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 #define DEBUG           /* Enable initcall_debug */
  13
  14 #include <linux/types.h>
  15 #include <linux/module.h>
  16 #include <linux/proc_fs.h>
  17 #include <linux/kernel.h>
  18 #include <linux/syscalls.h>
  19 #include <linux/stackprotector.h>
  20 #include <linux/string.h>
  21 #include <linux/ctype.h>
  22 #include <linux/delay.h>
  23 #include <linux/ioport.h>
  24 #include <linux/init.h>
  25 #include <linux/initrd.h>
  26 #include <linux/bootmem.h>
  27 #include <linux/acpi.h>
  28 #include <linux/tty.h>
  29 #include <linux/percpu.h>
  30 #include <linux/kmod.h>
  31 #include <linux/vmalloc.h>
  32 #include <linux/kernel_stat.h>
  33 #include <linux/start_kernel.h>
  34 #include <linux/security.h>
  35 #include <linux/smp.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/buffer_head.h>
  54 #include <linux/page_cgroup.h>
  55 #include <linux/debug_locks.h>
  56 #include <linux/debugobjects.h>
  57 #include <linux/lockdep.h>
  58 #include <linux/kmemleak.h>
  59 #include <linux/pid_namespace.h>
  60 #include <linux/device.h>
  61 #include <linux/kthread.h>
  62 #include <linux/sched.h>
  63 #include <linux/signal.h>
  64 #include <linux/idr.h>
  65 #include <linux/kgdb.h>
  66 #include <linux/ftrace.h>
  67 #include <linux/async.h>
  68 #include <linux/kmemcheck.h>
  69 #include <linux/sfi.h>
  70 #include <linux/shmem_fs.h>
  71 #include <linux/slab.h>
  72 #include <linux/perf_event.h>
  73 #include <linux/file.h>
  74 #include <linux/ptrace.h>
  75 #include <linux/blkdev.h>
  76 #include <linux/elevator.h>
  77 #include <linux/sched_clock.h>
  78 #include <linux/context_tracking.h>
  79 #include <linux/random.h>
  80
  81 #include <asm/io.h>
  82 #include <asm/bugs.h>
  83 #include <asm/setup.h>
  84 #include <asm/sections.h>
  85 #include <asm/cacheflush.h>
  86
  87 #ifdef CONFIG_X86_LOCAL_APIC
  88 #include <asm/smp.h>
  89 #endif
  90
  91 static int kernel_init(void *);
  92
  93 extern void init_IRQ(void);
  94 extern void fork_init(unsigned long);
  95 extern void mca_init(void);
  96 extern void sbus_init(void);
  97 extern void radix_tree_init(void);
  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 /*
 107  * Debug helper: via this flag we know that we are in 'early bootup code'
 108  * where only the boot processor is running with IRQ disabled.  This means
 109  * two things - IRQ must not be enabled before the flag is cleared and some
 110  * operations which are not allowed with IRQ disabled are allowed while the
 111  * flag is set.
 112  */
 113 bool early_boot_irqs_disabled __read_mostly;
 114
 115 enum system_states system_state __read_mostly;
 116 EXPORT_SYMBOL(system_state);
 117
 118 /*
 119  * Boot command-line arguments
 120  */
 121 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 122 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 123
 124 extern void time_init(void);
 125 /* Default late time init is NULL. archs can override this later. */
 126 void (*__initdata late_time_init)(void);
 127 extern void softirq_init(void);
 128
 129 /* Untouched command line saved by arch-specific code. */
 130 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 131 /* Untouched saved command line (eg. for /proc) */
 132 char *saved_command_line;
 133 /* Command line for parameter parsing */
 134 static char *static_command_line;
 135
 136 static char *execute_command;
 137 static char *ramdisk_execute_command;
 138
 139 /*
 140  * Used to generate warnings if static_key manipulation functions are used
 141  * before jump_label_init is called.
 142  */
 143 bool static_key_initialized __read_mostly = false;
 144 EXPORT_SYMBOL_GPL(static_key_initialized);
 145
 146 /*
 147  * If set, this is an indication to the drivers that reset the underlying
 148  * device before going ahead with the initialization otherwise driver might
 149  * rely on the BIOS and skip the reset operation.
 150  *
 151  * This is useful if kernel is booting in an unreliable environment.
 152  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 153  * skipped and devices will be in unknown state.
 154  */
 155 unsigned int reset_devices;
 156 EXPORT_SYMBOL(reset_devices);
 157
 158 static int __init set_reset_devices(char *str)
 159 {
 160         reset_devices = 1;
 161         return 1;
 162 }
 163
 164 __setup("reset_devices", set_reset_devices);
 165
 166 static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 167 const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 168 static const char *panic_later, *panic_param;
 169
 170 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 171
 172 static int __init obsolete_checksetup(char *line)
 173 {
 174         const struct obs_kernel_param *p;
 175         int had_early_param = 0;
 176
 177         p = __setup_start;
 178         do {
 179                 int n = strlen(p->str);
 180                 if (parameqn(line, p->str, n)) {
 181                         if (p->early) {
 182                                 /* Already done in parse_early_param?
 183                                  * (Needs exact match on param part).
 184                                  * Keep iterating, as we can have early
 185                                  * params and __setups of same names 8( */
 186                                 if (line[n] == '\0' || line[n] == '=')
 187                                         had_early_param = 1;
 188                         } else if (!p->setup_func) {
 189                                 pr_warn("Parameter %s is obsolete, ignored\n",
 190                                         p->str);
 191                                 return 1;
 192                         } else if (p->setup_func(line + n))
 193                                 return 1;
 194                 }
 195                 p++;
 196         } while (p < __setup_end);
 197
 198         return had_early_param;
 199 }
 200
 201 /*
 202  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 203  * still work even if initially too large, it will just take slightly longer
 204  */
 205 unsigned long loops_per_jiffy = (1<<12);
 206
 207 EXPORT_SYMBOL(loops_per_jiffy);
 208
 209 static int __init debug_kernel(char *str)
 210 {
 211         console_loglevel = 10;
 212         return 0;
 213 }
 214
 215 static int __init quiet_kernel(char *str)
 216 {
 217         console_loglevel = 4;
 218         return 0;
 219 }
 220
 221 early_param("debug", debug_kernel);
 222 early_param("quiet", quiet_kernel);
 223
 224 static int __init loglevel(char *str)
 225 {
 226         int newlevel;
 227
 228         /*
 229          * Only update loglevel value when a correct setting was passed,
 230          * to prevent blind crashes (when loglevel being set to 0) that
 231          * are quite hard to debug
 232          */
 233         if (get_option(&str, &newlevel)) {
 234                 console_loglevel = newlevel;
 235                 return 0;
 236         }
 237
 238         return -EINVAL;
 239 }
 240
 241 early_param("loglevel", loglevel);
 242
 243 /* Change NUL term back to "=", to make "param" the whole string. */
 244 static int __init repair_env_string(char *param, char *val, const char *unused)
 245 {
 246         if (val) {
 247                 /* param=val or param="val"? */
 248                 if (val == param+strlen(param)+1)
 249                         val[-1] = '=';
 250                 else if (val == param+strlen(param)+2) {
 251                         val[-2] = '=';
 252                         memmove(val-1, val, strlen(val)+1);
 253                         val--;
 254                 } else
 255                         BUG();
 256         }
 257         return 0;
 258 }
 259
 260 /*
 261  * Unknown boot options get handed to init, unless they look like
 262  * unused parameters (modprobe will find them in /proc/cmdline).
 263  */
 264 static int __init unknown_bootoption(char *param, char *val, const char *unused)
 265 {
 266         repair_env_string(param, val, unused);
 267
 268         /* Handle obsolete-style parameters */
 269         if (obsolete_checksetup(param))
 270                 return 0;
 271
 272         /* Unused module parameter. */
 273         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 274                 return 0;
 275
 276         if (panic_later)
 277                 return 0;
 278
 279         if (val) {
 280                 /* Environment option */
 281                 unsigned int i;
 282                 for (i = 0; envp_init[i]; i++) {
 283                         if (i == MAX_INIT_ENVS) {
 284                                 panic_later = "Too many boot env vars at `%s'";
 285                                 panic_param = param;
 286                         }
 287                         if (!strncmp(param, envp_init[i], val - param))
 288                                 break;
 289                 }
 290                 envp_init[i] = param;
 291         } else {
 292                 /* Command line option */
 293                 unsigned int i;
 294                 for (i = 0; argv_init[i]; i++) {
 295                         if (i == MAX_INIT_ARGS) {
 296                                 panic_later = "Too many boot init vars at `%s'";
 297                                 panic_param = param;
 298                         }
 299                 }
 300                 argv_init[i] = param;
 301         }
 302         return 0;
 303 }
 304
 305 static int __init init_setup(char *str)
 306 {
 307         unsigned int i;
 308
 309         execute_command = str;
 310         /*
 311          * In case LILO is going to boot us with default command line,
 312          * it prepends "auto" before the whole cmdline which makes
 313          * the shell think it should execute a script with such name.
 314          * So we ignore all arguments entered _before_ init=... [MJ]
 315          */
 316         for (i = 1; i < MAX_INIT_ARGS; i++)
 317                 argv_init[i] = NULL;
 318         return 1;
 319 }
 320 __setup("init=", init_setup);
 321
 322 static int __init rdinit_setup(char *str)
 323 {
 324         unsigned int i;
 325
 326         ramdisk_execute_command = str;
 327         /* See "auto" comment in init_setup */
 328         for (i = 1; i < MAX_INIT_ARGS; i++)
 329                 argv_init[i] = NULL;
 330         return 1;
 331 }
 332 __setup("rdinit=", rdinit_setup);
 333
 334 #ifndef CONFIG_SMP
 335 static const unsigned int setup_max_cpus = NR_CPUS;
 336 #ifdef CONFIG_X86_LOCAL_APIC
 337 static void __init smp_init(void)
 338 {
 339         APIC_init_uniprocessor();
 340 }
 341 #else
 342 #define smp_init()      do { } while (0)
 343 #endif
 344
 345 static inline void setup_nr_cpu_ids(void) { }
 346 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 347 #endif
 348
 349 /*
 350  * We need to store the untouched command line for future reference.
 351  * We also need to store the touched command line since the parameter
 352  * parsing is performed in place, and we should allow a component to
 353  * store reference of name/value for future reference.
 354  */
 355 static void __init setup_command_line(char *command_line)
 356 {
 357         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 358         static_command_line = alloc_bootmem(strlen (command_line)+1);
 359         strcpy (saved_command_line, boot_command_line);
 360         strcpy (static_command_line, command_line);
 361 }
 362
 363 /*
 364  * We need to finalize in a non-__init function or else race conditions
 365  * between the root thread and the init thread may cause start_kernel to
 366  * be reaped by free_initmem before the root thread has proceeded to
 367  * cpu_idle.
 368  *
 369  * gcc-3.4 accidentally inlines this function, so use noinline.
 370  */
 371
 372 static __initdata DECLARE_COMPLETION(kthreadd_done);
 373
 374 static noinline void __init_refok rest_init(void)
 375 {
 376         int pid;
 377
 378         rcu_scheduler_starting();
 379         /*
 380          * We need to spawn init first so that it obtains pid 1, however
 381          * the init task will end up wanting to create kthreads, which, if
 382          * we schedule it before we create kthreadd, will OOPS.
 383          */
 384         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 385         numa_default_policy();
 386         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 387         rcu_read_lock();
 388         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 389         rcu_read_unlock();
 390         complete(&kthreadd_done);
 391
 392         /*
 393          * The boot idle thread must execute schedule()
 394          * at least once to get things moving:
 395          */
 396         init_idle_bootup_task(current);
 397         schedule_preempt_disabled();
 398         /* Call into cpu_idle with preempt disabled */
 399         cpu_startup_entry(CPUHP_ONLINE);
 400 }
 401
 402 /* Check for early params. */
 403 static int __init do_early_param(char *param, char *val, const char *unused)
 404 {
 405         const struct obs_kernel_param *p;
 406
 407         for (p = __setup_start; p < __setup_end; p++) {
 408                 if ((p->early && parameq(param, p->str)) ||
 409                     (strcmp(param, "console") == 0 &&
 410                      strcmp(p->str, "earlycon") == 0)
 411                 ) {
 412                         if (p->setup_func(val) != 0)
 413                                 pr_warn("Malformed early option '%s'\n", param);
 414                 }
 415         }
 416         /* We accept everything at this stage. */
 417         return 0;
 418 }
 419
 420 void __init parse_early_options(char *cmdline)
 421 {
 422         parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
 423 }
 424
 425 /* Arch code calls this early on, or if not, just before other parsing. */
 426 void __init parse_early_param(void)
 427 {
 428         static __initdata int done = 0;
 429         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 430
 431         if (done)
 432                 return;
 433
 434         /* All fall through to do_early_param. */
 435         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 436         parse_early_options(tmp_cmdline);
 437         done = 1;
 438 }
 439
 440 /*
 441  *      Activate the first processor.
 442  */
 443
 444 static void __init boot_cpu_init(void)
 445 {
 446         int cpu = smp_processor_id();
 447         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 448         set_cpu_online(cpu, true);
 449         set_cpu_active(cpu, true);
 450         set_cpu_present(cpu, true);
 451         set_cpu_possible(cpu, true);
 452 }
 453
 454 void __init __weak smp_setup_processor_id(void)
 455 {
 456 }
 457
 458 # if THREAD_SIZE >= PAGE_SIZE
 459 void __init __weak thread_info_cache_init(void)
 460 {
 461 }
 462 #endif
 463
 464 /*
 465  * Set up kernel memory allocators
 466  */
 467 static void __init mm_init(void)
 468 {
 469         /*
 470          * page_cgroup requires contiguous pages,
 471          * bigger than MAX_ORDER unless SPARSEMEM.
 472          */
 473         page_cgroup_init_flatmem();
 474         mem_init();
 475         kmem_cache_init();
 476         percpu_init_late();
 477         pgtable_cache_init();
 478         vmalloc_init();
 479 }
 480
 481 asmlinkage void __init start_kernel(void)
 482 {
 483         char * command_line;
 484         extern const struct kernel_param __start___param[], __stop___param[];
 485
 486         /*
 487          * Need to run as early as possible, to initialize the
 488          * lockdep hash:
 489          */
 490         lockdep_init();
 491         smp_setup_processor_id();
 492         debug_objects_early_init();
 493
 494         /*
 495          * Set up the the initial canary ASAP:
 496          */
 497         boot_init_stack_canary();
 498
 499         cgroup_init_early();
 500
 501         local_irq_disable();
 502         early_boot_irqs_disabled = true;
 503
 504 /*
 505  * Interrupts are still disabled. Do necessary setups, then
 506  * enable them
 507  */
 508         boot_cpu_init();
 509         page_address_init();
 510         pr_notice("%s", linux_banner);
 511         setup_arch(&command_line);
 512         mm_init_owner(&init_mm, &init_task);
 513         mm_init_cpumask(&init_mm);
 514         setup_command_line(command_line);
 515         setup_nr_cpu_ids();
 516         setup_per_cpu_areas();
 517         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 518
 519         build_all_zonelists(NULL, NULL);
 520         page_alloc_init();
 521
 522         pr_notice("Kernel command line: %s\n", boot_command_line);
 523         parse_early_param();
 524         parse_args("Booting kernel", static_command_line, __start___param,
 525                    __stop___param - __start___param,
 526                    -1, -1, &unknown_bootoption);
 527
 528         jump_label_init();
 529
 530         /*
 531          * These use large bootmem allocations and must precede
 532          * kmem_cache_init()
 533          */
 534         setup_log_buf(0);
 535         pidhash_init();
 536         vfs_caches_init_early();
 537         sort_main_extable();
 538         trap_init();
 539         mm_init();
 540
 541         /*
 542          * Set up the scheduler prior starting any interrupts (such as the
 543          * timer interrupt). Full topology setup happens at smp_init()
 544          * time - but meanwhile we still have a functioning scheduler.
 545          */
 546         sched_init();
 547         /*
 548          * Disable preemption - early bootup scheduling is extremely
 549          * fragile until we cpu_idle() for the first time.
 550          */
 551         preempt_disable();
 552         if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
 553                 local_irq_disable();
 554         idr_init_cache();
 555         rcu_init();
 556         tick_nohz_init();
 557         context_tracking_init();
 558         radix_tree_init();
 559         /* init some links before init_ISA_irqs() */
 560         early_irq_init();
 561         init_IRQ();
 562         tick_init();
 563         init_timers();
 564         hrtimers_init();
 565         softirq_init();
 566         timekeeping_init();
 567         time_init();
 568         sched_clock_postinit();
 569         perf_event_init();
 570         profile_init();
 571         call_function_init();
 572         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 573         early_boot_irqs_disabled = false;
 574         local_irq_enable();
 575
 576         kmem_cache_init_late();
 577
 578         /*
 579          * HACK ALERT! This is early. We're enabling the console before
 580          * we've done PCI setups etc, and console_init() must be aware of
 581          * this. But we do want output early, in case something goes wrong.
 582          */
 583         console_init();
 584         if (panic_later)
 585                 panic(panic_later, panic_param);
 586
 587         lockdep_info();
 588
 589         /*
 590          * Need to run this when irqs are enabled, because it wants
 591          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 592          * too:
 593          */
 594         locking_selftest();
 595
 596 #ifdef CONFIG_BLK_DEV_INITRD
 597         if (initrd_start && !initrd_below_start_ok &&
 598             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 599                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
 600                     page_to_pfn(virt_to_page((void *)initrd_start)),
 601                     min_low_pfn);
 602                 initrd_start = 0;
 603         }
 604 #endif
 605         page_cgroup_init();
 606         debug_objects_mem_init();
 607         kmemleak_init();
 608         setup_per_cpu_pageset();
 609         numa_policy_init();
 610         if (late_time_init)
 611                 late_time_init();
 612         sched_clock_init();
 613         calibrate_delay();
 614         pidmap_init();
 615         anon_vma_init();
 616 #ifdef CONFIG_X86
 617         if (efi_enabled(EFI_RUNTIME_SERVICES))
 618                 efi_enter_virtual_mode();
 619 #endif
 620         thread_info_cache_init();
 621         cred_init();
 622         fork_init(totalram_pages);
 623         proc_caches_init();
 624         buffer_init();
 625         key_init();
 626         security_init();
 627         dbg_late_init();
 628         vfs_caches_init(totalram_pages);
 629         signals_init();
 630         /* rootfs populating might need page-writeback */
 631         page_writeback_init();
 632 #ifdef CONFIG_PROC_FS
 633         proc_root_init();
 634 #endif
 635         cgroup_init();
 636         cpuset_init();
 637         taskstats_init_early();
 638         delayacct_init();
 639
 640         check_bugs();
 641
 642         acpi_early_init(); /* before LAPIC and SMP init */
 643         sfi_init_late();
 644
 645         if (efi_enabled(EFI_RUNTIME_SERVICES)) {
 646                 efi_late_init();
 647                 efi_free_boot_services();
 648         }
 649
 650         ftrace_init();
 651
 652         /* Do the rest non-__init'ed, we're now alive */
 653         rest_init();
 654 }
 655
 656 /* Call all constructor functions linked into the kernel. */
 657 static void __init do_ctors(void)
 658 {
 659 #ifdef CONFIG_CONSTRUCTORS
 660         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 661
 662         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 663                 (*fn)();
 664 #endif
 665 }
 666
 667 bool initcall_debug;
 668 core_param(initcall_debug, initcall_debug, bool, 0644);
 669
 670 static int __init_or_module do_one_initcall_debug(initcall_t fn)
 671 {
 672         ktime_t calltime, delta, rettime;
 673         unsigned long long duration;
 674         int ret;
 675
 676         pr_debug("calling  %pF @ %i\n", fn, task_pid_nr(current));
 677         calltime = ktime_get();
 678         ret = fn();
 679         rettime = ktime_get();
 680         delta = ktime_sub(rettime, calltime);
 681         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 682         pr_debug("initcall %pF returned %d after %lld usecs\n",
 683                  fn, ret, duration);
 684
 685         return ret;
 686 }
 687
 688 int __init_or_module do_one_initcall(initcall_t fn)
 689 {
 690         int count = preempt_count();
 691         int ret;
 692         char msgbuf[64];
 693
 694         if (initcall_debug)
 695                 ret = do_one_initcall_debug(fn);
 696         else
 697                 ret = fn();
 698
 699         msgbuf[0] = 0;
 700
 701         if (preempt_count() != count) {
 702                 sprintf(msgbuf, "preemption imbalance ");
 703                 preempt_count() = count;
 704         }
 705         if (irqs_disabled()) {
 706                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 707                 local_irq_enable();
 708         }
 709         WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
 710
 711         return ret;
 712 }
 713
 714
 715 extern initcall_t __initcall_start[];
 716 extern initcall_t __initcall0_start[];
 717 extern initcall_t __initcall1_start[];
 718 extern initcall_t __initcall2_start[];
 719 extern initcall_t __initcall3_start[];
 720 extern initcall_t __initcall4_start[];
 721 extern initcall_t __initcall5_start[];
 722 extern initcall_t __initcall6_start[];
 723 extern initcall_t __initcall7_start[];
 724 extern initcall_t __initcall_end[];
 725
 726 static initcall_t *initcall_levels[] __initdata = {
 727         __initcall0_start,
 728         __initcall1_start,
 729         __initcall2_start,
 730         __initcall3_start,
 731         __initcall4_start,
 732         __initcall5_start,
 733         __initcall6_start,
 734         __initcall7_start,
 735         __initcall_end,
 736 };
 737
 738 /* Keep these in sync with initcalls in include/linux/init.h */
 739 static char *initcall_level_names[] __initdata = {
 740         "early",
 741         "core",
 742         "postcore",
 743         "arch",
 744         "subsys",
 745         "fs",
 746         "device",
 747         "late",
 748 };
 749
 750 static void __init do_initcall_level(int level)
 751 {
 752         extern const struct kernel_param __start___param[], __stop___param[];
 753         initcall_t *fn;
 754
 755         strcpy(static_command_line, saved_command_line);
 756         parse_args(initcall_level_names[level],
 757                    static_command_line, __start___param,
 758                    __stop___param - __start___param,
 759                    level, level,
 760                    &repair_env_string);
 761
 762         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
 763                 do_one_initcall(*fn);
 764 }
 765
 766 static void __init do_initcalls(void)
 767 {
 768         int level;
 769
 770         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
 771                 do_initcall_level(level);
 772 }
 773
 774 /*
 775  * Ok, the machine is now initialized. None of the devices
 776  * have been touched yet, but the CPU subsystem is up and
 777  * running, and memory and process management works.
 778  *
 779  * Now we can finally start doing some real work..
 780  */
 781 static void __init do_basic_setup(void)
 782 {
 783         cpuset_init_smp();
 784         usermodehelper_init();
 785         shmem_init();
 786         driver_init();
 787         init_irq_proc();
 788         do_ctors();
 789         usermodehelper_enable();
 790         do_initcalls();
 791         random_int_secret_init();
 792 }
 793
 794 static void __init do_pre_smp_initcalls(void)
 795 {
 796         initcall_t *fn;
 797
 798         for (fn = __initcall_start; fn < __initcall0_start; fn++)
 799                 do_one_initcall(*fn);
 800 }
 801
 802 /*
 803  * This function requests modules which should be loaded by default and is
 804  * called twice right after initrd is mounted and right before init is
 805  * exec'd.  If such modules are on either initrd or rootfs, they will be
 806  * loaded before control is passed to userland.
 807  */
 808 void __init load_default_modules(void)
 809 {
 810         load_default_elevator_module();
 811 }
 812
 813 static int run_init_process(const char *init_filename)
 814 {
 815         argv_init[0] = init_filename;
 816         return do_execve(init_filename,
 817                 (const char __user *const __user *)argv_init,
 818                 (const char __user *const __user *)envp_init);
 819 }
 820
 821 static noinline void __init kernel_init_freeable(void);
 822
 823 static int __ref kernel_init(void *unused)
 824 {
 825         kernel_init_freeable();
 826         /* need to finish all async __init code before freeing the memory */
 827         async_synchronize_full();
 828         free_initmem();
 829         mark_rodata_ro();
 830         system_state = SYSTEM_RUNNING;
 831         numa_default_policy();
 832
 833         flush_delayed_fput();
 834
 835         if (ramdisk_execute_command) {
 836                 if (!run_init_process(ramdisk_execute_command))
 837                         return 0;
 838                 pr_err("Failed to execute %s\n", ramdisk_execute_command);
 839         }
 840
 841         /*
 842          * We try each of these until one succeeds.
 843          *
 844          * The Bourne shell can be used instead of init if we are
 845          * trying to recover a really broken machine.
 846          */
 847         if (execute_command) {
 848                 if (!run_init_process(execute_command))
 849                         return 0;
 850                 pr_err("Failed to execute %s.  Attempting defaults...\n",
 851                         execute_command);
 852         }
 853         if (!run_init_process("/sbin/init") ||
 854             !run_init_process("/etc/init") ||
 855             !run_init_process("/bin/init") ||
 856             !run_init_process("/bin/sh"))
 857                 return 0;
 858
 859         panic("No init found.  Try passing init= option to kernel. "
 860               "See Linux Documentation/init.txt for guidance.");
 861 }
 862
 863 static noinline void __init kernel_init_freeable(void)
 864 {
 865         /*
 866          * Wait until kthreadd is all set-up.
 867          */
 868         wait_for_completion(&kthreadd_done);
 869
 870         /* Now the scheduler is fully set up and can do blocking allocations */
 871         gfp_allowed_mask = __GFP_BITS_MASK;
 872
 873         /*
 874          * init can allocate pages on any node
 875          */
 876         set_mems_allowed(node_states[N_MEMORY]);
 877         /*
 878          * init can run on any cpu.
 879          */
 880         set_cpus_allowed_ptr(current, cpu_all_mask);
 881
 882         cad_pid = task_pid(current);
 883
 884         smp_prepare_cpus(setup_max_cpus);
 885
 886         do_pre_smp_initcalls();
 887         lockup_detector_init();
 888
 889         smp_init();
 890         sched_init_smp();
 891
 892         do_basic_setup();
 893
 894         /* Open the /dev/console on the rootfs, this should never fail */
 895         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 896                 pr_err("Warning: unable to open an initial console.\n");
 897
 898         (void) sys_dup(0);
 899         (void) sys_dup(0);
 900         /*
 901          * check if there is an early userspace init.  If yes, let it do all
 902          * the work
 903          */
 904
 905         if (!ramdisk_execute_command)
 906                 ramdisk_execute_command = "/init";
 907
 908         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 909                 ramdisk_execute_command = NULL;
 910                 prepare_namespace();
 911         }
 912
 913         /*
 914          * Ok, we have completed the initial bootup, and
 915          * we're essentially up and running. Get rid of the
 916          * initmem segments and start the user-mode stuff..
 917          */
 918
 919         /* rootfs is available now, try loading default modules */
 920         load_default_modules();
 921 }