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 #ifdef CONFIG_SMP
 133 /* Setup configured maximum number of CPUs to activate */
 134 unsigned int setup_max_cpus = NR_CPUS;
 135 EXPORT_SYMBOL(setup_max_cpus);
 136
 137
 138 /*
 139  * Setup routine for controlling SMP activation
 140  *
 141  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 142  * activation entirely (the MPS table probe still happens, though).
 143  *
 144  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 145  * greater than 0, limits the maximum number of CPUs activated in
 146  * SMP mode to <NUM>.
 147  */
 148
 149 void __weak arch_disable_smp_support(void) { }
 150
 151 static int __init nosmp(char *str)
 152 {
 153         setup_max_cpus = 0;
 154         arch_disable_smp_support();
 155
 156         return 0;
 157 }
 158
 159 early_param("nosmp", nosmp);
 160
 161 /* this is hard limit */
 162 static int __init nrcpus(char *str)
 163 {
 164         int nr_cpus;
 165
 166         get_option(&str, &nr_cpus);
 167         if (nr_cpus > 0 && nr_cpus < nr_cpu_ids)
 168                 nr_cpu_ids = nr_cpus;
 169
 170         return 0;
 171 }
 172
 173 early_param("nr_cpus", nrcpus);
 174
 175 static int __init maxcpus(char *str)
 176 {
 177         get_option(&str, &setup_max_cpus);
 178         if (setup_max_cpus == 0)
 179                 arch_disable_smp_support();
 180
 181         return 0;
 182 }
 183
 184 early_param("maxcpus", maxcpus);
 185 #else
 186 static const unsigned int setup_max_cpus = NR_CPUS;
 187 #endif
 188
 189 /*
 190  * If set, this is an indication to the drivers that reset the underlying
 191  * device before going ahead with the initialization otherwise driver might
 192  * rely on the BIOS and skip the reset operation.
 193  *
 194  * This is useful if kernel is booting in an unreliable environment.
 195  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 196  * skipped and devices will be in unknown state.
 197  */
 198 unsigned int reset_devices;
 199 EXPORT_SYMBOL(reset_devices);
 200
 201 static int __init set_reset_devices(char *str)
 202 {
 203         reset_devices = 1;
 204         return 1;
 205 }
 206
 207 __setup("reset_devices", set_reset_devices);
 208
 209 static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 210 const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 211 static const char *panic_later, *panic_param;
 212
 213 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 214
 215 static int __init obsolete_checksetup(char *line)
 216 {
 217         const struct obs_kernel_param *p;
 218         int had_early_param = 0;
 219
 220         p = __setup_start;
 221         do {
 222                 int n = strlen(p->str);
 223                 if (!strncmp(line, p->str, n)) {
 224                         if (p->early) {
 225                                 /* Already done in parse_early_param?
 226                                  * (Needs exact match on param part).
 227                                  * Keep iterating, as we can have early
 228                                  * params and __setups of same names 8( */
 229                                 if (line[n] == '\0' || line[n] == '=')
 230                                         had_early_param = 1;
 231                         } else if (!p->setup_func) {
 232                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 233                                        " ignored\n", p->str);
 234                                 return 1;
 235                         } else if (p->setup_func(line + n))
 236                                 return 1;
 237                 }
 238                 p++;
 239         } while (p < __setup_end);
 240
 241         return had_early_param;
 242 }
 243
 244 /*
 245  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 246  * still work even if initially too large, it will just take slightly longer
 247  */
 248 unsigned long loops_per_jiffy = (1<<12);
 249
 250 EXPORT_SYMBOL(loops_per_jiffy);
 251
 252 static int __init debug_kernel(char *str)
 253 {
 254         console_loglevel = 10;
 255         return 0;
 256 }
 257
 258 static int __init quiet_kernel(char *str)
 259 {
 260         console_loglevel = 4;
 261         return 0;
 262 }
 263
 264 early_param("debug", debug_kernel);
 265 early_param("quiet", quiet_kernel);
 266
 267 static int __init loglevel(char *str)
 268 {
 269         get_option(&str, &console_loglevel);
 270         return 0;
 271 }
 272
 273 early_param("loglevel", loglevel);
 274
 275 /*
 276  * Unknown boot options get handed to init, unless they look like
 277  * unused parameters (modprobe will find them in /proc/cmdline).
 278  */
 279 static int __init unknown_bootoption(char *param, char *val)
 280 {
 281         /* Change NUL term back to "=", to make "param" the whole string. */
 282         if (val) {
 283                 /* param=val or param="val"? */
 284                 if (val == param+strlen(param)+1)
 285                         val[-1] = '=';
 286                 else if (val == param+strlen(param)+2) {
 287                         val[-2] = '=';
 288                         memmove(val-1, val, strlen(val)+1);
 289                         val--;
 290                 } else
 291                         BUG();
 292         }
 293
 294         /* Handle obsolete-style parameters */
 295         if (obsolete_checksetup(param))
 296                 return 0;
 297
 298         /* Unused module parameter. */
 299         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 300                 return 0;
 301
 302         if (panic_later)
 303                 return 0;
 304
 305         if (val) {
 306                 /* Environment option */
 307                 unsigned int i;
 308                 for (i = 0; envp_init[i]; i++) {
 309                         if (i == MAX_INIT_ENVS) {
 310                                 panic_later = "Too many boot env vars at `%s'";
 311                                 panic_param = param;
 312                         }
 313                         if (!strncmp(param, envp_init[i], val - param))
 314                                 break;
 315                 }
 316                 envp_init[i] = param;
 317         } else {
 318                 /* Command line option */
 319                 unsigned int i;
 320                 for (i = 0; argv_init[i]; i++) {
 321                         if (i == MAX_INIT_ARGS) {
 322                                 panic_later = "Too many boot init vars at `%s'";
 323                                 panic_param = param;
 324                         }
 325                 }
 326                 argv_init[i] = param;
 327         }
 328         return 0;
 329 }
 330
 331 #ifdef CONFIG_DEBUG_PAGEALLOC
 332 int __read_mostly debug_pagealloc_enabled = 0;
 333 #endif
 334
 335 static int __init init_setup(char *str)
 336 {
 337         unsigned int i;
 338
 339         execute_command = str;
 340         /*
 341          * In case LILO is going to boot us with default command line,
 342          * it prepends "auto" before the whole cmdline which makes
 343          * the shell think it should execute a script with such name.
 344          * So we ignore all arguments entered _before_ init=... [MJ]
 345          */
 346         for (i = 1; i < MAX_INIT_ARGS; i++)
 347                 argv_init[i] = NULL;
 348         return 1;
 349 }
 350 __setup("init=", init_setup);
 351
 352 static int __init rdinit_setup(char *str)
 353 {
 354         unsigned int i;
 355
 356         ramdisk_execute_command = str;
 357         /* See "auto" comment in init_setup */
 358         for (i = 1; i < MAX_INIT_ARGS; i++)
 359                 argv_init[i] = NULL;
 360         return 1;
 361 }
 362 __setup("rdinit=", rdinit_setup);
 363
 364 #ifndef CONFIG_SMP
 365
 366 #ifdef CONFIG_X86_LOCAL_APIC
 367 static void __init smp_init(void)
 368 {
 369         APIC_init_uniprocessor();
 370 }
 371 #else
 372 #define smp_init()      do { } while (0)
 373 #endif
 374
 375 static inline void setup_nr_cpu_ids(void) { }
 376 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 377
 378 #else
 379
 380 /* Setup number of possible processor ids */
 381 int nr_cpu_ids __read_mostly = NR_CPUS;
 382 EXPORT_SYMBOL(nr_cpu_ids);
 383
 384 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
 385 static void __init setup_nr_cpu_ids(void)
 386 {
 387         nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
 388 }
 389
 390 /* Called by boot processor to activate the rest. */
 391 static void __init smp_init(void)
 392 {
 393         unsigned int cpu;
 394
 395         /* FIXME: This should be done in userspace --RR */
 396         for_each_present_cpu(cpu) {
 397                 if (num_online_cpus() >= setup_max_cpus)
 398                         break;
 399                 if (!cpu_online(cpu))
 400                         cpu_up(cpu);
 401         }
 402
 403         /* Any cleanup work */
 404         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 405         smp_cpus_done(setup_max_cpus);
 406 }
 407
 408 #endif
 409
 410 /*
 411  * We need to store the untouched command line for future reference.
 412  * We also need to store the touched command line since the parameter
 413  * parsing is performed in place, and we should allow a component to
 414  * store reference of name/value for future reference.
 415  */
 416 static void __init setup_command_line(char *command_line)
 417 {
 418         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 419         static_command_line = alloc_bootmem(strlen (command_line)+1);
 420         strcpy (saved_command_line, boot_command_line);
 421         strcpy (static_command_line, command_line);
 422 }
 423
 424 /*
 425  * We need to finalize in a non-__init function or else race conditions
 426  * between the root thread and the init thread may cause start_kernel to
 427  * be reaped by free_initmem before the root thread has proceeded to
 428  * cpu_idle.
 429  *
 430  * gcc-3.4 accidentally inlines this function, so use noinline.
 431  */
 432
 433 static __initdata DECLARE_COMPLETION(kthreadd_done);
 434
 435 static noinline void __init_refok rest_init(void)
 436 {
 437         int pid;
 438
 439         rcu_scheduler_starting();
 440         /*
 441          * We need to spawn init first so that it obtains pid 1, however
 442          * the init task will end up wanting to create kthreads, which, if
 443          * we schedule it before we create kthreadd, will OOPS.
 444          */
 445         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 446         numa_default_policy();
 447         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 448         rcu_read_lock();
 449         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 450         rcu_read_unlock();
 451         complete(&kthreadd_done);
 452
 453         /*
 454          * The boot idle thread must execute schedule()
 455          * at least once to get things moving:
 456          */
 457         init_idle_bootup_task(current);
 458         preempt_enable_no_resched();
 459         schedule();
 460         preempt_disable();
 461
 462         /* Call into cpu_idle with preempt disabled */
 463         cpu_idle();
 464 }
 465
 466 /* Check for early params. */
 467 static int __init do_early_param(char *param, char *val)
 468 {
 469         const struct obs_kernel_param *p;
 470
 471         for (p = __setup_start; p < __setup_end; p++) {
 472                 if ((p->early && strcmp(param, p->str) == 0) ||
 473                     (strcmp(param, "console") == 0 &&
 474                      strcmp(p->str, "earlycon") == 0)
 475                 ) {
 476                         if (p->setup_func(val) != 0)
 477                                 printk(KERN_WARNING
 478                                        "Malformed early option '%s'\n", param);
 479                 }
 480         }
 481         /* We accept everything at this stage. */
 482         return 0;
 483 }
 484
 485 void __init parse_early_options(char *cmdline)
 486 {
 487         parse_args("early options", cmdline, NULL, 0, do_early_param);
 488 }
 489
 490 /* Arch code calls this early on, or if not, just before other parsing. */
 491 void __init parse_early_param(void)
 492 {
 493         static __initdata int done = 0;
 494         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 495
 496         if (done)
 497                 return;
 498
 499         /* All fall through to do_early_param. */
 500         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 501         parse_early_options(tmp_cmdline);
 502         done = 1;
 503 }
 504
 505 /*
 506  *      Activate the first processor.
 507  */
 508
 509 static void __init boot_cpu_init(void)
 510 {
 511         int cpu = smp_processor_id();
 512         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 513         set_cpu_online(cpu, true);
 514         set_cpu_active(cpu, true);
 515         set_cpu_present(cpu, true);
 516         set_cpu_possible(cpu, true);
 517 }
 518
 519 void __init __weak smp_setup_processor_id(void)
 520 {
 521 }
 522
 523 void __init __weak thread_info_cache_init(void)
 524 {
 525 }
 526
 527 /*
 528  * Set up kernel memory allocators
 529  */
 530 static void __init mm_init(void)
 531 {
 532         /*
 533          * page_cgroup requires countinous pages as memmap
 534          * and it's bigger than MAX_ORDER unless SPARSEMEM.
 535          */
 536         page_cgroup_init_flatmem();
 537         mem_init();
 538         kmem_cache_init();
 539         percpu_init_late();
 540         pgtable_cache_init();
 541         vmalloc_init();
 542 }
 543
 544 asmlinkage void __init start_kernel(void)
 545 {
 546         char * command_line;
 547         extern const struct kernel_param __start___param[], __stop___param[];
 548
 549         smp_setup_processor_id();
 550
 551         /*
 552          * Need to run as early as possible, to initialize the
 553          * lockdep hash:
 554          */
 555         lockdep_init();
 556         debug_objects_early_init();
 557
 558         /*
 559          * Set up the the initial canary ASAP:
 560          */
 561         boot_init_stack_canary();
 562
 563         cgroup_init_early();
 564
 565         local_irq_disable();
 566         early_boot_irqs_disabled = true;
 567
 568 /*
 569  * Interrupts are still disabled. Do necessary setups, then
 570  * enable them
 571  */
 572         tick_init();
 573         boot_cpu_init();
 574         page_address_init();
 575         printk(KERN_NOTICE "%s", linux_banner);
 576         setup_arch(&command_line);
 577         mm_init_owner(&init_mm, &init_task);
 578         setup_command_line(command_line);
 579         setup_nr_cpu_ids();
 580         setup_per_cpu_areas();
 581         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 582
 583         build_all_zonelists(NULL);
 584         page_alloc_init();
 585
 586         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 587         parse_early_param();
 588         parse_args("Booting kernel", static_command_line, __start___param,
 589                    __stop___param - __start___param,
 590                    &unknown_bootoption);
 591         /*
 592          * These use large bootmem allocations and must precede
 593          * kmem_cache_init()
 594          */
 595         pidhash_init();
 596         vfs_caches_init_early();
 597         sort_main_extable();
 598         trap_init();
 599         mm_init();
 600         /*
 601          * Set up the scheduler prior starting any interrupts (such as the
 602          * timer interrupt). Full topology setup happens at smp_init()
 603          * time - but meanwhile we still have a functioning scheduler.
 604          */
 605         sched_init();
 606         /*
 607          * Disable preemption - early bootup scheduling is extremely
 608          * fragile until we cpu_idle() for the first time.
 609          */
 610         preempt_disable();
 611         if (!irqs_disabled()) {
 612                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 613                                 "enabled *very* early, fixing it\n");
 614                 local_irq_disable();
 615         }
 616         idr_init_cache();
 617         perf_event_init();
 618         rcu_init();
 619         radix_tree_init();
 620         /* init some links before init_ISA_irqs() */
 621         early_irq_init();
 622         init_IRQ();
 623         prio_tree_init();
 624         init_timers();
 625         hrtimers_init();
 626         softirq_init();
 627         timekeeping_init();
 628         time_init();
 629         profile_init();
 630         if (!irqs_disabled())
 631                 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
 632                                  "enabled early\n");
 633         early_boot_irqs_disabled = false;
 634         local_irq_enable();
 635
 636         /* Interrupts are enabled now so all GFP allocations are safe. */
 637         gfp_allowed_mask = __GFP_BITS_MASK;
 638
 639         kmem_cache_init_late();
 640
 641         /*
 642          * HACK ALERT! This is early. We're enabling the console before
 643          * we've done PCI setups etc, and console_init() must be aware of
 644          * this. But we do want output early, in case something goes wrong.
 645          */
 646         console_init();
 647         if (panic_later)
 648                 panic(panic_later, panic_param);
 649
 650         lockdep_info();
 651
 652         /*
 653          * Need to run this when irqs are enabled, because it wants
 654          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 655          * too:
 656          */
 657         locking_selftest();
 658
 659 #ifdef CONFIG_BLK_DEV_INITRD
 660         if (initrd_start && !initrd_below_start_ok &&
 661             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 662                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 663                     "disabling it.\n",
 664                     page_to_pfn(virt_to_page((void *)initrd_start)),
 665                     min_low_pfn);
 666                 initrd_start = 0;
 667         }
 668 #endif
 669         page_cgroup_init();
 670         enable_debug_pagealloc();
 671         debug_objects_mem_init();
 672         kmemleak_init();
 673         setup_per_cpu_pageset();
 674         numa_policy_init();
 675         if (late_time_init)
 676                 late_time_init();
 677         sched_clock_init();
 678         calibrate_delay();
 679         pidmap_init();
 680         anon_vma_init();
 681 #ifdef CONFIG_X86
 682         if (efi_enabled)
 683                 efi_enter_virtual_mode();
 684 #endif
 685         thread_info_cache_init();
 686         cred_init();
 687         fork_init(totalram_pages);
 688         proc_caches_init();
 689         buffer_init();
 690         key_init();
 691         security_init();
 692         dbg_late_init();
 693         vfs_caches_init(totalram_pages);
 694         signals_init();
 695         /* rootfs populating might need page-writeback */
 696         page_writeback_init();
 697 #ifdef CONFIG_PROC_FS
 698         proc_root_init();
 699 #endif
 700         cgroup_init();
 701         cpuset_init();
 702         taskstats_init_early();
 703         delayacct_init();
 704
 705         check_bugs();
 706
 707         acpi_early_init(); /* before LAPIC and SMP init */
 708         sfi_init_late();
 709
 710         ftrace_init();
 711
 712         /* Do the rest non-__init'ed, we're now alive */
 713         rest_init();
 714 }
 715
 716 /* Call all constructor functions linked into the kernel. */
 717 static void __init do_ctors(void)
 718 {
 719 #ifdef CONFIG_CONSTRUCTORS
 720         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 721
 722         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 723                 (*fn)();
 724 #endif
 725 }
 726
 727 int initcall_debug;
 728 core_param(initcall_debug, initcall_debug, bool, 0644);
 729
 730 static char msgbuf[64];
 731
 732 static int __init_or_module do_one_initcall_debug(initcall_t fn)
 733 {
 734         ktime_t calltime, delta, rettime;
 735         unsigned long long duration;
 736         int ret;
 737
 738         printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
 739         calltime = ktime_get();
 740         ret = fn();
 741         rettime = ktime_get();
 742         delta = ktime_sub(rettime, calltime);
 743         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 744         printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn,
 745                 ret, duration);
 746
 747         return ret;
 748 }
 749
 750 int __init_or_module do_one_initcall(initcall_t fn)
 751 {
 752         int count = preempt_count();
 753         int ret;
 754
 755         if (initcall_debug)
 756                 ret = do_one_initcall_debug(fn);
 757         else
 758                 ret = fn();
 759
 760         msgbuf[0] = 0;
 761
 762         if (ret && ret != -ENODEV && initcall_debug)
 763                 sprintf(msgbuf, "error code %d ", ret);
 764
 765         if (preempt_count() != count) {
 766                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 767                 preempt_count() = count;
 768         }
 769         if (irqs_disabled()) {
 770                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 771                 local_irq_enable();
 772         }
 773         if (msgbuf[0]) {
 774                 printk("initcall %pF returned with %s\n", fn, msgbuf);
 775         }
 776
 777         return ret;
 778 }
 779
 780
 781 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
 782
 783 static void __init do_initcalls(void)
 784 {
 785         initcall_t *fn;
 786
 787         for (fn = __early_initcall_end; fn < __initcall_end; fn++)
 788                 do_one_initcall(*fn);
 789 }
 790
 791 /*
 792  * Ok, the machine is now initialized. None of the devices
 793  * have been touched yet, but the CPU subsystem is up and
 794  * running, and memory and process management works.
 795  *
 796  * Now we can finally start doing some real work..
 797  */
 798 static void __init do_basic_setup(void)
 799 {
 800         cpuset_init_smp();
 801         usermodehelper_init();
 802         init_tmpfs();
 803         driver_init();
 804         init_irq_proc();
 805         do_ctors();
 806         do_initcalls();
 807 }
 808
 809 static void __init do_pre_smp_initcalls(void)
 810 {
 811         initcall_t *fn;
 812
 813         for (fn = __initcall_start; fn < __early_initcall_end; fn++)
 814                 do_one_initcall(*fn);
 815 }
 816
 817 static void run_init_process(const char *init_filename)
 818 {
 819         argv_init[0] = init_filename;
 820         kernel_execve(init_filename, argv_init, envp_init);
 821 }
 822
 823 /* This is a non __init function. Force it to be noinline otherwise gcc
 824  * makes it inline to init() and it becomes part of init.text section
 825  */
 826 static noinline int init_post(void)
 827 {
 828         /* need to finish all async __init code before freeing the memory */
 829         async_synchronize_full();
 830         free_initmem();
 831         mark_rodata_ro();
 832         system_state = SYSTEM_RUNNING;
 833         numa_default_policy();
 834
 835
 836         current->signal->flags |= SIGNAL_UNKILLABLE;
 837
 838         if (ramdisk_execute_command) {
 839                 run_init_process(ramdisk_execute_command);
 840                 printk(KERN_WARNING "Failed to execute %s\n",
 841                                 ramdisk_execute_command);
 842         }
 843
 844         /*
 845          * We try each of these until one succeeds.
 846          *
 847          * The Bourne shell can be used instead of init if we are
 848          * trying to recover a really broken machine.
 849          */
 850         if (execute_command) {
 851                 run_init_process(execute_command);
 852                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 853                                         "defaults...\n", execute_command);
 854         }
 855         run_init_process("/sbin/init");
 856         run_init_process("/etc/init");
 857         run_init_process("/bin/init");
 858         run_init_process("/bin/sh");
 859
 860         panic("No init found.  Try passing init= option to kernel. "
 861               "See Linux Documentation/init.txt for guidance.");
 862 }
 863
 864 static int __init kernel_init(void * unused)
 865 {
 866         /*
 867          * Wait until kthreadd is all set-up.
 868          */
 869         wait_for_completion(&kthreadd_done);
 870         /*
 871          * init can allocate pages on any node
 872          */
 873         set_mems_allowed(node_states[N_HIGH_MEMORY]);
 874         /*
 875          * init can run on any cpu.
 876          */
 877         set_cpus_allowed_ptr(current, cpu_all_mask);
 878         /*
 879          * Tell the world that we're going to be the grim
 880          * reaper of innocent orphaned children.
 881          *
 882          * We don't want people to have to make incorrect
 883          * assumptions about where in the task array this
 884          * can be found.
 885          */
 886         init_pid_ns.child_reaper = current;
 887
 888         cad_pid = task_pid(current);
 889
 890         smp_prepare_cpus(setup_max_cpus);
 891
 892         do_pre_smp_initcalls();
 893         lockup_detector_init();
 894
 895         smp_init();
 896         sched_init_smp();
 897
 898         do_basic_setup();
 899
 900         /* Open the /dev/console on the rootfs, this should never fail */
 901         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 902                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 903
 904         (void) sys_dup(0);
 905         (void) sys_dup(0);
 906         /*
 907          * check if there is an early userspace init.  If yes, let it do all
 908          * the work
 909          */
 910
 911         if (!ramdisk_execute_command)
 912                 ramdisk_execute_command = "/init";
 913
 914         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 915                 ramdisk_execute_command = NULL;
 916                 prepare_namespace();
 917         }
 918
 919         /*
 920          * Ok, we have completed the initial bootup, and
 921          * we're essentially up and running. Get rid of the
 922          * initmem segments and start the user-mode stuff..
 923          */
 924
 925         init_post();
 926         return 0;
 927 }