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/file.h>
  72 #include <linux/ptrace.h>
  73
  74 #include <asm/io.h>
  75 #include <asm/bugs.h>
  76 #include <asm/setup.h>
  77 #include <asm/sections.h>
  78 #include <asm/cacheflush.h>
  79
  80 #ifdef CONFIG_X86_LOCAL_APIC
  81 #include <asm/smp.h>
  82 #endif
  83
  84 static int kernel_init(void *);
  85
  86 extern void init_IRQ(void);
  87 extern void fork_init(unsigned long);
  88 extern void mca_init(void);
  89 extern void sbus_init(void);
  90 extern void prio_tree_init(void);
  91 extern void radix_tree_init(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 /* Change NUL term back to "=", to make "param" the whole string. */
 231 static int __init repair_env_string(char *param, char *val, const char *unused)
 232 {
 233         if (val) {
 234                 /* param=val or param="val"? */
 235                 if (val == param+strlen(param)+1)
 236                         val[-1] = '=';
 237                 else if (val == param+strlen(param)+2) {
 238                         val[-2] = '=';
 239                         memmove(val-1, val, strlen(val)+1);
 240                         val--;
 241                 } else
 242                         BUG();
 243         }
 244         return 0;
 245 }
 246
 247 /*
 248  * Unknown boot options get handed to init, unless they look like
 249  * unused parameters (modprobe will find them in /proc/cmdline).
 250  */
 251 static int __init unknown_bootoption(char *param, char *val, const char *unused)
 252 {
 253         repair_env_string(param, val, unused);
 254
 255         /* Handle obsolete-style parameters */
 256         if (obsolete_checksetup(param))
 257                 return 0;
 258
 259         /* Unused module parameter. */
 260         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 261                 return 0;
 262
 263         if (panic_later)
 264                 return 0;
 265
 266         if (val) {
 267                 /* Environment option */
 268                 unsigned int i;
 269                 for (i = 0; envp_init[i]; i++) {
 270                         if (i == MAX_INIT_ENVS) {
 271                                 panic_later = "Too many boot env vars at `%s'";
 272                                 panic_param = param;
 273                         }
 274                         if (!strncmp(param, envp_init[i], val - param))
 275                                 break;
 276                 }
 277                 envp_init[i] = param;
 278         } else {
 279                 /* Command line option */
 280                 unsigned int i;
 281                 for (i = 0; argv_init[i]; i++) {
 282                         if (i == MAX_INIT_ARGS) {
 283                                 panic_later = "Too many boot init vars at `%s'";
 284                                 panic_param = param;
 285                         }
 286                 }
 287                 argv_init[i] = param;
 288         }
 289         return 0;
 290 }
 291
 292 static int __init init_setup(char *str)
 293 {
 294         unsigned int i;
 295
 296         execute_command = str;
 297         /*
 298          * In case LILO is going to boot us with default command line,
 299          * it prepends "auto" before the whole cmdline which makes
 300          * the shell think it should execute a script with such name.
 301          * So we ignore all arguments entered _before_ init=... [MJ]
 302          */
 303         for (i = 1; i < MAX_INIT_ARGS; i++)
 304                 argv_init[i] = NULL;
 305         return 1;
 306 }
 307 __setup("init=", init_setup);
 308
 309 static int __init rdinit_setup(char *str)
 310 {
 311         unsigned int i;
 312
 313         ramdisk_execute_command = str;
 314         /* See "auto" comment in init_setup */
 315         for (i = 1; i < MAX_INIT_ARGS; i++)
 316                 argv_init[i] = NULL;
 317         return 1;
 318 }
 319 __setup("rdinit=", rdinit_setup);
 320
 321 #ifndef CONFIG_SMP
 322 static const unsigned int setup_max_cpus = NR_CPUS;
 323 #ifdef CONFIG_X86_LOCAL_APIC
 324 static void __init smp_init(void)
 325 {
 326         APIC_init_uniprocessor();
 327 }
 328 #else
 329 #define smp_init()      do { } while (0)
 330 #endif
 331
 332 static inline void setup_nr_cpu_ids(void) { }
 333 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 334 #endif
 335
 336 /*
 337  * We need to store the untouched command line for future reference.
 338  * We also need to store the touched command line since the parameter
 339  * parsing is performed in place, and we should allow a component to
 340  * store reference of name/value for future reference.
 341  */
 342 static void __init setup_command_line(char *command_line)
 343 {
 344         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 345         static_command_line = alloc_bootmem(strlen (command_line)+1);
 346         strcpy (saved_command_line, boot_command_line);
 347         strcpy (static_command_line, command_line);
 348 }
 349
 350 /*
 351  * We need to finalize in a non-__init function or else race conditions
 352  * between the root thread and the init thread may cause start_kernel to
 353  * be reaped by free_initmem before the root thread has proceeded to
 354  * cpu_idle.
 355  *
 356  * gcc-3.4 accidentally inlines this function, so use noinline.
 357  */
 358
 359 static __initdata DECLARE_COMPLETION(kthreadd_done);
 360
 361 static noinline void __init_refok rest_init(void)
 362 {
 363         int pid;
 364
 365         rcu_scheduler_starting();
 366         /*
 367          * We need to spawn init first so that it obtains pid 1, however
 368          * the init task will end up wanting to create kthreads, which, if
 369          * we schedule it before we create kthreadd, will OOPS.
 370          */
 371         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 372         numa_default_policy();
 373         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 374         rcu_read_lock();
 375         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 376         rcu_read_unlock();
 377         complete(&kthreadd_done);
 378
 379         /*
 380          * The boot idle thread must execute schedule()
 381          * at least once to get things moving:
 382          */
 383         init_idle_bootup_task(current);
 384         schedule_preempt_disabled();
 385         /* Call into cpu_idle with preempt disabled */
 386         cpu_idle();
 387 }
 388
 389 /* Check for early params. */
 390 static int __init do_early_param(char *param, char *val, const char *unused)
 391 {
 392         const struct obs_kernel_param *p;
 393
 394         for (p = __setup_start; p < __setup_end; p++) {
 395                 if ((p->early && parameq(param, p->str)) ||
 396                     (strcmp(param, "console") == 0 &&
 397                      strcmp(p->str, "earlycon") == 0)
 398                 ) {
 399                         if (p->setup_func(val) != 0)
 400                                 printk(KERN_WARNING
 401                                        "Malformed early option '%s'\n", param);
 402                 }
 403         }
 404         /* We accept everything at this stage. */
 405         return 0;
 406 }
 407
 408 void __init parse_early_options(char *cmdline)
 409 {
 410         parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
 411 }
 412
 413 /* Arch code calls this early on, or if not, just before other parsing. */
 414 void __init parse_early_param(void)
 415 {
 416         static __initdata int done = 0;
 417         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 418
 419         if (done)
 420                 return;
 421
 422         /* All fall through to do_early_param. */
 423         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 424         parse_early_options(tmp_cmdline);
 425         done = 1;
 426 }
 427
 428 /*
 429  *      Activate the first processor.
 430  */
 431
 432 static void __init boot_cpu_init(void)
 433 {
 434         int cpu = smp_processor_id();
 435         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 436         set_cpu_online(cpu, true);
 437         set_cpu_active(cpu, true);
 438         set_cpu_present(cpu, true);
 439         set_cpu_possible(cpu, true);
 440 }
 441
 442 void __init __weak smp_setup_processor_id(void)
 443 {
 444 }
 445
 446 void __init __weak thread_info_cache_init(void)
 447 {
 448 }
 449
 450 /*
 451  * Set up kernel memory allocators
 452  */
 453 static void __init mm_init(void)
 454 {
 455         /*
 456          * page_cgroup requires contiguous pages,
 457          * bigger than MAX_ORDER unless SPARSEMEM.
 458          */
 459         page_cgroup_init_flatmem();
 460         mem_init();
 461         kmem_cache_init();
 462         percpu_init_late();
 463         pgtable_cache_init();
 464         vmalloc_init();
 465 }
 466
 467 asmlinkage void __init start_kernel(void)
 468 {
 469         char * command_line;
 470         extern const struct kernel_param __start___param[], __stop___param[];
 471
 472         /*
 473          * Need to run as early as possible, to initialize the
 474          * lockdep hash:
 475          */
 476         lockdep_init();
 477         smp_setup_processor_id();
 478         debug_objects_early_init();
 479
 480         /*
 481          * Set up the the initial canary ASAP:
 482          */
 483         boot_init_stack_canary();
 484
 485         cgroup_init_early();
 486
 487         local_irq_disable();
 488         early_boot_irqs_disabled = true;
 489
 490 /*
 491  * Interrupts are still disabled. Do necessary setups, then
 492  * enable them
 493  */
 494         tick_init();
 495         boot_cpu_init();
 496         page_address_init();
 497         printk(KERN_NOTICE "%s", linux_banner);
 498         setup_arch(&command_line);
 499         mm_init_owner(&init_mm, &init_task);
 500         mm_init_cpumask(&init_mm);
 501         setup_command_line(command_line);
 502         setup_nr_cpu_ids();
 503         setup_per_cpu_areas();
 504         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 505
 506         build_all_zonelists(NULL, NULL);
 507         page_alloc_init();
 508
 509         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 510         parse_early_param();
 511         parse_args("Booting kernel", static_command_line, __start___param,
 512                    __stop___param - __start___param,
 513                    -1, -1, &unknown_bootoption);
 514
 515         jump_label_init();
 516
 517         /*
 518          * These use large bootmem allocations and must precede
 519          * kmem_cache_init()
 520          */
 521         setup_log_buf(0);
 522         pidhash_init();
 523         vfs_caches_init_early();
 524         sort_main_extable();
 525         trap_init();
 526         mm_init();
 527
 528         /*
 529          * Set up the scheduler prior starting any interrupts (such as the
 530          * timer interrupt). Full topology setup happens at smp_init()
 531          * time - but meanwhile we still have a functioning scheduler.
 532          */
 533         sched_init();
 534         /*
 535          * Disable preemption - early bootup scheduling is extremely
 536          * fragile until we cpu_idle() for the first time.
 537          */
 538         preempt_disable();
 539         if (!irqs_disabled()) {
 540                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 541                                 "enabled *very* early, fixing it\n");
 542                 local_irq_disable();
 543         }
 544         idr_init_cache();
 545         perf_event_init();
 546         rcu_init();
 547         radix_tree_init();
 548         /* init some links before init_ISA_irqs() */
 549         early_irq_init();
 550         init_IRQ();
 551         prio_tree_init();
 552         init_timers();
 553         hrtimers_init();
 554         softirq_init();
 555         timekeeping_init();
 556         time_init();
 557         profile_init();
 558         call_function_init();
 559         if (!irqs_disabled())
 560                 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
 561                                  "enabled early\n");
 562         early_boot_irqs_disabled = false;
 563         local_irq_enable();
 564
 565         kmem_cache_init_late();
 566
 567         /*
 568          * HACK ALERT! This is early. We're enabling the console before
 569          * we've done PCI setups etc, and console_init() must be aware of
 570          * this. But we do want output early, in case something goes wrong.
 571          */
 572         console_init();
 573         if (panic_later)
 574                 panic(panic_later, panic_param);
 575
 576         lockdep_info();
 577
 578         /*
 579          * Need to run this when irqs are enabled, because it wants
 580          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 581          * too:
 582          */
 583         locking_selftest();
 584
 585 #ifdef CONFIG_BLK_DEV_INITRD
 586         if (initrd_start && !initrd_below_start_ok &&
 587             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 588                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 589                     "disabling it.\n",
 590                     page_to_pfn(virt_to_page((void *)initrd_start)),
 591                     min_low_pfn);
 592                 initrd_start = 0;
 593         }
 594 #endif
 595         page_cgroup_init();
 596         debug_objects_mem_init();
 597         kmemleak_init();
 598         setup_per_cpu_pageset();
 599         numa_policy_init();
 600         if (late_time_init)
 601                 late_time_init();
 602         sched_clock_init();
 603         calibrate_delay();
 604         pidmap_init();
 605         anon_vma_init();
 606 #ifdef CONFIG_X86
 607         if (efi_enabled)
 608                 efi_enter_virtual_mode();
 609 #endif
 610         thread_info_cache_init();
 611         cred_init();
 612         fork_init(totalram_pages);
 613         proc_caches_init();
 614         buffer_init();
 615         key_init();
 616         security_init();
 617         dbg_late_init();
 618         vfs_caches_init(totalram_pages);
 619         signals_init();
 620         /* rootfs populating might need page-writeback */
 621         page_writeback_init();
 622 #ifdef CONFIG_PROC_FS
 623         proc_root_init();
 624 #endif
 625         cgroup_init();
 626         cpuset_init();
 627         taskstats_init_early();
 628         delayacct_init();
 629
 630         check_bugs();
 631
 632         acpi_early_init(); /* before LAPIC and SMP init */
 633         sfi_init_late();
 634
 635         ftrace_init();
 636
 637         /* Do the rest non-__init'ed, we're now alive */
 638         rest_init();
 639 }
 640
 641 /* Call all constructor functions linked into the kernel. */
 642 static void __init do_ctors(void)
 643 {
 644 #ifdef CONFIG_CONSTRUCTORS
 645         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 646
 647         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 648                 (*fn)();
 649 #endif
 650 }
 651
 652 bool initcall_debug;
 653 core_param(initcall_debug, initcall_debug, bool, 0644);
 654
 655 static char msgbuf[64];
 656
 657 static int __init_or_module do_one_initcall_debug(initcall_t fn)
 658 {
 659         ktime_t calltime, delta, rettime;
 660         unsigned long long duration;
 661         int ret;
 662
 663         printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
 664         calltime = ktime_get();
 665         ret = fn();
 666         rettime = ktime_get();
 667         delta = ktime_sub(rettime, calltime);
 668         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 669         printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn,
 670                 ret, duration);
 671
 672         return ret;
 673 }
 674
 675 int __init_or_module do_one_initcall(initcall_t fn)
 676 {
 677         int count = preempt_count();
 678         int ret;
 679
 680         if (initcall_debug)
 681                 ret = do_one_initcall_debug(fn);
 682         else
 683                 ret = fn();
 684
 685         msgbuf[0] = 0;
 686
 687         if (ret && ret != -ENODEV && initcall_debug)
 688                 sprintf(msgbuf, "error code %d ", ret);
 689
 690         if (preempt_count() != count) {
 691                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 692                 preempt_count() = count;
 693         }
 694         if (irqs_disabled()) {
 695                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 696                 local_irq_enable();
 697         }
 698         if (msgbuf[0]) {
 699                 printk("initcall %pF returned with %s\n", fn, msgbuf);
 700         }
 701
 702         return ret;
 703 }
 704
 705
 706 extern initcall_t __initcall_start[];
 707 extern initcall_t __initcall0_start[];
 708 extern initcall_t __initcall1_start[];
 709 extern initcall_t __initcall2_start[];
 710 extern initcall_t __initcall3_start[];
 711 extern initcall_t __initcall4_start[];
 712 extern initcall_t __initcall5_start[];
 713 extern initcall_t __initcall6_start[];
 714 extern initcall_t __initcall7_start[];
 715 extern initcall_t __initcall_end[];
 716
 717 static initcall_t *initcall_levels[] __initdata = {
 718         __initcall0_start,
 719         __initcall1_start,
 720         __initcall2_start,
 721         __initcall3_start,
 722         __initcall4_start,
 723         __initcall5_start,
 724         __initcall6_start,
 725         __initcall7_start,
 726         __initcall_end,
 727 };
 728
 729 /* Keep these in sync with initcalls in include/linux/init.h */
 730 static char *initcall_level_names[] __initdata = {
 731         "early",
 732         "core",
 733         "postcore",
 734         "arch",
 735         "subsys",
 736         "fs",
 737         "device",
 738         "late",
 739 };
 740
 741 static void __init do_initcall_level(int level)
 742 {
 743         extern const struct kernel_param __start___param[], __stop___param[];
 744         initcall_t *fn;
 745
 746         strcpy(static_command_line, saved_command_line);
 747         parse_args(initcall_level_names[level],
 748                    static_command_line, __start___param,
 749                    __stop___param - __start___param,
 750                    level, level,
 751                    &repair_env_string);
 752
 753         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
 754                 do_one_initcall(*fn);
 755 }
 756
 757 static void __init do_initcalls(void)
 758 {
 759         int level;
 760
 761         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
 762                 do_initcall_level(level);
 763 }
 764
 765 /*
 766  * Ok, the machine is now initialized. None of the devices
 767  * have been touched yet, but the CPU subsystem is up and
 768  * running, and memory and process management works.
 769  *
 770  * Now we can finally start doing some real work..
 771  */
 772 static void __init do_basic_setup(void)
 773 {
 774         cpuset_init_smp();
 775         usermodehelper_init();
 776         shmem_init();
 777         driver_init();
 778         init_irq_proc();
 779         do_ctors();
 780         usermodehelper_enable();
 781         do_initcalls();
 782 }
 783
 784 static void __init do_pre_smp_initcalls(void)
 785 {
 786         initcall_t *fn;
 787
 788         for (fn = __initcall_start; fn < __initcall0_start; fn++)
 789                 do_one_initcall(*fn);
 790 }
 791
 792 static int run_init_process(const char *init_filename)
 793 {
 794         argv_init[0] = init_filename;
 795         return kernel_execve(init_filename, argv_init, envp_init);
 796 }
 797
 798 static void __init kernel_init_freeable(void);
 799
 800 static int __ref kernel_init(void *unused)
 801 {
 802         kernel_init_freeable();
 803         /* need to finish all async __init code before freeing the memory */
 804         async_synchronize_full();
 805         free_initmem();
 806         mark_rodata_ro();
 807         system_state = SYSTEM_RUNNING;
 808         numa_default_policy();
 809
 810         current->signal->flags |= SIGNAL_UNKILLABLE;
 811         flush_delayed_fput();
 812
 813         if (ramdisk_execute_command) {
 814                 if (!run_init_process(ramdisk_execute_command))
 815                         return 0;
 816                 printk(KERN_WARNING "Failed to execute %s\n",
 817                                 ramdisk_execute_command);
 818         }
 819
 820         /*
 821          * We try each of these until one succeeds.
 822          *
 823          * The Bourne shell can be used instead of init if we are
 824          * trying to recover a really broken machine.
 825          */
 826         if (execute_command) {
 827                 if (!run_init_process(execute_command))
 828                         return 0;
 829                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 830                                         "defaults...\n", execute_command);
 831         }
 832         if (!run_init_process("/sbin/init") ||
 833             !run_init_process("/etc/init") ||
 834             !run_init_process("/bin/init") ||
 835             !run_init_process("/bin/sh"))
 836                 return 0;
 837
 838         panic("No init found.  Try passing init= option to kernel. "
 839               "See Linux Documentation/init.txt for guidance.");
 840 }
 841
 842 static void __init kernel_init_freeable(void)
 843 {
 844         /*
 845          * Wait until kthreadd is all set-up.
 846          */
 847         wait_for_completion(&kthreadd_done);
 848
 849         /* Now the scheduler is fully set up and can do blocking allocations */
 850         gfp_allowed_mask = __GFP_BITS_MASK;
 851
 852         /*
 853          * init can allocate pages on any node
 854          */
 855         set_mems_allowed(node_states[N_HIGH_MEMORY]);
 856         /*
 857          * init can run on any cpu.
 858          */
 859         set_cpus_allowed_ptr(current, cpu_all_mask);
 860
 861         cad_pid = task_pid(current);
 862
 863         smp_prepare_cpus(setup_max_cpus);
 864
 865         do_pre_smp_initcalls();
 866         lockup_detector_init();
 867
 868         smp_init();
 869         sched_init_smp();
 870
 871         do_basic_setup();
 872
 873         /* Open the /dev/console on the rootfs, this should never fail */
 874         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 875                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 876
 877         (void) sys_dup(0);
 878         (void) sys_dup(0);
 879         /*
 880          * check if there is an early userspace init.  If yes, let it do all
 881          * the work
 882          */
 883
 884         if (!ramdisk_execute_command)
 885                 ramdisk_execute_command = "/init";
 886
 887         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 888                 ramdisk_execute_command = NULL;
 889                 prepare_namespace();
 890         }
 891
 892         /*
 893          * Ok, we have completed the initial bootup, and
 894          * we're essentially up and running. Get rid of the
 895          * initmem segments and start the user-mode stuff..
 896          */
 897 }