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/random.h>
  79 #include <linux/time_log.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 #include <linux/bootprof.h>
  87
  88 #ifdef CONFIG_X86_LOCAL_APIC
  89 #include <asm/smp.h>
  90 #endif
  91
  92 static int kernel_init(void *);
  93
  94 extern void init_IRQ(void);
  95 extern void fork_init(unsigned long);
  96 extern void mca_init(void);
  97 extern void sbus_init(void);
  98 extern void radix_tree_init(void);
  99 #ifndef CONFIG_DEBUG_RODATA
 100 static inline void mark_rodata_ro(void) { }
 101 #endif
 102
 103 #ifdef CONFIG_TC
 104 extern void tc_init(void);
 105 #endif
 106
 107 /*
 108  * Debug helper: via this flag we know that we are in 'early bootup code'
 109  * where only the boot processor is running with IRQ disabled.  This means
 110  * two things - IRQ must not be enabled before the flag is cleared and some
 111  * operations which are not allowed with IRQ disabled are allowed while the
 112  * flag is set.
 113  */
 114 bool early_boot_irqs_disabled __read_mostly;
 115
 116 enum system_states system_state __read_mostly;
 117 EXPORT_SYMBOL(system_state);
 118
 119 /*
 120  * Boot command-line arguments
 121  */
 122 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 123 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 124
 125 extern void time_init(void);
 126 /* Default late time init is NULL. archs can override this later. */
 127 void (*__initdata late_time_init)(void);
 128 extern void softirq_init(void);
 129
 130 /* Untouched command line saved by arch-specific code. */
 131 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 132 /* Untouched saved command line (eg. for /proc) */
 133 char *saved_command_line;
 134 EXPORT_SYMBOL_GPL(saved_command_line);
 135 /* Command line for parameter parsing */
 136 static char *static_command_line;
 137
 138 static char *execute_command;
 139 static char *ramdisk_execute_command;
 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
 202 EXPORT_SYMBOL(loops_per_jiffy);
 203
 204 static int __init debug_kernel(char *str)
 205 {
 206         console_loglevel = 10;
 207         return 0;
 208 }
 209
 210 static int __init quiet_kernel(char *str)
 211 {
 212         console_loglevel = 4;
 213         return 0;
 214 }
 215
 216 early_param("debug", debug_kernel);
 217 early_param("quiet", quiet_kernel);
 218
 219 static int __init loglevel(char *str)
 220 {
 221         int newlevel;
 222
 223         /*
 224          * Only update loglevel value when a correct setting was passed,
 225          * to prevent blind crashes (when loglevel being set to 0) that
 226          * are quite hard to debug
 227          */
 228         if (get_option(&str, &newlevel)) {
 229                 console_loglevel = newlevel;
 230                 return 0;
 231         }
 232
 233         return -EINVAL;
 234 }
 235
 236 early_param("loglevel", loglevel);
 237
 238 /* Change NUL term back to "=", to make "param" the whole string. */
 239 static int __init repair_env_string(char *param, char *val, const char *unused)
 240 {
 241         if (val) {
 242                 /* param=val or param="val"? */
 243                 if (val == param+strlen(param)+1)
 244                         val[-1] = '=';
 245                 else if (val == param+strlen(param)+2) {
 246                         val[-2] = '=';
 247                         memmove(val-1, val, strlen(val)+1);
 248                         val--;
 249                 } else
 250                         BUG();
 251         }
 252         return 0;
 253 }
 254
 255 /*
 256  * Unknown boot options get handed to init, unless they look like
 257  * unused parameters (modprobe will find them in /proc/cmdline).
 258  */
 259 static int __init unknown_bootoption(char *param, char *val, const char *unused)
 260 {
 261         repair_env_string(param, val, unused);
 262
 263         /* Handle obsolete-style parameters */
 264         if (obsolete_checksetup(param))
 265                 return 0;
 266
 267         /* Unused module parameter. */
 268         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 269                 return 0;
 270
 271         if (panic_later)
 272                 return 0;
 273
 274         if (val) {
 275                 /* Environment option */
 276                 unsigned int i;
 277                 for (i = 0; envp_init[i]; i++) {
 278                         if (i == MAX_INIT_ENVS) {
 279                                 panic_later = "Too many boot env vars at `%s'";
 280                                 panic_param = param;
 281                         }
 282                         if (!strncmp(param, envp_init[i], val - param))
 283                                 break;
 284                 }
 285                 envp_init[i] = param;
 286         } else {
 287                 /* Command line option */
 288                 unsigned int i;
 289                 for (i = 0; argv_init[i]; i++) {
 290                         if (i == MAX_INIT_ARGS) {
 291                                 panic_later = "Too many boot init vars at `%s'";
 292                                 panic_param = param;
 293                         }
 294                 }
 295                 argv_init[i] = param;
 296         }
 297         return 0;
 298 }
 299
 300 static int __init init_setup(char *str)
 301 {
 302         unsigned int i;
 303
 304         execute_command = str;
 305         /*
 306          * In case LILO is going to boot us with default command line,
 307          * it prepends "auto" before the whole cmdline which makes
 308          * the shell think it should execute a script with such name.
 309          * So we ignore all arguments entered _before_ init=... [MJ]
 310          */
 311         for (i = 1; i < MAX_INIT_ARGS; i++)
 312                 argv_init[i] = NULL;
 313         return 1;
 314 }
 315 __setup("init=", init_setup);
 316
 317 static int __init rdinit_setup(char *str)
 318 {
 319         unsigned int i;
 320
 321         ramdisk_execute_command = str;
 322         /* See "auto" comment in init_setup */
 323         for (i = 1; i < MAX_INIT_ARGS; i++)
 324                 argv_init[i] = NULL;
 325         return 1;
 326 }
 327 __setup("rdinit=", rdinit_setup);
 328
 329 #ifndef CONFIG_SMP
 330 static const unsigned int setup_max_cpus = NR_CPUS;
 331 #ifdef CONFIG_X86_LOCAL_APIC
 332 static void __init smp_init(void)
 333 {
 334         APIC_init_uniprocessor();
 335 }
 336 #else
 337 #define smp_init()      do { } while (0)
 338 #endif
 339
 340 static inline void setup_nr_cpu_ids(void) { }
 341 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 342 #endif
 343
 344 /*
 345  * We need to store the untouched command line for future reference.
 346  * We also need to store the touched command line since the parameter
 347  * parsing is performed in place, and we should allow a component to
 348  * store reference of name/value for future reference.
 349  */
 350 static void __init setup_command_line(char *command_line)
 351 {
 352         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 353         static_command_line = alloc_bootmem(strlen (command_line)+1);
 354         strcpy (saved_command_line, boot_command_line);
 355         strcpy (static_command_line, command_line);
 356 }
 357
 358 /*
 359  * We need to finalize in a non-__init function or else race conditions
 360  * between the root thread and the init thread may cause start_kernel to
 361  * be reaped by free_initmem before the root thread has proceeded to
 362  * cpu_idle.
 363  *
 364  * gcc-3.4 accidentally inlines this function, so use noinline.
 365  */
 366
 367 static __initdata DECLARE_COMPLETION(kthreadd_done);
 368
 369 static noinline void __init_refok rest_init(void)
 370 {
 371         int pid;
 372
 373         rcu_scheduler_starting();
 374         /*
 375          * We need to spawn init first so that it obtains pid 1, however
 376          * the init task will end up wanting to create kthreads, which, if
 377          * we schedule it before we create kthreadd, will OOPS.
 378          */
 379         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 380         numa_default_policy();
 381         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 382         rcu_read_lock();
 383         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 384         rcu_read_unlock();
 385         complete(&kthreadd_done);
 386
 387         /*
 388          * The boot idle thread must execute schedule()
 389          * at least once to get things moving:
 390          */
 391         init_idle_bootup_task(current);
 392         schedule_preempt_disabled();
 393         /* Call into cpu_idle with preempt disabled */
 394         cpu_startup_entry(CPUHP_ONLINE);
 395 }
 396
 397 /* Check for early params. */
 398 static int __init do_early_param(char *param, char *val, const char *unused)
 399 {
 400         const struct obs_kernel_param *p;
 401
 402         for (p = __setup_start; p < __setup_end; p++) {
 403                 if ((p->early && parameq(param, p->str)) ||
 404                     (strcmp(param, "console") == 0 &&
 405                      strcmp(p->str, "earlycon") == 0)
 406                 ) {
 407                         if (p->setup_func(val) != 0)
 408                                 pr_warn("Malformed early option '%s'\n", param);
 409                 }
 410         }
 411         /* We accept everything at this stage. */
 412         return 0;
 413 }
 414
 415 void __init parse_early_options(char *cmdline)
 416 {
 417         parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
 418 }
 419
 420 /* Arch code calls this early on, or if not, just before other parsing. */
 421 void __init parse_early_param(void)
 422 {
 423         static __initdata int done = 0;
 424         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 425
 426         if (done)
 427                 return;
 428
 429         /* All fall through to do_early_param. */
 430         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 431         parse_early_options(tmp_cmdline);
 432         done = 1;
 433 }
 434
 435 /*
 436  *      Activate the first processor.
 437  */
 438
 439 static void __init boot_cpu_init(void)
 440 {
 441         int cpu = smp_processor_id();
 442         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 443         set_cpu_online(cpu, true);
 444         set_cpu_active(cpu, true);
 445         set_cpu_present(cpu, true);
 446         set_cpu_possible(cpu, true);
 447 }
 448
 449 void __init __weak smp_setup_processor_id(void)
 450 {
 451 }
 452
 453 # if THREAD_SIZE >= PAGE_SIZE
 454 void __init __weak thread_info_cache_init(void)
 455 {
 456 }
 457 #endif
 458
 459 /*
 460  * Set up kernel memory allocators
 461  */
 462 static void __init mm_init(void)
 463 {
 464         /*
 465          * page_cgroup requires contiguous pages,
 466          * bigger than MAX_ORDER unless SPARSEMEM.
 467          */
 468         page_cgroup_init_flatmem();
 469         mem_init();
 470         kmem_cache_init();
 471         percpu_init_late();
 472         pgtable_cache_init();
 473         vmalloc_init();
 474 }
 475
 476 asmlinkage void __init start_kernel(void)
 477 {
 478         char * command_line;
 479         extern const struct kernel_param __start___param[], __stop___param[];
 480
 481         /*
 482          * Need to run as early as possible, to initialize the
 483          * lockdep hash:
 484          */
 485         lockdep_init();
 486         smp_setup_processor_id();
 487         debug_objects_early_init();
 488
 489         /*
 490          * Set up the the initial canary ASAP:
 491          */
 492         boot_init_stack_canary();
 493
 494         cgroup_init_early();
 495
 496         local_irq_disable();
 497         early_boot_irqs_disabled = true;
 498
 499 /*
 500  * Interrupts are still disabled. Do necessary setups, then
 501  * enable them
 502  */
 503         boot_cpu_init();
 504         page_address_init();
 505         pr_notice("%s", linux_banner);
 506         setup_arch(&command_line);
 507         mm_init_owner(&init_mm, &init_task);
 508         mm_init_cpumask(&init_mm);
 509         setup_command_line(command_line);
 510         setup_nr_cpu_ids();
 511         setup_per_cpu_areas();
 512         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 513
 514         build_all_zonelists(NULL, NULL);
 515         page_alloc_init();
 516
 517         pr_notice("Kernel command line: %s\n", boot_command_line);
 518         parse_early_param();
 519         parse_args("Booting kernel", static_command_line, __start___param,
 520                    __stop___param - __start___param,
 521                    -1, -1, &unknown_bootoption);
 522
 523         jump_label_init();
 524
 525         /*
 526          * These use large bootmem allocations and must precede
 527          * kmem_cache_init()
 528          */
 529         setup_log_buf(0);
 530         pidhash_init();
 531         vfs_caches_init_early();
 532         sort_main_extable();
 533         trap_init();
 534         mm_init();
 535
 536         /*
 537          * Set up the scheduler prior starting any interrupts (such as the
 538          * timer interrupt). Full topology setup happens at smp_init()
 539          * time - but meanwhile we still have a functioning scheduler.
 540          */
 541         sched_init();
 542         /*
 543          * Disable preemption - early bootup scheduling is extremely
 544          * fragile until we cpu_idle() for the first time.
 545          */
 546         preempt_disable();
 547         if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
 548                 local_irq_disable();
 549         idr_init_cache();
 550         perf_event_init();
 551         rcu_init();
 552         tick_nohz_init();
 553         radix_tree_init();
 554         /* init some links before init_ISA_irqs() */
 555         early_irq_init();
 556         init_IRQ();
 557         tick_init();
 558         init_timers();
 559         hrtimers_init();
 560         softirq_init();
 561         timekeeping_init();
 562         time_init();
 563         sched_clock_postinit();
 564         profile_init();
 565         call_function_init();
 566         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 567         early_boot_irqs_disabled = false;
 568         local_irq_enable();
 569
 570         kmem_cache_init_late();
 571
 572         /*
 573          * HACK ALERT! This is early. We're enabling the console before
 574          * we've done PCI setups etc, and console_init() must be aware of
 575          * this. But we do want output early, in case something goes wrong.
 576          */
 577         console_init();
 578         if (panic_later)
 579                 panic(panic_later, panic_param);
 580
 581         lockdep_info();
 582
 583         /*
 584          * Need to run this when irqs are enabled, because it wants
 585          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 586          * too:
 587          */
 588         locking_selftest();
 589
 590 #ifdef CONFIG_BLK_DEV_INITRD
 591         if (initrd_start && !initrd_below_start_ok &&
 592             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 593                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
 594                     page_to_pfn(virt_to_page((void *)initrd_start)),
 595                     min_low_pfn);
 596                 initrd_start = 0;
 597         }
 598 #endif
 599         page_cgroup_init();
 600         debug_objects_mem_init();
 601         kmemleak_init();
 602         setup_per_cpu_pageset();
 603         numa_policy_init();
 604         if (late_time_init)
 605                 late_time_init();
 606         sched_clock_init();
 607         calibrate_delay();
 608         pidmap_init();
 609         anon_vma_init();
 610 #ifdef CONFIG_X86
 611         if (efi_enabled(EFI_RUNTIME_SERVICES))
 612                 efi_enter_virtual_mode();
 613 #endif
 614 #ifdef CONFIG_X86_ESPFIX64
 615         /* Should be run before the first non-init thread is created */
 616         init_espfix_bsp();
 617 #endif
 618         thread_info_cache_init();
 619         cred_init();
 620         fork_init(totalram_pages);
 621         proc_caches_init();
 622         buffer_init();
 623         key_init();
 624         security_init();
 625         dbg_late_init();
 626         vfs_caches_init(totalram_pages);
 627         signals_init();
 628         /* rootfs populating might need page-writeback */
 629         page_writeback_init();
 630 #ifdef CONFIG_PROC_FS
 631         proc_root_init();
 632 #endif
 633         cgroup_init();
 634         cpuset_init();
 635         taskstats_init_early();
 636         delayacct_init();
 637
 638         check_bugs();
 639
 640         acpi_early_init(); /* before LAPIC and SMP init */
 641         sfi_init_late();
 642
 643         if (efi_enabled(EFI_RUNTIME_SERVICES)) {
 644                 efi_late_init();
 645                 efi_free_boot_services();
 646         }
 647
 648         ftrace_init();
 649
 650         /* Do the rest non-__init'ed, we're now alive */
 651         rest_init();
 652 }
 653
 654 /* Call all constructor functions linked into the kernel. */
 655 static void __init do_ctors(void)
 656 {
 657 #ifdef CONFIG_CONSTRUCTORS
 658         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 659
 660         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 661                 (*fn)();
 662 #endif
 663 }
 664
 665 bool initcall_debug;
 666 core_param(initcall_debug, initcall_debug, bool, 0644);
 667
 668 static char msgbuf[64];
 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         TIME_LOG_START();
 693         if (initcall_debug)
 694                 ret = do_one_initcall_debug(fn);
 695         else
 696                 ret = fn();
 697
 698         msgbuf[0] = 0;
 699
 700         if (preempt_count() != count) {
 701                 sprintf(msgbuf, "preemption imbalance ");
 702                 preempt_count() = count;
 703         }
 704         if (irqs_disabled()) {
 705                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 706                 local_irq_enable();
 707         }
 708         WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
 709
 710         TIME_LOG_END("[init] f(%p)\n", fn);
 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         log_boot("Kernel_init_done");
 836
 837         if (ramdisk_execute_command) {
 838                 if (!run_init_process(ramdisk_execute_command))
 839                         return 0;
 840                 pr_err("Failed to execute %s\n", ramdisk_execute_command);
 841         }
 842
 843         /*
 844          * We try each of these until one succeeds.
 845          *
 846          * The Bourne shell can be used instead of init if we are
 847          * trying to recover a really broken machine.
 848          */
 849         if (execute_command) {
 850                 if (!run_init_process(execute_command))
 851                         return 0;
 852                 pr_err("Failed to execute %s.  Attempting defaults...\n",
 853                         execute_command);
 854         }
 855         if (!run_init_process("/sbin/init") ||
 856             !run_init_process("/etc/init") ||
 857             !run_init_process("/bin/init") ||
 858             !run_init_process("/bin/sh"))
 859                 return 0;
 860
 861         panic("No init found.  Try passing init= option to kernel. "
 862               "See Linux Documentation/init.txt for guidance.");
 863 }
 864
 865
 866 #ifdef CONFIG_MTK_HIBERNATION
 867 // IPO-H, move here for console ok after hibernaton
 868 extern int software_resume(void);
 869 #endif
 870 static noinline void __init kernel_init_freeable(void)
 871 {
 872         /*
 873          * Wait until kthreadd is all set-up.
 874          */
 875         wait_for_completion(&kthreadd_done);
 876
 877         /* Now the scheduler is fully set up and can do blocking allocations */
 878         gfp_allowed_mask = __GFP_BITS_MASK;
 879
 880         /*
 881          * init can allocate pages on any node
 882          */
 883         set_mems_allowed(node_states[N_MEMORY]);
 884         /*
 885          * init can run on any cpu.
 886          */
 887         set_cpus_allowed_ptr(current, cpu_all_mask);
 888
 889         cad_pid = task_pid(current);
 890
 891         smp_prepare_cpus(setup_max_cpus);
 892
 893         do_pre_smp_initcalls();
 894         lockup_detector_init();
 895
 896         smp_init();
 897         sched_init_smp();
 898
 899         do_basic_setup();
 900
 901         /* Open the /dev/console on the rootfs, this should never fail */
 902         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 903                 pr_err("Warning: unable to open an initial console.\n");
 904
 905         (void) sys_dup(0);
 906         (void) sys_dup(0);
 907
 908 #ifdef CONFIG_MTK_HIBERNATION
 909 // IPO-H, move here for console ok after hibernaton resume
 910     software_resume();
 911 #endif
 912         /*
 913          * check if there is an early userspace init.  If yes, let it do all
 914          * the work
 915          */
 916
 917         if (!ramdisk_execute_command)
 918                 ramdisk_execute_command = "/init";
 919
 920         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 921                 ramdisk_execute_command = NULL;
 922                 prepare_namespace();
 923         }
 924
 925         /*
 926          * Ok, we have completed the initial bootup, and
 927          * we're essentially up and running. Get rid of the
 928          * initmem segments and start the user-mode stuff..
 929          */
 930
 931         /* rootfs is available now, try loading default modules */
 932         load_default_modules();
 933 }