init/main.c

   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/string.h>
  18 #include <linux/ctype.h>
  19 #include <linux/delay.h>
  20 #include <linux/utsname.h>
  21 #include <linux/ioport.h>
  22 #include <linux/init.h>
  23 #include <linux/smp_lock.h>
  24 #include <linux/initrd.h>
  25 #include <linux/hdreg.h>
  26 #include <linux/bootmem.h>
  27 #include <linux/tty.h>
  28 #include <linux/gfp.h>
  29 #include <linux/percpu.h>
  30 #include <linux/kmod.h>
  31 #include <linux/kernel_stat.h>
  32 #include <linux/start_kernel.h>
  33 #include <linux/security.h>
  34 #include <linux/workqueue.h>
  35 #include <linux/profile.h>
  36 #include <linux/rcupdate.h>
  37 #include <linux/moduleparam.h>
  38 #include <linux/kallsyms.h>
  39 #include <linux/writeback.h>
  40 #include <linux/cpu.h>
  41 #include <linux/cpuset.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/unwind.h>
  52 #include <linux/buffer_head.h>
  53 #include <linux/debug_locks.h>
  54 #include <linux/lockdep.h>
  55 #include <linux/pid_namespace.h>
  56 #include <linux/device.h>
  57
  58 #include <asm/io.h>
  59 #include <asm/bugs.h>
  60 #include <asm/setup.h>
  61 #include <asm/sections.h>
  62 #include <asm/cacheflush.h>
  63
  64 #ifdef CONFIG_X86_LOCAL_APIC
  65 #include <asm/smp.h>
  66 #endif
  67
  68 /*
  69  * This is one of the first .c files built. Error out early if we have compiler
  70  * trouble.
  71  *
  72  * Versions of gcc older than that listed below may actually compile and link
  73  * okay, but the end product can have subtle run time bugs.  To avoid associated
  74  * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
  75  * too old from the very beginning.
  76  */
  77 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
  78 #error Sorry, your GCC is too old. It builds incorrect kernels.
  79 #endif
  80
  81 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
  82 #warning gcc-4.1.0 is known to miscompile the kernel.  A different compiler version is recommended.
  83 #endif
  84
  85 static int kernel_init(void *);
  86
  87 extern void init_IRQ(void);
  88 extern void fork_init(unsigned long);
  89 extern void mca_init(void);
  90 extern void sbus_init(void);
  91 extern void signals_init(void);
  92 extern void pidhash_init(void);
  93 extern void pidmap_init(void);
  94 extern void prio_tree_init(void);
  95 extern void radix_tree_init(void);
  96 extern void free_initmem(void);
  97 #ifdef  CONFIG_ACPI
  98 extern void acpi_early_init(void);
  99 #else
 100 static inline void acpi_early_init(void) { }
 101 #endif
 102 #ifndef CONFIG_DEBUG_RODATA
 103 static inline void mark_rodata_ro(void) { }
 104 #endif
 105
 106 #ifdef CONFIG_TC
 107 extern void tc_init(void);
 108 #endif
 109
 110 enum system_states system_state;
 111 EXPORT_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
 119 extern void time_init(void);
 120 /* Default late time init is NULL. archs can override this later. */
 121 void (*late_time_init)(void);
 122 extern void softirq_init(void);
 123
 124 /* Untouched command line saved by arch-specific code. */
 125 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 126 /* Untouched saved command line (eg. for /proc) */
 127 char *saved_command_line;
 128 /* Command line for parameter parsing */
 129 static char *static_command_line;
 130
 131 static char *execute_command;
 132 static char *ramdisk_execute_command;
 133
 134 /* Setup configured maximum number of CPUs to activate */
 135 static unsigned int max_cpus = NR_CPUS;
 136
 137 /*
 138  * If set, this is an indication to the drivers that reset the underlying
 139  * device before going ahead with the initialization otherwise driver might
 140  * rely on the BIOS and skip the reset operation.
 141  *
 142  * This is useful if kernel is booting in an unreliable environment.
 143  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 144  * skipped and devices will be in unknown state.
 145  */
 146 unsigned int reset_devices;
 147 EXPORT_SYMBOL(reset_devices);
 148
 149 /*
 150  * Setup routine for controlling SMP activation
 151  *
 152  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 153  * activation entirely (the MPS table probe still happens, though).
 154  *
 155  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 156  * greater than 0, limits the maximum number of CPUs activated in
 157  * SMP mode to <NUM>.
 158  */
 159 static int __init nosmp(char *str)
 160 {
 161         max_cpus = 0;
 162         return 1;
 163 }
 164
 165 __setup("nosmp", nosmp);
 166
 167 static int __init maxcpus(char *str)
 168 {
 169         get_option(&str, &max_cpus);
 170         return 1;
 171 }
 172
 173 __setup("maxcpus=", maxcpus);
 174
 175 static int __init set_reset_devices(char *str)
 176 {
 177         reset_devices = 1;
 178         return 1;
 179 }
 180
 181 __setup("reset_devices", set_reset_devices);
 182
 183 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 184 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 185 static const char *panic_later, *panic_param;
 186
 187 extern struct obs_kernel_param __setup_start[], __setup_end[];
 188
 189 static int __init obsolete_checksetup(char *line)
 190 {
 191         struct obs_kernel_param *p;
 192         int had_early_param = 0;
 193
 194         p = __setup_start;
 195         do {
 196                 int n = strlen(p->str);
 197                 if (!strncmp(line, p->str, n)) {
 198                         if (p->early) {
 199                                 /* Already done in parse_early_param?
 200                                  * (Needs exact match on param part).
 201                                  * Keep iterating, as we can have early
 202                                  * params and __setups of same names 8( */
 203                                 if (line[n] == '\0' || line[n] == '=')
 204                                         had_early_param = 1;
 205                         } else if (!p->setup_func) {
 206                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 207                                        " ignored\n", p->str);
 208                                 return 1;
 209                         } else if (p->setup_func(line + n))
 210                                 return 1;
 211                 }
 212                 p++;
 213         } while (p < __setup_end);
 214
 215         return had_early_param;
 216 }
 217
 218 /*
 219  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 220  * still work even if initially too large, it will just take slightly longer
 221  */
 222 unsigned long loops_per_jiffy = (1<<12);
 223
 224 EXPORT_SYMBOL(loops_per_jiffy);
 225
 226 static int __init debug_kernel(char *str)
 227 {
 228         if (*str)
 229                 return 0;
 230         console_loglevel = 10;
 231         return 1;
 232 }
 233
 234 static int __init quiet_kernel(char *str)
 235 {
 236         if (*str)
 237                 return 0;
 238         console_loglevel = 4;
 239         return 1;
 240 }
 241
 242 __setup("debug", debug_kernel);
 243 __setup("quiet", quiet_kernel);
 244
 245 static int __init loglevel(char *str)
 246 {
 247         get_option(&str, &console_loglevel);
 248         return 1;
 249 }
 250
 251 __setup("loglevel=", loglevel);
 252
 253 /*
 254  * Unknown boot options get handed to init, unless they look like
 255  * failed parameters
 256  */
 257 static int __init unknown_bootoption(char *param, char *val)
 258 {
 259         /* Change NUL term back to "=", to make "param" the whole string. */
 260         if (val) {
 261                 /* param=val or param="val"? */
 262                 if (val == param+strlen(param)+1)
 263                         val[-1] = '=';
 264                 else if (val == param+strlen(param)+2) {
 265                         val[-2] = '=';
 266                         memmove(val-1, val, strlen(val)+1);
 267                         val--;
 268                 } else
 269                         BUG();
 270         }
 271
 272         /* Handle obsolete-style parameters */
 273         if (obsolete_checksetup(param))
 274                 return 0;
 275
 276         /*
 277          * Preemptive maintenance for "why didn't my mispelled command
 278          * line work?"
 279          */
 280         if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
 281                 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
 282                 return 0;
 283         }
 284
 285         if (panic_later)
 286                 return 0;
 287
 288         if (val) {
 289                 /* Environment option */
 290                 unsigned int i;
 291                 for (i = 0; envp_init[i]; i++) {
 292                         if (i == MAX_INIT_ENVS) {
 293                                 panic_later = "Too many boot env vars at `%s'";
 294                                 panic_param = param;
 295                         }
 296                         if (!strncmp(param, envp_init[i], val - param))
 297                                 break;
 298                 }
 299                 envp_init[i] = param;
 300         } else {
 301                 /* Command line option */
 302                 unsigned int i;
 303                 for (i = 0; argv_init[i]; i++) {
 304                         if (i == MAX_INIT_ARGS) {
 305                                 panic_later = "Too many boot init vars at `%s'";
 306                                 panic_param = param;
 307                         }
 308                 }
 309                 argv_init[i] = param;
 310         }
 311         return 0;
 312 }
 313
 314 static int __init init_setup(char *str)
 315 {
 316         unsigned int i;
 317
 318         execute_command = str;
 319         /*
 320          * In case LILO is going to boot us with default command line,
 321          * it prepends "auto" before the whole cmdline which makes
 322          * the shell think it should execute a script with such name.
 323          * So we ignore all arguments entered _before_ init=... [MJ]
 324          */
 325         for (i = 1; i < MAX_INIT_ARGS; i++)
 326                 argv_init[i] = NULL;
 327         return 1;
 328 }
 329 __setup("init=", init_setup);
 330
 331 static int __init rdinit_setup(char *str)
 332 {
 333         unsigned int i;
 334
 335         ramdisk_execute_command = str;
 336         /* See "auto" comment in init_setup */
 337         for (i = 1; i < MAX_INIT_ARGS; i++)
 338                 argv_init[i] = NULL;
 339         return 1;
 340 }
 341 __setup("rdinit=", rdinit_setup);
 342
 343 #ifndef CONFIG_SMP
 344
 345 #ifdef CONFIG_X86_LOCAL_APIC
 346 static void __init smp_init(void)
 347 {
 348         APIC_init_uniprocessor();
 349 }
 350 #else
 351 #define smp_init()      do { } while (0)
 352 #endif
 353
 354 static inline void setup_per_cpu_areas(void) { }
 355 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 356
 357 #else
 358
 359 #ifdef __GENERIC_PER_CPU
 360 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
 361
 362 EXPORT_SYMBOL(__per_cpu_offset);
 363
 364 static void __init setup_per_cpu_areas(void)
 365 {
 366         unsigned long size, i;
 367         char *ptr;
 368         unsigned long nr_possible_cpus = num_possible_cpus();
 369
 370         /* Copy section for each CPU (we discard the original) */
 371         size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
 372         ptr = alloc_bootmem_pages(size * nr_possible_cpus);
 373
 374         for_each_possible_cpu(i) {
 375                 __per_cpu_offset[i] = ptr - __per_cpu_start;
 376                 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 377                 ptr += size;
 378         }
 379 }
 380 #endif /* !__GENERIC_PER_CPU */
 381
 382 /* Called by boot processor to activate the rest. */
 383 static void __init smp_init(void)
 384 {
 385         unsigned int cpu;
 386
 387         /* FIXME: This should be done in userspace --RR */
 388         for_each_present_cpu(cpu) {
 389                 if (num_online_cpus() >= max_cpus)
 390                         break;
 391                 if (!cpu_online(cpu))
 392                         cpu_up(cpu);
 393         }
 394
 395         /* Any cleanup work */
 396         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 397         smp_cpus_done(max_cpus);
 398 }
 399
 400 #endif
 401
 402 /*
 403  * We need to store the untouched command line for future reference.
 404  * We also need to store the touched command line since the parameter
 405  * parsing is performed in place, and we should allow a component to
 406  * store reference of name/value for future reference.
 407  */
 408 static void __init setup_command_line(char *command_line)
 409 {
 410         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 411         static_command_line = alloc_bootmem(strlen (command_line)+1);
 412         strcpy (saved_command_line, boot_command_line);
 413         strcpy (static_command_line, command_line);
 414 }
 415
 416 /*
 417  * We need to finalize in a non-__init function or else race conditions
 418  * between the root thread and the init thread may cause start_kernel to
 419  * be reaped by free_initmem before the root thread has proceeded to
 420  * cpu_idle.
 421  *
 422  * gcc-3.4 accidentally inlines this function, so use noinline.
 423  */
 424
 425 static void noinline rest_init(void)
 426         __releases(kernel_lock)
 427 {
 428         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 429         numa_default_policy();
 430         unlock_kernel();
 431
 432         /*
 433          * The boot idle thread must execute schedule()
 434          * at least one to get things moving:
 435          */
 436         preempt_enable_no_resched();
 437         schedule();
 438         preempt_disable();
 439
 440         /* Call into cpu_idle with preempt disabled */
 441         cpu_idle();
 442 }
 443
 444 /* Check for early params. */
 445 static int __init do_early_param(char *param, char *val)
 446 {
 447         struct obs_kernel_param *p;
 448
 449         for (p = __setup_start; p < __setup_end; p++) {
 450                 if (p->early && strcmp(param, p->str) == 0) {
 451                         if (p->setup_func(val) != 0)
 452                                 printk(KERN_WARNING
 453                                        "Malformed early option '%s'\n", param);
 454                 }
 455         }
 456         /* We accept everything at this stage. */
 457         return 0;
 458 }
 459
 460 /* Arch code calls this early on, or if not, just before other parsing. */
 461 void __init parse_early_param(void)
 462 {
 463         static __initdata int done = 0;
 464         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 465
 466         if (done)
 467                 return;
 468
 469         /* All fall through to do_early_param. */
 470         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 471         parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
 472         done = 1;
 473 }
 474
 475 /*
 476  *      Activate the first processor.
 477  */
 478
 479 static void __init boot_cpu_init(void)
 480 {
 481         int cpu = smp_processor_id();
 482         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 483         cpu_set(cpu, cpu_online_map);
 484         cpu_set(cpu, cpu_present_map);
 485         cpu_set(cpu, cpu_possible_map);
 486 }
 487
 488 void __init __attribute__((weak)) smp_setup_processor_id(void)
 489 {
 490 }
 491
 492 asmlinkage void __init start_kernel(void)
 493 {
 494         char * command_line;
 495         extern struct kernel_param __start___param[], __stop___param[];
 496
 497         smp_setup_processor_id();
 498
 499         /*
 500          * Need to run as early as possible, to initialize the
 501          * lockdep hash:
 502          */
 503         unwind_init();
 504         lockdep_init();
 505
 506         local_irq_disable();
 507         early_boot_irqs_off();
 508         early_init_irq_lock_class();
 509
 510 /*
 511  * Interrupts are still disabled. Do necessary setups, then
 512  * enable them
 513  */
 514         lock_kernel();
 515         tick_init();
 516         boot_cpu_init();
 517         page_address_init();
 518         printk(KERN_NOTICE);
 519         printk(linux_banner);
 520         setup_arch(&command_line);
 521         setup_command_line(command_line);
 522         unwind_setup();
 523         setup_per_cpu_areas();
 524         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 525
 526         /*
 527          * Set up the scheduler prior starting any interrupts (such as the
 528          * timer interrupt). Full topology setup happens at smp_init()
 529          * time - but meanwhile we still have a functioning scheduler.
 530          */
 531         sched_init();
 532         /*
 533          * Disable preemption - early bootup scheduling is extremely
 534          * fragile until we cpu_idle() for the first time.
 535          */
 536         preempt_disable();
 537         build_all_zonelists();
 538         page_alloc_init();
 539         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 540         parse_early_param();
 541         parse_args("Booting kernel", static_command_line, __start___param,
 542                    __stop___param - __start___param,
 543                    &unknown_bootoption);
 544         if (!irqs_disabled()) {
 545                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 546                                 "enabled *very* early, fixing it\n");
 547                 local_irq_disable();
 548         }
 549         sort_main_extable();
 550         trap_init();
 551         rcu_init();
 552         init_IRQ();
 553         pidhash_init();
 554         init_timers();
 555         hrtimers_init();
 556         softirq_init();
 557         timekeeping_init();
 558         time_init();
 559         profile_init();
 560         if (!irqs_disabled())
 561                 printk("start_kernel(): bug: interrupts were enabled early\n");
 562         early_boot_irqs_on();
 563         local_irq_enable();
 564
 565         /*
 566          * HACK ALERT! This is early. We're enabling the console before
 567          * we've done PCI setups etc, and console_init() must be aware of
 568          * this. But we do want output early, in case something goes wrong.
 569          */
 570         console_init();
 571         if (panic_later)
 572                 panic(panic_later, panic_param);
 573
 574         lockdep_info();
 575
 576         /*
 577          * Need to run this when irqs are enabled, because it wants
 578          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 579          * too:
 580          */
 581         locking_selftest();
 582
 583 #ifdef CONFIG_BLK_DEV_INITRD
 584         if (initrd_start && !initrd_below_start_ok &&
 585                         initrd_start < min_low_pfn << PAGE_SHIFT) {
 586                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 587                     "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
 588                 initrd_start = 0;
 589         }
 590 #endif
 591         vfs_caches_init_early();
 592         cpuset_init_early();
 593         mem_init();
 594         kmem_cache_init();
 595         setup_per_cpu_pageset();
 596         numa_policy_init();
 597         if (late_time_init)
 598                 late_time_init();
 599         calibrate_delay();
 600         pidmap_init();
 601         pgtable_cache_init();
 602         prio_tree_init();
 603         anon_vma_init();
 604 #ifdef CONFIG_X86
 605         if (efi_enabled)
 606                 efi_enter_virtual_mode();
 607 #endif
 608         fork_init(num_physpages);
 609         proc_caches_init();
 610         buffer_init();
 611         unnamed_dev_init();
 612         key_init();
 613         security_init();
 614         vfs_caches_init(num_physpages);
 615         radix_tree_init();
 616         signals_init();
 617         /* rootfs populating might need page-writeback */
 618         page_writeback_init();
 619 #ifdef CONFIG_PROC_FS
 620         proc_root_init();
 621 #endif
 622         cpuset_init();
 623         taskstats_init_early();
 624         delayacct_init();
 625
 626         check_bugs();
 627
 628         acpi_early_init(); /* before LAPIC and SMP init */
 629
 630         /* Do the rest non-__init'ed, we're now alive */
 631         rest_init();
 632 }
 633
 634 static int __initdata initcall_debug;
 635
 636 static int __init initcall_debug_setup(char *str)
 637 {
 638         initcall_debug = 1;
 639         return 1;
 640 }
 641 __setup("initcall_debug", initcall_debug_setup);
 642
 643 extern initcall_t __initcall_start[], __initcall_end[];
 644
 645 static void __init do_initcalls(void)
 646 {
 647         initcall_t *call;
 648         int count = preempt_count();
 649
 650         for (call = __initcall_start; call < __initcall_end; call++) {
 651                 ktime_t t0, t1, delta;
 652                 char *msg = NULL;
 653                 char msgbuf[40];
 654                 int result;
 655
 656                 if (initcall_debug) {
 657                         printk("Calling initcall 0x%p", *call);
 658                         print_fn_descriptor_symbol(": %s()",
 659                                         (unsigned long) *call);
 660                         printk("\n");
 661                         t0 = ktime_get();
 662                 }
 663
 664                 result = (*call)();
 665
 666                 if (initcall_debug) {
 667                         t1 = ktime_get();
 668                         delta = ktime_sub(t1, t0);
 669
 670                         printk("initcall 0x%p", *call);
 671                         print_fn_descriptor_symbol(": %s()",
 672                                         (unsigned long) *call);
 673                         printk(" returned %d.\n", result);
 674
 675                         printk("initcall 0x%p ran for %Ld msecs: ",
 676                                 *call, (unsigned long long)delta.tv64 >> 20);
 677                         print_fn_descriptor_symbol("%s()\n",
 678                                 (unsigned long) *call);
 679                 }
 680
 681                 if (result && result != -ENODEV && initcall_debug) {
 682                         sprintf(msgbuf, "error code %d", result);
 683                         msg = msgbuf;
 684                 }
 685                 if (preempt_count() != count) {
 686                         msg = "preemption imbalance";
 687                         preempt_count() = count;
 688                 }
 689                 if (irqs_disabled()) {
 690                         msg = "disabled interrupts";
 691                         local_irq_enable();
 692                 }
 693                 if (msg) {
 694                         printk(KERN_WARNING "initcall at 0x%p", *call);
 695                         print_fn_descriptor_symbol(": %s()",
 696                                         (unsigned long) *call);
 697                         printk(": returned with %s\n", msg);
 698                 }
 699         }
 700
 701         /* Make sure there is no pending stuff from the initcall sequence */
 702         flush_scheduled_work();
 703 }
 704
 705 /*
 706  * Ok, the machine is now initialized. None of the devices
 707  * have been touched yet, but the CPU subsystem is up and
 708  * running, and memory and process management works.
 709  *
 710  * Now we can finally start doing some real work..
 711  */
 712 static void __init do_basic_setup(void)
 713 {
 714         /* drivers will send hotplug events */
 715         init_workqueues();
 716         usermodehelper_init();
 717         driver_init();
 718         init_irq_proc();
 719         do_initcalls();
 720 }
 721
 722 static void __init do_pre_smp_initcalls(void)
 723 {
 724         extern int spawn_ksoftirqd(void);
 725 #ifdef CONFIG_SMP
 726         extern int migration_init(void);
 727
 728         migration_init();
 729 #endif
 730         spawn_ksoftirqd();
 731         spawn_softlockup_task();
 732 }
 733
 734 static void run_init_process(char *init_filename)
 735 {
 736         argv_init[0] = init_filename;
 737         kernel_execve(init_filename, argv_init, envp_init);
 738 }
 739
 740 /* This is a non __init function. Force it to be noinline otherwise gcc
 741  * makes it inline to init() and it becomes part of init.text section
 742  */
 743 static int noinline init_post(void)
 744 {
 745         free_initmem();
 746         unlock_kernel();
 747         mark_rodata_ro();
 748         system_state = SYSTEM_RUNNING;
 749         numa_default_policy();
 750
 751         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 752                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 753
 754         (void) sys_dup(0);
 755         (void) sys_dup(0);
 756
 757         if (ramdisk_execute_command) {
 758                 run_init_process(ramdisk_execute_command);
 759                 printk(KERN_WARNING "Failed to execute %s\n",
 760                                 ramdisk_execute_command);
 761         }
 762
 763         /*
 764          * We try each of these until one succeeds.
 765          *
 766          * The Bourne shell can be used instead of init if we are
 767          * trying to recover a really broken machine.
 768          */
 769         if (execute_command) {
 770                 run_init_process(execute_command);
 771                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 772                                         "defaults...\n", execute_command);
 773         }
 774         run_init_process("/sbin/init");
 775         run_init_process("/etc/init");
 776         run_init_process("/bin/init");
 777         run_init_process("/bin/sh");
 778
 779         panic("No init found.  Try passing init= option to kernel.");
 780 }
 781
 782 static int __init kernel_init(void * unused)
 783 {
 784         lock_kernel();
 785         /*
 786          * init can run on any cpu.
 787          */
 788         set_cpus_allowed(current, CPU_MASK_ALL);
 789         /*
 790          * Tell the world that we're going to be the grim
 791          * reaper of innocent orphaned children.
 792          *
 793          * We don't want people to have to make incorrect
 794          * assumptions about where in the task array this
 795          * can be found.
 796          */
 797         init_pid_ns.child_reaper = current;
 798
 799         cad_pid = task_pid(current);
 800
 801         smp_prepare_cpus(max_cpus);
 802
 803         do_pre_smp_initcalls();
 804
 805         smp_init();
 806         sched_init_smp();
 807
 808         cpuset_init_smp();
 809
 810         do_basic_setup();
 811
 812         /*
 813          * check if there is an early userspace init.  If yes, let it do all
 814          * the work
 815          */
 816
 817         if (!ramdisk_execute_command)
 818                 ramdisk_execute_command = "/init";
 819
 820         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 821                 ramdisk_execute_command = NULL;
 822                 prepare_namespace();
 823         }
 824
 825         /*
 826          * Ok, we have completed the initial bootup, and
 827          * we're essentially up and running. Get rid of the
 828          * initmem segments and start the user-mode stuff..
 829          */
 830         init_post();
 831         return 0;
 832 }