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