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 #ifndef CONFIG_DEBUG_RODATA
  91 static inline void mark_rodata_ro(void) { }
  92 #endif
  93
  94 #ifdef CONFIG_TC
  95 extern void tc_init(void);
  96 #endif
  97
  98 /*
  99  * Debug helper: via this flag we know that we are in 'early bootup code'
 100  * where only the boot processor is running with IRQ disabled.  This means
 101  * two things - IRQ must not be enabled before the flag is cleared and some
 102  * operations which are not allowed with IRQ disabled are allowed while the
 103  * flag is set.
 104  */
 105 bool early_boot_irqs_disabled __read_mostly;
 106
 107 enum system_states system_state __read_mostly;
 108 EXPORT_SYMBOL(system_state);
 109
 110 /*
 111  * Boot command-line arguments
 112  */
 113 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 114 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 115
 116 extern void time_init(void);
 117 /* Default late time init is NULL. archs can override this later. */
 118 void (*__initdata late_time_init)(void);
 119 extern void softirq_init(void);
 120
 121 /* Untouched command line saved by arch-specific code. */
 122 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 123 /* Untouched saved command line (eg. for /proc) */
 124 char *saved_command_line;
 125 /* Command line for parameter parsing */
 126 static char *static_command_line;
 127
 128 static char *execute_command;
 129 static char *ramdisk_execute_command;
 130
 131 /*
 132  * If set, this is an indication to the drivers that reset the underlying
 133  * device before going ahead with the initialization otherwise driver might
 134  * rely on the BIOS and skip the reset operation.
 135  *
 136  * This is useful if kernel is booting in an unreliable environment.
 137  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 138  * skipped and devices will be in unknown state.
 139  */
 140 unsigned int reset_devices;
 141 EXPORT_SYMBOL(reset_devices);
 142
 143 static int __init set_reset_devices(char *str)
 144 {
 145         reset_devices = 1;
 146         return 1;
 147 }
 148
 149 __setup("reset_devices", set_reset_devices);
 150
 151 static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 152 const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 153 static const char *panic_later, *panic_param;
 154
 155 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 156
 157 static int __init obsolete_checksetup(char *line)
 158 {
 159         const struct obs_kernel_param *p;
 160         int had_early_param = 0;
 161
 162         p = __setup_start;
 163         do {
 164                 int n = strlen(p->str);
 165                 if (parameqn(line, p->str, n)) {
 166                         if (p->early) {
 167                                 /* Already done in parse_early_param?
 168                                  * (Needs exact match on param part).
 169                                  * Keep iterating, as we can have early
 170                                  * params and __setups of same names 8( */
 171                                 if (line[n] == '\0' || line[n] == '=')
 172                                         had_early_param = 1;
 173                         } else if (!p->setup_func) {
 174                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 175                                        " ignored\n", p->str);
 176                                 return 1;
 177                         } else if (p->setup_func(line + n))
 178                                 return 1;
 179                 }
 180                 p++;
 181         } while (p < __setup_end);
 182
 183         return had_early_param;
 184 }
 185
 186 /*
 187  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 188  * still work even if initially too large, it will just take slightly longer
 189  */
 190 unsigned long loops_per_jiffy = (1<<12);
 191
 192 EXPORT_SYMBOL(loops_per_jiffy);
 193
 194 static int __init debug_kernel(char *str)
 195 {
 196         console_loglevel = 10;
 197         return 0;
 198 }
 199
 200 static int __init quiet_kernel(char *str)
 201 {
 202         console_loglevel = 4;
 203         return 0;
 204 }
 205
 206 early_param("debug", debug_kernel);
 207 early_param("quiet", quiet_kernel);
 208
 209 static int __init loglevel(char *str)
 210 {
 211         int newlevel;
 212
 213         /*
 214          * Only update loglevel value when a correct setting was passed,
 215          * to prevent blind crashes (when loglevel being set to 0) that
 216          * are quite hard to debug
 217          */
 218         if (get_option(&str, &newlevel)) {
 219                 console_loglevel = newlevel;
 220                 return 0;
 221         }
 222
 223         return -EINVAL;
 224 }
 225
 226 early_param("loglevel", loglevel);
 227
 228 /*
 229  * Unknown boot options get handed to init, unless they look like
 230  * unused parameters (modprobe will find them in /proc/cmdline).
 231  */
 232 static int __init unknown_bootoption(char *param, char *val)
 233 {
 234         /* Change NUL term back to "=", to make "param" the whole string. */
 235         if (val) {
 236                 /* param=val or param="val"? */
 237                 if (val == param+strlen(param)+1)
 238                         val[-1] = '=';
 239                 else if (val == param+strlen(param)+2) {
 240                         val[-2] = '=';
 241                         memmove(val-1, val, strlen(val)+1);
 242                         val--;
 243                 } else
 244                         BUG();
 245         }
 246
 247         /* Handle obsolete-style parameters */
 248         if (obsolete_checksetup(param))
 249                 return 0;
 250
 251         /* Unused module parameter. */
 252         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 253                 return 0;
 254
 255         if (panic_later)
 256                 return 0;
 257
 258         if (val) {
 259                 /* Environment option */
 260                 unsigned int i;
 261                 for (i = 0; envp_init[i]; i++) {
 262                         if (i == MAX_INIT_ENVS) {
 263                                 panic_later = "Too many boot env vars at `%s'";
 264                                 panic_param = param;
 265                         }
 266                         if (!strncmp(param, envp_init[i], val - param))
 267                                 break;
 268                 }
 269                 envp_init[i] = param;
 270         } else {
 271                 /* Command line option */
 272                 unsigned int i;
 273                 for (i = 0; argv_init[i]; i++) {
 274                         if (i == MAX_INIT_ARGS) {
 275                                 panic_later = "Too many boot init vars at `%s'";
 276                                 panic_param = param;
 277                         }
 278                 }
 279                 argv_init[i] = param;
 280         }
 281         return 0;
 282 }
 283
 284 static int __init init_setup(char *str)
 285 {
 286         unsigned int i;
 287
 288         execute_command = str;
 289         /*
 290          * In case LILO is going to boot us with default command line,
 291          * it prepends "auto" before the whole cmdline which makes
 292          * the shell think it should execute a script with such name.
 293          * So we ignore all arguments entered _before_ init=... [MJ]
 294          */
 295         for (i = 1; i < MAX_INIT_ARGS; i++)
 296                 argv_init[i] = NULL;
 297         return 1;
 298 }
 299 __setup("init=", init_setup);
 300
 301 static int __init rdinit_setup(char *str)
 302 {
 303         unsigned int i;
 304
 305         ramdisk_execute_command = str;
 306         /* See "auto" comment in init_setup */
 307         for (i = 1; i < MAX_INIT_ARGS; i++)
 308                 argv_init[i] = NULL;
 309         return 1;
 310 }
 311 __setup("rdinit=", rdinit_setup);
 312
 313 #ifndef CONFIG_SMP
 314 static const unsigned int setup_max_cpus = NR_CPUS;
 315 #ifdef CONFIG_X86_LOCAL_APIC
 316 static void __init smp_init(void)
 317 {
 318         APIC_init_uniprocessor();
 319 }
 320 #else
 321 #define smp_init()      do { } while (0)
 322 #endif
 323
 324 static inline void setup_nr_cpu_ids(void) { }
 325 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 326 #endif
 327
 328 /*
 329  * We need to store the untouched command line for future reference.
 330  * We also need to store the touched command line since the parameter
 331  * parsing is performed in place, and we should allow a component to
 332  * store reference of name/value for future reference.
 333  */
 334 static void __init setup_command_line(char *command_line)
 335 {
 336         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 337         static_command_line = alloc_bootmem(strlen (command_line)+1);
 338         strcpy (saved_command_line, boot_command_line);
 339         strcpy (static_command_line, command_line);
 340 }
 341
 342 /*
 343  * We need to finalize in a non-__init function or else race conditions
 344  * between the root thread and the init thread may cause start_kernel to
 345  * be reaped by free_initmem before the root thread has proceeded to
 346  * cpu_idle.
 347  *
 348  * gcc-3.4 accidentally inlines this function, so use noinline.
 349  */
 350
 351 static __initdata DECLARE_COMPLETION(kthreadd_done);
 352
 353 static noinline void __init_refok rest_init(void)
 354 {
 355         int pid;
 356
 357         rcu_scheduler_starting();
 358         /*
 359          * We need to spawn init first so that it obtains pid 1, however
 360          * the init task will end up wanting to create kthreads, which, if
 361          * we schedule it before we create kthreadd, will OOPS.
 362          */
 363         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 364         numa_default_policy();
 365         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 366         rcu_read_lock();
 367         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 368         rcu_read_unlock();
 369         complete(&kthreadd_done);
 370
 371         /*
 372          * The boot idle thread must execute schedule()
 373          * at least once to get things moving:
 374          */
 375         init_idle_bootup_task(current);
 376         schedule_preempt_disabled();
 377         /* Call into cpu_idle with preempt disabled */
 378         cpu_idle();
 379 }
 380
 381 /* Check for early params. */
 382 static int __init do_early_param(char *param, char *val)
 383 {
 384         const struct obs_kernel_param *p;
 385
 386         for (p = __setup_start; p < __setup_end; p++) {
 387                 if ((p->early && parameq(param, p->str)) ||
 388                     (strcmp(param, "console") == 0 &&
 389                      strcmp(p->str, "earlycon") == 0)
 390                 ) {
 391                         if (p->setup_func(val) != 0)
 392                                 printk(KERN_WARNING
 393                                        "Malformed early option '%s'\n", param);
 394                 }
 395         }
 396         /* We accept everything at this stage. */
 397         return 0;
 398 }
 399
 400 void __init parse_early_options(char *cmdline)
 401 {
 402         parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
 403 }
 404
 405 /* Arch code calls this early on, or if not, just before other parsing. */
 406 void __init parse_early_param(void)
 407 {
 408         static __initdata int done = 0;
 409         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 410
 411         if (done)
 412                 return;
 413
 414         /* All fall through to do_early_param. */
 415         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 416         parse_early_options(tmp_cmdline);
 417         done = 1;
 418 }
 419
 420 /*
 421  *      Activate the first processor.
 422  */
 423
 424 static void __init boot_cpu_init(void)
 425 {
 426         int cpu = smp_processor_id();
 427         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 428         set_cpu_online(cpu, true);
 429         set_cpu_active(cpu, true);
 430         set_cpu_present(cpu, true);
 431         set_cpu_possible(cpu, true);
 432 }
 433
 434 void __init __weak smp_setup_processor_id(void)
 435 {
 436 }
 437
 438 void __init __weak thread_info_cache_init(void)
 439 {
 440 }
 441
 442 /*
 443  * Set up kernel memory allocators
 444  */
 445 static void __init mm_init(void)
 446 {
 447         /*
 448          * page_cgroup requires contiguous pages,
 449          * bigger than MAX_ORDER unless SPARSEMEM.
 450          */
 451         page_cgroup_init_flatmem();
 452         mem_init();
 453         kmem_cache_init();
 454         percpu_init_late();
 455         pgtable_cache_init();
 456         vmalloc_init();
 457 }
 458
 459 asmlinkage void __init start_kernel(void)
 460 {
 461         char * command_line;
 462         extern const struct kernel_param __start___param[], __stop___param[];
 463
 464         /*
 465          * Need to run as early as possible, to initialize the
 466          * lockdep hash:
 467          */
 468         lockdep_init();
 469         smp_setup_processor_id();
 470         debug_objects_early_init();
 471
 472         /*
 473          * Set up the the initial canary ASAP:
 474          */
 475         boot_init_stack_canary();
 476
 477         cgroup_init_early();
 478
 479         local_irq_disable();
 480         early_boot_irqs_disabled = true;
 481
 482 /*
 483  * Interrupts are still disabled. Do necessary setups, then
 484  * enable them
 485  */
 486         tick_init();
 487         boot_cpu_init();
 488         page_address_init();
 489         printk(KERN_NOTICE "%s", linux_banner);
 490         setup_arch(&command_line);
 491         mm_init_owner(&init_mm, &init_task);
 492         mm_init_cpumask(&init_mm);
 493         setup_command_line(command_line);
 494         setup_nr_cpu_ids();
 495         setup_per_cpu_areas();
 496         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 497
 498         build_all_zonelists(NULL);
 499         page_alloc_init();
 500
 501         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 502         parse_early_param();
 503         parse_args("Booting kernel", static_command_line, __start___param,
 504                    __stop___param - __start___param,
 505                    0, 0, &unknown_bootoption);
 506
 507         jump_label_init();
 508
 509         /*
 510          * These use large bootmem allocations and must precede
 511          * kmem_cache_init()
 512          */
 513         setup_log_buf(0);
 514         pidhash_init();
 515         vfs_caches_init_early();
 516         sort_main_extable();
 517         trap_init();
 518         mm_init();
 519
 520         /*
 521          * Set up the scheduler prior starting any interrupts (such as the
 522          * timer interrupt). Full topology setup happens at smp_init()
 523          * time - but meanwhile we still have a functioning scheduler.
 524          */
 525         sched_init();
 526         /*
 527          * Disable preemption - early bootup scheduling is extremely
 528          * fragile until we cpu_idle() for the first time.
 529          */
 530         preempt_disable();
 531         if (!irqs_disabled()) {
 532                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 533                                 "enabled *very* early, fixing it\n");
 534                 local_irq_disable();
 535         }
 536         idr_init_cache();
 537         perf_event_init();
 538         rcu_init();
 539         radix_tree_init();
 540         /* init some links before init_ISA_irqs() */
 541         early_irq_init();
 542         init_IRQ();
 543         prio_tree_init();
 544         init_timers();
 545         hrtimers_init();
 546         softirq_init();
 547         timekeeping_init();
 548         time_init();
 549         profile_init();
 550         call_function_init();
 551         if (!irqs_disabled())
 552                 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
 553                                  "enabled early\n");
 554         early_boot_irqs_disabled = false;
 555         local_irq_enable();
 556
 557         /* Interrupts are enabled now so all GFP allocations are safe. */
 558         gfp_allowed_mask = __GFP_BITS_MASK;
 559
 560         kmem_cache_init_late();
 561
 562         /*
 563          * HACK ALERT! This is early. We're enabling the console before
 564          * we've done PCI setups etc, and console_init() must be aware of
 565          * this. But we do want output early, in case something goes wrong.
 566          */
 567         console_init();
 568         if (panic_later)
 569                 panic(panic_later, panic_param);
 570
 571         lockdep_info();
 572
 573         /*
 574          * Need to run this when irqs are enabled, because it wants
 575          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 576          * too:
 577          */
 578         locking_selftest();
 579
 580 #ifdef CONFIG_BLK_DEV_INITRD
 581         if (initrd_start && !initrd_below_start_ok &&
 582             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 583                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 584                     "disabling it.\n",
 585                     page_to_pfn(virt_to_page((void *)initrd_start)),
 586                     min_low_pfn);
 587                 initrd_start = 0;
 588         }
 589 #endif
 590         page_cgroup_init();
 591         debug_objects_mem_init();
 592         kmemleak_init();
 593         setup_per_cpu_pageset();
 594         numa_policy_init();
 595         if (late_time_init)
 596                 late_time_init();
 597         sched_clock_init();
 598         calibrate_delay();
 599         pidmap_init();
 600         anon_vma_init();
 601 #ifdef CONFIG_X86
 602         if (efi_enabled)
 603                 efi_enter_virtual_mode();
 604 #endif
 605         thread_info_cache_init();
 606         cred_init();
 607         fork_init(totalram_pages);
 608         proc_caches_init();
 609         buffer_init();
 610         key_init();
 611         security_init();
 612         dbg_late_init();
 613         vfs_caches_init(totalram_pages);
 614         signals_init();
 615         /* rootfs populating might need page-writeback */
 616         page_writeback_init();
 617 #ifdef CONFIG_PROC_FS
 618         proc_root_init();
 619 #endif
 620         cgroup_init();
 621         cpuset_init();
 622         taskstats_init_early();
 623         delayacct_init();
 624
 625         check_bugs();
 626
 627         acpi_early_init(); /* before LAPIC and SMP init */
 628         sfi_init_late();
 629
 630         ftrace_init();
 631
 632         /* Do the rest non-__init'ed, we're now alive */
 633         rest_init();
 634 }
 635
 636 /* Call all constructor functions linked into the kernel. */
 637 static void __init do_ctors(void)
 638 {
 639 #ifdef CONFIG_CONSTRUCTORS
 640         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 641
 642         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 643                 (*fn)();
 644 #endif
 645 }
 646
 647 bool initcall_debug;
 648 core_param(initcall_debug, initcall_debug, bool, 0644);
 649
 650 static char msgbuf[64];
 651
 652 static int __init_or_module do_one_initcall_debug(initcall_t fn)
 653 {
 654         ktime_t calltime, delta, rettime;
 655         unsigned long long duration;
 656         int ret;
 657
 658         printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
 659         calltime = ktime_get();
 660         ret = fn();
 661         rettime = ktime_get();
 662         delta = ktime_sub(rettime, calltime);
 663         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 664         printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn,
 665                 ret, duration);
 666
 667         return ret;
 668 }
 669
 670 int __init_or_module do_one_initcall(initcall_t fn)
 671 {
 672         int count = preempt_count();
 673         int ret;
 674
 675         if (initcall_debug)
 676                 ret = do_one_initcall_debug(fn);
 677         else
 678                 ret = fn();
 679
 680         msgbuf[0] = 0;
 681
 682         if (ret && ret != -ENODEV && initcall_debug)
 683                 sprintf(msgbuf, "error code %d ", ret);
 684
 685         if (preempt_count() != count) {
 686                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 687                 preempt_count() = count;
 688         }
 689         if (irqs_disabled()) {
 690                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 691                 local_irq_enable();
 692         }
 693         if (msgbuf[0]) {
 694                 printk("initcall %pF returned with %s\n", fn, msgbuf);
 695         }
 696
 697         return ret;
 698 }
 699
 700
 701 extern initcall_t __initcall_start[];
 702 extern initcall_t __initcall0_start[];
 703 extern initcall_t __initcall1_start[];
 704 extern initcall_t __initcall2_start[];
 705 extern initcall_t __initcall3_start[];
 706 extern initcall_t __initcall4_start[];
 707 extern initcall_t __initcall5_start[];
 708 extern initcall_t __initcall6_start[];
 709 extern initcall_t __initcall7_start[];
 710 extern initcall_t __initcall_end[];
 711
 712 static initcall_t *initcall_levels[] __initdata = {
 713         __initcall0_start,
 714         __initcall1_start,
 715         __initcall2_start,
 716         __initcall3_start,
 717         __initcall4_start,
 718         __initcall5_start,
 719         __initcall6_start,
 720         __initcall7_start,
 721         __initcall_end,
 722 };
 723
 724 static char *initcall_level_names[] __initdata = {
 725         "early parameters",
 726         "core parameters",
 727         "postcore parameters",
 728         "arch parameters",
 729         "subsys parameters",
 730         "fs parameters",
 731         "device parameters",
 732         "late parameters",
 733 };
 734
 735 static int __init ignore_unknown_bootoption(char *param, char *val)
 736 {
 737         return 0;
 738 }
 739
 740 static void __init do_initcall_level(int level)
 741 {
 742         extern const struct kernel_param __start___param[], __stop___param[];
 743         initcall_t *fn;
 744
 745         strcpy(static_command_line, saved_command_line);
 746         parse_args(initcall_level_names[level],
 747                    static_command_line, __start___param,
 748                    __stop___param - __start___param,
 749                    level, level,
 750                    ignore_unknown_bootoption);
 751
 752         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
 753                 do_one_initcall(*fn);
 754 }
 755
 756 static void __init do_initcalls(void)
 757 {
 758         int level;
 759
 760         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
 761                 do_initcall_level(level);
 762 }
 763
 764 /*
 765  * Ok, the machine is now initialized. None of the devices
 766  * have been touched yet, but the CPU subsystem is up and
 767  * running, and memory and process management works.
 768  *
 769  * Now we can finally start doing some real work..
 770  */
 771 static void __init do_basic_setup(void)
 772 {
 773         cpuset_init_smp();
 774         usermodehelper_init();
 775         shmem_init();
 776         driver_init();
 777         init_irq_proc();
 778         do_ctors();
 779         usermodehelper_enable();
 780         do_initcalls();
 781 }
 782
 783 static void __init do_pre_smp_initcalls(void)
 784 {
 785         initcall_t *fn;
 786
 787         for (fn = __initcall_start; fn < __initcall0_start; fn++)
 788                 do_one_initcall(*fn);
 789 }
 790
 791 static void run_init_process(const char *init_filename)
 792 {
 793         argv_init[0] = init_filename;
 794         kernel_execve(init_filename, argv_init, envp_init);
 795 }
 796
 797 /* This is a non __init function. Force it to be noinline otherwise gcc
 798  * makes it inline to init() and it becomes part of init.text section
 799  */
 800 static noinline int init_post(void)
 801 {
 802         /* need to finish all async __init code before freeing the memory */
 803         async_synchronize_full();
 804         free_initmem();
 805         mark_rodata_ro();
 806         system_state = SYSTEM_RUNNING;
 807         numa_default_policy();
 808
 809
 810         current->signal->flags |= SIGNAL_UNKILLABLE;
 811
 812         if (ramdisk_execute_command) {
 813                 run_init_process(ramdisk_execute_command);
 814                 printk(KERN_WARNING "Failed to execute %s\n",
 815                                 ramdisk_execute_command);
 816         }
 817
 818         /*
 819          * We try each of these until one succeeds.
 820          *
 821          * The Bourne shell can be used instead of init if we are
 822          * trying to recover a really broken machine.
 823          */
 824         if (execute_command) {
 825                 run_init_process(execute_command);
 826                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 827                                         "defaults...\n", execute_command);
 828         }
 829         run_init_process("/sbin/init");
 830         run_init_process("/etc/init");
 831         run_init_process("/bin/init");
 832         run_init_process("/bin/sh");
 833
 834         panic("No init found.  Try passing init= option to kernel. "
 835               "See Linux Documentation/init.txt for guidance.");
 836 }
 837
 838 static int __init kernel_init(void * unused)
 839 {
 840         /*
 841          * Wait until kthreadd is all set-up.
 842          */
 843         wait_for_completion(&kthreadd_done);
 844         /*
 845          * init can allocate pages on any node
 846          */
 847         set_mems_allowed(node_states[N_HIGH_MEMORY]);
 848         /*
 849          * init can run on any cpu.
 850          */
 851         set_cpus_allowed_ptr(current, cpu_all_mask);
 852
 853         cad_pid = task_pid(current);
 854
 855         smp_prepare_cpus(setup_max_cpus);
 856
 857         do_pre_smp_initcalls();
 858         lockup_detector_init();
 859
 860         smp_init();
 861         sched_init_smp();
 862
 863         do_basic_setup();
 864
 865         /* Open the /dev/console on the rootfs, this should never fail */
 866         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 867                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 868
 869         (void) sys_dup(0);
 870         (void) sys_dup(0);
 871         /*
 872          * check if there is an early userspace init.  If yes, let it do all
 873          * the work
 874          */
 875
 876         if (!ramdisk_execute_command)
 877                 ramdisk_execute_command = "/init";
 878
 879         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 880                 ramdisk_execute_command = NULL;
 881                 prepare_namespace();
 882         }
 883
 884         /*
 885          * Ok, we have completed the initial bootup, and
 886          * we're essentially up and running. Get rid of the
 887          * initmem segments and start the user-mode stuff..
 888          */
 889
 890         init_post();
 891         return 0;
 892 }