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