init/main.c

   1 /*
   2  *  linux/init/main.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  *
   6  *  GK 2/5/95  -  Changed to support mounting root fs via NFS
   7  *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
   8  *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
   9  *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
  10  */
  11
  12 #include <linux/types.h>
  13 #include <linux/module.h>
  14 #include <linux/proc_fs.h>
  15 #include <linux/kernel.h>
  16 #include <linux/syscalls.h>
  17 #include <linux/stackprotector.h>
  18 #include <linux/string.h>
  19 #include <linux/ctype.h>
  20 #include <linux/delay.h>
  21 #include <linux/utsname.h>
  22 #include <linux/ioport.h>
  23 #include <linux/init.h>
  24 #include <linux/smp_lock.h>
  25 #include <linux/initrd.h>
  26 #include <linux/bootmem.h>
  27 #include <linux/tty.h>
  28 #include <linux/gfp.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/workqueue.h>
  37 #include <linux/profile.h>
  38 #include <linux/rcupdate.h>
  39 #include <linux/moduleparam.h>
  40 #include <linux/kallsyms.h>
  41 #include <linux/writeback.h>
  42 #include <linux/cpu.h>
  43 #include <linux/cpuset.h>
  44 #include <linux/cgroup.h>
  45 #include <linux/efi.h>
  46 #include <linux/tick.h>
  47 #include <linux/interrupt.h>
  48 #include <linux/taskstats_kern.h>
  49 #include <linux/delayacct.h>
  50 #include <linux/unistd.h>
  51 #include <linux/rmap.h>
  52 #include <linux/mempolicy.h>
  53 #include <linux/key.h>
  54 #include <linux/buffer_head.h>
  55 #include <linux/page_cgroup.h>
  56 #include <linux/debug_locks.h>
  57 #include <linux/debugobjects.h>
  58 #include <linux/lockdep.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/ftrace.h>
  66 #include <linux/async.h>
  67 #include <trace/boot.h>
  68
  69 #include <asm/io.h>
  70 #include <asm/bugs.h>
  71 #include <asm/setup.h>
  72 #include <asm/sections.h>
  73 #include <asm/cacheflush.h>
  74
  75 #ifdef CONFIG_X86_LOCAL_APIC
  76 #include <asm/smp.h>
  77 #endif
  78
  79 static int kernel_init(void *);
  80
  81 extern void init_IRQ(void);
  82 extern void fork_init(unsigned long);
  83 extern void mca_init(void);
  84 extern void sbus_init(void);
  85 extern void prio_tree_init(void);
  86 extern void radix_tree_init(void);
  87 extern void free_initmem(void);
  88 #ifdef  CONFIG_ACPI
  89 extern void acpi_early_init(void);
  90 #else
  91 static inline void acpi_early_init(void) { }
  92 #endif
  93 #ifndef CONFIG_DEBUG_RODATA
  94 static inline void mark_rodata_ro(void) { }
  95 #endif
  96
  97 #ifdef CONFIG_TC
  98 extern void tc_init(void);
  99 #endif
 100
 101 enum system_states system_state __read_mostly;
 102 EXPORT_SYMBOL(system_state);
 103
 104 /*
 105  * Boot command-line arguments
 106  */
 107 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 108 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 109
 110 extern void time_init(void);
 111 /* Default late time init is NULL. archs can override this later. */
 112 void (*__initdata late_time_init)(void);
 113 extern void softirq_init(void);
 114
 115 /* Untouched command line saved by arch-specific code. */
 116 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 117 /* Untouched saved command line (eg. for /proc) */
 118 char *saved_command_line;
 119 /* Command line for parameter parsing */
 120 static char *static_command_line;
 121
 122 static char *execute_command;
 123 static char *ramdisk_execute_command;
 124
 125 #ifdef CONFIG_SMP
 126 /* Setup configured maximum number of CPUs to activate */
 127 unsigned int __initdata setup_max_cpus = NR_CPUS;
 128
 129 /*
 130  * Setup routine for controlling SMP activation
 131  *
 132  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 133  * activation entirely (the MPS table probe still happens, though).
 134  *
 135  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 136  * greater than 0, limits the maximum number of CPUs activated in
 137  * SMP mode to <NUM>.
 138  */
 139
 140 void __weak arch_disable_smp_support(void) { }
 141
 142 static int __init nosmp(char *str)
 143 {
 144         setup_max_cpus = 0;
 145         arch_disable_smp_support();
 146
 147         return 0;
 148 }
 149
 150 early_param("nosmp", nosmp);
 151
 152 static int __init maxcpus(char *str)
 153 {
 154         get_option(&str, &setup_max_cpus);
 155         if (setup_max_cpus == 0)
 156                 arch_disable_smp_support();
 157
 158         return 0;
 159 }
 160
 161 early_param("maxcpus", maxcpus);
 162 #else
 163 const unsigned int setup_max_cpus = NR_CPUS;
 164 #endif
 165
 166 /*
 167  * If set, this is an indication to the drivers that reset the underlying
 168  * device before going ahead with the initialization otherwise driver might
 169  * rely on the BIOS and skip the reset operation.
 170  *
 171  * This is useful if kernel is booting in an unreliable environment.
 172  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 173  * skipped and devices will be in unknown state.
 174  */
 175 unsigned int reset_devices;
 176 EXPORT_SYMBOL(reset_devices);
 177
 178 static int __init set_reset_devices(char *str)
 179 {
 180         reset_devices = 1;
 181         return 1;
 182 }
 183
 184 __setup("reset_devices", set_reset_devices);
 185
 186 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 187 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 188 static const char *panic_later, *panic_param;
 189
 190 extern struct obs_kernel_param __setup_start[], __setup_end[];
 191
 192 static int __init obsolete_checksetup(char *line)
 193 {
 194         struct obs_kernel_param *p;
 195         int had_early_param = 0;
 196
 197         p = __setup_start;
 198         do {
 199                 int n = strlen(p->str);
 200                 if (!strncmp(line, p->str, n)) {
 201                         if (p->early) {
 202                                 /* Already done in parse_early_param?
 203                                  * (Needs exact match on param part).
 204                                  * Keep iterating, as we can have early
 205                                  * params and __setups of same names 8( */
 206                                 if (line[n] == '\0' || line[n] == '=')
 207                                         had_early_param = 1;
 208                         } else if (!p->setup_func) {
 209                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 210                                        " ignored\n", p->str);
 211                                 return 1;
 212                         } else if (p->setup_func(line + n))
 213                                 return 1;
 214                 }
 215                 p++;
 216         } while (p < __setup_end);
 217
 218         return had_early_param;
 219 }
 220
 221 /*
 222  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 223  * still work even if initially too large, it will just take slightly longer
 224  */
 225 unsigned long loops_per_jiffy = (1<<12);
 226
 227 EXPORT_SYMBOL(loops_per_jiffy);
 228
 229 static int __init debug_kernel(char *str)
 230 {
 231         console_loglevel = 10;
 232         return 0;
 233 }
 234
 235 static int __init quiet_kernel(char *str)
 236 {
 237         console_loglevel = 4;
 238         return 0;
 239 }
 240
 241 early_param("debug", debug_kernel);
 242 early_param("quiet", quiet_kernel);
 243
 244 static int __init loglevel(char *str)
 245 {
 246         get_option(&str, &console_loglevel);
 247         return 0;
 248 }
 249
 250 early_param("loglevel", loglevel);
 251
 252 /*
 253  * Unknown boot options get handed to init, unless they look like
 254  * failed parameters
 255  */
 256 static int __init unknown_bootoption(char *param, char *val)
 257 {
 258         /* Change NUL term back to "=", to make "param" the whole string. */
 259         if (val) {
 260                 /* param=val or param="val"? */
 261                 if (val == param+strlen(param)+1)
 262                         val[-1] = '=';
 263                 else if (val == param+strlen(param)+2) {
 264                         val[-2] = '=';
 265                         memmove(val-1, val, strlen(val)+1);
 266                         val--;
 267                 } else
 268                         BUG();
 269         }
 270
 271         /* Handle obsolete-style parameters */
 272         if (obsolete_checksetup(param))
 273                 return 0;
 274
 275         /*
 276          * Preemptive maintenance for "why didn't my misspelled command
 277          * line work?"
 278          */
 279         if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
 280                 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
 281                 return 0;
 282         }
 283
 284         if (panic_later)
 285                 return 0;
 286
 287         if (val) {
 288                 /* Environment option */
 289                 unsigned int i;
 290                 for (i = 0; envp_init[i]; i++) {
 291                         if (i == MAX_INIT_ENVS) {
 292                                 panic_later = "Too many boot env vars at `%s'";
 293                                 panic_param = param;
 294                         }
 295                         if (!strncmp(param, envp_init[i], val - param))
 296                                 break;
 297                 }
 298                 envp_init[i] = param;
 299         } else {
 300                 /* Command line option */
 301                 unsigned int i;
 302                 for (i = 0; argv_init[i]; i++) {
 303                         if (i == MAX_INIT_ARGS) {
 304                                 panic_later = "Too many boot init vars at `%s'";
 305                                 panic_param = param;
 306                         }
 307                 }
 308                 argv_init[i] = param;
 309         }
 310         return 0;
 311 }
 312
 313 #ifdef CONFIG_DEBUG_PAGEALLOC
 314 int __read_mostly debug_pagealloc_enabled = 0;
 315 #endif
 316
 317 static int __init init_setup(char *str)
 318 {
 319         unsigned int i;
 320
 321         execute_command = str;
 322         /*
 323          * In case LILO is going to boot us with default command line,
 324          * it prepends "auto" before the whole cmdline which makes
 325          * the shell think it should execute a script with such name.
 326          * So we ignore all arguments entered _before_ init=... [MJ]
 327          */
 328         for (i = 1; i < MAX_INIT_ARGS; i++)
 329                 argv_init[i] = NULL;
 330         return 1;
 331 }
 332 __setup("init=", init_setup);
 333
 334 static int __init rdinit_setup(char *str)
 335 {
 336         unsigned int i;
 337
 338         ramdisk_execute_command = str;
 339         /* See "auto" comment in init_setup */
 340         for (i = 1; i < MAX_INIT_ARGS; i++)
 341                 argv_init[i] = NULL;
 342         return 1;
 343 }
 344 __setup("rdinit=", rdinit_setup);
 345
 346 #ifndef CONFIG_SMP
 347
 348 #ifdef CONFIG_X86_LOCAL_APIC
 349 static void __init smp_init(void)
 350 {
 351         APIC_init_uniprocessor();
 352 }
 353 #else
 354 #define smp_init()      do { } while (0)
 355 #endif
 356
 357 static inline void setup_per_cpu_areas(void) { }
 358 static inline void setup_nr_cpu_ids(void) { }
 359 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 360
 361 #else
 362
 363 #if NR_CPUS > BITS_PER_LONG
 364 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
 365 EXPORT_SYMBOL(cpu_mask_all);
 366 #endif
 367
 368 /* Setup number of possible processor ids */
 369 int nr_cpu_ids __read_mostly = NR_CPUS;
 370 EXPORT_SYMBOL(nr_cpu_ids);
 371
 372 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
 373 static void __init setup_nr_cpu_ids(void)
 374 {
 375         nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
 376 }
 377
 378 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
 379 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
 380
 381 EXPORT_SYMBOL(__per_cpu_offset);
 382
 383 static void __init setup_per_cpu_areas(void)
 384 {
 385         unsigned long size, i;
 386         char *ptr;
 387         unsigned long nr_possible_cpus = num_possible_cpus();
 388
 389         /* Copy section for each CPU (we discard the original) */
 390         size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
 391         ptr = alloc_bootmem_pages(size * nr_possible_cpus);
 392
 393         for_each_possible_cpu(i) {
 394                 __per_cpu_offset[i] = ptr - __per_cpu_start;
 395                 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 396                 ptr += size;
 397         }
 398 }
 399 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
 400
 401 /* Called by boot processor to activate the rest. */
 402 static void __init smp_init(void)
 403 {
 404         unsigned int cpu;
 405
 406         /*
 407          * Set up the current CPU as possible to migrate to.
 408          * The other ones will be done by cpu_up/cpu_down()
 409          */
 410         cpu = smp_processor_id();
 411         cpu_set(cpu, cpu_active_map);
 412
 413         /* FIXME: This should be done in userspace --RR */
 414         for_each_present_cpu(cpu) {
 415                 if (num_online_cpus() >= setup_max_cpus)
 416                         break;
 417                 if (!cpu_online(cpu))
 418                         cpu_up(cpu);
 419         }
 420
 421         /* Any cleanup work */
 422         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 423         smp_cpus_done(setup_max_cpus);
 424 }
 425
 426 #endif
 427
 428 /*
 429  * We need to store the untouched command line for future reference.
 430  * We also need to store the touched command line since the parameter
 431  * parsing is performed in place, and we should allow a component to
 432  * store reference of name/value for future reference.
 433  */
 434 static void __init setup_command_line(char *command_line)
 435 {
 436         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 437         static_command_line = alloc_bootmem(strlen (command_line)+1);
 438         strcpy (saved_command_line, boot_command_line);
 439         strcpy (static_command_line, command_line);
 440 }
 441
 442 /*
 443  * We need to finalize in a non-__init function or else race conditions
 444  * between the root thread and the init thread may cause start_kernel to
 445  * be reaped by free_initmem before the root thread has proceeded to
 446  * cpu_idle.
 447  *
 448  * gcc-3.4 accidentally inlines this function, so use noinline.
 449  */
 450
 451 static noinline void __init_refok rest_init(void)
 452         __releases(kernel_lock)
 453 {
 454         int pid;
 455
 456         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 457         numa_default_policy();
 458         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 459         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 460         unlock_kernel();
 461
 462         /*
 463          * The boot idle thread must execute schedule()
 464          * at least once to get things moving:
 465          */
 466         init_idle_bootup_task(current);
 467         rcu_scheduler_starting();
 468         preempt_enable_no_resched();
 469         schedule();
 470         preempt_disable();
 471
 472         /* Call into cpu_idle with preempt disabled */
 473         cpu_idle();
 474 }
 475
 476 /* Check for early params. */
 477 static int __init do_early_param(char *param, char *val)
 478 {
 479         struct obs_kernel_param *p;
 480
 481         for (p = __setup_start; p < __setup_end; p++) {
 482                 if ((p->early && strcmp(param, p->str) == 0) ||
 483                     (strcmp(param, "console") == 0 &&
 484                      strcmp(p->str, "earlycon") == 0)
 485                 ) {
 486                         if (p->setup_func(val) != 0)
 487                                 printk(KERN_WARNING
 488                                        "Malformed early option '%s'\n", param);
 489                 }
 490         }
 491         /* We accept everything at this stage. */
 492         return 0;
 493 }
 494
 495 /* Arch code calls this early on, or if not, just before other parsing. */
 496 void __init parse_early_param(void)
 497 {
 498         static __initdata int done = 0;
 499         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 500
 501         if (done)
 502                 return;
 503
 504         /* All fall through to do_early_param. */
 505         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 506         parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
 507         done = 1;
 508 }
 509
 510 /*
 511  *      Activate the first processor.
 512  */
 513
 514 static void __init boot_cpu_init(void)
 515 {
 516         int cpu = smp_processor_id();
 517         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 518         set_cpu_online(cpu, true);
 519         set_cpu_present(cpu, true);
 520         set_cpu_possible(cpu, true);
 521 }
 522
 523 void __init __weak smp_setup_processor_id(void)
 524 {
 525 }
 526
 527 void __init __weak thread_info_cache_init(void)
 528 {
 529 }
 530
 531 asmlinkage void __init start_kernel(void)
 532 {
 533         char * command_line;
 534         extern struct kernel_param __start___param[], __stop___param[];
 535
 536         smp_setup_processor_id();
 537
 538         /*
 539          * Need to run as early as possible, to initialize the
 540          * lockdep hash:
 541          */
 542         lockdep_init();
 543         debug_objects_early_init();
 544
 545         /*
 546          * Set up the the initial canary ASAP:
 547          */
 548         boot_init_stack_canary();
 549
 550         cgroup_init_early();
 551
 552         local_irq_disable();
 553         early_boot_irqs_off();
 554         early_init_irq_lock_class();
 555
 556 /*
 557  * Interrupts are still disabled. Do necessary setups, then
 558  * enable them
 559  */
 560         lock_kernel();
 561         tick_init();
 562         boot_cpu_init();
 563         page_address_init();
 564         printk(KERN_NOTICE);
 565         printk(linux_banner);
 566         setup_arch(&command_line);
 567         mm_init_owner(&init_mm, &init_task);
 568         setup_command_line(command_line);
 569         setup_per_cpu_areas();
 570         setup_nr_cpu_ids();
 571         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 572
 573         /*
 574          * Set up the scheduler prior starting any interrupts (such as the
 575          * timer interrupt). Full topology setup happens at smp_init()
 576          * time - but meanwhile we still have a functioning scheduler.
 577          */
 578         sched_init();
 579         /*
 580          * Disable preemption - early bootup scheduling is extremely
 581          * fragile until we cpu_idle() for the first time.
 582          */
 583         preempt_disable();
 584         build_all_zonelists();
 585         page_alloc_init();
 586         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 587         parse_early_param();
 588         parse_args("Booting kernel", static_command_line, __start___param,
 589                    __stop___param - __start___param,
 590                    &unknown_bootoption);
 591         if (!irqs_disabled()) {
 592                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 593                                 "enabled *very* early, fixing it\n");
 594                 local_irq_disable();
 595         }
 596         sort_main_extable();
 597         trap_init();
 598         rcu_init();
 599         /* init some links before init_ISA_irqs() */
 600         early_irq_init();
 601         init_IRQ();
 602         pidhash_init();
 603         init_timers();
 604         hrtimers_init();
 605         softirq_init();
 606         timekeeping_init();
 607         time_init();
 608         sched_clock_init();
 609         profile_init();
 610         if (!irqs_disabled())
 611                 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
 612                                  "enabled early\n");
 613         early_boot_irqs_on();
 614         local_irq_enable();
 615
 616         /*
 617          * HACK ALERT! This is early. We're enabling the console before
 618          * we've done PCI setups etc, and console_init() must be aware of
 619          * this. But we do want output early, in case something goes wrong.
 620          */
 621         console_init();
 622         if (panic_later)
 623                 panic(panic_later, panic_param);
 624
 625         lockdep_info();
 626
 627         /*
 628          * Need to run this when irqs are enabled, because it wants
 629          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 630          * too:
 631          */
 632         locking_selftest();
 633
 634 #ifdef CONFIG_BLK_DEV_INITRD
 635         if (initrd_start && !initrd_below_start_ok &&
 636             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 637                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 638                     "disabling it.\n",
 639                     page_to_pfn(virt_to_page((void *)initrd_start)),
 640                     min_low_pfn);
 641                 initrd_start = 0;
 642         }
 643 #endif
 644         vmalloc_init();
 645         vfs_caches_init_early();
 646         cpuset_init_early();
 647         page_cgroup_init();
 648         mem_init();
 649         enable_debug_pagealloc();
 650         cpu_hotplug_init();
 651         kmem_cache_init();
 652         debug_objects_mem_init();
 653         idr_init_cache();
 654         setup_per_cpu_pageset();
 655         numa_policy_init();
 656         if (late_time_init)
 657                 late_time_init();
 658         calibrate_delay();
 659         pidmap_init();
 660         pgtable_cache_init();
 661         prio_tree_init();
 662         anon_vma_init();
 663 #ifdef CONFIG_X86
 664         if (efi_enabled)
 665                 efi_enter_virtual_mode();
 666 #endif
 667         thread_info_cache_init();
 668         cred_init();
 669         fork_init(num_physpages);
 670         proc_caches_init();
 671         buffer_init();
 672         key_init();
 673         security_init();
 674         vfs_caches_init(num_physpages);
 675         radix_tree_init();
 676         signals_init();
 677         /* rootfs populating might need page-writeback */
 678         page_writeback_init();
 679 #ifdef CONFIG_PROC_FS
 680         proc_root_init();
 681 #endif
 682         cgroup_init();
 683         cpuset_init();
 684         taskstats_init_early();
 685         delayacct_init();
 686
 687         check_bugs();
 688
 689         acpi_early_init(); /* before LAPIC and SMP init */
 690
 691         ftrace_init();
 692
 693         /* Do the rest non-__init'ed, we're now alive */
 694         rest_init();
 695 }
 696
 697 int initcall_debug;
 698 core_param(initcall_debug, initcall_debug, bool, 0644);
 699
 700 int do_one_initcall(initcall_t fn)
 701 {
 702         int count = preempt_count();
 703         ktime_t calltime, delta, rettime;
 704         char msgbuf[64];
 705         struct boot_trace_call call;
 706         struct boot_trace_ret ret;
 707
 708         if (initcall_debug) {
 709                 call.caller = task_pid_nr(current);
 710                 printk("calling  %pF @ %i\n", fn, call.caller);
 711                 calltime = ktime_get();
 712                 trace_boot_call(&call, fn);
 713                 enable_boot_trace();
 714         }
 715
 716         ret.result = fn();
 717
 718         if (initcall_debug) {
 719                 disable_boot_trace();
 720                 rettime = ktime_get();
 721                 delta = ktime_sub(rettime, calltime);
 722                 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 723                 trace_boot_ret(&ret, fn);
 724                 printk("initcall %pF returned %d after %Ld usecs\n", fn,
 725                         ret.result, ret.duration);
 726         }
 727
 728         msgbuf[0] = 0;
 729
 730         if (ret.result && ret.result != -ENODEV && initcall_debug)
 731                 sprintf(msgbuf, "error code %d ", ret.result);
 732
 733         if (preempt_count() != count) {
 734                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 735                 preempt_count() = count;
 736         }
 737         if (irqs_disabled()) {
 738                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 739                 local_irq_enable();
 740         }
 741         if (msgbuf[0]) {
 742                 printk("initcall %pF returned with %s\n", fn, msgbuf);
 743         }
 744
 745         return ret.result;
 746 }
 747
 748
 749 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
 750
 751 static void __init do_initcalls(void)
 752 {
 753         initcall_t *call;
 754
 755         for (call = __early_initcall_end; call < __initcall_end; call++)
 756                 do_one_initcall(*call);
 757
 758         /* Make sure there is no pending stuff from the initcall sequence */
 759         flush_scheduled_work();
 760 }
 761
 762 /*
 763  * Ok, the machine is now initialized. None of the devices
 764  * have been touched yet, but the CPU subsystem is up and
 765  * running, and memory and process management works.
 766  *
 767  * Now we can finally start doing some real work..
 768  */
 769 static void __init do_basic_setup(void)
 770 {
 771         rcu_init_sched(); /* needed by module_init stage. */
 772         init_workqueues();
 773         usermodehelper_init();
 774         driver_init();
 775         init_irq_proc();
 776         do_initcalls();
 777 }
 778
 779 static void __init do_pre_smp_initcalls(void)
 780 {
 781         initcall_t *call;
 782
 783         for (call = __initcall_start; call < __early_initcall_end; call++)
 784                 do_one_initcall(*call);
 785 }
 786
 787 static void run_init_process(char *init_filename)
 788 {
 789         argv_init[0] = init_filename;
 790         kernel_execve(init_filename, argv_init, envp_init);
 791 }
 792
 793 /* This is a non __init function. Force it to be noinline otherwise gcc
 794  * makes it inline to init() and it becomes part of init.text section
 795  */
 796 static noinline int init_post(void)
 797 {
 798         /* need to finish all async __init code before freeing the memory */
 799         async_synchronize_full();
 800         free_initmem();
 801         unlock_kernel();
 802         mark_rodata_ro();
 803         system_state = SYSTEM_RUNNING;
 804         numa_default_policy();
 805
 806         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 807                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 808
 809         (void) sys_dup(0);
 810         (void) sys_dup(0);
 811
 812         current->signal->flags |= SIGNAL_UNKILLABLE;
 813
 814         if (ramdisk_execute_command) {
 815                 run_init_process(ramdisk_execute_command);
 816                 printk(KERN_WARNING "Failed to execute %s\n",
 817                                 ramdisk_execute_command);
 818         }
 819
 820         /*
 821          * We try each of these until one succeeds.
 822          *
 823          * The Bourne shell can be used instead of init if we are
 824          * trying to recover a really broken machine.
 825          */
 826         if (execute_command) {
 827                 run_init_process(execute_command);
 828                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 829                                         "defaults...\n", execute_command);
 830         }
 831         run_init_process("/sbin/init");
 832         run_init_process("/etc/init");
 833         run_init_process("/bin/init");
 834         run_init_process("/bin/sh");
 835
 836         panic("No init found.  Try passing init= option to kernel.");
 837 }
 838
 839 static int __init kernel_init(void * unused)
 840 {
 841         lock_kernel();
 842         /*
 843          * init can run on any cpu.
 844          */
 845         set_cpus_allowed_ptr(current, CPU_MASK_ALL_PTR);
 846         /*
 847          * Tell the world that we're going to be the grim
 848          * reaper of innocent orphaned children.
 849          *
 850          * We don't want people to have to make incorrect
 851          * assumptions about where in the task array this
 852          * can be found.
 853          */
 854         init_pid_ns.child_reaper = current;
 855
 856         cad_pid = task_pid(current);
 857
 858         smp_prepare_cpus(setup_max_cpus);
 859
 860         do_pre_smp_initcalls();
 861         start_boot_trace();
 862
 863         smp_init();
 864         sched_init_smp();
 865
 866         cpuset_init_smp();
 867
 868         do_basic_setup();
 869
 870         /*
 871          * check if there is an early userspace init.  If yes, let it do all
 872          * the work
 873          */
 874
 875         if (!ramdisk_execute_command)
 876                 ramdisk_execute_command = "/init";
 877
 878         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 879                 ramdisk_execute_command = NULL;
 880                 prepare_namespace();
 881         }
 882
 883         /*
 884          * Ok, we have completed the initial bootup, and
 885          * we're essentially up and running. Get rid of the
 886          * initmem segments and start the user-mode stuff..
 887          */
 888
 889         init_post();
 890         return 0;
 891 }