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/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/acpi.h>
  27 #include <linux/tty.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/profile.h>
  36 #include <linux/rcupdate.h>
  37 #include <linux/moduleparam.h>
  38 #include <linux/kallsyms.h>
  39 #include <linux/writeback.h>
  40 #include <linux/cpu.h>
  41 #include <linux/cpuset.h>
  42 #include <linux/cgroup.h>
  43 #include <linux/efi.h>
  44 #include <linux/tick.h>
  45 #include <linux/interrupt.h>
  46 #include <linux/taskstats_kern.h>
  47 #include <linux/delayacct.h>
  48 #include <linux/unistd.h>
  49 #include <linux/rmap.h>
  50 #include <linux/mempolicy.h>
  51 #include <linux/key.h>
  52 #include <linux/buffer_head.h>
  53 #include <linux/page_cgroup.h>
  54 #include <linux/debug_locks.h>
  55 #include <linux/debugobjects.h>
  56 #include <linux/lockdep.h>
  57 #include <linux/kmemleak.h>
  58 #include <linux/pid_namespace.h>
  59 #include <linux/device.h>
  60 #include <linux/kthread.h>
  61 #include <linux/sched.h>
  62 #include <linux/signal.h>
  63 #include <linux/idr.h>
  64 #include <linux/kgdb.h>
  65 #include <linux/ftrace.h>
  66 #include <linux/async.h>
  67 #include <linux/kmemcheck.h>
  68 #include <linux/sfi.h>
  69 #include <linux/shmem_fs.h>
  70 #include <linux/slab.h>
  71
  72 #include <asm/io.h>
  73 #include <asm/bugs.h>
  74 #include <asm/setup.h>
  75 #include <asm/sections.h>
  76 #include <asm/cacheflush.h>
  77
  78 #ifdef CONFIG_X86_LOCAL_APIC
  79 #include <asm/smp.h>
  80 #endif
  81
  82 static int kernel_init(void *);
  83
  84 extern void init_IRQ(void);
  85 extern void fork_init(unsigned long);
  86 extern void mca_init(void);
  87 extern void sbus_init(void);
  88 extern void prio_tree_init(void);
  89 extern void radix_tree_init(void);
  90 extern void free_initmem(void);
  91 #ifndef CONFIG_DEBUG_RODATA
  92 static inline void mark_rodata_ro(void) { }
  93 #endif
  94
  95 #ifdef CONFIG_TC
  96 extern void tc_init(void);
  97 #endif
  98
  99 enum system_states system_state __read_mostly;
 100 EXPORT_SYMBOL(system_state);
 101
 102 /*
 103  * Boot command-line arguments
 104  */
 105 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 106 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 107
 108 extern void time_init(void);
 109 /* Default late time init is NULL. archs can override this later. */
 110 void (*__initdata late_time_init)(void);
 111 extern void softirq_init(void);
 112
 113 /* Untouched command line saved by arch-specific code. */
 114 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 115 /* Untouched saved command line (eg. for /proc) */
 116 char *saved_command_line;
 117 /* Command line for parameter parsing */
 118 static char *static_command_line;
 119
 120 static char *execute_command;
 121 static char *ramdisk_execute_command;
 122
 123 #ifdef CONFIG_SMP
 124 /* Setup configured maximum number of CPUs to activate */
 125 unsigned int setup_max_cpus = NR_CPUS;
 126 EXPORT_SYMBOL(setup_max_cpus);
 127
 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 /* this is hard limit */
 153 static int __init nrcpus(char *str)
 154 {
 155         int nr_cpus;
 156
 157         get_option(&str, &nr_cpus);
 158         if (nr_cpus > 0 && nr_cpus < nr_cpu_ids)
 159                 nr_cpu_ids = nr_cpus;
 160
 161         return 0;
 162 }
 163
 164 early_param("nr_cpus", nrcpus);
 165
 166 static int __init maxcpus(char *str)
 167 {
 168         get_option(&str, &setup_max_cpus);
 169         if (setup_max_cpus == 0)
 170                 arch_disable_smp_support();
 171
 172         return 0;
 173 }
 174
 175 early_param("maxcpus", maxcpus);
 176 #else
 177 static const unsigned int setup_max_cpus = NR_CPUS;
 178 #endif
 179
 180 /*
 181  * If set, this is an indication to the drivers that reset the underlying
 182  * device before going ahead with the initialization otherwise driver might
 183  * rely on the BIOS and skip the reset operation.
 184  *
 185  * This is useful if kernel is booting in an unreliable environment.
 186  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 187  * skipped and devices will be in unknown state.
 188  */
 189 unsigned int reset_devices;
 190 EXPORT_SYMBOL(reset_devices);
 191
 192 static int __init set_reset_devices(char *str)
 193 {
 194         reset_devices = 1;
 195         return 1;
 196 }
 197
 198 __setup("reset_devices", set_reset_devices);
 199
 200 static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 201 const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 202 static const char *panic_later, *panic_param;
 203
 204 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 205
 206 static int __init obsolete_checksetup(char *line)
 207 {
 208         const struct obs_kernel_param *p;
 209         int had_early_param = 0;
 210
 211         p = __setup_start;
 212         do {
 213                 int n = strlen(p->str);
 214                 if (!strncmp(line, p->str, n)) {
 215                         if (p->early) {
 216                                 /* Already done in parse_early_param?
 217                                  * (Needs exact match on param part).
 218                                  * Keep iterating, as we can have early
 219                                  * params and __setups of same names 8( */
 220                                 if (line[n] == '\0' || line[n] == '=')
 221                                         had_early_param = 1;
 222                         } else if (!p->setup_func) {
 223                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 224                                        " ignored\n", p->str);
 225                                 return 1;
 226                         } else if (p->setup_func(line + n))
 227                                 return 1;
 228                 }
 229                 p++;
 230         } while (p < __setup_end);
 231
 232         return had_early_param;
 233 }
 234
 235 /*
 236  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 237  * still work even if initially too large, it will just take slightly longer
 238  */
 239 unsigned long loops_per_jiffy = (1<<12);
 240
 241 EXPORT_SYMBOL(loops_per_jiffy);
 242
 243 static int __init debug_kernel(char *str)
 244 {
 245         console_loglevel = 10;
 246         return 0;
 247 }
 248
 249 static int __init quiet_kernel(char *str)
 250 {
 251         console_loglevel = 4;
 252         return 0;
 253 }
 254
 255 early_param("debug", debug_kernel);
 256 early_param("quiet", quiet_kernel);
 257
 258 static int __init loglevel(char *str)
 259 {
 260         get_option(&str, &console_loglevel);
 261         return 0;
 262 }
 263
 264 early_param("loglevel", loglevel);
 265
 266 /*
 267  * Unknown boot options get handed to init, unless they look like
 268  * unused parameters (modprobe will find them in /proc/cmdline).
 269  */
 270 static int __init unknown_bootoption(char *param, char *val)
 271 {
 272         /* Change NUL term back to "=", to make "param" the whole string. */
 273         if (val) {
 274                 /* param=val or param="val"? */
 275                 if (val == param+strlen(param)+1)
 276                         val[-1] = '=';
 277                 else if (val == param+strlen(param)+2) {
 278                         val[-2] = '=';
 279                         memmove(val-1, val, strlen(val)+1);
 280                         val--;
 281                 } else
 282                         BUG();
 283         }
 284
 285         /* Handle obsolete-style parameters */
 286         if (obsolete_checksetup(param))
 287                 return 0;
 288
 289         /* Unused module parameter. */
 290         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 291                 return 0;
 292
 293         if (panic_later)
 294                 return 0;
 295
 296         if (val) {
 297                 /* Environment option */
 298                 unsigned int i;
 299                 for (i = 0; envp_init[i]; i++) {
 300                         if (i == MAX_INIT_ENVS) {
 301                                 panic_later = "Too many boot env vars at `%s'";
 302                                 panic_param = param;
 303                         }
 304                         if (!strncmp(param, envp_init[i], val - param))
 305                                 break;
 306                 }
 307                 envp_init[i] = param;
 308         } else {
 309                 /* Command line option */
 310                 unsigned int i;
 311                 for (i = 0; argv_init[i]; i++) {
 312                         if (i == MAX_INIT_ARGS) {
 313                                 panic_later = "Too many boot init vars at `%s'";
 314                                 panic_param = param;
 315                         }
 316                 }
 317                 argv_init[i] = param;
 318         }
 319         return 0;
 320 }
 321
 322 #ifdef CONFIG_DEBUG_PAGEALLOC
 323 int __read_mostly debug_pagealloc_enabled = 0;
 324 #endif
 325
 326 static int __init init_setup(char *str)
 327 {
 328         unsigned int i;
 329
 330         execute_command = str;
 331         /*
 332          * In case LILO is going to boot us with default command line,
 333          * it prepends "auto" before the whole cmdline which makes
 334          * the shell think it should execute a script with such name.
 335          * So we ignore all arguments entered _before_ init=... [MJ]
 336          */
 337         for (i = 1; i < MAX_INIT_ARGS; i++)
 338                 argv_init[i] = NULL;
 339         return 1;
 340 }
 341 __setup("init=", init_setup);
 342
 343 static int __init rdinit_setup(char *str)
 344 {
 345         unsigned int i;
 346
 347         ramdisk_execute_command = str;
 348         /* See "auto" comment in init_setup */
 349         for (i = 1; i < MAX_INIT_ARGS; i++)
 350                 argv_init[i] = NULL;
 351         return 1;
 352 }
 353 __setup("rdinit=", rdinit_setup);
 354
 355 #ifndef CONFIG_SMP
 356
 357 #ifdef CONFIG_X86_LOCAL_APIC
 358 static void __init smp_init(void)
 359 {
 360         APIC_init_uniprocessor();
 361 }
 362 #else
 363 #define smp_init()      do { } while (0)
 364 #endif
 365
 366 static inline void setup_nr_cpu_ids(void) { }
 367 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 368
 369 #else
 370
 371 /* Setup number of possible processor ids */
 372 int nr_cpu_ids __read_mostly = NR_CPUS;
 373 EXPORT_SYMBOL(nr_cpu_ids);
 374
 375 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
 376 static void __init setup_nr_cpu_ids(void)
 377 {
 378         nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
 379 }
 380
 381 /* Called by boot processor to activate the rest. */
 382 static void __init smp_init(void)
 383 {
 384         unsigned int cpu;
 385
 386         /* FIXME: This should be done in userspace --RR */
 387         for_each_present_cpu(cpu) {
 388                 if (num_online_cpus() >= setup_max_cpus)
 389                         break;
 390                 if (!cpu_online(cpu))
 391                         cpu_up(cpu);
 392         }
 393
 394         /* Any cleanup work */
 395         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 396         smp_cpus_done(setup_max_cpus);
 397 }
 398
 399 #endif
 400
 401 /*
 402  * We need to store the untouched command line for future reference.
 403  * We also need to store the touched command line since the parameter
 404  * parsing is performed in place, and we should allow a component to
 405  * store reference of name/value for future reference.
 406  */
 407 static void __init setup_command_line(char *command_line)
 408 {
 409         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 410         static_command_line = alloc_bootmem(strlen (command_line)+1);
 411         strcpy (saved_command_line, boot_command_line);
 412         strcpy (static_command_line, command_line);
 413 }
 414
 415 /*
 416  * We need to finalize in a non-__init function or else race conditions
 417  * between the root thread and the init thread may cause start_kernel to
 418  * be reaped by free_initmem before the root thread has proceeded to
 419  * cpu_idle.
 420  *
 421  * gcc-3.4 accidentally inlines this function, so use noinline.
 422  */
 423
 424 static __initdata DECLARE_COMPLETION(kthreadd_done);
 425
 426 static noinline void __init_refok rest_init(void)
 427 {
 428         int pid;
 429
 430         rcu_scheduler_starting();
 431         /*
 432          * We need to spawn init first so that it obtains pid 1, however
 433          * the init task will end up wanting to create kthreads, which, if
 434          * we schedule it before we create kthreadd, will OOPS.
 435          */
 436         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 437         numa_default_policy();
 438         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 439         rcu_read_lock();
 440         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 441         rcu_read_unlock();
 442         complete(&kthreadd_done);
 443
 444         /*
 445          * The boot idle thread must execute schedule()
 446          * at least once to get things moving:
 447          */
 448         init_idle_bootup_task(current);
 449         preempt_enable_no_resched();
 450         schedule();
 451         preempt_disable();
 452
 453         /* Call into cpu_idle with preempt disabled */
 454         cpu_idle();
 455 }
 456
 457 /* Check for early params. */
 458 static int __init do_early_param(char *param, char *val)
 459 {
 460         const struct obs_kernel_param *p;
 461
 462         for (p = __setup_start; p < __setup_end; p++) {
 463                 if ((p->early && strcmp(param, p->str) == 0) ||
 464                     (strcmp(param, "console") == 0 &&
 465                      strcmp(p->str, "earlycon") == 0)
 466                 ) {
 467                         if (p->setup_func(val) != 0)
 468                                 printk(KERN_WARNING
 469                                        "Malformed early option '%s'\n", param);
 470                 }
 471         }
 472         /* We accept everything at this stage. */
 473         return 0;
 474 }
 475
 476 void __init parse_early_options(char *cmdline)
 477 {
 478         parse_args("early options", cmdline, NULL, 0, do_early_param);
 479 }
 480
 481 /* Arch code calls this early on, or if not, just before other parsing. */
 482 void __init parse_early_param(void)
 483 {
 484         static __initdata int done = 0;
 485         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 486
 487         if (done)
 488                 return;
 489
 490         /* All fall through to do_early_param. */
 491         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 492         parse_early_options(tmp_cmdline);
 493         done = 1;
 494 }
 495
 496 /*
 497  *      Activate the first processor.
 498  */
 499
 500 static void __init boot_cpu_init(void)
 501 {
 502         int cpu = smp_processor_id();
 503         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 504         set_cpu_online(cpu, true);
 505         set_cpu_active(cpu, true);
 506         set_cpu_present(cpu, true);
 507         set_cpu_possible(cpu, true);
 508 }
 509
 510 void __init __weak smp_setup_processor_id(void)
 511 {
 512 }
 513
 514 void __init __weak thread_info_cache_init(void)
 515 {
 516 }
 517
 518 /*
 519  * Set up kernel memory allocators
 520  */
 521 static void __init mm_init(void)
 522 {
 523         /*
 524          * page_cgroup requires countinous pages as memmap
 525          * and it's bigger than MAX_ORDER unless SPARSEMEM.
 526          */
 527         page_cgroup_init_flatmem();
 528         mem_init();
 529         kmem_cache_init();
 530         percpu_init_late();
 531         pgtable_cache_init();
 532         vmalloc_init();
 533 }
 534
 535 asmlinkage void __init start_kernel(void)
 536 {
 537         char * command_line;
 538         extern const struct kernel_param __start___param[], __stop___param[];
 539
 540         smp_setup_processor_id();
 541
 542         /*
 543          * Need to run as early as possible, to initialize the
 544          * lockdep hash:
 545          */
 546         lockdep_init();
 547         debug_objects_early_init();
 548
 549         /*
 550          * Set up the the initial canary ASAP:
 551          */
 552         boot_init_stack_canary();
 553
 554         cgroup_init_early();
 555
 556         local_irq_disable();
 557         early_boot_irqs_off();
 558
 559 /*
 560  * Interrupts are still disabled. Do necessary setups, then
 561  * enable them
 562  */
 563         tick_init();
 564         boot_cpu_init();
 565         page_address_init();
 566         printk(KERN_NOTICE "%s", linux_banner);
 567         setup_arch(&command_line);
 568         mm_init_owner(&init_mm, &init_task);
 569         setup_command_line(command_line);
 570         setup_nr_cpu_ids();
 571         setup_per_cpu_areas();
 572         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 573
 574         build_all_zonelists(NULL);
 575         page_alloc_init();
 576
 577         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 578         parse_early_param();
 579         parse_args("Booting kernel", static_command_line, __start___param,
 580                    __stop___param - __start___param,
 581                    &unknown_bootoption);
 582         /*
 583          * These use large bootmem allocations and must precede
 584          * kmem_cache_init()
 585          */
 586         pidhash_init();
 587         vfs_caches_init_early();
 588         sort_main_extable();
 589         trap_init();
 590         mm_init();
 591         /*
 592          * Set up the scheduler prior starting any interrupts (such as the
 593          * timer interrupt). Full topology setup happens at smp_init()
 594          * time - but meanwhile we still have a functioning scheduler.
 595          */
 596         sched_init();
 597         /*
 598          * Disable preemption - early bootup scheduling is extremely
 599          * fragile until we cpu_idle() for the first time.
 600          */
 601         preempt_disable();
 602         if (!irqs_disabled()) {
 603                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 604                                 "enabled *very* early, fixing it\n");
 605                 local_irq_disable();
 606         }
 607         rcu_init();
 608         radix_tree_init();
 609         /* init some links before init_ISA_irqs() */
 610         early_irq_init();
 611         init_IRQ();
 612         prio_tree_init();
 613         init_timers();
 614         hrtimers_init();
 615         softirq_init();
 616         timekeeping_init();
 617         time_init();
 618         profile_init();
 619         if (!irqs_disabled())
 620                 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
 621                                  "enabled early\n");
 622         early_boot_irqs_on();
 623         local_irq_enable();
 624
 625         /* Interrupts are enabled now so all GFP allocations are safe. */
 626         gfp_allowed_mask = __GFP_BITS_MASK;
 627
 628         kmem_cache_init_late();
 629
 630         /*
 631          * HACK ALERT! This is early. We're enabling the console before
 632          * we've done PCI setups etc, and console_init() must be aware of
 633          * this. But we do want output early, in case something goes wrong.
 634          */
 635         console_init();
 636         if (panic_later)
 637                 panic(panic_later, panic_param);
 638
 639         lockdep_info();
 640
 641         /*
 642          * Need to run this when irqs are enabled, because it wants
 643          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 644          * too:
 645          */
 646         locking_selftest();
 647
 648 #ifdef CONFIG_BLK_DEV_INITRD
 649         if (initrd_start && !initrd_below_start_ok &&
 650             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 651                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 652                     "disabling it.\n",
 653                     page_to_pfn(virt_to_page((void *)initrd_start)),
 654                     min_low_pfn);
 655                 initrd_start = 0;
 656         }
 657 #endif
 658         page_cgroup_init();
 659         enable_debug_pagealloc();
 660         kmemleak_init();
 661         debug_objects_mem_init();
 662         idr_init_cache();
 663         setup_per_cpu_pageset();
 664         numa_policy_init();
 665         if (late_time_init)
 666                 late_time_init();
 667         sched_clock_init();
 668         calibrate_delay();
 669         pidmap_init();
 670         anon_vma_init();
 671 #ifdef CONFIG_X86
 672         if (efi_enabled)
 673                 efi_enter_virtual_mode();
 674 #endif
 675         thread_info_cache_init();
 676         cred_init();
 677         fork_init(totalram_pages);
 678         proc_caches_init();
 679         buffer_init();
 680         key_init();
 681         security_init();
 682         dbg_late_init();
 683         vfs_caches_init(totalram_pages);
 684         signals_init();
 685         /* rootfs populating might need page-writeback */
 686         page_writeback_init();
 687 #ifdef CONFIG_PROC_FS
 688         proc_root_init();
 689 #endif
 690         cgroup_init();
 691         cpuset_init();
 692         taskstats_init_early();
 693         delayacct_init();
 694
 695         check_bugs();
 696
 697         acpi_early_init(); /* before LAPIC and SMP init */
 698         sfi_init_late();
 699
 700         ftrace_init();
 701
 702         /* Do the rest non-__init'ed, we're now alive */
 703         rest_init();
 704 }
 705
 706 /* Call all constructor functions linked into the kernel. */
 707 static void __init do_ctors(void)
 708 {
 709 #ifdef CONFIG_CONSTRUCTORS
 710         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 711
 712         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 713                 (*fn)();
 714 #endif
 715 }
 716
 717 int initcall_debug;
 718 core_param(initcall_debug, initcall_debug, bool, 0644);
 719
 720 static char msgbuf[64];
 721
 722 static int __init_or_module do_one_initcall_debug(initcall_t fn)
 723 {
 724         ktime_t calltime, delta, rettime;
 725         unsigned long long duration;
 726         int ret;
 727
 728         printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
 729         calltime = ktime_get();
 730         ret = fn();
 731         rettime = ktime_get();
 732         delta = ktime_sub(rettime, calltime);
 733         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 734         printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn,
 735                 ret, duration);
 736
 737         return ret;
 738 }
 739
 740 int __init_or_module do_one_initcall(initcall_t fn)
 741 {
 742         int count = preempt_count();
 743         int ret;
 744
 745         if (initcall_debug)
 746                 ret = do_one_initcall_debug(fn);
 747         else
 748                 ret = fn();
 749
 750         msgbuf[0] = 0;
 751
 752         if (ret && ret != -ENODEV && initcall_debug)
 753                 sprintf(msgbuf, "error code %d ", ret);
 754
 755         if (preempt_count() != count) {
 756                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 757                 preempt_count() = count;
 758         }
 759         if (irqs_disabled()) {
 760                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 761                 local_irq_enable();
 762         }
 763         if (msgbuf[0]) {
 764                 printk("initcall %pF returned with %s\n", fn, msgbuf);
 765         }
 766
 767         return ret;
 768 }
 769
 770
 771 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
 772
 773 static void __init do_initcalls(void)
 774 {
 775         initcall_t *fn;
 776
 777         for (fn = __early_initcall_end; fn < __initcall_end; fn++)
 778                 do_one_initcall(*fn);
 779
 780         /* Make sure there is no pending stuff from the initcall sequence */
 781         flush_scheduled_work();
 782 }
 783
 784 /*
 785  * Ok, the machine is now initialized. None of the devices
 786  * have been touched yet, but the CPU subsystem is up and
 787  * running, and memory and process management works.
 788  *
 789  * Now we can finally start doing some real work..
 790  */
 791 static void __init do_basic_setup(void)
 792 {
 793         cpuset_init_smp();
 794         usermodehelper_init();
 795         init_tmpfs();
 796         driver_init();
 797         init_irq_proc();
 798         do_ctors();
 799         do_initcalls();
 800 }
 801
 802 static void __init do_pre_smp_initcalls(void)
 803 {
 804         initcall_t *fn;
 805
 806         for (fn = __initcall_start; fn < __early_initcall_end; fn++)
 807                 do_one_initcall(*fn);
 808 }
 809
 810 static void run_init_process(const char *init_filename)
 811 {
 812         argv_init[0] = init_filename;
 813         kernel_execve(init_filename, argv_init, envp_init);
 814 }
 815
 816 /* This is a non __init function. Force it to be noinline otherwise gcc
 817  * makes it inline to init() and it becomes part of init.text section
 818  */
 819 static noinline int init_post(void)
 820 {
 821         /* need to finish all async __init code before freeing the memory */
 822         async_synchronize_full();
 823         free_initmem();
 824         mark_rodata_ro();
 825         system_state = SYSTEM_RUNNING;
 826         numa_default_policy();
 827
 828
 829         current->signal->flags |= SIGNAL_UNKILLABLE;
 830
 831         if (ramdisk_execute_command) {
 832                 run_init_process(ramdisk_execute_command);
 833                 printk(KERN_WARNING "Failed to execute %s\n",
 834                                 ramdisk_execute_command);
 835         }
 836
 837         /*
 838          * We try each of these until one succeeds.
 839          *
 840          * The Bourne shell can be used instead of init if we are
 841          * trying to recover a really broken machine.
 842          */
 843         if (execute_command) {
 844                 run_init_process(execute_command);
 845                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 846                                         "defaults...\n", execute_command);
 847         }
 848         run_init_process("/sbin/init");
 849         run_init_process("/etc/init");
 850         run_init_process("/bin/init");
 851         run_init_process("/bin/sh");
 852
 853         panic("No init found.  Try passing init= option to kernel. "
 854               "See Linux Documentation/init.txt for guidance.");
 855 }
 856
 857 static int __init kernel_init(void * unused)
 858 {
 859         /*
 860          * Wait until kthreadd is all set-up.
 861          */
 862         wait_for_completion(&kthreadd_done);
 863         /*
 864          * init can allocate pages on any node
 865          */
 866         set_mems_allowed(node_states[N_HIGH_MEMORY]);
 867         /*
 868          * init can run on any cpu.
 869          */
 870         set_cpus_allowed_ptr(current, cpu_all_mask);
 871         /*
 872          * Tell the world that we're going to be the grim
 873          * reaper of innocent orphaned children.
 874          *
 875          * We don't want people to have to make incorrect
 876          * assumptions about where in the task array this
 877          * can be found.
 878          */
 879         init_pid_ns.child_reaper = current;
 880
 881         cad_pid = task_pid(current);
 882
 883         smp_prepare_cpus(setup_max_cpus);
 884
 885         do_pre_smp_initcalls();
 886
 887         smp_init();
 888         sched_init_smp();
 889
 890         do_basic_setup();
 891
 892         /* Open the /dev/console on the rootfs, this should never fail */
 893         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 894                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 895
 896         (void) sys_dup(0);
 897         (void) sys_dup(0);
 898         /*
 899          * check if there is an early userspace init.  If yes, let it do all
 900          * the work
 901          */
 902
 903         if (!ramdisk_execute_command)
 904                 ramdisk_execute_command = "/init";
 905
 906         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 907                 ramdisk_execute_command = NULL;
 908                 prepare_namespace();
 909         }
 910
 911         /*
 912          * Ok, we have completed the initial bootup, and
 913          * we're essentially up and running. Get rid of the
 914          * initmem segments and start the user-mode stuff..
 915          */
 916
 917         init_post();
 918         return 0;
 919 }