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/string.h>
  18 #include <linux/ctype.h>
  19 #include <linux/delay.h>
  20 #include <linux/utsname.h>
  21 #include <linux/ioport.h>
  22 #include <linux/init.h>
  23 #include <linux/smp_lock.h>
  24 #include <linux/initrd.h>
  25 #include <linux/hdreg.h>
  26 #include <linux/bootmem.h>
  27 #include <linux/tty.h>
  28 #include <linux/gfp.h>
  29 #include <linux/percpu.h>
  30 #include <linux/kmod.h>
  31 #include <linux/kernel_stat.h>
  32 #include <linux/start_kernel.h>
  33 #include <linux/security.h>
  34 #include <linux/smp.h>
  35 #include <linux/workqueue.h>
  36 #include <linux/profile.h>
  37 #include <linux/rcupdate.h>
  38 #include <linux/moduleparam.h>
  39 #include <linux/kallsyms.h>
  40 #include <linux/writeback.h>
  41 #include <linux/cpu.h>
  42 #include <linux/cpuset.h>
  43 #include <linux/cgroup.h>
  44 #include <linux/efi.h>
  45 #include <linux/tick.h>
  46 #include <linux/interrupt.h>
  47 #include <linux/taskstats_kern.h>
  48 #include <linux/delayacct.h>
  49 #include <linux/unistd.h>
  50 #include <linux/rmap.h>
  51 #include <linux/mempolicy.h>
  52 #include <linux/key.h>
  53 #include <linux/unwind.h>
  54 #include <linux/buffer_head.h>
  55 #include <linux/debug_locks.h>
  56 #include <linux/debugobjects.h>
  57 #include <linux/lockdep.h>
  58 #include <linux/pid_namespace.h>
  59 #include <linux/device.h>
  60 #include <linux/kthread.h>
  61 #include <linux/sched.h>
  62 #include <linux/signal.h>
  63 #include <linux/idr.h>
  64
  65 #include <asm/io.h>
  66 #include <asm/bugs.h>
  67 #include <asm/setup.h>
  68 #include <asm/sections.h>
  69 #include <asm/cacheflush.h>
  70
  71 #ifdef CONFIG_X86_LOCAL_APIC
  72 #include <asm/smp.h>
  73 #endif
  74
  75 /*
  76  * This is one of the first .c files built. Error out early if we have compiler
  77  * trouble.
  78  */
  79
  80 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
  81 #warning gcc-4.1.0 is known to miscompile the kernel.  A different compiler version is recommended.
  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 extern void free_initmem(void);
  93 #ifdef  CONFIG_ACPI
  94 extern void acpi_early_init(void);
  95 #else
  96 static inline void acpi_early_init(void) { }
  97 #endif
  98 #ifndef CONFIG_DEBUG_RODATA
  99 static inline void mark_rodata_ro(void) { }
 100 #endif
 101
 102 #ifdef CONFIG_TC
 103 extern void tc_init(void);
 104 #endif
 105
 106 enum system_states system_state;
 107 EXPORT_SYMBOL(system_state);
 108
 109 /*
 110  * Boot command-line arguments
 111  */
 112 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 113 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 114
 115 extern void time_init(void);
 116 /* Default late time init is NULL. archs can override this later. */
 117 void (*late_time_init)(void);
 118 extern void softirq_init(void);
 119
 120 /* Untouched command line saved by arch-specific code. */
 121 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 122 /* Untouched saved command line (eg. for /proc) */
 123 char *saved_command_line;
 124 /* Command line for parameter parsing */
 125 static char *static_command_line;
 126
 127 static char *execute_command;
 128 static char *ramdisk_execute_command;
 129
 130 #ifdef CONFIG_SMP
 131 /* Setup configured maximum number of CPUs to activate */
 132 unsigned int __initdata setup_max_cpus = NR_CPUS;
 133
 134 /*
 135  * Setup routine for controlling SMP activation
 136  *
 137  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 138  * activation entirely (the MPS table probe still happens, though).
 139  *
 140  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 141  * greater than 0, limits the maximum number of CPUs activated in
 142  * SMP mode to <NUM>.
 143  */
 144 #ifndef CONFIG_X86_IO_APIC
 145 static inline void disable_ioapic_setup(void) {};
 146 #endif
 147
 148 static int __init nosmp(char *str)
 149 {
 150         setup_max_cpus = 0;
 151         disable_ioapic_setup();
 152         return 0;
 153 }
 154
 155 early_param("nosmp", nosmp);
 156
 157 static int __init maxcpus(char *str)
 158 {
 159         get_option(&str, &setup_max_cpus);
 160         if (setup_max_cpus == 0)
 161                 disable_ioapic_setup();
 162
 163         return 0;
 164 }
 165
 166 early_param("maxcpus", maxcpus);
 167 #else
 168 #define setup_max_cpus NR_CPUS
 169 #endif
 170
 171 /*
 172  * If set, this is an indication to the drivers that reset the underlying
 173  * device before going ahead with the initialization otherwise driver might
 174  * rely on the BIOS and skip the reset operation.
 175  *
 176  * This is useful if kernel is booting in an unreliable environment.
 177  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 178  * skipped and devices will be in unknown state.
 179  */
 180 unsigned int reset_devices;
 181 EXPORT_SYMBOL(reset_devices);
 182
 183 static int __init set_reset_devices(char *str)
 184 {
 185         reset_devices = 1;
 186         return 1;
 187 }
 188
 189 __setup("reset_devices", set_reset_devices);
 190
 191 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 192 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 193 static const char *panic_later, *panic_param;
 194
 195 extern struct obs_kernel_param __setup_start[], __setup_end[];
 196
 197 static int __init obsolete_checksetup(char *line)
 198 {
 199         struct obs_kernel_param *p;
 200         int had_early_param = 0;
 201
 202         p = __setup_start;
 203         do {
 204                 int n = strlen(p->str);
 205                 if (!strncmp(line, p->str, n)) {
 206                         if (p->early) {
 207                                 /* Already done in parse_early_param?
 208                                  * (Needs exact match on param part).
 209                                  * Keep iterating, as we can have early
 210                                  * params and __setups of same names 8( */
 211                                 if (line[n] == '\0' || line[n] == '=')
 212                                         had_early_param = 1;
 213                         } else if (!p->setup_func) {
 214                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 215                                        " ignored\n", p->str);
 216                                 return 1;
 217                         } else if (p->setup_func(line + n))
 218                                 return 1;
 219                 }
 220                 p++;
 221         } while (p < __setup_end);
 222
 223         return had_early_param;
 224 }
 225
 226 /*
 227  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 228  * still work even if initially too large, it will just take slightly longer
 229  */
 230 unsigned long loops_per_jiffy = (1<<12);
 231
 232 EXPORT_SYMBOL(loops_per_jiffy);
 233
 234 static int __init debug_kernel(char *str)
 235 {
 236         console_loglevel = 10;
 237         return 0;
 238 }
 239
 240 static int __init quiet_kernel(char *str)
 241 {
 242         console_loglevel = 4;
 243         return 0;
 244 }
 245
 246 early_param("debug", debug_kernel);
 247 early_param("quiet", quiet_kernel);
 248
 249 static int __init loglevel(char *str)
 250 {
 251         get_option(&str, &console_loglevel);
 252         return 0;
 253 }
 254
 255 early_param("loglevel", loglevel);
 256
 257 /*
 258  * Unknown boot options get handed to init, unless they look like
 259  * failed parameters
 260  */
 261 static int __init unknown_bootoption(char *param, char *val)
 262 {
 263         /* Change NUL term back to "=", to make "param" the whole string. */
 264         if (val) {
 265                 /* param=val or param="val"? */
 266                 if (val == param+strlen(param)+1)
 267                         val[-1] = '=';
 268                 else if (val == param+strlen(param)+2) {
 269                         val[-2] = '=';
 270                         memmove(val-1, val, strlen(val)+1);
 271                         val--;
 272                 } else
 273                         BUG();
 274         }
 275
 276         /* Handle obsolete-style parameters */
 277         if (obsolete_checksetup(param))
 278                 return 0;
 279
 280         /*
 281          * Preemptive maintenance for "why didn't my misspelled command
 282          * line work?"
 283          */
 284         if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
 285                 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
 286                 return 0;
 287         }
 288
 289         if (panic_later)
 290                 return 0;
 291
 292         if (val) {
 293                 /* Environment option */
 294                 unsigned int i;
 295                 for (i = 0; envp_init[i]; i++) {
 296                         if (i == MAX_INIT_ENVS) {
 297                                 panic_later = "Too many boot env vars at `%s'";
 298                                 panic_param = param;
 299                         }
 300                         if (!strncmp(param, envp_init[i], val - param))
 301                                 break;
 302                 }
 303                 envp_init[i] = param;
 304         } else {
 305                 /* Command line option */
 306                 unsigned int i;
 307                 for (i = 0; argv_init[i]; i++) {
 308                         if (i == MAX_INIT_ARGS) {
 309                                 panic_later = "Too many boot init vars at `%s'";
 310                                 panic_param = param;
 311                         }
 312                 }
 313                 argv_init[i] = param;
 314         }
 315         return 0;
 316 }
 317
 318 #ifdef CONFIG_DEBUG_PAGEALLOC
 319 int __read_mostly debug_pagealloc_enabled = 0;
 320 #endif
 321
 322 static int __init init_setup(char *str)
 323 {
 324         unsigned int i;
 325
 326         execute_command = str;
 327         /*
 328          * In case LILO is going to boot us with default command line,
 329          * it prepends "auto" before the whole cmdline which makes
 330          * the shell think it should execute a script with such name.
 331          * So we ignore all arguments entered _before_ init=... [MJ]
 332          */
 333         for (i = 1; i < MAX_INIT_ARGS; i++)
 334                 argv_init[i] = NULL;
 335         return 1;
 336 }
 337 __setup("init=", init_setup);
 338
 339 static int __init rdinit_setup(char *str)
 340 {
 341         unsigned int i;
 342
 343         ramdisk_execute_command = str;
 344         /* See "auto" comment in init_setup */
 345         for (i = 1; i < MAX_INIT_ARGS; i++)
 346                 argv_init[i] = NULL;
 347         return 1;
 348 }
 349 __setup("rdinit=", rdinit_setup);
 350
 351 #ifndef CONFIG_SMP
 352
 353 #ifdef CONFIG_X86_LOCAL_APIC
 354 static void __init smp_init(void)
 355 {
 356         APIC_init_uniprocessor();
 357 }
 358 #else
 359 #define smp_init()      do { } while (0)
 360 #endif
 361
 362 static inline void setup_per_cpu_areas(void) { }
 363 static inline void setup_nr_cpu_ids(void) { }
 364 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 365
 366 #else
 367
 368 #if NR_CPUS > BITS_PER_LONG
 369 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
 370 EXPORT_SYMBOL(cpu_mask_all);
 371 #endif
 372
 373 /* Setup number of possible processor ids */
 374 int nr_cpu_ids __read_mostly = NR_CPUS;
 375 EXPORT_SYMBOL(nr_cpu_ids);
 376
 377 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
 378 static void __init setup_nr_cpu_ids(void)
 379 {
 380         int cpu, highest_cpu = 0;
 381
 382         for_each_possible_cpu(cpu)
 383                 highest_cpu = cpu;
 384
 385         nr_cpu_ids = highest_cpu + 1;
 386 }
 387
 388 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
 389 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
 390
 391 EXPORT_SYMBOL(__per_cpu_offset);
 392
 393 static void __init setup_per_cpu_areas(void)
 394 {
 395         unsigned long size, i;
 396         char *ptr;
 397         unsigned long nr_possible_cpus = num_possible_cpus();
 398
 399         /* Copy section for each CPU (we discard the original) */
 400         size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
 401         ptr = alloc_bootmem_pages(size * nr_possible_cpus);
 402
 403         for_each_possible_cpu(i) {
 404                 __per_cpu_offset[i] = ptr - __per_cpu_start;
 405                 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 406                 ptr += size;
 407         }
 408 }
 409 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
 410
 411 /* Called by boot processor to activate the rest. */
 412 static void __init smp_init(void)
 413 {
 414         unsigned int cpu;
 415
 416         /*
 417          * Set up the current CPU as possible to migrate to.
 418          * The other ones will be done by cpu_up/cpu_down()
 419          */
 420         cpu = smp_processor_id();
 421         cpu_set(cpu, cpu_active_map);
 422
 423         /* FIXME: This should be done in userspace --RR */
 424         for_each_present_cpu(cpu) {
 425                 if (num_online_cpus() >= setup_max_cpus)
 426                         break;
 427                 if (!cpu_online(cpu))
 428                         cpu_up(cpu);
 429         }
 430
 431         /* Any cleanup work */
 432         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 433         smp_cpus_done(setup_max_cpus);
 434 }
 435
 436 #endif
 437
 438 /*
 439  * We need to store the untouched command line for future reference.
 440  * We also need to store the touched command line since the parameter
 441  * parsing is performed in place, and we should allow a component to
 442  * store reference of name/value for future reference.
 443  */
 444 static void __init setup_command_line(char *command_line)
 445 {
 446         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 447         static_command_line = alloc_bootmem(strlen (command_line)+1);
 448         strcpy (saved_command_line, boot_command_line);
 449         strcpy (static_command_line, command_line);
 450 }
 451
 452 /*
 453  * We need to finalize in a non-__init function or else race conditions
 454  * between the root thread and the init thread may cause start_kernel to
 455  * be reaped by free_initmem before the root thread has proceeded to
 456  * cpu_idle.
 457  *
 458  * gcc-3.4 accidentally inlines this function, so use noinline.
 459  */
 460
 461 static void noinline __init_refok rest_init(void)
 462         __releases(kernel_lock)
 463 {
 464         int pid;
 465
 466         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 467         numa_default_policy();
 468         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 469         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 470         unlock_kernel();
 471
 472         /*
 473          * The boot idle thread must execute schedule()
 474          * at least once to get things moving:
 475          */
 476         init_idle_bootup_task(current);
 477         preempt_enable_no_resched();
 478         schedule();
 479         preempt_disable();
 480
 481         /* Call into cpu_idle with preempt disabled */
 482         cpu_idle();
 483 }
 484
 485 /* Check for early params. */
 486 static int __init do_early_param(char *param, char *val)
 487 {
 488         struct obs_kernel_param *p;
 489
 490         for (p = __setup_start; p < __setup_end; p++) {
 491                 if ((p->early && strcmp(param, p->str) == 0) ||
 492                     (strcmp(param, "console") == 0 &&
 493                      strcmp(p->str, "earlycon") == 0)
 494                 ) {
 495                         if (p->setup_func(val) != 0)
 496                                 printk(KERN_WARNING
 497                                        "Malformed early option '%s'\n", param);
 498                 }
 499         }
 500         /* We accept everything at this stage. */
 501         return 0;
 502 }
 503
 504 /* Arch code calls this early on, or if not, just before other parsing. */
 505 void __init parse_early_param(void)
 506 {
 507         static __initdata int done = 0;
 508         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 509
 510         if (done)
 511                 return;
 512
 513         /* All fall through to do_early_param. */
 514         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 515         parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
 516         done = 1;
 517 }
 518
 519 /*
 520  *      Activate the first processor.
 521  */
 522
 523 static void __init boot_cpu_init(void)
 524 {
 525         int cpu = smp_processor_id();
 526         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 527         cpu_set(cpu, cpu_online_map);
 528         cpu_set(cpu, cpu_present_map);
 529         cpu_set(cpu, cpu_possible_map);
 530 }
 531
 532 void __init __weak smp_setup_processor_id(void)
 533 {
 534 }
 535
 536 void __init __weak thread_info_cache_init(void)
 537 {
 538 }
 539
 540 asmlinkage void __init start_kernel(void)
 541 {
 542         char * command_line;
 543         extern struct kernel_param __start___param[], __stop___param[];
 544
 545         smp_setup_processor_id();
 546
 547         /*
 548          * Need to run as early as possible, to initialize the
 549          * lockdep hash:
 550          */
 551         unwind_init();
 552         lockdep_init();
 553         debug_objects_early_init();
 554         cgroup_init_early();
 555
 556         local_irq_disable();
 557         early_boot_irqs_off();
 558         early_init_irq_lock_class();
 559
 560 /*
 561  * Interrupts are still disabled. Do necessary setups, then
 562  * enable them
 563  */
 564         lock_kernel();
 565         tick_init();
 566         boot_cpu_init();
 567         page_address_init();
 568         printk(KERN_NOTICE);
 569         printk(linux_banner);
 570         setup_arch(&command_line);
 571         mm_init_owner(&init_mm, &init_task);
 572         setup_command_line(command_line);
 573         unwind_setup();
 574         setup_per_cpu_areas();
 575         setup_nr_cpu_ids();
 576         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 577
 578         /*
 579          * Set up the scheduler prior starting any interrupts (such as the
 580          * timer interrupt). Full topology setup happens at smp_init()
 581          * time - but meanwhile we still have a functioning scheduler.
 582          */
 583         sched_init();
 584         /*
 585          * Disable preemption - early bootup scheduling is extremely
 586          * fragile until we cpu_idle() for the first time.
 587          */
 588         preempt_disable();
 589         build_all_zonelists();
 590         page_alloc_init();
 591         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 592         parse_early_param();
 593         parse_args("Booting kernel", static_command_line, __start___param,
 594                    __stop___param - __start___param,
 595                    &unknown_bootoption);
 596         if (!irqs_disabled()) {
 597                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 598                                 "enabled *very* early, fixing it\n");
 599                 local_irq_disable();
 600         }
 601         sort_main_extable();
 602         trap_init();
 603         rcu_init();
 604         init_IRQ();
 605         pidhash_init();
 606         init_timers();
 607         hrtimers_init();
 608         softirq_init();
 609         timekeeping_init();
 610         time_init();
 611         sched_clock_init();
 612         profile_init();
 613         if (!irqs_disabled())
 614                 printk("start_kernel(): bug: interrupts were enabled early\n");
 615         early_boot_irqs_on();
 616         local_irq_enable();
 617
 618         /*
 619          * HACK ALERT! This is early. We're enabling the console before
 620          * we've done PCI setups etc, and console_init() must be aware of
 621          * this. But we do want output early, in case something goes wrong.
 622          */
 623         console_init();
 624         if (panic_later)
 625                 panic(panic_later, panic_param);
 626
 627         lockdep_info();
 628
 629         /*
 630          * Need to run this when irqs are enabled, because it wants
 631          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 632          * too:
 633          */
 634         locking_selftest();
 635
 636 #ifdef CONFIG_BLK_DEV_INITRD
 637         if (initrd_start && !initrd_below_start_ok &&
 638             page_to_pfn(virt_to_page(initrd_start)) < min_low_pfn) {
 639                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 640                     "disabling it.\n",
 641                     page_to_pfn(virt_to_page(initrd_start)), min_low_pfn);
 642                 initrd_start = 0;
 643         }
 644 #endif
 645         vfs_caches_init_early();
 646         cpuset_init_early();
 647         mem_init();
 648         enable_debug_pagealloc();
 649         cpu_hotplug_init();
 650         kmem_cache_init();
 651         debug_objects_mem_init();
 652         idr_init_cache();
 653         setup_per_cpu_pageset();
 654         numa_policy_init();
 655         if (late_time_init)
 656                 late_time_init();
 657         calibrate_delay();
 658         pidmap_init();
 659         pgtable_cache_init();
 660         prio_tree_init();
 661         anon_vma_init();
 662 #ifdef CONFIG_X86
 663         if (efi_enabled)
 664                 efi_enter_virtual_mode();
 665 #endif
 666         thread_info_cache_init();
 667         fork_init(num_physpages);
 668         proc_caches_init();
 669         buffer_init();
 670         unnamed_dev_init();
 671         key_init();
 672         security_init();
 673         vfs_caches_init(num_physpages);
 674         radix_tree_init();
 675         signals_init();
 676         /* rootfs populating might need page-writeback */
 677         page_writeback_init();
 678 #ifdef CONFIG_PROC_FS
 679         proc_root_init();
 680 #endif
 681         cgroup_init();
 682         cpuset_init();
 683         taskstats_init_early();
 684         delayacct_init();
 685
 686         check_bugs();
 687
 688         acpi_early_init(); /* before LAPIC and SMP init */
 689
 690         /* Do the rest non-__init'ed, we're now alive */
 691         rest_init();
 692 }
 693
 694 static int __initdata initcall_debug;
 695
 696 static int __init initcall_debug_setup(char *str)
 697 {
 698         initcall_debug = 1;
 699         return 1;
 700 }
 701 __setup("initcall_debug", initcall_debug_setup);
 702
 703 static void __init do_one_initcall(initcall_t fn)
 704 {
 705         int count = preempt_count();
 706         ktime_t t0, t1, delta;
 707         char msgbuf[64];
 708         int result;
 709
 710         if (initcall_debug) {
 711                 print_fn_descriptor_symbol("calling  %s\n", fn);
 712                 t0 = ktime_get();
 713         }
 714
 715         result = fn();
 716
 717         if (initcall_debug) {
 718                 t1 = ktime_get();
 719                 delta = ktime_sub(t1, t0);
 720
 721                 print_fn_descriptor_symbol("initcall %s", fn);
 722                 printk(" returned %d after %Ld msecs\n", result,
 723                         (unsigned long long) delta.tv64 >> 20);
 724         }
 725
 726         msgbuf[0] = 0;
 727
 728         if (result && result != -ENODEV && initcall_debug)
 729                 sprintf(msgbuf, "error code %d ", result);
 730
 731         if (preempt_count() != count) {
 732                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 733                 preempt_count() = count;
 734         }
 735         if (irqs_disabled()) {
 736                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 737                 local_irq_enable();
 738         }
 739         if (msgbuf[0]) {
 740                 print_fn_descriptor_symbol(KERN_WARNING "initcall %s", fn);
 741                 printk(" returned with %s\n", msgbuf);
 742         }
 743 }
 744
 745
 746 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
 747
 748 static void __init do_initcalls(void)
 749 {
 750         initcall_t *call;
 751
 752         for (call = __early_initcall_end; call < __initcall_end; call++)
 753                 do_one_initcall(*call);
 754
 755         /* Make sure there is no pending stuff from the initcall sequence */
 756         flush_scheduled_work();
 757 }
 758
 759 /*
 760  * Ok, the machine is now initialized. None of the devices
 761  * have been touched yet, but the CPU subsystem is up and
 762  * running, and memory and process management works.
 763  *
 764  * Now we can finally start doing some real work..
 765  */
 766 static void __init do_basic_setup(void)
 767 {
 768         rcu_init_sched(); /* needed by module_init stage. */
 769         /* drivers will send hotplug events */
 770         init_workqueues();
 771         usermodehelper_init();
 772         driver_init();
 773         init_irq_proc();
 774         do_initcalls();
 775 }
 776
 777 static void __init do_pre_smp_initcalls(void)
 778 {
 779         initcall_t *call;
 780
 781         for (call = __initcall_start; call < __early_initcall_end; call++)
 782                 do_one_initcall(*call);
 783 }
 784
 785 static void run_init_process(char *init_filename)
 786 {
 787         argv_init[0] = init_filename;
 788         kernel_execve(init_filename, argv_init, envp_init);
 789 }
 790
 791 /* This is a non __init function. Force it to be noinline otherwise gcc
 792  * makes it inline to init() and it becomes part of init.text section
 793  */
 794 static int noinline init_post(void)
 795 {
 796         free_initmem();
 797         unlock_kernel();
 798         mark_rodata_ro();
 799         system_state = SYSTEM_RUNNING;
 800         numa_default_policy();
 801
 802         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 803                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 804
 805         (void) sys_dup(0);
 806         (void) sys_dup(0);
 807
 808         current->signal->flags |= SIGNAL_UNKILLABLE;
 809
 810         if (ramdisk_execute_command) {
 811                 run_init_process(ramdisk_execute_command);
 812                 printk(KERN_WARNING "Failed to execute %s\n",
 813                                 ramdisk_execute_command);
 814         }
 815
 816         /*
 817          * We try each of these until one succeeds.
 818          *
 819          * The Bourne shell can be used instead of init if we are
 820          * trying to recover a really broken machine.
 821          */
 822         if (execute_command) {
 823                 run_init_process(execute_command);
 824                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 825                                         "defaults...\n", execute_command);
 826         }
 827         run_init_process("/sbin/init");
 828         run_init_process("/etc/init");
 829         run_init_process("/bin/init");
 830         run_init_process("/bin/sh");
 831
 832         panic("No init found.  Try passing init= option to kernel.");
 833 }
 834
 835 static int __init kernel_init(void * unused)
 836 {
 837         lock_kernel();
 838         /*
 839          * init can run on any cpu.
 840          */
 841         set_cpus_allowed_ptr(current, CPU_MASK_ALL_PTR);
 842         /*
 843          * Tell the world that we're going to be the grim
 844          * reaper of innocent orphaned children.
 845          *
 846          * We don't want people to have to make incorrect
 847          * assumptions about where in the task array this
 848          * can be found.
 849          */
 850         init_pid_ns.child_reaper = current;
 851
 852         cad_pid = task_pid(current);
 853
 854         smp_prepare_cpus(setup_max_cpus);
 855
 856         do_pre_smp_initcalls();
 857
 858         smp_init();
 859         sched_init_smp();
 860
 861         cpuset_init_smp();
 862
 863         do_basic_setup();
 864
 865         /*
 866          * check if there is an early userspace init.  If yes, let it do all
 867          * the work
 868          */
 869
 870         if (!ramdisk_execute_command)
 871                 ramdisk_execute_command = "/init";
 872
 873         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 874                 ramdisk_execute_command = NULL;
 875                 prepare_namespace();
 876         }
 877
 878         /*
 879          * Ok, we have completed the initial bootup, and
 880          * we're essentially up and running. Get rid of the
 881          * initmem segments and start the user-mode stuff..
 882          */
 883         init_post();
 884         return 0;
 885 }