
 697        key_init();
 698        security_init();
 699        vfs_caches_init(num_physpages);
 700        radix_tree_init();
 701        signals_init();
 702        /* rootfs populating might need page-writeback */
 703        page_writeback_init();
 704#ifdef CONFIG_PROC_FS
 705        proc_root_init();
 706#endif
 707        cgroup_init();
 708        cpuset_init();
 709        taskstats_init_early();
 710        delayacct_init();
 711
 712        check_bugs();
 713
 714        acpi_early_init(); /* before LAPIC and SMP init */
 715
 716        ftrace_init();
 717
 718        /* Do the rest non-__init'ed, we're now alive */
 719        rest_init();
 720}
 721
 722/* Call all constructor functions linked into the kernel. */
 723static void __init do_ctors(void)
 724{
 725#ifdef CONFIG_CONSTRUCTORS
 726        ctor_fn_t *call = (ctor_fn_t *) __ctors_start;
 727
 728        for (; call < (ctor_fn_t *) __ctors_end; call++)
 729                (*call)();
 730#endif
 731}
 732
 733int initcall_debug;
 734core_param(initcall_debug, initcall_debug, bool, 0644);
 735
 736static char msgbuf[64];
 737static struct boot_trace_call call;
 738static struct boot_trace_ret ret;
 739
 740int do_one_initcall(initcall_t fn)
 741{
 742        int count = preempt_count();
 743        ktime_t calltime, delta, rettime;
 744
 745        if (initcall_debug) {
 746                call.caller = task_pid_nr(current);
 747                printk("calling  %pF @ %i\n", fn, call.caller);
 748                calltime = ktime_get();
 749                trace_boot_call(&call, fn);
 750                enable_boot_trace();
 751        }
 752
 753        ret.result = fn();
 754
 755        if (initcall_debug) {
 756                disable_boot_trace();
 757                rettime = ktime_get();
 758                delta = ktime_sub(rettime, calltime);
 759                ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 760                trace_boot_ret(&ret, fn);
 761                printk("initcall %pF returned %d after %Ld usecs\n", fn,
 762                        ret.result, ret.duration);
 763        }
 764
 765        msgbuf[0] = 0;
 766
 767        if (ret.result && ret.result != -ENODEV && initcall_debug)
 768                sprintf(msgbuf, "error code %d ", ret.result);
 769
 770        if (preempt_count() != count) {
 771                strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 772                preempt_count() = count;
 773        }
 774        if (irqs_disabled()) {
 775                strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 776                local_irq_enable();
 777        }
 778        if (msgbuf[0]) {
 779                printk("initcall %pF returned with %s\n", fn, msgbuf);
 780        }
 781
 782        return ret.result;
 783}
 784
 785
 786extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
 787
 788static void __init do_initcalls(void)
 789{
 790        initcall_t *call;
 791
 792        for (call = __early_initcall_end; call < __initcall_end; call++)
 793                do_one_initcall(*call);
 794
 795        /* Make sure there is no pending stuff from the initcall sequence */
 796        flush_scheduled_work();
 797}
 798
 799/*
 800 * Ok, the machine is now initialized. None of the devices
 801 * have been touched yet, but the CPU subsystem is up and
 802 * running, and memory and process management works.
 803 *
 804 * Now we can finally start doing some real work..
 805 */
 806static void __init do_basic_setup(void)
 807{
 808        rcu_init_sched(); /* needed by module_init stage. */
 809        init_workqueues();
 810        cpuset_init_smp();
 811        usermodehelper_init();
 812        driver_init();
 813        init_irq_proc();
 814        do_ctors();
 815        do_initcalls();
 816}
 817
 818static void __init do_pre_smp_initcalls(void)
 819{
 820        initcall_t *call;
 821
 822        for (call = __initcall_start; call < __early_initcall_end; call++)
 823                do_one_initcall(*call);
 824}
 825
 826static void run_init_process(char *init_filename)
 827{
 828        argv_init[0] = init_filename;
 829        kernel_execve(init_filename, argv_init, envp_init);
 830}
 831
 832/* This is a non __init function. Force it to be noinline otherwise gcc
 833 * makes it inline to init() and it becomes part of init.text section
 834 */
 835static noinline int init_post(void)
 836        __releases(kernel_lock)
 837{
 838        /* need to finish all async __init code before freeing the memory */
 839        async_synchronize_full();
 840        free_initmem();
 841        unlock_kernel();
 842        mark_rodata_ro();
 843        system_state = SYSTEM_RUNNING;
 844        numa_default_policy();
 845
 846        if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 847                printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 848
 849        (void) sys_dup(0);
 850        (void) sys_dup(0);
 851
 852        current->signal->flags |= SIGNAL_UNKILLABLE;
 853
 854        if (ramdisk_execute_command) {
 855                run_init_process(ramdisk_execute_command);
 856                printk(KERN_WARNING "Failed to execute %s\n",
 857                                ramdisk_execute_command);
 858        }
 859
 860        /*
 861         * We try each of these until one succeeds.
 862         *
 863         * The Bourne shell can be used instead of init if we are
 864         * trying to recover a really broken machine.
 865         */
 866        if (execute_command) {
 867                run_init_process(execute_command);
 868                printk(KERN_WARNING "Failed to execute %s.  Attempting "
 869                                        "defaults...\n", execute_command);
 870        }
 871        run_init_process("/sbin/init");
 872        run_init_process("/etc/init");
 873        run_init_process("/bin/init");
 874        run_init_process("/bin/sh");
 875
 876        panic("No init found.  Try passing init= option to kernel.");
 877}
 878
 879static int __init kernel_init(void * unused)
 880{
 881        lock_kernel();
 882
 883        /*
 884         * init can allocate pages on any node
 885         */
 886        set_mems_allowed(node_possible_map);
 887        /*
 888         * init can run on any cpu.
 889         */
 890        set_cpus_allowed_ptr(current, cpu_all_mask);
 891        /*
 892         * Tell the world that we're going to be the grim
 893         * reaper of innocent orphaned children.
 894         *
 895         * We don't want people to have to make incorrect
 896         * assumptions about where in the task array this
 897         * can be found.
 898         */
 899        init_pid_ns.child_reaper = current;
 900
 901        cad_pid = task_pid(current);
 902
 903        smp_prepare_cpus(setup_max_cpus);
 904
 905        do_pre_smp_initcalls();
 906        start_boot_trace();
 907
 908        smp_init();
 909        sched_init_smp();
 910
 911        do_basic_setup();
 912
 913        /*
 914         * check if there is an early userspace init.  If yes, let it do all
 915         * the work
 916         */
 917
 918        if (!ramdisk_execute_command)
 919                ramdisk_execute_command = "/init";
 920
 921        if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 922                ramdisk_execute_command = NULL;
 923                prepare_namespace();
 924        }
 925
 926        /*
 927         * Ok, we have completed the initial bootup, and
 928         * we're essentially up and running. Get rid of the
 929         * initmem segments and start the user-mode stuff..
 930         */
 931
 932        init_post();
 933        return 0;
 934}
 935
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.