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/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/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 #include <linux/async.h>
  66 #include <trace/boot.h>
  67
  68 #include <asm/io.h>
  69 #include <asm/bugs.h>
  70 #include <asm/setup.h>
  71 #include <asm/sections.h>
  72 #include <asm/cacheflush.h>
  73
  74 #ifdef CONFIG_X86_LOCAL_APIC
  75 #include <asm/smp.h>
  76 #endif
  77
  78 static int kernel_init(void *);
  79
  80 extern void init_IRQ(void);
  81 extern void fork_init(unsigned long);
  82 extern void mca_init(void);
  83 extern void sbus_init(void);
  84 extern void prio_tree_init(void);
  85 extern void radix_tree_init(void);
  86 extern void free_initmem(void);
  87 #ifdef  CONFIG_ACPI
  88 extern void acpi_early_init(void);
  89 #else
  90 static inline void acpi_early_init(void) { }
  91 #endif
  92 #ifndef CONFIG_DEBUG_RODATA
  93 static inline void mark_rodata_ro(void) { }
  94 #endif
  95
  96 #ifdef CONFIG_TC
  97 extern void tc_init(void);
  98 #endif
  99
 100 enum system_states system_state __read_mostly;
 101 EXPORT_SYMBOL(system_state);
 102
 103 /*
 104  * Boot command-line arguments
 105  */
 106 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 107 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 108
 109 extern void time_init(void);
 110 /* Default late time init is NULL. archs can override this later. */
 111 void (*__initdata late_time_init)(void);
 112 extern void softirq_init(void);
 113
 114 /* Untouched command line saved by arch-specific code. */
 115 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 116 /* Untouched saved command line (eg. for /proc) */
 117 char *saved_command_line;
 118 /* Command line for parameter parsing */
 119 static char *static_command_line;
 120
 121 static char *execute_command;
 122 static char *ramdisk_execute_command;
 123
 124 #ifdef CONFIG_SMP
 125 /* Setup configured maximum number of CPUs to activate */
 126 unsigned int __initdata setup_max_cpus = NR_CPUS;
 127
 128 /*
 129  * Setup routine for controlling SMP activation
 130  *
 131  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 132  * activation entirely (the MPS table probe still happens, though).
 133  *
 134  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 135  * greater than 0, limits the maximum number of CPUs activated in
 136  * SMP mode to <NUM>.
 137  */
 138 #ifndef CONFIG_X86_IO_APIC
 139 static inline void disable_ioapic_setup(void) {};
 140 #endif
 141
 142 static int __init nosmp(char *str)
 143 {
 144         setup_max_cpus = 0;
 145         disable_ioapic_setup();
 146         return 0;
 147 }
 148
 149 early_param("nosmp", nosmp);
 150
 151 static int __init maxcpus(char *str)
 152 {
 153         get_option(&str, &setup_max_cpus);
 154         if (setup_max_cpus == 0)
 155                 disable_ioapic_setup();
 156
 157         return 0;
 158 }
 159
 160 early_param("maxcpus", maxcpus);
 161 #else
 162 #define setup_max_cpus NR_CPUS
 163 #endif
 164
 165 /*
 166  * If set, this is an indication to the drivers that reset the underlying
 167  * device before going ahead with the initialization otherwise driver might
 168  * rely on the BIOS and skip the reset operation.
 169  *
 170  * This is useful if kernel is booting in an unreliable environment.
 171  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 172  * skipped and devices will be in unknown state.
 173  */
 174 unsigned int reset_devices;
 175 EXPORT_SYMBOL(reset_devices);
 176
 177 static int __init set_reset_devices(char *str)
 178 {
 179         reset_devices = 1;
 180         return 1;
 181 }
 182
 183 __setup("reset_devices", set_reset_devices);
 184
 185 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 186 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 187 static const char *panic_later, *panic_param;
 188
 189 extern struct obs_kernel_param __setup_start[], __setup_end[];
 190
 191 static int __init obsolete_checksetup(char *line)
 192 {
 193         struct obs_kernel_param *p;
 194         int had_early_param = 0;
 195
 196         p = __setup_start;
 197         do {
 198                 int n = strlen(p->str);
 199                 if (!strncmp(line, p->str, n)) {
 200                         if (p->early) {
 201                                 /* Already done in parse_early_param?
 202                                  * (Needs exact match on param part).
 203                                  * Keep iterating, as we can have early
 204                                  * params and __setups of same names 8( */
 205                                 if (line[n] == '\0' || line[n] == '=')
 206                                         had_early_param = 1;
 207                         } else if (!p->setup_func) {
 208                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 209                                        " ignored\n", p->str);
 210                                 return 1;
 211                         } else if (p->setup_func(line + n))
 212                                 return 1;
 213                 }
 214                 p++;
 215         } while (p < __setup_end);
 216
 217         return had_early_param;
 218 }
 219
 220 /*
 221  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 222  * still work even if initially too large, it will just take slightly longer
 223  */
 224 unsigned long loops_per_jiffy = (1<<12);
 225
 226 EXPORT_SYMBOL(loops_per_jiffy);
 227
 228 static int __init debug_kernel(char *str)
 229 {
 230         console_loglevel = 10;
 231         return 0;
 232 }
 233
 234 static int __init quiet_kernel(char *str)
 235 {
 236         console_loglevel = 4;
 237         return 0;
 238 }
 239
 240 early_param("debug", debug_kernel);
 241 early_param("quiet", quiet_kernel);
 242
 243 static int __init loglevel(char *str)
 244 {
 245         get_option(&str, &console_loglevel);
 246         return 0;
 247 }
 248
 249 early_param("loglevel", loglevel);
 250
 251 /*
 252  * Unknown boot options get handed to init, unless they look like
 253  * failed parameters
 254  */
 255 static int __init unknown_bootoption(char *param, char *val)
 256 {
 257         /* Change NUL term back to "=", to make "param" the whole string. */
 258         if (val) {
 259                 /* param=val or param="val"? */
 260                 if (val == param+strlen(param)+1)
 261                         val[-1] = '=';
 262                 else if (val == param+strlen(param)+2) {
 263                         val[-2] = '=';
 264                         memmove(val-1, val, strlen(val)+1);
 265                         val--;
 266                 } else
 267                         BUG();
 268         }
 269
 270         /* Handle obsolete-style parameters */
 271         if (obsolete_checksetup(param))
 272                 return 0;
 273
 274         /*
 275          * Preemptive maintenance for "why didn't my misspelled command
 276          * line work?"
 277          */
 278         if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
 279                 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
 280                 return 0;
 281         }
 282
 283         if (panic_later)
 284                 return 0;
 285
 286         if (val) {
 287                 /* Environment option */
 288                 unsigned int i;
 289                 for (i = 0; envp_init[i]; i++) {
 290                         if (i == MAX_INIT_ENVS) {
 291                                 panic_later = "Too many boot env vars at `%s'";
 292                                 panic_param = param;
 293                         }
 294                         if (!strncmp(param, envp_init[i], val - param))
 295                                 break;
 296                 }
 297                 envp_init[i] = param;
 298         } else {
 299                 /* Command line option */
 300                 unsigned int i;
 301                 for (i = 0; argv_init[i]; i++) {
 302                         if (i == MAX_INIT_ARGS) {
 303                                 panic_later = "Too many boot init vars at `%s'";
 304                                 panic_param = param;
 305                         }
 306                 }
 307                 argv_init[i] = param;
 308         }
 309         return 0;
 310 }
 311
 312 #ifdef CONFIG_DEBUG_PAGEALLOC
 313 int __read_mostly debug_pagealloc_enabled = 0;
 314 #endif
 315
 316 static int __init init_setup(char *str)
 317 {
 318         unsigned int i;
 319
 320         execute_command = str;
 321         /*
 322          * In case LILO is going to boot us with default command line,
 323          * it prepends "auto" before the whole cmdline which makes
 324          * the shell think it should execute a script with such name.
 325          * So we ignore all arguments entered _before_ init=... [MJ]
 326          */
 327         for (i = 1; i < MAX_INIT_ARGS; i++)
 328                 argv_init[i] = NULL;
 329         return 1;
 330 }
 331 __setup("init=", init_setup);
 332
 333 static int __init rdinit_setup(char *str)
 334 {
 335         unsigned int i;
 336
 337         ramdisk_execute_command = str;
 338         /* See "auto" comment in init_setup */
 339         for (i = 1; i < MAX_INIT_ARGS; i++)
 340                 argv_init[i] = NULL;
 341         return 1;
 342 }
 343 __setup("rdinit=", rdinit_setup);
 344
 345 #ifndef CONFIG_SMP
 346
 347 #ifdef CONFIG_X86_LOCAL_APIC
 348 static void __init smp_init(void)
 349 {
 350         APIC_init_uniprocessor();
 351 }
 352 #else
 353 #define smp_init()      do { } while (0)
 354 #endif
 355
 356 static inline void setup_per_cpu_areas(void) { }
 357 static inline void setup_nr_cpu_ids(void) { }
 358 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 359
 360 #else
 361
 362 #if NR_CPUS > BITS_PER_LONG
 363 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
 364 EXPORT_SYMBOL(cpu_mask_all);
 365 #endif
 366
 367 /* Setup number of possible processor ids */
 368 int nr_cpu_ids __read_mostly = NR_CPUS;
 369 EXPORT_SYMBOL(nr_cpu_ids);
 370
 371 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
 372 static void __init setup_nr_cpu_ids(void)
 373 {
 374         nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
 375 }
 376
 377 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
 378 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
 379
 380 EXPORT_SYMBOL(__per_cpu_offset);
 381
 382 static void __init setup_per_cpu_areas(void)
 383 {
 384         unsigned long size, i;
 385         char *ptr;
 386         unsigned long nr_possible_cpus = num_possible_cpus();
 387
 388         /* Copy section for each CPU (we discard the original) */
 389         size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
 390         ptr = alloc_bootmem_pages(size * nr_possible_cpus);
 391
 392         for_each_possible_cpu(i) {
 393                 __per_cpu_offset[i] = ptr - __per_cpu_start;
 394                 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 395                 ptr += size;
 396         }
 397 }
 398 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
 399
 400 /* Called by boot processor to activate the rest. */
 401 static void __init smp_init(void)
 402 {
 403         unsigned int cpu;
 404
 405         /*
 406          * Set up the current CPU as possible to migrate to.
 407          * The other ones will be done by cpu_up/cpu_down()
 408          */
 409         cpu = smp_processor_id();
 410         cpu_set(cpu, cpu_active_map);
 411
 412         /* FIXME: This should be done in userspace --RR */
 413         for_each_present_cpu(cpu) {
 414                 if (num_online_cpus() >= setup_max_cpus)
 415                         break;
 416                 if (!cpu_online(cpu))
 417                         cpu_up(cpu);
 418         }
 419
 420         /* Any cleanup work */
 421         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 422         smp_cpus_done(setup_max_cpus);
 423 }
 424
 425 #endif
 426
 427 /*
 428  * We need to store the untouched command line for future reference.
 429  * We also need to store the touched command line since the parameter
 430  * parsing is performed in place, and we should allow a component to
 431  * store reference of name/value for future reference.
 432  */
 433 static void __init setup_command_line(char *command_line)
 434 {
 435         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 436         static_command_line = alloc_bootmem(strlen (command_line)+1);
 437         strcpy (saved_command_line, boot_command_line);
 438         strcpy (static_command_line, command_line);
 439 }
 440
 441 /*
 442  * We need to finalize in a non-__init function or else race conditions
 443  * between the root thread and the init thread may cause start_kernel to
 444  * be reaped by free_initmem before the root thread has proceeded to
 445  * cpu_idle.
 446  *
 447  * gcc-3.4 accidentally inlines this function, so use noinline.
 448  */
 449
 450 static noinline void __init_refok rest_init(void)
 451         __releases(kernel_lock)
 452 {
 453         int pid;
 454
 455         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 456         numa_default_policy();
 457         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 458         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 459         unlock_kernel();
 460
 461         /*
 462          * The boot idle thread must execute schedule()
 463          * at least once to get things moving:
 464          */
 465         init_idle_bootup_task(current);
 466         rcu_scheduler_starting();
 467         preempt_enable_no_resched();
 468         schedule();
 469         preempt_disable();
 470
 471         /* Call into cpu_idle with preempt disabled */
 472         cpu_idle();
 473 }
 474
 475 /* Check for early params. */
 476 static int __init do_early_param(char *param, char *val)
 477 {
 478         struct obs_kernel_param *p;
 479
 480         for (p = __setup_start; p < __setup_end; p++) {
 481                 if ((p->early && strcmp(param, p->str) == 0) ||
 482                     (strcmp(param, "console") == 0 &&
 483                      strcmp(p->str, "earlycon") == 0)
 484                 ) {
 485                         if (p->setup_func(val) != 0)
 486                                 printk(KERN_WARNING
 487                                        "Malformed early option '%s'\n", param);
 488                 }
 489         }
 490         /* We accept everything at this stage. */
 491         return 0;
 492 }
 493
 494 /* Arch code calls this early on, or if not, just before other parsing. */
 495 void __init parse_early_param(void)
 496 {
 497         static __initdata int done = 0;
 498         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 499
 500         if (done)
 501                 return;
 502
 503         /* All fall through to do_early_param. */
 504         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 505         parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
 506         done = 1;
 507 }
 508
 509 /*
 510  *      Activate the first processor.
 511  */
 512
 513 static void __init boot_cpu_init(void)
 514 {
 515         int cpu = smp_processor_id();
 516         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 517         set_cpu_online(cpu, true);
 518         set_cpu_present(cpu, true);
 519         set_cpu_possible(cpu, true);
 520 }
 521
 522 void __init __weak smp_setup_processor_id(void)
 523 {
 524 }
 525
 526 void __init __weak thread_info_cache_init(void)
 527 {
 528 }
 529
 530 asmlinkage void __init start_kernel(void)
 531 {
 532         char * command_line;
 533         extern struct kernel_param __start___param[], __stop___param[];
 534
 535         smp_setup_processor_id();
 536
 537         /*
 538          * Need to run as early as possible, to initialize the
 539          * lockdep hash:
 540          */
 541         lockdep_init();
 542         debug_objects_early_init();
 543         cgroup_init_early();
 544
 545         local_irq_disable();
 546         early_boot_irqs_off();
 547         early_init_irq_lock_class();
 548
 549 /*
 550  * Interrupts are still disabled. Do necessary setups, then
 551  * enable them
 552  */
 553         lock_kernel();
 554         tick_init();
 555         boot_cpu_init();
 556         page_address_init();
 557         printk(KERN_NOTICE);
 558         printk(linux_banner);
 559         setup_arch(&command_line);
 560         mm_init_owner(&init_mm, &init_task);
 561         setup_command_line(command_line);
 562         setup_per_cpu_areas();
 563         setup_nr_cpu_ids();
 564         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 565
 566         /*
 567          * Set up the scheduler prior starting any interrupts (such as the
 568          * timer interrupt). Full topology setup happens at smp_init()
 569          * time - but meanwhile we still have a functioning scheduler.
 570          */
 571         sched_init();
 572         /*
 573          * Disable preemption - early bootup scheduling is extremely
 574          * fragile until we cpu_idle() for the first time.
 575          */
 576         preempt_disable();
 577         build_all_zonelists();
 578         page_alloc_init();
 579         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 580         parse_early_param();
 581         parse_args("Booting kernel", static_command_line, __start___param,
 582                    __stop___param - __start___param,
 583                    &unknown_bootoption);
 584         if (!irqs_disabled()) {
 585                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 586                                 "enabled *very* early, fixing it\n");
 587                 local_irq_disable();
 588         }
 589         sort_main_extable();
 590         trap_init();
 591         rcu_init();
 592         /* init some links before init_ISA_irqs() */
 593         early_irq_init();
 594         init_IRQ();
 595         pidhash_init();
 596         init_timers();
 597         hrtimers_init();
 598         softirq_init();
 599         timekeeping_init();
 600         time_init();
 601         sched_clock_init();
 602         profile_init();
 603         if (!irqs_disabled())
 604                 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
 605                                  "enabled early\n");
 606         early_boot_irqs_on();
 607         local_irq_enable();
 608
 609         /*
 610          * HACK ALERT! This is early. We're enabling the console before
 611          * we've done PCI setups etc, and console_init() must be aware of
 612          * this. But we do want output early, in case something goes wrong.
 613          */
 614         console_init();
 615         if (panic_later)
 616                 panic(panic_later, panic_param);
 617
 618         lockdep_info();
 619
 620         /*
 621          * Need to run this when irqs are enabled, because it wants
 622          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 623          * too:
 624          */
 625         locking_selftest();
 626
 627 #ifdef CONFIG_BLK_DEV_INITRD
 628         if (initrd_start && !initrd_below_start_ok &&
 629             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 630                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 631                     "disabling it.\n",
 632                     page_to_pfn(virt_to_page((void *)initrd_start)),
 633                     min_low_pfn);
 634                 initrd_start = 0;
 635         }
 636 #endif
 637         vmalloc_init();
 638         vfs_caches_init_early();
 639         cpuset_init_early();
 640         page_cgroup_init();
 641         mem_init();
 642         enable_debug_pagealloc();
 643         cpu_hotplug_init();
 644         kmem_cache_init();
 645         debug_objects_mem_init();
 646         idr_init_cache();
 647         setup_per_cpu_pageset();
 648         numa_policy_init();
 649         if (late_time_init)
 650                 late_time_init();
 651         calibrate_delay();
 652         pidmap_init();
 653         pgtable_cache_init();
 654         prio_tree_init();
 655         anon_vma_init();
 656 #ifdef CONFIG_X86
 657         if (efi_enabled)
 658                 efi_enter_virtual_mode();
 659 #endif
 660         thread_info_cache_init();
 661         cred_init();
 662         fork_init(num_physpages);
 663         proc_caches_init();
 664         buffer_init();
 665         key_init();
 666         security_init();
 667         vfs_caches_init(num_physpages);
 668         radix_tree_init();
 669         signals_init();
 670         /* rootfs populating might need page-writeback */
 671         page_writeback_init();
 672 #ifdef CONFIG_PROC_FS
 673         proc_root_init();
 674 #endif
 675         cgroup_init();
 676         cpuset_init();
 677         taskstats_init_early();
 678         delayacct_init();
 679
 680         check_bugs();
 681
 682         acpi_early_init(); /* before LAPIC and SMP init */
 683
 684         ftrace_init();
 685
 686         /* Do the rest non-__init'ed, we're now alive */
 687         rest_init();
 688 }
 689
 690 int initcall_debug;
 691 core_param(initcall_debug, initcall_debug, bool, 0644);
 692
 693 int do_one_initcall(initcall_t fn)
 694 {
 695         int count = preempt_count();
 696         ktime_t calltime, delta, rettime;
 697         char msgbuf[64];
 698         struct boot_trace_call call;
 699         struct boot_trace_ret ret;
 700
 701         if (initcall_debug) {
 702                 call.caller = task_pid_nr(current);
 703                 printk("calling  %pF @ %i\n", fn, call.caller);
 704                 calltime = ktime_get();
 705                 trace_boot_call(&call, fn);
 706                 enable_boot_trace();
 707         }
 708
 709         ret.result = fn();
 710
 711         if (initcall_debug) {
 712                 disable_boot_trace();
 713                 rettime = ktime_get();
 714                 delta = ktime_sub(rettime, calltime);
 715                 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 716                 trace_boot_ret(&ret, fn);
 717                 printk("initcall %pF returned %d after %Ld usecs\n", fn,
 718                         ret.result, ret.duration);
 719         }
 720
 721         msgbuf[0] = 0;
 722
 723         if (ret.result && ret.result != -ENODEV && initcall_debug)
 724                 sprintf(msgbuf, "error code %d ", ret.result);
 725
 726         if (preempt_count() != count) {
 727                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 728                 preempt_count() = count;
 729         }
 730         if (irqs_disabled()) {
 731                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 732                 local_irq_enable();
 733         }
 734         if (msgbuf[0]) {
 735                 printk("initcall %pF returned with %s\n", fn, msgbuf);
 736         }
 737
 738         return ret.result;
 739 }
 740
 741
 742 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
 743
 744 static void __init do_initcalls(void)
 745 {
 746         initcall_t *call;
 747
 748         for (call = __early_initcall_end; call < __initcall_end; call++)
 749                 do_one_initcall(*call);
 750
 751         /* Make sure there is no pending stuff from the initcall sequence */
 752         flush_scheduled_work();
 753 }
 754
 755 /*
 756  * Ok, the machine is now initialized. None of the devices
 757  * have been touched yet, but the CPU subsystem is up and
 758  * running, and memory and process management works.
 759  *
 760  * Now we can finally start doing some real work..
 761  */
 762 static void __init do_basic_setup(void)
 763 {
 764         rcu_init_sched(); /* needed by module_init stage. */
 765         init_workqueues();
 766         usermodehelper_init();
 767         driver_init();
 768         init_irq_proc();
 769         do_initcalls();
 770 }
 771
 772 static void __init do_pre_smp_initcalls(void)
 773 {
 774         initcall_t *call;
 775
 776         for (call = __initcall_start; call < __early_initcall_end; call++)
 777                 do_one_initcall(*call);
 778 }
 779
 780 static void run_init_process(char *init_filename)
 781 {
 782         argv_init[0] = init_filename;
 783         kernel_execve(init_filename, argv_init, envp_init);
 784 }
 785
 786 /* This is a non __init function. Force it to be noinline otherwise gcc
 787  * makes it inline to init() and it becomes part of init.text section
 788  */
 789 static noinline int init_post(void)
 790 {
 791         /* need to finish all async __init code before freeing the memory */
 792         async_synchronize_full();
 793         free_initmem();
 794         unlock_kernel();
 795         mark_rodata_ro();
 796         system_state = SYSTEM_RUNNING;
 797         numa_default_policy();
 798
 799         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 800                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 801
 802         (void) sys_dup(0);
 803         (void) sys_dup(0);
 804
 805         current->signal->flags |= SIGNAL_UNKILLABLE;
 806
 807         if (ramdisk_execute_command) {
 808                 run_init_process(ramdisk_execute_command);
 809                 printk(KERN_WARNING "Failed to execute %s\n",
 810                                 ramdisk_execute_command);
 811         }
 812
 813         /*
 814          * We try each of these until one succeeds.
 815          *
 816          * The Bourne shell can be used instead of init if we are
 817          * trying to recover a really broken machine.
 818          */
 819         if (execute_command) {
 820                 run_init_process(execute_command);
 821                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 822                                         "defaults...\n", execute_command);
 823         }
 824         run_init_process("/sbin/init");
 825         run_init_process("/etc/init");
 826         run_init_process("/bin/init");
 827         run_init_process("/bin/sh");
 828
 829         panic("No init found.  Try passing init= option to kernel.");
 830 }
 831
 832 static int __init kernel_init(void * unused)
 833 {
 834         lock_kernel();
 835         /*
 836          * init can run on any cpu.
 837          */
 838         set_cpus_allowed_ptr(current, CPU_MASK_ALL_PTR);
 839         /*
 840          * Tell the world that we're going to be the grim
 841          * reaper of innocent orphaned children.
 842          *
 843          * We don't want people to have to make incorrect
 844          * assumptions about where in the task array this
 845          * can be found.
 846          */
 847         init_pid_ns.child_reaper = current;
 848
 849         cad_pid = task_pid(current);
 850
 851         smp_prepare_cpus(setup_max_cpus);
 852
 853         do_pre_smp_initcalls();
 854         start_boot_trace();
 855
 856         smp_init();
 857         sched_init_smp();
 858
 859         cpuset_init_smp();
 860
 861         do_basic_setup();
 862
 863         /*
 864          * check if there is an early userspace init.  If yes, let it do all
 865          * the work
 866          */
 867
 868         if (!ramdisk_execute_command)
 869                 ramdisk_execute_command = "/init";
 870
 871         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 872                 ramdisk_execute_command = NULL;
 873                 prepare_namespace();
 874         }
 875
 876         /*
 877          * Ok, we have completed the initial bootup, and
 878          * we're essentially up and running. Get rid of the
 879          * initmem segments and start the user-mode stuff..
 880          */
 881
 882         init_post();
 883         return 0;
 884 }