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/debug_locks.h>
  55 #include <linux/debugobjects.h>
  56 #include <linux/lockdep.h>
  57 #include <linux/kmemleak.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 #include <linux/kgdb.h>
  65 #include <linux/ftrace.h>
  66 #include <linux/async.h>
  67 #include <linux/kmemcheck.h>
  68 #include <linux/sfi.h>
  69 #include <linux/shmem_fs.h>
  70 #include <linux/slab.h>
  71 #include <linux/perf_event.h>
  72 #include <linux/file.h>
  73 #include <linux/ptrace.h>
  74 #include <linux/blkdev.h>
  75 #include <linux/elevator.h>
  76 #include <linux/sched_clock.h>
  77 #include <linux/context_tracking.h>
  78 #include <linux/random.h>
  79 #include <linux/list.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 radix_tree_init(void);
  96 #ifndef CONFIG_DEBUG_RODATA
  97 static inline void mark_rodata_ro(void) { }
  98 #endif
  99
 100 /*
 101  * Debug helper: via this flag we know that we are in 'early bootup code'
 102  * where only the boot processor is running with IRQ disabled.  This means
 103  * two things - IRQ must not be enabled before the flag is cleared and some
 104  * operations which are not allowed with IRQ disabled are allowed while the
 105  * flag is set.
 106  */
 107 bool early_boot_irqs_disabled __read_mostly;
 108
 109 enum system_states system_state __read_mostly;
 110 EXPORT_SYMBOL(system_state);
 111
 112 /*
 113  * Boot command-line arguments
 114  */
 115 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 116 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 117
 118 extern void time_init(void);
 119 /* Default late time init is NULL. archs can override this later. */
 120 void (*__initdata late_time_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 /* Command line for per-initcall parameter parsing */
 129 static char *initcall_command_line;
 130
 131 static char *execute_command;
 132 static char *ramdisk_execute_command;
 133
 134 /*
 135  * Used to generate warnings if static_key manipulation functions are used
 136  * before jump_label_init is called.
 137  */
 138 bool static_key_initialized __read_mostly;
 139 EXPORT_SYMBOL_GPL(static_key_initialized);
 140
 141 /*
 142  * If set, this is an indication to the drivers that reset the underlying
 143  * device before going ahead with the initialization otherwise driver might
 144  * rely on the BIOS and skip the reset operation.
 145  *
 146  * This is useful if kernel is booting in an unreliable environment.
 147  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 148  * skipped and devices will be in unknown state.
 149  */
 150 unsigned int reset_devices;
 151 EXPORT_SYMBOL(reset_devices);
 152
 153 static int __init set_reset_devices(char *str)
 154 {
 155         reset_devices = 1;
 156         return 1;
 157 }
 158
 159 __setup("reset_devices", set_reset_devices);
 160
 161 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 162 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 163 static const char *panic_later, *panic_param;
 164
 165 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 166
 167 static int __init obsolete_checksetup(char *line)
 168 {
 169         const struct obs_kernel_param *p;
 170         int had_early_param = 0;
 171
 172         p = __setup_start;
 173         do {
 174                 int n = strlen(p->str);
 175                 if (parameqn(line, p->str, n)) {
 176                         if (p->early) {
 177                                 /* Already done in parse_early_param?
 178                                  * (Needs exact match on param part).
 179                                  * Keep iterating, as we can have early
 180                                  * params and __setups of same names 8( */
 181                                 if (line[n] == '\0' || line[n] == '=')
 182                                         had_early_param = 1;
 183                         } else if (!p->setup_func) {
 184                                 pr_warn("Parameter %s is obsolete, ignored\n",
 185                                         p->str);
 186                                 return 1;
 187                         } else if (p->setup_func(line + n))
 188                                 return 1;
 189                 }
 190                 p++;
 191         } while (p < __setup_end);
 192
 193         return had_early_param;
 194 }
 195
 196 /*
 197  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 198  * still work even if initially too large, it will just take slightly longer
 199  */
 200 unsigned long loops_per_jiffy = (1<<12);
 201 EXPORT_SYMBOL(loops_per_jiffy);
 202
 203 static int __init debug_kernel(char *str)
 204 {
 205         console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
 206         return 0;
 207 }
 208
 209 static int __init quiet_kernel(char *str)
 210 {
 211         console_loglevel = CONSOLE_LOGLEVEL_QUIET;
 212         return 0;
 213 }
 214
 215 early_param("debug", debug_kernel);
 216 early_param("quiet", quiet_kernel);
 217
 218 static int __init loglevel(char *str)
 219 {
 220         int newlevel;
 221
 222         /*
 223          * Only update loglevel value when a correct setting was passed,
 224          * to prevent blind crashes (when loglevel being set to 0) that
 225          * are quite hard to debug
 226          */
 227         if (get_option(&str, &newlevel)) {
 228                 console_loglevel = newlevel;
 229                 return 0;
 230         }
 231
 232         return -EINVAL;
 233 }
 234
 235 early_param("loglevel", loglevel);
 236
 237 /* Change NUL term back to "=", to make "param" the whole string. */
 238 static int __init repair_env_string(char *param, char *val, const char *unused)
 239 {
 240         if (val) {
 241                 /* param=val or param="val"? */
 242                 if (val == param+strlen(param)+1)
 243                         val[-1] = '=';
 244                 else if (val == param+strlen(param)+2) {
 245                         val[-2] = '=';
 246                         memmove(val-1, val, strlen(val)+1);
 247                         val--;
 248                 } else
 249                         BUG();
 250         }
 251         return 0;
 252 }
 253
 254 /* Anything after -- gets handed straight to init. */
 255 static int __init set_init_arg(char *param, char *val, const char *unused)
 256 {
 257         unsigned int i;
 258
 259         if (panic_later)
 260                 return 0;
 261
 262         repair_env_string(param, val, unused);
 263
 264         for (i = 0; argv_init[i]; i++) {
 265                 if (i == MAX_INIT_ARGS) {
 266                         panic_later = "init";
 267                         panic_param = param;
 268                         return 0;
 269                 }
 270         }
 271         argv_init[i] = param;
 272         return 0;
 273 }
 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, const char *unused)
 280 {
 281         repair_env_string(param, val, unused);
 282
 283         /* Handle obsolete-style parameters */
 284         if (obsolete_checksetup(param))
 285                 return 0;
 286
 287         /* Unused module parameter. */
 288         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 289                 return 0;
 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 = "env";
 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 = "init";
 312                                 panic_param = param;
 313                         }
 314                 }
 315                 argv_init[i] = param;
 316         }
 317         return 0;
 318 }
 319
 320 static int __init init_setup(char *str)
 321 {
 322         unsigned int i;
 323
 324         execute_command = str;
 325         /*
 326          * In case LILO is going to boot us with default command line,
 327          * it prepends "auto" before the whole cmdline which makes
 328          * the shell think it should execute a script with such name.
 329          * So we ignore all arguments entered _before_ init=... [MJ]
 330          */
 331         for (i = 1; i < MAX_INIT_ARGS; i++)
 332                 argv_init[i] = NULL;
 333         return 1;
 334 }
 335 __setup("init=", init_setup);
 336
 337 static int __init rdinit_setup(char *str)
 338 {
 339         unsigned int i;
 340
 341         ramdisk_execute_command = str;
 342         /* See "auto" comment in init_setup */
 343         for (i = 1; i < MAX_INIT_ARGS; i++)
 344                 argv_init[i] = NULL;
 345         return 1;
 346 }
 347 __setup("rdinit=", rdinit_setup);
 348
 349 #ifndef CONFIG_SMP
 350 static const unsigned int setup_max_cpus = NR_CPUS;
 351 #ifdef CONFIG_X86_LOCAL_APIC
 352 static void __init smp_init(void)
 353 {
 354         APIC_init_uniprocessor();
 355 }
 356 #else
 357 #define smp_init()      do { } while (0)
 358 #endif
 359
 360 static inline void setup_nr_cpu_ids(void) { }
 361 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 362 #endif
 363
 364 /*
 365  * We need to store the untouched command line for future reference.
 366  * We also need to store the touched command line since the parameter
 367  * parsing is performed in place, and we should allow a component to
 368  * store reference of name/value for future reference.
 369  */
 370 static void __init setup_command_line(char *command_line)
 371 {
 372         saved_command_line =
 373                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 374         initcall_command_line =
 375                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 376         static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
 377         strcpy(saved_command_line, boot_command_line);
 378         strcpy(static_command_line, command_line);
 379 }
 380
 381 /*
 382  * We need to finalize in a non-__init function or else race conditions
 383  * between the root thread and the init thread may cause start_kernel to
 384  * be reaped by free_initmem before the root thread has proceeded to
 385  * cpu_idle.
 386  *
 387  * gcc-3.4 accidentally inlines this function, so use noinline.
 388  */
 389
 390 static __initdata DECLARE_COMPLETION(kthreadd_done);
 391
 392 static noinline void __init_refok rest_init(void)
 393 {
 394         int pid;
 395
 396         rcu_scheduler_starting();
 397         /*
 398          * We need to spawn init first so that it obtains pid 1, however
 399          * the init task will end up wanting to create kthreads, which, if
 400          * we schedule it before we create kthreadd, will OOPS.
 401          */
 402         kernel_thread(kernel_init, NULL, CLONE_FS);
 403         numa_default_policy();
 404         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 405         rcu_read_lock();
 406         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 407         rcu_read_unlock();
 408         complete(&kthreadd_done);
 409
 410         /*
 411          * The boot idle thread must execute schedule()
 412          * at least once to get things moving:
 413          */
 414         init_idle_bootup_task(current);
 415         schedule_preempt_disabled();
 416         /* Call into cpu_idle with preempt disabled */
 417         cpu_startup_entry(CPUHP_ONLINE);
 418 }
 419
 420 /* Check for early params. */
 421 static int __init do_early_param(char *param, char *val, const char *unused)
 422 {
 423         const struct obs_kernel_param *p;
 424
 425         for (p = __setup_start; p < __setup_end; p++) {
 426                 if ((p->early && parameq(param, p->str)) ||
 427                     (strcmp(param, "console") == 0 &&
 428                      strcmp(p->str, "earlycon") == 0)
 429                 ) {
 430                         if (p->setup_func(val) != 0)
 431                                 pr_warn("Malformed early option '%s'\n", param);
 432                 }
 433         }
 434         /* We accept everything at this stage. */
 435         return 0;
 436 }
 437
 438 void __init parse_early_options(char *cmdline)
 439 {
 440         parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
 441 }
 442
 443 /* Arch code calls this early on, or if not, just before other parsing. */
 444 void __init parse_early_param(void)
 445 {
 446         static int done __initdata;
 447         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 448
 449         if (done)
 450                 return;
 451
 452         /* All fall through to do_early_param. */
 453         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 454         parse_early_options(tmp_cmdline);
 455         done = 1;
 456 }
 457
 458 /*
 459  *      Activate the first processor.
 460  */
 461
 462 static void __init boot_cpu_init(void)
 463 {
 464         int cpu = smp_processor_id();
 465         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 466         set_cpu_online(cpu, true);
 467         set_cpu_active(cpu, true);
 468         set_cpu_present(cpu, true);
 469         set_cpu_possible(cpu, true);
 470 }
 471
 472 void __init __weak smp_setup_processor_id(void)
 473 {
 474 }
 475
 476 # if THREAD_SIZE >= PAGE_SIZE
 477 void __init __weak thread_info_cache_init(void)
 478 {
 479 }
 480 #endif
 481
 482 /*
 483  * Set up kernel memory allocators
 484  */
 485 static void __init mm_init(void)
 486 {
 487         mem_init();
 488         kmem_cache_init();
 489         percpu_init_late();
 490         pgtable_init();
 491         vmalloc_init();
 492 }
 493
 494 asmlinkage __visible void __init start_kernel(void)
 495 {
 496         char *command_line;
 497         char *after_dashes;
 498
 499         /*
 500          * Need to run as early as possible, to initialize the
 501          * lockdep hash:
 502          */
 503         lockdep_init();
 504         set_task_stack_end_magic(&init_task);
 505         smp_setup_processor_id();
 506         debug_objects_early_init();
 507
 508         /*
 509          * Set up the the initial canary ASAP:
 510          */
 511         boot_init_stack_canary();
 512
 513         cgroup_init_early();
 514
 515         local_irq_disable();
 516         early_boot_irqs_disabled = true;
 517
 518 /*
 519  * Interrupts are still disabled. Do necessary setups, then
 520  * enable them
 521  */
 522         boot_cpu_init();
 523         page_address_init();
 524         pr_notice("%s", linux_banner);
 525         setup_arch(&command_line);
 526         mm_init_cpumask(&init_mm);
 527         setup_command_line(command_line);
 528         setup_nr_cpu_ids();
 529         setup_per_cpu_areas();
 530         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 531
 532         build_all_zonelists(NULL, NULL);
 533         page_alloc_init();
 534
 535         pr_notice("Kernel command line: %s\n", boot_command_line);
 536         parse_early_param();
 537         after_dashes = parse_args("Booting kernel",
 538                                   static_command_line, __start___param,
 539                                   __stop___param - __start___param,
 540                                   -1, -1, &unknown_bootoption);
 541         if (!IS_ERR_OR_NULL(after_dashes))
 542                 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
 543                            set_init_arg);
 544
 545         jump_label_init();
 546
 547         /*
 548          * These use large bootmem allocations and must precede
 549          * kmem_cache_init()
 550          */
 551         setup_log_buf(0);
 552         pidhash_init();
 553         vfs_caches_init_early();
 554         sort_main_extable();
 555         trap_init();
 556         mm_init();
 557
 558         /*
 559          * Set up the scheduler prior starting any interrupts (such as the
 560          * timer interrupt). Full topology setup happens at smp_init()
 561          * time - but meanwhile we still have a functioning scheduler.
 562          */
 563         sched_init();
 564         /*
 565          * Disable preemption - early bootup scheduling is extremely
 566          * fragile until we cpu_idle() for the first time.
 567          */
 568         preempt_disable();
 569         if (WARN(!irqs_disabled(),
 570                  "Interrupts were enabled *very* early, fixing it\n"))
 571                 local_irq_disable();
 572         idr_init_cache();
 573         rcu_init();
 574         context_tracking_init();
 575         radix_tree_init();
 576         /* init some links before init_ISA_irqs() */
 577         early_irq_init();
 578         init_IRQ();
 579         tick_init();
 580         rcu_init_nohz();
 581         init_timers();
 582         hrtimers_init();
 583         softirq_init();
 584         timekeeping_init();
 585         time_init();
 586         sched_clock_postinit();
 587         perf_event_init();
 588         profile_init();
 589         call_function_init();
 590         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 591         early_boot_irqs_disabled = false;
 592         local_irq_enable();
 593
 594         kmem_cache_init_late();
 595
 596         /*
 597          * HACK ALERT! This is early. We're enabling the console before
 598          * we've done PCI setups etc, and console_init() must be aware of
 599          * this. But we do want output early, in case something goes wrong.
 600          */
 601         console_init();
 602         if (panic_later)
 603                 panic("Too many boot %s vars at `%s'", panic_later,
 604                       panic_param);
 605
 606         lockdep_info();
 607
 608         /*
 609          * Need to run this when irqs are enabled, because it wants
 610          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 611          * too:
 612          */
 613         locking_selftest();
 614
 615 #ifdef CONFIG_BLK_DEV_INITRD
 616         if (initrd_start && !initrd_below_start_ok &&
 617             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 618                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
 619                     page_to_pfn(virt_to_page((void *)initrd_start)),
 620                     min_low_pfn);
 621                 initrd_start = 0;
 622         }
 623 #endif
 624         debug_objects_mem_init();
 625         kmemleak_init();
 626         setup_per_cpu_pageset();
 627         numa_policy_init();
 628         if (late_time_init)
 629                 late_time_init();
 630         sched_clock_init();
 631         calibrate_delay();
 632         pidmap_init();
 633         anon_vma_init();
 634         acpi_early_init();
 635 #ifdef CONFIG_X86
 636         if (efi_enabled(EFI_RUNTIME_SERVICES))
 637                 efi_enter_virtual_mode();
 638 #endif
 639 #ifdef CONFIG_X86_ESPFIX64
 640         /* Should be run before the first non-init thread is created */
 641         init_espfix_bsp();
 642 #endif
 643         thread_info_cache_init();
 644         cred_init();
 645         fork_init(totalram_pages);
 646         proc_caches_init();
 647         buffer_init();
 648         key_init();
 649         security_init();
 650         dbg_late_init();
 651         vfs_caches_init(totalram_pages);
 652         signals_init();
 653         /* rootfs populating might need page-writeback */
 654         page_writeback_init();
 655         proc_root_init();
 656         cgroup_init();
 657         cpuset_init();
 658         taskstats_init_early();
 659         delayacct_init();
 660
 661         check_bugs();
 662
 663         sfi_init_late();
 664
 665         if (efi_enabled(EFI_RUNTIME_SERVICES)) {
 666                 efi_late_init();
 667                 efi_free_boot_services();
 668         }
 669
 670         ftrace_init();
 671
 672         /* Do the rest non-__init'ed, we're now alive */
 673         rest_init();
 674 }
 675
 676 /* Call all constructor functions linked into the kernel. */
 677 static void __init do_ctors(void)
 678 {
 679 #ifdef CONFIG_CONSTRUCTORS
 680         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 681
 682         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 683                 (*fn)();
 684 #endif
 685 }
 686
 687 bool initcall_debug;
 688 core_param(initcall_debug, initcall_debug, bool, 0644);
 689
 690 #ifdef CONFIG_KALLSYMS
 691 struct blacklist_entry {
 692         struct list_head next;
 693         char *buf;
 694 };
 695
 696 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
 697
 698 static int __init initcall_blacklist(char *str)
 699 {
 700         char *str_entry;
 701         struct blacklist_entry *entry;
 702
 703         /* str argument is a comma-separated list of functions */
 704         do {
 705                 str_entry = strsep(&str, ",");
 706                 if (str_entry) {
 707                         pr_debug("blacklisting initcall %s\n", str_entry);
 708                         entry = alloc_bootmem(sizeof(*entry));
 709                         entry->buf = alloc_bootmem(strlen(str_entry) + 1);
 710                         strcpy(entry->buf, str_entry);
 711                         list_add(&entry->next, &blacklisted_initcalls);
 712                 }
 713         } while (str_entry);
 714
 715         return 0;
 716 }
 717
 718 static bool __init_or_module initcall_blacklisted(initcall_t fn)
 719 {
 720         struct list_head *tmp;
 721         struct blacklist_entry *entry;
 722         char *fn_name;
 723
 724         fn_name = kasprintf(GFP_KERNEL, "%pf", fn);
 725         if (!fn_name)
 726                 return false;
 727
 728         list_for_each(tmp, &blacklisted_initcalls) {
 729                 entry = list_entry(tmp, struct blacklist_entry, next);
 730                 if (!strcmp(fn_name, entry->buf)) {
 731                         pr_debug("initcall %s blacklisted\n", fn_name);
 732                         kfree(fn_name);
 733                         return true;
 734                 }
 735         }
 736
 737         kfree(fn_name);
 738         return false;
 739 }
 740 #else
 741 static int __init initcall_blacklist(char *str)
 742 {
 743         pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
 744         return 0;
 745 }
 746
 747 static bool __init_or_module initcall_blacklisted(initcall_t fn)
 748 {
 749         return false;
 750 }
 751 #endif
 752 __setup("initcall_blacklist=", initcall_blacklist);
 753
 754 static int __init_or_module do_one_initcall_debug(initcall_t fn)
 755 {
 756         ktime_t calltime, delta, rettime;
 757         unsigned long long duration;
 758         int ret;
 759
 760         printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
 761         calltime = ktime_get();
 762         ret = fn();
 763         rettime = ktime_get();
 764         delta = ktime_sub(rettime, calltime);
 765         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 766         printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
 767                  fn, ret, duration);
 768
 769         return ret;
 770 }
 771
 772 int __init_or_module do_one_initcall(initcall_t fn)
 773 {
 774         int count = preempt_count();
 775         int ret;
 776         char msgbuf[64];
 777
 778         if (initcall_blacklisted(fn))
 779                 return -EPERM;
 780
 781         if (initcall_debug)
 782                 ret = do_one_initcall_debug(fn);
 783         else
 784                 ret = fn();
 785
 786         msgbuf[0] = 0;
 787
 788         if (preempt_count() != count) {
 789                 sprintf(msgbuf, "preemption imbalance ");
 790                 preempt_count_set(count);
 791         }
 792         if (irqs_disabled()) {
 793                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 794                 local_irq_enable();
 795         }
 796         WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
 797
 798         return ret;
 799 }
 800
 801
 802 extern initcall_t __initcall_start[];
 803 extern initcall_t __initcall0_start[];
 804 extern initcall_t __initcall1_start[];
 805 extern initcall_t __initcall2_start[];
 806 extern initcall_t __initcall3_start[];
 807 extern initcall_t __initcall4_start[];
 808 extern initcall_t __initcall5_start[];
 809 extern initcall_t __initcall6_start[];
 810 extern initcall_t __initcall7_start[];
 811 extern initcall_t __initcall_end[];
 812
 813 static initcall_t *initcall_levels[] __initdata = {
 814         __initcall0_start,
 815         __initcall1_start,
 816         __initcall2_start,
 817         __initcall3_start,
 818         __initcall4_start,
 819         __initcall5_start,
 820         __initcall6_start,
 821         __initcall7_start,
 822         __initcall_end,
 823 };
 824
 825 /* Keep these in sync with initcalls in include/linux/init.h */
 826 static char *initcall_level_names[] __initdata = {
 827         "early",
 828         "core",
 829         "postcore",
 830         "arch",
 831         "subsys",
 832         "fs",
 833         "device",
 834         "late",
 835 };
 836
 837 static void __init do_initcall_level(int level)
 838 {
 839         initcall_t *fn;
 840
 841         strcpy(initcall_command_line, saved_command_line);
 842         parse_args(initcall_level_names[level],
 843                    initcall_command_line, __start___param,
 844                    __stop___param - __start___param,
 845                    level, level,
 846                    &repair_env_string);
 847
 848         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
 849                 do_one_initcall(*fn);
 850 }
 851
 852 static void __init do_initcalls(void)
 853 {
 854         int level;
 855
 856         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
 857                 do_initcall_level(level);
 858 }
 859
 860 /*
 861  * Ok, the machine is now initialized. None of the devices
 862  * have been touched yet, but the CPU subsystem is up and
 863  * running, and memory and process management works.
 864  *
 865  * Now we can finally start doing some real work..
 866  */
 867 static void __init do_basic_setup(void)
 868 {
 869         cpuset_init_smp();
 870         usermodehelper_init();
 871         shmem_init();
 872         driver_init();
 873         init_irq_proc();
 874         do_ctors();
 875         usermodehelper_enable();
 876         do_initcalls();
 877         random_int_secret_init();
 878 }
 879
 880 static void __init do_pre_smp_initcalls(void)
 881 {
 882         initcall_t *fn;
 883
 884         for (fn = __initcall_start; fn < __initcall0_start; fn++)
 885                 do_one_initcall(*fn);
 886 }
 887
 888 /*
 889  * This function requests modules which should be loaded by default and is
 890  * called twice right after initrd is mounted and right before init is
 891  * exec'd.  If such modules are on either initrd or rootfs, they will be
 892  * loaded before control is passed to userland.
 893  */
 894 void __init load_default_modules(void)
 895 {
 896         load_default_elevator_module();
 897 }
 898
 899 static int run_init_process(const char *init_filename)
 900 {
 901         argv_init[0] = init_filename;
 902         return do_execve(getname_kernel(init_filename),
 903                 (const char __user *const __user *)argv_init,
 904                 (const char __user *const __user *)envp_init);
 905 }
 906
 907 static int try_to_run_init_process(const char *init_filename)
 908 {
 909         int ret;
 910
 911         ret = run_init_process(init_filename);
 912
 913         if (ret && ret != -ENOENT) {
 914                 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
 915                        init_filename, ret);
 916         }
 917
 918         return ret;
 919 }
 920
 921 static noinline void __init kernel_init_freeable(void);
 922
 923 static int __ref kernel_init(void *unused)
 924 {
 925         int ret;
 926
 927         kernel_init_freeable();
 928         /* need to finish all async __init code before freeing the memory */
 929         async_synchronize_full();
 930         free_initmem();
 931         mark_rodata_ro();
 932         system_state = SYSTEM_RUNNING;
 933         numa_default_policy();
 934
 935         flush_delayed_fput();
 936
 937         if (ramdisk_execute_command) {
 938                 ret = run_init_process(ramdisk_execute_command);
 939                 if (!ret)
 940                         return 0;
 941                 pr_err("Failed to execute %s (error %d)\n",
 942                        ramdisk_execute_command, ret);
 943         }
 944
 945         /*
 946          * We try each of these until one succeeds.
 947          *
 948          * The Bourne shell can be used instead of init if we are
 949          * trying to recover a really broken machine.
 950          */
 951         if (execute_command) {
 952                 ret = run_init_process(execute_command);
 953                 if (!ret)
 954                         return 0;
 955                 pr_err("Failed to execute %s (error %d).  Attempting defaults...\n",
 956                         execute_command, ret);
 957         }
 958         if (!try_to_run_init_process("/sbin/init") ||
 959             !try_to_run_init_process("/etc/init") ||
 960             !try_to_run_init_process("/bin/init") ||
 961             !try_to_run_init_process("/bin/sh"))
 962                 return 0;
 963
 964         panic("No working init found.  Try passing init= option to kernel. "
 965               "See Linux Documentation/init.txt for guidance.");
 966 }
 967
 968 static noinline void __init kernel_init_freeable(void)
 969 {
 970         /*
 971          * Wait until kthreadd is all set-up.
 972          */
 973         wait_for_completion(&kthreadd_done);
 974
 975         /* Now the scheduler is fully set up and can do blocking allocations */
 976         gfp_allowed_mask = __GFP_BITS_MASK;
 977
 978         /*
 979          * init can allocate pages on any node
 980          */
 981         set_mems_allowed(node_states[N_MEMORY]);
 982         /*
 983          * init can run on any cpu.
 984          */
 985         set_cpus_allowed_ptr(current, cpu_all_mask);
 986
 987         cad_pid = task_pid(current);
 988
 989         smp_prepare_cpus(setup_max_cpus);
 990
 991         do_pre_smp_initcalls();
 992         lockup_detector_init();
 993
 994         smp_init();
 995         sched_init_smp();
 996
 997         do_basic_setup();
 998
 999         /* Open the /dev/console on the rootfs, this should never fail */
1000         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1001                 pr_err("Warning: unable to open an initial console.\n");
1002
1003         (void) sys_dup(0);
1004         (void) sys_dup(0);
1005         /*
1006          * check if there is an early userspace init.  If yes, let it do all
1007          * the work
1008          */
1009
1010         if (!ramdisk_execute_command)
1011                 ramdisk_execute_command = "/init";
1012
1013         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1014                 ramdisk_execute_command = NULL;
1015                 prepare_namespace();
1016         }
1017
1018         /*
1019          * Ok, we have completed the initial bootup, and
1020          * we're essentially up and running. Get rid of the
1021          * initmem segments and start the user-mode stuff..
1022          */
1023
1024         /* rootfs is available now, try loading default modules */
1025         load_default_modules();
1026 }