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 pidhash_init(void);
  91 extern void pidmap_init(void);
  92 extern void prio_tree_init(void);
  93 extern void radix_tree_init(void);
  94 extern void free_initmem(void);
  95 #ifdef  CONFIG_ACPI
  96 extern void acpi_early_init(void);
  97 #else
  98 static inline void acpi_early_init(void) { }
  99 #endif
 100 #ifndef CONFIG_DEBUG_RODATA
 101 static inline void mark_rodata_ro(void) { }
 102 #endif
 103
 104 #ifdef CONFIG_TC
 105 extern void tc_init(void);
 106 #endif
 107
 108 enum system_states system_state;
 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 (*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 #ifdef CONFIG_SMP
 133 /* Setup configured maximum number of CPUs to activate */
 134 unsigned int __initdata setup_max_cpus = NR_CPUS;
 135
 136 /*
 137  * Setup routine for controlling SMP activation
 138  *
 139  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 140  * activation entirely (the MPS table probe still happens, though).
 141  *
 142  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 143  * greater than 0, limits the maximum number of CPUs activated in
 144  * SMP mode to <NUM>.
 145  */
 146 #ifndef CONFIG_X86_IO_APIC
 147 static inline void disable_ioapic_setup(void) {};
 148 #endif
 149
 150 static int __init nosmp(char *str)
 151 {
 152         setup_max_cpus = 0;
 153         disable_ioapic_setup();
 154         return 0;
 155 }
 156
 157 early_param("nosmp", nosmp);
 158
 159 static int __init maxcpus(char *str)
 160 {
 161         get_option(&str, &setup_max_cpus);
 162         if (setup_max_cpus == 0)
 163                 disable_ioapic_setup();
 164
 165         return 0;
 166 }
 167
 168 early_param("maxcpus", maxcpus);
 169 #else
 170 #define setup_max_cpus NR_CPUS
 171 #endif
 172
 173 /*
 174  * If set, this is an indication to the drivers that reset the underlying
 175  * device before going ahead with the initialization otherwise driver might
 176  * rely on the BIOS and skip the reset operation.
 177  *
 178  * This is useful if kernel is booting in an unreliable environment.
 179  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 180  * skipped and devices will be in unknown state.
 181  */
 182 unsigned int reset_devices;
 183 EXPORT_SYMBOL(reset_devices);
 184
 185 static int __init set_reset_devices(char *str)
 186 {
 187         reset_devices = 1;
 188         return 1;
 189 }
 190
 191 __setup("reset_devices", set_reset_devices);
 192
 193 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 194 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 195 static const char *panic_later, *panic_param;
 196
 197 extern struct obs_kernel_param __setup_start[], __setup_end[];
 198
 199 static int __init obsolete_checksetup(char *line)
 200 {
 201         struct obs_kernel_param *p;
 202         int had_early_param = 0;
 203
 204         p = __setup_start;
 205         do {
 206                 int n = strlen(p->str);
 207                 if (!strncmp(line, p->str, n)) {
 208                         if (p->early) {
 209                                 /* Already done in parse_early_param?
 210                                  * (Needs exact match on param part).
 211                                  * Keep iterating, as we can have early
 212                                  * params and __setups of same names 8( */
 213                                 if (line[n] == '\0' || line[n] == '=')
 214                                         had_early_param = 1;
 215                         } else if (!p->setup_func) {
 216                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 217                                        " ignored\n", p->str);
 218                                 return 1;
 219                         } else if (p->setup_func(line + n))
 220                                 return 1;
 221                 }
 222                 p++;
 223         } while (p < __setup_end);
 224
 225         return had_early_param;
 226 }
 227
 228 /*
 229  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 230  * still work even if initially too large, it will just take slightly longer
 231  */
 232 unsigned long loops_per_jiffy = (1<<12);
 233
 234 EXPORT_SYMBOL(loops_per_jiffy);
 235
 236 static int __init debug_kernel(char *str)
 237 {
 238         console_loglevel = 10;
 239         return 0;
 240 }
 241
 242 static int __init quiet_kernel(char *str)
 243 {
 244         console_loglevel = 4;
 245         return 0;
 246 }
 247
 248 early_param("debug", debug_kernel);
 249 early_param("quiet", quiet_kernel);
 250
 251 static int __init loglevel(char *str)
 252 {
 253         get_option(&str, &console_loglevel);
 254         return 0;
 255 }
 256
 257 early_param("loglevel", loglevel);
 258
 259 /*
 260  * Unknown boot options get handed to init, unless they look like
 261  * failed parameters
 262  */
 263 static int __init unknown_bootoption(char *param, char *val)
 264 {
 265         /* Change NUL term back to "=", to make "param" the whole string. */
 266         if (val) {
 267                 /* param=val or param="val"? */
 268                 if (val == param+strlen(param)+1)
 269                         val[-1] = '=';
 270                 else if (val == param+strlen(param)+2) {
 271                         val[-2] = '=';
 272                         memmove(val-1, val, strlen(val)+1);
 273                         val--;
 274                 } else
 275                         BUG();
 276         }
 277
 278         /* Handle obsolete-style parameters */
 279         if (obsolete_checksetup(param))
 280                 return 0;
 281
 282         /*
 283          * Preemptive maintenance for "why didn't my misspelled command
 284          * line work?"
 285          */
 286         if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
 287                 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
 288                 return 0;
 289         }
 290
 291         if (panic_later)
 292                 return 0;
 293
 294         if (val) {
 295                 /* Environment option */
 296                 unsigned int i;
 297                 for (i = 0; envp_init[i]; i++) {
 298                         if (i == MAX_INIT_ENVS) {
 299                                 panic_later = "Too many boot env vars at `%s'";
 300                                 panic_param = param;
 301                         }
 302                         if (!strncmp(param, envp_init[i], val - param))
 303                                 break;
 304                 }
 305                 envp_init[i] = param;
 306         } else {
 307                 /* Command line option */
 308                 unsigned int i;
 309                 for (i = 0; argv_init[i]; i++) {
 310                         if (i == MAX_INIT_ARGS) {
 311                                 panic_later = "Too many boot init vars at `%s'";
 312                                 panic_param = param;
 313                         }
 314                 }
 315                 argv_init[i] = param;
 316         }
 317         return 0;
 318 }
 319
 320 #ifdef CONFIG_DEBUG_PAGEALLOC
 321 int __read_mostly debug_pagealloc_enabled = 0;
 322 #endif
 323
 324 static int __init init_setup(char *str)
 325 {
 326         unsigned int i;
 327
 328         execute_command = str;
 329         /*
 330          * In case LILO is going to boot us with default command line,
 331          * it prepends "auto" before the whole cmdline which makes
 332          * the shell think it should execute a script with such name.
 333          * So we ignore all arguments entered _before_ init=... [MJ]
 334          */
 335         for (i = 1; i < MAX_INIT_ARGS; i++)
 336                 argv_init[i] = NULL;
 337         return 1;
 338 }
 339 __setup("init=", init_setup);
 340
 341 static int __init rdinit_setup(char *str)
 342 {
 343         unsigned int i;
 344
 345         ramdisk_execute_command = str;
 346         /* See "auto" comment in init_setup */
 347         for (i = 1; i < MAX_INIT_ARGS; i++)
 348                 argv_init[i] = NULL;
 349         return 1;
 350 }
 351 __setup("rdinit=", rdinit_setup);
 352
 353 #ifndef CONFIG_SMP
 354
 355 #ifdef CONFIG_X86_LOCAL_APIC
 356 static void __init smp_init(void)
 357 {
 358         APIC_init_uniprocessor();
 359 }
 360 #else
 361 #define smp_init()      do { } while (0)
 362 #endif
 363
 364 static inline void setup_per_cpu_areas(void) { }
 365 static inline void setup_nr_cpu_ids(void) { }
 366 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 367
 368 #else
 369
 370 #if NR_CPUS > BITS_PER_LONG
 371 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
 372 EXPORT_SYMBOL(cpu_mask_all);
 373 #endif
 374
 375 /* Setup number of possible processor ids */
 376 int nr_cpu_ids __read_mostly = NR_CPUS;
 377 EXPORT_SYMBOL(nr_cpu_ids);
 378
 379 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
 380 static void __init setup_nr_cpu_ids(void)
 381 {
 382         int cpu, highest_cpu = 0;
 383
 384         for_each_possible_cpu(cpu)
 385                 highest_cpu = cpu;
 386
 387         nr_cpu_ids = highest_cpu + 1;
 388 }
 389
 390 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
 391 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
 392
 393 EXPORT_SYMBOL(__per_cpu_offset);
 394
 395 static void __init setup_per_cpu_areas(void)
 396 {
 397         unsigned long size, i;
 398         char *ptr;
 399         unsigned long nr_possible_cpus = num_possible_cpus();
 400
 401         /* Copy section for each CPU (we discard the original) */
 402         size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
 403         ptr = alloc_bootmem_pages(size * nr_possible_cpus);
 404
 405         for_each_possible_cpu(i) {
 406                 __per_cpu_offset[i] = ptr - __per_cpu_start;
 407                 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 408                 ptr += size;
 409         }
 410 }
 411 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
 412
 413 /* Called by boot processor to activate the rest. */
 414 static void __init smp_init(void)
 415 {
 416         unsigned int cpu;
 417
 418         /*
 419          * Set up the current CPU as possible to migrate to.
 420          * The other ones will be done by cpu_up/cpu_down()
 421          */
 422         cpu = smp_processor_id();
 423         cpu_set(cpu, cpu_active_map);
 424
 425         /* FIXME: This should be done in userspace --RR */
 426         for_each_present_cpu(cpu) {
 427                 if (num_online_cpus() >= setup_max_cpus)
 428                         break;
 429                 if (!cpu_online(cpu))
 430                         cpu_up(cpu);
 431         }
 432
 433         /* Any cleanup work */
 434         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 435         smp_cpus_done(setup_max_cpus);
 436 }
 437
 438 #endif
 439
 440 /*
 441  * We need to store the untouched command line for future reference.
 442  * We also need to store the touched command line since the parameter
 443  * parsing is performed in place, and we should allow a component to
 444  * store reference of name/value for future reference.
 445  */
 446 static void __init setup_command_line(char *command_line)
 447 {
 448         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 449         static_command_line = alloc_bootmem(strlen (command_line)+1);
 450         strcpy (saved_command_line, boot_command_line);
 451         strcpy (static_command_line, command_line);
 452 }
 453
 454 /*
 455  * We need to finalize in a non-__init function or else race conditions
 456  * between the root thread and the init thread may cause start_kernel to
 457  * be reaped by free_initmem before the root thread has proceeded to
 458  * cpu_idle.
 459  *
 460  * gcc-3.4 accidentally inlines this function, so use noinline.
 461  */
 462
 463 static void noinline __init_refok rest_init(void)
 464         __releases(kernel_lock)
 465 {
 466         int pid;
 467
 468         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 469         numa_default_policy();
 470         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 471         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 472         unlock_kernel();
 473
 474         /*
 475          * The boot idle thread must execute schedule()
 476          * at least once to get things moving:
 477          */
 478         init_idle_bootup_task(current);
 479         preempt_enable_no_resched();
 480         schedule();
 481         preempt_disable();
 482
 483         /* Call into cpu_idle with preempt disabled */
 484         cpu_idle();
 485 }
 486
 487 /* Check for early params. */
 488 static int __init do_early_param(char *param, char *val)
 489 {
 490         struct obs_kernel_param *p;
 491
 492         for (p = __setup_start; p < __setup_end; p++) {
 493                 if ((p->early && strcmp(param, p->str) == 0) ||
 494                     (strcmp(param, "console") == 0 &&
 495                      strcmp(p->str, "earlycon") == 0)
 496                 ) {
 497                         if (p->setup_func(val) != 0)
 498                                 printk(KERN_WARNING
 499                                        "Malformed early option '%s'\n", param);
 500                 }
 501         }
 502         /* We accept everything at this stage. */
 503         return 0;
 504 }
 505
 506 /* Arch code calls this early on, or if not, just before other parsing. */
 507 void __init parse_early_param(void)
 508 {
 509         static __initdata int done = 0;
 510         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 511
 512         if (done)
 513                 return;
 514
 515         /* All fall through to do_early_param. */
 516         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 517         parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
 518         done = 1;
 519 }
 520
 521 /*
 522  *      Activate the first processor.
 523  */
 524
 525 static void __init boot_cpu_init(void)
 526 {
 527         int cpu = smp_processor_id();
 528         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 529         cpu_set(cpu, cpu_online_map);
 530         cpu_set(cpu, cpu_present_map);
 531         cpu_set(cpu, cpu_possible_map);
 532 }
 533
 534 void __init __weak smp_setup_processor_id(void)
 535 {
 536 }
 537
 538 void __init __weak thread_info_cache_init(void)
 539 {
 540 }
 541
 542 asmlinkage void __init start_kernel(void)
 543 {
 544         char * command_line;
 545         extern struct kernel_param __start___param[], __stop___param[];
 546
 547         smp_setup_processor_id();
 548
 549         /*
 550          * Need to run as early as possible, to initialize the
 551          * lockdep hash:
 552          */
 553         unwind_init();
 554         lockdep_init();
 555         debug_objects_early_init();
 556         cgroup_init_early();
 557
 558         local_irq_disable();
 559         early_boot_irqs_off();
 560         early_init_irq_lock_class();
 561
 562 /*
 563  * Interrupts are still disabled. Do necessary setups, then
 564  * enable them
 565  */
 566         lock_kernel();
 567         tick_init();
 568         boot_cpu_init();
 569         page_address_init();
 570         printk(KERN_NOTICE);
 571         printk(linux_banner);
 572         setup_arch(&command_line);
 573         mm_init_owner(&init_mm, &init_task);
 574         setup_command_line(command_line);
 575         unwind_setup();
 576         setup_per_cpu_areas();
 577         setup_nr_cpu_ids();
 578         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 579
 580         /*
 581          * Set up the scheduler prior starting any interrupts (such as the
 582          * timer interrupt). Full topology setup happens at smp_init()
 583          * time - but meanwhile we still have a functioning scheduler.
 584          */
 585         sched_init();
 586         /*
 587          * Disable preemption - early bootup scheduling is extremely
 588          * fragile until we cpu_idle() for the first time.
 589          */
 590         preempt_disable();
 591         build_all_zonelists();
 592         page_alloc_init();
 593         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 594         parse_early_param();
 595         parse_args("Booting kernel", static_command_line, __start___param,
 596                    __stop___param - __start___param,
 597                    &unknown_bootoption);
 598         if (!irqs_disabled()) {
 599                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 600                                 "enabled *very* early, fixing it\n");
 601                 local_irq_disable();
 602         }
 603         sort_main_extable();
 604         trap_init();
 605         rcu_init();
 606         init_IRQ();
 607         pidhash_init();
 608         init_timers();
 609         hrtimers_init();
 610         softirq_init();
 611         timekeeping_init();
 612         time_init();
 613         sched_clock_init();
 614         profile_init();
 615         if (!irqs_disabled())
 616                 printk("start_kernel(): bug: interrupts were enabled early\n");
 617         early_boot_irqs_on();
 618         local_irq_enable();
 619
 620         /*
 621          * HACK ALERT! This is early. We're enabling the console before
 622          * we've done PCI setups etc, and console_init() must be aware of
 623          * this. But we do want output early, in case something goes wrong.
 624          */
 625         console_init();
 626         if (panic_later)
 627                 panic(panic_later, panic_param);
 628
 629         lockdep_info();
 630
 631         /*
 632          * Need to run this when irqs are enabled, because it wants
 633          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 634          * too:
 635          */
 636         locking_selftest();
 637
 638 #ifdef CONFIG_BLK_DEV_INITRD
 639         if (initrd_start && !initrd_below_start_ok &&
 640             page_to_pfn(virt_to_page(initrd_start)) < min_low_pfn) {
 641                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 642                     "disabling it.\n",
 643                     page_to_pfn(virt_to_page(initrd_start)), min_low_pfn);
 644                 initrd_start = 0;
 645         }
 646 #endif
 647         vfs_caches_init_early();
 648         cpuset_init_early();
 649         mem_init();
 650         enable_debug_pagealloc();
 651         cpu_hotplug_init();
 652         kmem_cache_init();
 653         debug_objects_mem_init();
 654         idr_init_cache();
 655         setup_per_cpu_pageset();
 656         numa_policy_init();
 657         if (late_time_init)
 658                 late_time_init();
 659         calibrate_delay();
 660         pidmap_init();
 661         pgtable_cache_init();
 662         prio_tree_init();
 663         anon_vma_init();
 664 #ifdef CONFIG_X86
 665         if (efi_enabled)
 666                 efi_enter_virtual_mode();
 667 #endif
 668         thread_info_cache_init();
 669         fork_init(num_physpages);
 670         proc_caches_init();
 671         buffer_init();
 672         unnamed_dev_init();
 673         key_init();
 674         security_init();
 675         vfs_caches_init(num_physpages);
 676         radix_tree_init();
 677         signals_init();
 678         /* rootfs populating might need page-writeback */
 679         page_writeback_init();
 680 #ifdef CONFIG_PROC_FS
 681         proc_root_init();
 682 #endif
 683         cgroup_init();
 684         cpuset_init();
 685         taskstats_init_early();
 686         delayacct_init();
 687
 688         check_bugs();
 689
 690         acpi_early_init(); /* before LAPIC and SMP init */
 691
 692         /* Do the rest non-__init'ed, we're now alive */
 693         rest_init();
 694 }
 695
 696 static int __initdata initcall_debug;
 697
 698 static int __init initcall_debug_setup(char *str)
 699 {
 700         initcall_debug = 1;
 701         return 1;
 702 }
 703 __setup("initcall_debug", initcall_debug_setup);
 704
 705 static void __init do_one_initcall(initcall_t fn)
 706 {
 707         int count = preempt_count();
 708         ktime_t t0, t1, delta;
 709         char msgbuf[64];
 710         int result;
 711
 712         if (initcall_debug) {
 713                 print_fn_descriptor_symbol("calling  %s\n", fn);
 714                 t0 = ktime_get();
 715         }
 716
 717         result = fn();
 718
 719         if (initcall_debug) {
 720                 t1 = ktime_get();
 721                 delta = ktime_sub(t1, t0);
 722
 723                 print_fn_descriptor_symbol("initcall %s", fn);
 724                 printk(" returned %d after %Ld msecs\n", result,
 725                         (unsigned long long) delta.tv64 >> 20);
 726         }
 727
 728         msgbuf[0] = 0;
 729
 730         if (result && result != -ENODEV && initcall_debug)
 731                 sprintf(msgbuf, "error code %d ", result);
 732
 733         if (preempt_count() != count) {
 734                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 735                 preempt_count() = count;
 736         }
 737         if (irqs_disabled()) {
 738                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 739                 local_irq_enable();
 740         }
 741         if (msgbuf[0]) {
 742                 print_fn_descriptor_symbol(KERN_WARNING "initcall %s", fn);
 743                 printk(" returned with %s\n", msgbuf);
 744         }
 745 }
 746
 747
 748 extern initcall_t __initcall_start[], __initcall_end[];
 749
 750 static void __init do_initcalls(void)
 751 {
 752         initcall_t *call;
 753
 754         for (call = __initcall_start; call < __initcall_end; call++)
 755                 do_one_initcall(*call);
 756
 757         /* Make sure there is no pending stuff from the initcall sequence */
 758         flush_scheduled_work();
 759 }
 760
 761 /*
 762  * Ok, the machine is now initialized. None of the devices
 763  * have been touched yet, but the CPU subsystem is up and
 764  * running, and memory and process management works.
 765  *
 766  * Now we can finally start doing some real work..
 767  */
 768 static void __init do_basic_setup(void)
 769 {
 770         rcu_init_sched(); /* needed by module_init stage. */
 771         /* drivers will send hotplug events */
 772         init_workqueues();
 773         usermodehelper_init();
 774         driver_init();
 775         init_irq_proc();
 776         do_initcalls();
 777 }
 778
 779 static int __initdata nosoftlockup;
 780
 781 static int __init nosoftlockup_setup(char *str)
 782 {
 783         nosoftlockup = 1;
 784         return 1;
 785 }
 786 __setup("nosoftlockup", nosoftlockup_setup);
 787
 788 static void __init do_pre_smp_initcalls(void)
 789 {
 790         extern int spawn_ksoftirqd(void);
 791
 792         init_call_single_data();
 793         migration_init();
 794         spawn_ksoftirqd();
 795         if (!nosoftlockup)
 796                 spawn_softlockup_task();
 797 }
 798
 799 static void run_init_process(char *init_filename)
 800 {
 801         argv_init[0] = init_filename;
 802         kernel_execve(init_filename, argv_init, envp_init);
 803 }
 804
 805 /* This is a non __init function. Force it to be noinline otherwise gcc
 806  * makes it inline to init() and it becomes part of init.text section
 807  */
 808 static int noinline init_post(void)
 809 {
 810         free_initmem();
 811         unlock_kernel();
 812         mark_rodata_ro();
 813         system_state = SYSTEM_RUNNING;
 814         numa_default_policy();
 815
 816         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 817                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 818
 819         (void) sys_dup(0);
 820         (void) sys_dup(0);
 821
 822         current->signal->flags |= SIGNAL_UNKILLABLE;
 823
 824         if (ramdisk_execute_command) {
 825                 run_init_process(ramdisk_execute_command);
 826                 printk(KERN_WARNING "Failed to execute %s\n",
 827                                 ramdisk_execute_command);
 828         }
 829
 830         /*
 831          * We try each of these until one succeeds.
 832          *
 833          * The Bourne shell can be used instead of init if we are
 834          * trying to recover a really broken machine.
 835          */
 836         if (execute_command) {
 837                 run_init_process(execute_command);
 838                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 839                                         "defaults...\n", execute_command);
 840         }
 841         run_init_process("/sbin/init");
 842         run_init_process("/etc/init");
 843         run_init_process("/bin/init");
 844         run_init_process("/bin/sh");
 845
 846         panic("No init found.  Try passing init= option to kernel.");
 847 }
 848
 849 static int __init kernel_init(void * unused)
 850 {
 851         lock_kernel();
 852         /*
 853          * init can run on any cpu.
 854          */
 855         set_cpus_allowed_ptr(current, CPU_MASK_ALL_PTR);
 856         /*
 857          * Tell the world that we're going to be the grim
 858          * reaper of innocent orphaned children.
 859          *
 860          * We don't want people to have to make incorrect
 861          * assumptions about where in the task array this
 862          * can be found.
 863          */
 864         init_pid_ns.child_reaper = current;
 865
 866         cad_pid = task_pid(current);
 867
 868         smp_prepare_cpus(setup_max_cpus);
 869
 870         do_pre_smp_initcalls();
 871
 872         smp_init();
 873         sched_init_smp();
 874
 875         cpuset_init_smp();
 876
 877         do_basic_setup();
 878
 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         init_post();
 898         return 0;
 899 }