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/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/random.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
  83 static int kernel_init(void *);
  84
  85 extern void init_IRQ(void);
  86 extern void fork_init(unsigned long);
  87 extern void mca_init(void);
  88 extern void sbus_init(void);
  89 extern void prio_tree_init(void);
  90 extern void radix_tree_init(void);
  91 extern void free_initmem(void);
  92 #ifndef CONFIG_DEBUG_RODATA
  93 static inline void mark_rodata_ro(void) { }
  94 #endif
  95
  96 #ifdef CONFIG_TC
  97 extern void tc_init(void);
  98 #endif
  99
 100 /*
 101  * Debug helper: via this flag we know that we are in 'early bootup code'
 102  * where only the boot processor is running with IRQ disabled.  This means
 103  * two things - IRQ must not be enabled before the flag is cleared and some
 104  * operations which are not allowed with IRQ disabled are allowed while the
 105  * flag is set.
 106  */
 107 bool early_boot_irqs_disabled __read_mostly;
 108
 109 enum system_states system_state __read_mostly;
 110 EXPORT_SYMBOL(system_state);
 111
 112 /*
 113  * Boot command-line arguments
 114  */
 115 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 116 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 117
 118 extern void time_init(void);
 119 /* Default late time init is NULL. archs can override this later. */
 120 void (*__initdata late_time_init)(void);
 121 extern void softirq_init(void);
 122
 123 /* Untouched command line saved by arch-specific code. */
 124 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 125 /* Untouched saved command line (eg. for /proc) */
 126 char *saved_command_line;
 127 /* Command line for parameter parsing */
 128 static char *static_command_line;
 129
 130 static char *execute_command;
 131 static char *ramdisk_execute_command;
 132
 133 /*
 134  * If set, this is an indication to the drivers that reset the underlying
 135  * device before going ahead with the initialization otherwise driver might
 136  * rely on the BIOS and skip the reset operation.
 137  *
 138  * This is useful if kernel is booting in an unreliable environment.
 139  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 140  * skipped and devices will be in unknown state.
 141  */
 142 unsigned int reset_devices;
 143 EXPORT_SYMBOL(reset_devices);
 144
 145 static int __init set_reset_devices(char *str)
 146 {
 147         reset_devices = 1;
 148         return 1;
 149 }
 150
 151 __setup("reset_devices", set_reset_devices);
 152
 153 static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 154 const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 155 static const char *panic_later, *panic_param;
 156
 157 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 158
 159 static int __init obsolete_checksetup(char *line)
 160 {
 161         const struct obs_kernel_param *p;
 162         int had_early_param = 0;
 163
 164         p = __setup_start;
 165         do {
 166                 int n = strlen(p->str);
 167                 if (parameqn(line, p->str, n)) {
 168                         if (p->early) {
 169                                 /* Already done in parse_early_param?
 170                                  * (Needs exact match on param part).
 171                                  * Keep iterating, as we can have early
 172                                  * params and __setups of same names 8( */
 173                                 if (line[n] == '\0' || line[n] == '=')
 174                                         had_early_param = 1;
 175                         } else if (!p->setup_func) {
 176                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 177                                        " ignored\n", p->str);
 178                                 return 1;
 179                         } else if (p->setup_func(line + n))
 180                                 return 1;
 181                 }
 182                 p++;
 183         } while (p < __setup_end);
 184
 185         return had_early_param;
 186 }
 187
 188 /*
 189  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 190  * still work even if initially too large, it will just take slightly longer
 191  */
 192 unsigned long loops_per_jiffy = (1<<12);
 193
 194 EXPORT_SYMBOL(loops_per_jiffy);
 195
 196 static int __init debug_kernel(char *str)
 197 {
 198         console_loglevel = 10;
 199         return 0;
 200 }
 201
 202 static int __init quiet_kernel(char *str)
 203 {
 204         console_loglevel = 4;
 205         return 0;
 206 }
 207
 208 early_param("debug", debug_kernel);
 209 early_param("quiet", quiet_kernel);
 210
 211 static int __init loglevel(char *str)
 212 {
 213         int newlevel;
 214
 215         /*
 216          * Only update loglevel value when a correct setting was passed,
 217          * to prevent blind crashes (when loglevel being set to 0) that
 218          * are quite hard to debug
 219          */
 220         if (get_option(&str, &newlevel)) {
 221                 console_loglevel = newlevel;
 222                 return 0;
 223         }
 224
 225         return -EINVAL;
 226 }
 227
 228 early_param("loglevel", loglevel);
 229
 230 /*
 231  * Unknown boot options get handed to init, unless they look like
 232  * unused parameters (modprobe will find them in /proc/cmdline).
 233  */
 234 static int __init unknown_bootoption(char *param, char *val)
 235 {
 236         /* Change NUL term back to "=", to make "param" the whole string. */
 237         if (val) {
 238                 /* param=val or param="val"? */
 239                 if (val == param+strlen(param)+1)
 240                         val[-1] = '=';
 241                 else if (val == param+strlen(param)+2) {
 242                         val[-2] = '=';
 243                         memmove(val-1, val, strlen(val)+1);
 244                         val--;
 245                 } else
 246                         BUG();
 247         }
 248
 249         /* Handle obsolete-style parameters */
 250         if (obsolete_checksetup(param))
 251                 return 0;
 252
 253         /* Unused module parameter. */
 254         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 255                 return 0;
 256
 257         if (panic_later)
 258                 return 0;
 259
 260         if (val) {
 261                 /* Environment option */
 262                 unsigned int i;
 263                 for (i = 0; envp_init[i]; i++) {
 264                         if (i == MAX_INIT_ENVS) {
 265                                 panic_later = "Too many boot env vars at `%s'";
 266                                 panic_param = param;
 267                         }
 268                         if (!strncmp(param, envp_init[i], val - param))
 269                                 break;
 270                 }
 271                 envp_init[i] = param;
 272         } else {
 273                 /* Command line option */
 274                 unsigned int i;
 275                 for (i = 0; argv_init[i]; i++) {
 276                         if (i == MAX_INIT_ARGS) {
 277                                 panic_later = "Too many boot init vars at `%s'";
 278                                 panic_param = param;
 279                         }
 280                 }
 281                 argv_init[i] = param;
 282         }
 283         return 0;
 284 }
 285
 286 #ifdef CONFIG_DEBUG_PAGEALLOC
 287 int __read_mostly debug_pagealloc_enabled = 0;
 288 #endif
 289
 290 static int __init init_setup(char *str)
 291 {
 292         unsigned int i;
 293
 294         execute_command = str;
 295         /*
 296          * In case LILO is going to boot us with default command line,
 297          * it prepends "auto" before the whole cmdline which makes
 298          * the shell think it should execute a script with such name.
 299          * So we ignore all arguments entered _before_ init=... [MJ]
 300          */
 301         for (i = 1; i < MAX_INIT_ARGS; i++)
 302                 argv_init[i] = NULL;
 303         return 1;
 304 }
 305 __setup("init=", init_setup);
 306
 307 static int __init rdinit_setup(char *str)
 308 {
 309         unsigned int i;
 310
 311         ramdisk_execute_command = str;
 312         /* See "auto" comment in init_setup */
 313         for (i = 1; i < MAX_INIT_ARGS; i++)
 314                 argv_init[i] = NULL;
 315         return 1;
 316 }
 317 __setup("rdinit=", rdinit_setup);
 318
 319 #ifndef CONFIG_SMP
 320 static const unsigned int setup_max_cpus = NR_CPUS;
 321 #ifdef CONFIG_X86_LOCAL_APIC
 322 static void __init smp_init(void)
 323 {
 324         APIC_init_uniprocessor();
 325 }
 326 #else
 327 #define smp_init()      do { } while (0)
 328 #endif
 329
 330 static inline void setup_nr_cpu_ids(void) { }
 331 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 332 #endif
 333
 334 /*
 335  * We need to store the untouched command line for future reference.
 336  * We also need to store the touched command line since the parameter
 337  * parsing is performed in place, and we should allow a component to
 338  * store reference of name/value for future reference.
 339  */
 340 static void __init setup_command_line(char *command_line)
 341 {
 342         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 343         static_command_line = alloc_bootmem(strlen (command_line)+1);
 344         strcpy (saved_command_line, boot_command_line);
 345         strcpy (static_command_line, command_line);
 346 }
 347
 348 /*
 349  * We need to finalize in a non-__init function or else race conditions
 350  * between the root thread and the init thread may cause start_kernel to
 351  * be reaped by free_initmem before the root thread has proceeded to
 352  * cpu_idle.
 353  *
 354  * gcc-3.4 accidentally inlines this function, so use noinline.
 355  */
 356
 357 static __initdata DECLARE_COMPLETION(kthreadd_done);
 358
 359 static noinline void __init_refok rest_init(void)
 360 {
 361         int pid;
 362
 363         rcu_scheduler_starting();
 364         /*
 365          * We need to spawn init first so that it obtains pid 1, however
 366          * the init task will end up wanting to create kthreads, which, if
 367          * we schedule it before we create kthreadd, will OOPS.
 368          */
 369         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 370         numa_default_policy();
 371         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 372         rcu_read_lock();
 373         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 374         rcu_read_unlock();
 375         complete(&kthreadd_done);
 376
 377         /*
 378          * The boot idle thread must execute schedule()
 379          * at least once to get things moving:
 380          */
 381         init_idle_bootup_task(current);
 382         preempt_enable_no_resched();
 383         schedule();
 384
 385         /* Call into cpu_idle with preempt disabled */
 386         preempt_disable();
 387         cpu_idle();
 388 }
 389
 390 /* Check for early params. */
 391 static int __init do_early_param(char *param, char *val)
 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
 409 void __init parse_early_options(char *cmdline)
 410 {
 411         parse_args("early options", cmdline, NULL, 0, do_early_param);
 412 }
 413
 414 /* Arch code calls this early on, or if not, just before other parsing. */
 415 void __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
 433 static 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
 443 void __init __weak smp_setup_processor_id(void)
 444 {
 445 }
 446
 447 void __init __weak thread_info_cache_init(void)
 448 {
 449 }
 450
 451 /*
 452  * Set up kernel memory allocators
 453  */
 454 static void __init mm_init(void)
 455 {
 456         /*
 457          * page_cgroup requires countinous pages as memmap
 458          * and it's bigger than MAX_ORDER unless SPARSEMEM.
 459          */
 460         page_cgroup_init_flatmem();
 461         mem_init();
 462         kmem_cache_init();
 463         percpu_init_late();
 464         pgtable_cache_init();
 465         vmalloc_init();
 466 }
 467
 468 asmlinkage void __init start_kernel(void)
 469 {
 470         char * command_line;
 471         extern const struct kernel_param __start___param[], __stop___param[];
 472
 473         smp_setup_processor_id();
 474
 475         /*
 476          * Need to run as early as possible, to initialize the
 477          * lockdep hash:
 478          */
 479         lockdep_init();
 480         debug_objects_early_init();
 481
 482         /*
 483          * Set up the the initial canary ASAP:
 484          */
 485         boot_init_stack_canary();
 486
 487         cgroup_init_early();
 488
 489         local_irq_disable();
 490         early_boot_irqs_disabled = true;
 491
 492 /*
 493  * Interrupts are still disabled. Do necessary setups, then
 494  * enable them
 495  */
 496         tick_init();
 497         boot_cpu_init();
 498         page_address_init();
 499         printk(KERN_NOTICE "%s", linux_banner);
 500         setup_arch(&command_line);
 501         mm_init_owner(&init_mm, &init_task);
 502         mm_init_cpumask(&init_mm);
 503         setup_command_line(command_line);
 504         setup_nr_cpu_ids();
 505         setup_per_cpu_areas();
 506         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 507
 508         build_all_zonelists(NULL);
 509         page_alloc_init();
 510
 511         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 512         parse_early_param();
 513         parse_args("Booting kernel", static_command_line, __start___param,
 514                    __stop___param - __start___param,
 515                    &unknown_bootoption);
 516
 517         jump_label_init();
 518
 519         /*
 520          * These use large bootmem allocations and must precede
 521          * kmem_cache_init()
 522          */
 523         setup_log_buf(0);
 524         pidhash_init();
 525         vfs_caches_init_early();
 526         sort_main_extable();
 527         trap_init();
 528         mm_init();
 529
 530         /*
 531          * Set up the scheduler prior starting any interrupts (such as the
 532          * timer interrupt). Full topology setup happens at smp_init()
 533          * time - but meanwhile we still have a functioning scheduler.
 534          */
 535         sched_init();
 536         /*
 537          * Disable preemption - early bootup scheduling is extremely
 538          * fragile until we cpu_idle() for the first time.
 539          */
 540         preempt_disable();
 541         if (!irqs_disabled()) {
 542                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 543                                 "enabled *very* early, fixing it\n");
 544                 local_irq_disable();
 545         }
 546         idr_init_cache();
 547         perf_event_init();
 548         rcu_init();
 549         radix_tree_init();
 550         /* init some links before init_ISA_irqs() */
 551         early_irq_init();
 552         init_IRQ();
 553         prio_tree_init();
 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         enable_debug_pagealloc();
 599         debug_objects_mem_init();
 600         kmemleak_init();
 601         setup_per_cpu_pageset();
 602         numa_policy_init();
 603         if (late_time_init)
 604                 late_time_init();
 605         sched_clock_init();
 606         calibrate_delay();
 607         pidmap_init();
 608         anon_vma_init();
 609 #ifdef CONFIG_X86
 610         if (efi_enabled(EFI_RUNTIME_SERVICES))
 611                 efi_enter_virtual_mode();
 612 #endif
 613         thread_info_cache_init();
 614         cred_init();
 615         fork_init(totalram_pages);
 616         proc_caches_init();
 617         buffer_init();
 618         key_init();
 619         security_init();
 620         dbg_late_init();
 621         vfs_caches_init(totalram_pages);
 622         signals_init();
 623         /* rootfs populating might need page-writeback */
 624         page_writeback_init();
 625 #ifdef CONFIG_PROC_FS
 626         proc_root_init();
 627 #endif
 628         cgroup_init();
 629         cpuset_init();
 630         taskstats_init_early();
 631         delayacct_init();
 632
 633         check_bugs();
 634
 635         acpi_early_init(); /* before LAPIC and SMP init */
 636         sfi_init_late();
 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. */
 645 static 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
 655 int initcall_debug;
 656 core_param(initcall_debug, initcall_debug, bool, 0644);
 657
 658 static char msgbuf[64];
 659
 660 static 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         printk(KERN_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         printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn,
 673                 ret, duration);
 674
 675         return ret;
 676 }
 677
 678 int __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 (ret && ret != -ENODEV && initcall_debug)
 691                 sprintf(msgbuf, "error code %d ", ret);
 692
 693         if (preempt_count() != count) {
 694                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 695                 preempt_count() = count;
 696         }
 697         if (irqs_disabled()) {
 698                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 699                 local_irq_enable();
 700         }
 701         if (msgbuf[0]) {
 702                 printk("initcall %pF returned with %s\n", fn, msgbuf);
 703         }
 704
 705         return ret;
 706 }
 707
 708
 709 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
 710
 711 static void __init do_initcalls(void)
 712 {
 713         initcall_t *fn;
 714
 715         for (fn = __early_initcall_end; fn < __initcall_end; fn++)
 716                 do_one_initcall(*fn);
 717 }
 718
 719 /*
 720  * Ok, the machine is now initialized. None of the devices
 721  * have been touched yet, but the CPU subsystem is up and
 722  * running, and memory and process management works.
 723  *
 724  * Now we can finally start doing some real work..
 725  */
 726 static void __init do_basic_setup(void)
 727 {
 728         cpuset_init_smp();
 729         usermodehelper_init();
 730         shmem_init();
 731         driver_init();
 732         init_irq_proc();
 733         do_ctors();
 734         usermodehelper_enable();
 735         do_initcalls();
 736         random_int_secret_init();
 737 }
 738
 739 static void __init do_pre_smp_initcalls(void)
 740 {
 741         initcall_t *fn;
 742
 743         for (fn = __initcall_start; fn < __early_initcall_end; fn++)
 744                 do_one_initcall(*fn);
 745 }
 746
 747 static void run_init_process(const char *init_filename)
 748 {
 749         argv_init[0] = init_filename;
 750         kernel_execve(init_filename, argv_init, envp_init);
 751 }
 752
 753 /* This is a non __init function. Force it to be noinline otherwise gcc
 754  * makes it inline to init() and it becomes part of init.text section
 755  */
 756 static noinline int init_post(void)
 757 {
 758         /* need to finish all async __init code before freeing the memory */
 759         async_synchronize_full();
 760         free_initmem();
 761         mark_rodata_ro();
 762         system_state = SYSTEM_RUNNING;
 763         numa_default_policy();
 764
 765
 766         current->signal->flags |= SIGNAL_UNKILLABLE;
 767
 768         if (ramdisk_execute_command) {
 769                 run_init_process(ramdisk_execute_command);
 770                 printk(KERN_WARNING "Failed to execute %s\n",
 771                                 ramdisk_execute_command);
 772         }
 773
 774         /*
 775          * We try each of these until one succeeds.
 776          *
 777          * The Bourne shell can be used instead of init if we are
 778          * trying to recover a really broken machine.
 779          */
 780         if (execute_command) {
 781                 run_init_process(execute_command);
 782                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 783                                         "defaults...\n", execute_command);
 784         }
 785         run_init_process("/sbin/init");
 786         run_init_process("/etc/init");
 787         run_init_process("/bin/init");
 788         run_init_process("/bin/sh");
 789
 790         panic("No init found.  Try passing init= option to kernel. "
 791               "See Linux Documentation/init.txt for guidance.");
 792 }
 793
 794 static int __init kernel_init(void * unused)
 795 {
 796         /*
 797          * Wait until kthreadd is all set-up.
 798          */
 799         wait_for_completion(&kthreadd_done);
 800
 801         /* Now the scheduler is fully set up and can do blocking allocations */
 802         gfp_allowed_mask = __GFP_BITS_MASK;
 803
 804         /*
 805          * init can allocate pages on any node
 806          */
 807         set_mems_allowed(node_states[N_HIGH_MEMORY]);
 808         /*
 809          * init can run on any cpu.
 810          */
 811         set_cpus_allowed_ptr(current, cpu_all_mask);
 812
 813         cad_pid = task_pid(current);
 814
 815         smp_prepare_cpus(setup_max_cpus);
 816
 817         do_pre_smp_initcalls();
 818         lockup_detector_init();
 819
 820         smp_init();
 821         sched_init_smp();
 822
 823         do_basic_setup();
 824
 825         /* Open the /dev/console on the rootfs, this should never fail */
 826         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 827                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 828
 829         (void) sys_dup(0);
 830         (void) sys_dup(0);
 831         /*
 832          * check if there is an early userspace init.  If yes, let it do all
 833          * the work
 834          */
 835
 836         if (!ramdisk_execute_command)
 837                 ramdisk_execute_command = "/init";
 838
 839         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 840                 ramdisk_execute_command = NULL;
 841                 prepare_namespace();
 842         }
 843
 844         /*
 845          * Ok, we have completed the initial bootup, and
 846          * we're essentially up and running. Get rid of the
 847          * initmem segments and start the user-mode stuff..
 848          */
 849
 850         init_post();
 851         return 0;
 852 }