init/main.c

   1 /*
   2  *  linux/init/main.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  *
   6  *  GK 2/5/95  -  Changed to support mounting root fs via NFS
   7  *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
   8  *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
   9  *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
  10  */
  11
  12 #include <linux/types.h>
  13 #include <linux/module.h>
  14 #include <linux/proc_fs.h>
  15 #include <linux/kernel.h>
  16 #include <linux/syscalls.h>
  17 #include <linux/string.h>
  18 #include <linux/ctype.h>
  19 #include <linux/delay.h>
  20 #include <linux/utsname.h>
  21 #include <linux/ioport.h>
  22 #include <linux/init.h>
  23 #include <linux/smp_lock.h>
  24 #include <linux/initrd.h>
  25 #include <linux/bootmem.h>
  26 #include <linux/tty.h>
  27 #include <linux/gfp.h>
  28 #include <linux/percpu.h>
  29 #include <linux/kmod.h>
  30 #include <linux/vmalloc.h>
  31 #include <linux/kernel_stat.h>
  32 #include <linux/start_kernel.h>
  33 #include <linux/security.h>
  34 #include <linux/smp.h>
  35 #include <linux/workqueue.h>
  36 #include <linux/profile.h>
  37 #include <linux/rcupdate.h>
  38 #include <linux/moduleparam.h>
  39 #include <linux/kallsyms.h>
  40 #include <linux/writeback.h>
  41 #include <linux/cpu.h>
  42 #include <linux/cpuset.h>
  43 #include <linux/cgroup.h>
  44 #include <linux/efi.h>
  45 #include <linux/tick.h>
  46 #include <linux/interrupt.h>
  47 #include <linux/taskstats_kern.h>
  48 #include <linux/delayacct.h>
  49 #include <linux/unistd.h>
  50 #include <linux/rmap.h>
  51 #include <linux/mempolicy.h>
  52 #include <linux/key.h>
  53 #include <linux/unwind.h>
  54 #include <linux/buffer_head.h>
  55 #include <linux/debug_locks.h>
  56 #include <linux/debugobjects.h>
  57 #include <linux/lockdep.h>
  58 #include <linux/pid_namespace.h>
  59 #include <linux/device.h>
  60 #include <linux/kthread.h>
  61 #include <linux/sched.h>
  62 #include <linux/signal.h>
  63 #include <linux/idr.h>
  64 #include <linux/ftrace.h>
  65
  66 #include <asm/io.h>
  67 #include <asm/bugs.h>
  68 #include <asm/setup.h>
  69 #include <asm/sections.h>
  70 #include <asm/cacheflush.h>
  71
  72 #ifdef CONFIG_X86_LOCAL_APIC
  73 #include <asm/smp.h>
  74 #endif
  75
  76 /*
  77  * This is one of the first .c files built. Error out early if we have compiler
  78  * trouble.
  79  */
  80
  81 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
  82 #warning gcc-4.1.0 is known to miscompile the kernel.  A different compiler version is recommended.
  83 #endif
  84
  85 static int kernel_init(void *);
  86
  87 extern void init_IRQ(void);
  88 extern void fork_init(unsigned long);
  89 extern void mca_init(void);
  90 extern void sbus_init(void);
  91 extern void prio_tree_init(void);
  92 extern void radix_tree_init(void);
  93 extern void free_initmem(void);
  94 #ifdef  CONFIG_ACPI
  95 extern void acpi_early_init(void);
  96 #else
  97 static inline void acpi_early_init(void) { }
  98 #endif
  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 enum system_states system_state;
 108 EXPORT_SYMBOL(system_state);
 109
 110 /*
 111  * Boot command-line arguments
 112  */
 113 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 114 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 115
 116 extern void time_init(void);
 117 /* Default late time init is NULL. archs can override this later. */
 118 void (*late_time_init)(void);
 119 extern void softirq_init(void);
 120
 121 /* Untouched command line saved by arch-specific code. */
 122 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 123 /* Untouched saved command line (eg. for /proc) */
 124 char *saved_command_line;
 125 /* Command line for parameter parsing */
 126 static char *static_command_line;
 127
 128 static char *execute_command;
 129 static char *ramdisk_execute_command;
 130
 131 #ifdef CONFIG_SMP
 132 /* Setup configured maximum number of CPUs to activate */
 133 unsigned int __initdata setup_max_cpus = NR_CPUS;
 134
 135 /*
 136  * Setup routine for controlling SMP activation
 137  *
 138  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 139  * activation entirely (the MPS table probe still happens, though).
 140  *
 141  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 142  * greater than 0, limits the maximum number of CPUs activated in
 143  * SMP mode to <NUM>.
 144  */
 145 #ifndef CONFIG_X86_IO_APIC
 146 static inline void disable_ioapic_setup(void) {};
 147 #endif
 148
 149 static int __init nosmp(char *str)
 150 {
 151         setup_max_cpus = 0;
 152         disable_ioapic_setup();
 153         return 0;
 154 }
 155
 156 early_param("nosmp", nosmp);
 157
 158 static int __init maxcpus(char *str)
 159 {
 160         get_option(&str, &setup_max_cpus);
 161         if (setup_max_cpus == 0)
 162                 disable_ioapic_setup();
 163
 164         return 0;
 165 }
 166
 167 early_param("maxcpus", maxcpus);
 168 #else
 169 #define setup_max_cpus NR_CPUS
 170 #endif
 171
 172 /*
 173  * If set, this is an indication to the drivers that reset the underlying
 174  * device before going ahead with the initialization otherwise driver might
 175  * rely on the BIOS and skip the reset operation.
 176  *
 177  * This is useful if kernel is booting in an unreliable environment.
 178  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 179  * skipped and devices will be in unknown state.
 180  */
 181 unsigned int reset_devices;
 182 EXPORT_SYMBOL(reset_devices);
 183
 184 static int __init set_reset_devices(char *str)
 185 {
 186         reset_devices = 1;
 187         return 1;
 188 }
 189
 190 __setup("reset_devices", set_reset_devices);
 191
 192 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 193 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 194 static const char *panic_later, *panic_param;
 195
 196 extern struct obs_kernel_param __setup_start[], __setup_end[];
 197
 198 static int __init obsolete_checksetup(char *line)
 199 {
 200         struct obs_kernel_param *p;
 201         int had_early_param = 0;
 202
 203         p = __setup_start;
 204         do {
 205                 int n = strlen(p->str);
 206                 if (!strncmp(line, p->str, n)) {
 207                         if (p->early) {
 208                                 /* Already done in parse_early_param?
 209                                  * (Needs exact match on param part).
 210                                  * Keep iterating, as we can have early
 211                                  * params and __setups of same names 8( */
 212                                 if (line[n] == '\0' || line[n] == '=')
 213                                         had_early_param = 1;
 214                         } else if (!p->setup_func) {
 215                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 216                                        " ignored\n", p->str);
 217                                 return 1;
 218                         } else if (p->setup_func(line + n))
 219                                 return 1;
 220                 }
 221                 p++;
 222         } while (p < __setup_end);
 223
 224         return had_early_param;
 225 }
 226
 227 /*
 228  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 229  * still work even if initially too large, it will just take slightly longer
 230  */
 231 unsigned long loops_per_jiffy = (1<<12);
 232
 233 EXPORT_SYMBOL(loops_per_jiffy);
 234
 235 static int __init debug_kernel(char *str)
 236 {
 237         console_loglevel = 10;
 238         return 0;
 239 }
 240
 241 static int __init quiet_kernel(char *str)
 242 {
 243         console_loglevel = 4;
 244         return 0;
 245 }
 246
 247 early_param("debug", debug_kernel);
 248 early_param("quiet", quiet_kernel);
 249
 250 static int __init loglevel(char *str)
 251 {
 252         get_option(&str, &console_loglevel);
 253         return 0;
 254 }
 255
 256 early_param("loglevel", loglevel);
 257
 258 /*
 259  * Unknown boot options get handed to init, unless they look like
 260  * failed parameters
 261  */
 262 static int __init unknown_bootoption(char *param, char *val)
 263 {
 264         /* Change NUL term back to "=", to make "param" the whole string. */
 265         if (val) {
 266                 /* param=val or param="val"? */
 267                 if (val == param+strlen(param)+1)
 268                         val[-1] = '=';
 269                 else if (val == param+strlen(param)+2) {
 270                         val[-2] = '=';
 271                         memmove(val-1, val, strlen(val)+1);
 272                         val--;
 273                 } else
 274                         BUG();
 275         }
 276
 277         /* Handle obsolete-style parameters */
 278         if (obsolete_checksetup(param))
 279                 return 0;
 280
 281         /*
 282          * Preemptive maintenance for "why didn't my misspelled command
 283          * line work?"
 284          */
 285         if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
 286                 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
 287                 return 0;
 288         }
 289
 290         if (panic_later)
 291                 return 0;
 292
 293         if (val) {
 294                 /* Environment option */
 295                 unsigned int i;
 296                 for (i = 0; envp_init[i]; i++) {
 297                         if (i == MAX_INIT_ENVS) {
 298                                 panic_later = "Too many boot env vars at `%s'";
 299                                 panic_param = param;
 300                         }
 301                         if (!strncmp(param, envp_init[i], val - param))
 302                                 break;
 303                 }
 304                 envp_init[i] = param;
 305         } else {
 306                 /* Command line option */
 307                 unsigned int i;
 308                 for (i = 0; argv_init[i]; i++) {
 309                         if (i == MAX_INIT_ARGS) {
 310                                 panic_later = "Too many boot init vars at `%s'";
 311                                 panic_param = param;
 312                         }
 313                 }
 314                 argv_init[i] = param;
 315         }
 316         return 0;
 317 }
 318
 319 #ifdef CONFIG_DEBUG_PAGEALLOC
 320 int __read_mostly debug_pagealloc_enabled = 0;
 321 #endif
 322
 323 static int __init init_setup(char *str)
 324 {
 325         unsigned int i;
 326
 327         execute_command = str;
 328         /*
 329          * In case LILO is going to boot us with default command line,
 330          * it prepends "auto" before the whole cmdline which makes
 331          * the shell think it should execute a script with such name.
 332          * So we ignore all arguments entered _before_ init=... [MJ]
 333          */
 334         for (i = 1; i < MAX_INIT_ARGS; i++)
 335                 argv_init[i] = NULL;
 336         return 1;
 337 }
 338 __setup("init=", init_setup);
 339
 340 static int __init rdinit_setup(char *str)
 341 {
 342         unsigned int i;
 343
 344         ramdisk_execute_command = str;
 345         /* See "auto" comment in init_setup */
 346         for (i = 1; i < MAX_INIT_ARGS; i++)
 347                 argv_init[i] = NULL;
 348         return 1;
 349 }
 350 __setup("rdinit=", rdinit_setup);
 351
 352 #ifndef CONFIG_SMP
 353
 354 #ifdef CONFIG_X86_LOCAL_APIC
 355 static void __init smp_init(void)
 356 {
 357         APIC_init_uniprocessor();
 358 }
 359 #else
 360 #define smp_init()      do { } while (0)
 361 #endif
 362
 363 static inline void setup_per_cpu_areas(void) { }
 364 static inline void setup_nr_cpu_ids(void) { }
 365 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 366
 367 #else
 368
 369 #if NR_CPUS > BITS_PER_LONG
 370 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
 371 EXPORT_SYMBOL(cpu_mask_all);
 372 #endif
 373
 374 /* Setup number of possible processor ids */
 375 int nr_cpu_ids __read_mostly = NR_CPUS;
 376 EXPORT_SYMBOL(nr_cpu_ids);
 377
 378 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
 379 static void __init setup_nr_cpu_ids(void)
 380 {
 381         int cpu, highest_cpu = 0;
 382
 383         for_each_possible_cpu(cpu)
 384                 highest_cpu = cpu;
 385
 386         nr_cpu_ids = highest_cpu + 1;
 387 }
 388
 389 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
 390 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
 391
 392 EXPORT_SYMBOL(__per_cpu_offset);
 393
 394 static void __init setup_per_cpu_areas(void)
 395 {
 396         unsigned long size, i;
 397         char *ptr;
 398         unsigned long nr_possible_cpus = num_possible_cpus();
 399
 400         /* Copy section for each CPU (we discard the original) */
 401         size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
 402         ptr = alloc_bootmem_pages(size * nr_possible_cpus);
 403
 404         for_each_possible_cpu(i) {
 405                 __per_cpu_offset[i] = ptr - __per_cpu_start;
 406                 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 407                 ptr += size;
 408         }
 409 }
 410 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
 411
 412 /* Called by boot processor to activate the rest. */
 413 static void __init smp_init(void)
 414 {
 415         unsigned int cpu;
 416
 417         /*
 418          * Set up the current CPU as possible to migrate to.
 419          * The other ones will be done by cpu_up/cpu_down()
 420          */
 421         cpu = smp_processor_id();
 422         cpu_set(cpu, cpu_active_map);
 423
 424         /* FIXME: This should be done in userspace --RR */
 425         for_each_present_cpu(cpu) {
 426                 if (num_online_cpus() >= setup_max_cpus)
 427                         break;
 428                 if (!cpu_online(cpu))
 429                         cpu_up(cpu);
 430         }
 431
 432         /* Any cleanup work */
 433         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 434         smp_cpus_done(setup_max_cpus);
 435 }
 436
 437 #endif
 438
 439 /*
 440  * We need to store the untouched command line for future reference.
 441  * We also need to store the touched command line since the parameter
 442  * parsing is performed in place, and we should allow a component to
 443  * store reference of name/value for future reference.
 444  */
 445 static void __init setup_command_line(char *command_line)
 446 {
 447         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 448         static_command_line = alloc_bootmem(strlen (command_line)+1);
 449         strcpy (saved_command_line, boot_command_line);
 450         strcpy (static_command_line, command_line);
 451 }
 452
 453 /*
 454  * We need to finalize in a non-__init function or else race conditions
 455  * between the root thread and the init thread may cause start_kernel to
 456  * be reaped by free_initmem before the root thread has proceeded to
 457  * cpu_idle.
 458  *
 459  * gcc-3.4 accidentally inlines this function, so use noinline.
 460  */
 461
 462 static void noinline __init_refok rest_init(void)
 463         __releases(kernel_lock)
 464 {
 465         int pid;
 466
 467         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 468         numa_default_policy();
 469         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 470         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 471         unlock_kernel();
 472
 473         /*
 474          * The boot idle thread must execute schedule()
 475          * at least once to get things moving:
 476          */
 477         init_idle_bootup_task(current);
 478         preempt_enable_no_resched();
 479         schedule();
 480         preempt_disable();
 481
 482         /* Call into cpu_idle with preempt disabled */
 483         cpu_idle();
 484 }
 485
 486 /* Check for early params. */
 487 static int __init do_early_param(char *param, char *val)
 488 {
 489         struct obs_kernel_param *p;
 490
 491         for (p = __setup_start; p < __setup_end; p++) {
 492                 if ((p->early && strcmp(param, p->str) == 0) ||
 493                     (strcmp(param, "console") == 0 &&
 494                      strcmp(p->str, "earlycon") == 0)
 495                 ) {
 496                         if (p->setup_func(val) != 0)
 497                                 printk(KERN_WARNING
 498                                        "Malformed early option '%s'\n", param);
 499                 }
 500         }
 501         /* We accept everything at this stage. */
 502         return 0;
 503 }
 504
 505 /* Arch code calls this early on, or if not, just before other parsing. */
 506 void __init parse_early_param(void)
 507 {
 508         static __initdata int done = 0;
 509         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 510
 511         if (done)
 512                 return;
 513
 514         /* All fall through to do_early_param. */
 515         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 516         parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
 517         done = 1;
 518 }
 519
 520 /*
 521  *      Activate the first processor.
 522  */
 523
 524 static void __init boot_cpu_init(void)
 525 {
 526         int cpu = smp_processor_id();
 527         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 528         cpu_set(cpu, cpu_online_map);
 529         cpu_set(cpu, cpu_present_map);
 530         cpu_set(cpu, cpu_possible_map);
 531 }
 532
 533 void __init __weak smp_setup_processor_id(void)
 534 {
 535 }
 536
 537 void __init __weak thread_info_cache_init(void)
 538 {
 539 }
 540
 541 asmlinkage void __init start_kernel(void)
 542 {
 543         char * command_line;
 544         extern struct kernel_param __start___param[], __stop___param[];
 545
 546         smp_setup_processor_id();
 547
 548         /*
 549          * Need to run as early as possible, to initialize the
 550          * lockdep hash:
 551          */
 552         unwind_init();
 553         lockdep_init();
 554         debug_objects_early_init();
 555         cgroup_init_early();
 556
 557         local_irq_disable();
 558         early_boot_irqs_off();
 559         early_init_irq_lock_class();
 560
 561 /*
 562  * Interrupts are still disabled. Do necessary setups, then
 563  * enable them
 564  */
 565         lock_kernel();
 566         tick_init();
 567         boot_cpu_init();
 568         page_address_init();
 569         printk(KERN_NOTICE);
 570         printk(linux_banner);
 571         setup_arch(&command_line);
 572         mm_init_owner(&init_mm, &init_task);
 573         setup_command_line(command_line);
 574         unwind_setup();
 575         setup_per_cpu_areas();
 576         setup_nr_cpu_ids();
 577         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 578
 579         /*
 580          * Set up the scheduler prior starting any interrupts (such as the
 581          * timer interrupt). Full topology setup happens at smp_init()
 582          * time - but meanwhile we still have a functioning scheduler.
 583          */
 584         sched_init();
 585         /*
 586          * Disable preemption - early bootup scheduling is extremely
 587          * fragile until we cpu_idle() for the first time.
 588          */
 589         preempt_disable();
 590         build_all_zonelists();
 591         page_alloc_init();
 592         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 593         parse_early_param();
 594         parse_args("Booting kernel", static_command_line, __start___param,
 595                    __stop___param - __start___param,
 596                    &unknown_bootoption);
 597         if (!irqs_disabled()) {
 598                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 599                                 "enabled *very* early, fixing it\n");
 600                 local_irq_disable();
 601         }
 602         sort_main_extable();
 603         trap_init();
 604         rcu_init();
 605         init_IRQ();
 606         pidhash_init();
 607         init_timers();
 608         hrtimers_init();
 609         softirq_init();
 610         timekeeping_init();
 611         time_init();
 612         sched_clock_init();
 613         profile_init();
 614         if (!irqs_disabled())
 615                 printk("start_kernel(): bug: interrupts were enabled early\n");
 616         early_boot_irqs_on();
 617         local_irq_enable();
 618
 619         /*
 620          * HACK ALERT! This is early. We're enabling the console before
 621          * we've done PCI setups etc, and console_init() must be aware of
 622          * this. But we do want output early, in case something goes wrong.
 623          */
 624         console_init();
 625         if (panic_later)
 626                 panic(panic_later, panic_param);
 627
 628         lockdep_info();
 629
 630         /*
 631          * Need to run this when irqs are enabled, because it wants
 632          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 633          * too:
 634          */
 635         locking_selftest();
 636
 637 #ifdef CONFIG_BLK_DEV_INITRD
 638         if (initrd_start && !initrd_below_start_ok &&
 639             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 640                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 641                     "disabling it.\n",
 642                     page_to_pfn(virt_to_page((void *)initrd_start)),
 643                     min_low_pfn);
 644                 initrd_start = 0;
 645         }
 646 #endif
 647         vmalloc_init();
 648         vfs_caches_init_early();
 649         cpuset_init_early();
 650         mem_init();
 651         enable_debug_pagealloc();
 652         cpu_hotplug_init();
 653         kmem_cache_init();
 654         debug_objects_mem_init();
 655         idr_init_cache();
 656         setup_per_cpu_pageset();
 657         numa_policy_init();
 658         if (late_time_init)
 659                 late_time_init();
 660         calibrate_delay();
 661         pidmap_init();
 662         pgtable_cache_init();
 663         prio_tree_init();
 664         anon_vma_init();
 665 #ifdef CONFIG_X86
 666         if (efi_enabled)
 667                 efi_enter_virtual_mode();
 668 #endif
 669         thread_info_cache_init();
 670         fork_init(num_physpages);
 671         proc_caches_init();
 672         buffer_init();
 673         unnamed_dev_init();
 674         key_init();
 675         security_init();
 676         vfs_caches_init(num_physpages);
 677         radix_tree_init();
 678         signals_init();
 679         /* rootfs populating might need page-writeback */
 680         page_writeback_init();
 681 #ifdef CONFIG_PROC_FS
 682         proc_root_init();
 683 #endif
 684         cgroup_init();
 685         cpuset_init();
 686         taskstats_init_early();
 687         delayacct_init();
 688
 689         check_bugs();
 690
 691         acpi_early_init(); /* before LAPIC and SMP init */
 692
 693         ftrace_init();
 694
 695         /* Do the rest non-__init'ed, we're now alive */
 696         rest_init();
 697 }
 698
 699 static int initcall_debug;
 700
 701 static int __init initcall_debug_setup(char *str)
 702 {
 703         initcall_debug = 1;
 704         return 1;
 705 }
 706 __setup("initcall_debug", initcall_debug_setup);
 707
 708 int do_one_initcall(initcall_t fn)
 709 {
 710         int count = preempt_count();
 711         ktime_t delta;
 712         char msgbuf[64];
 713         struct boot_trace it;
 714
 715         if (initcall_debug) {
 716                 it.caller = task_pid_nr(current);
 717                 printk("calling  %pF @ %i\n", fn, it.caller);
 718                 it.calltime = ktime_get();
 719         }
 720
 721         it.result = fn();
 722
 723         if (initcall_debug) {
 724                 it.rettime = ktime_get();
 725                 delta = ktime_sub(it.rettime, it.calltime);
 726                 it.duration = (unsigned long long) delta.tv64 >> 10;
 727                 printk("initcall %pF returned %d after %Ld usecs\n", fn,
 728                         it.result, it.duration);
 729                 trace_boot(&it, fn);
 730         }
 731
 732         msgbuf[0] = 0;
 733
 734         if (it.result && it.result != -ENODEV && initcall_debug)
 735                 sprintf(msgbuf, "error code %d ", it.result);
 736
 737         if (preempt_count() != count) {
 738                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 739                 preempt_count() = count;
 740         }
 741         if (irqs_disabled()) {
 742                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 743                 local_irq_enable();
 744         }
 745         if (msgbuf[0]) {
 746                 printk("initcall %pF returned with %s\n", fn, msgbuf);
 747         }
 748
 749         return it.result;
 750 }
 751
 752
 753 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
 754
 755 static void __init do_initcalls(void)
 756 {
 757         initcall_t *call;
 758
 759         for (call = __early_initcall_end; call < __initcall_end; call++)
 760                 do_one_initcall(*call);
 761
 762         /* Make sure there is no pending stuff from the initcall sequence */
 763         flush_scheduled_work();
 764 }
 765
 766 /*
 767  * Ok, the machine is now initialized. None of the devices
 768  * have been touched yet, but the CPU subsystem is up and
 769  * running, and memory and process management works.
 770  *
 771  * Now we can finally start doing some real work..
 772  */
 773 static void __init do_basic_setup(void)
 774 {
 775         rcu_init_sched(); /* needed by module_init stage. */
 776         /* drivers will send hotplug events */
 777         init_workqueues();
 778         usermodehelper_init();
 779         driver_init();
 780         init_irq_proc();
 781         do_initcalls();
 782 }
 783
 784 static void __init do_pre_smp_initcalls(void)
 785 {
 786         initcall_t *call;
 787
 788         for (call = __initcall_start; call < __early_initcall_end; call++)
 789                 do_one_initcall(*call);
 790 }
 791
 792 static void run_init_process(char *init_filename)
 793 {
 794         argv_init[0] = init_filename;
 795         kernel_execve(init_filename, argv_init, envp_init);
 796 }
 797
 798 /* This is a non __init function. Force it to be noinline otherwise gcc
 799  * makes it inline to init() and it becomes part of init.text section
 800  */
 801 static int noinline init_post(void)
 802 {
 803         free_initmem();
 804         unlock_kernel();
 805         mark_rodata_ro();
 806         system_state = SYSTEM_RUNNING;
 807         numa_default_policy();
 808
 809         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 810                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 811
 812         (void) sys_dup(0);
 813         (void) sys_dup(0);
 814
 815         current->signal->flags |= SIGNAL_UNKILLABLE;
 816
 817         if (ramdisk_execute_command) {
 818                 run_init_process(ramdisk_execute_command);
 819                 printk(KERN_WARNING "Failed to execute %s\n",
 820                                 ramdisk_execute_command);
 821         }
 822
 823         /*
 824          * We try each of these until one succeeds.
 825          *
 826          * The Bourne shell can be used instead of init if we are
 827          * trying to recover a really broken machine.
 828          */
 829         if (execute_command) {
 830                 run_init_process(execute_command);
 831                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 832                                         "defaults...\n", execute_command);
 833         }
 834         run_init_process("/sbin/init");
 835         run_init_process("/etc/init");
 836         run_init_process("/bin/init");
 837         run_init_process("/bin/sh");
 838
 839         panic("No init found.  Try passing init= option to kernel.");
 840 }
 841
 842 static int __init kernel_init(void * unused)
 843 {
 844         lock_kernel();
 845         /*
 846          * init can run on any cpu.
 847          */
 848         set_cpus_allowed_ptr(current, CPU_MASK_ALL_PTR);
 849         /*
 850          * Tell the world that we're going to be the grim
 851          * reaper of innocent orphaned children.
 852          *
 853          * We don't want people to have to make incorrect
 854          * assumptions about where in the task array this
 855          * can be found.
 856          */
 857         init_pid_ns.child_reaper = current;
 858
 859         cad_pid = task_pid(current);
 860
 861         smp_prepare_cpus(setup_max_cpus);
 862
 863         do_pre_smp_initcalls();
 864         start_boot_trace();
 865
 866         smp_init();
 867         sched_init_smp();
 868
 869         cpuset_init_smp();
 870
 871         do_basic_setup();
 872
 873         /*
 874          * check if there is an early userspace init.  If yes, let it do all
 875          * the work
 876          */
 877
 878         if (!ramdisk_execute_command)
 879                 ramdisk_execute_command = "/init";
 880
 881         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 882                 ramdisk_execute_command = NULL;
 883                 prepare_namespace();
 884         }
 885
 886         /*
 887          * Ok, we have completed the initial bootup, and
 888          * we're essentially up and running. Get rid of the
 889          * initmem segments and start the user-mode stuff..
 890          */
 891         stop_boot_trace();
 892         init_post();
 893         return 0;
 894 }