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/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/kmemleak.h>
  59 #include <linux/pid_namespace.h>
  60 #include <linux/device.h>
  61 #include <linux/kthread.h>
  62 #include <linux/sched.h>
  63 #include <linux/signal.h>
  64 #include <linux/idr.h>
  65 #include <linux/kgdb.h>
  66 #include <linux/ftrace.h>
  67 #include <linux/async.h>
  68 #include <linux/kmemcheck.h>
  69 #include <linux/sfi.h>
  70 #include <linux/shmem_fs.h>
  71 #include <linux/slab.h>
  72
  73 #include <asm/io.h>
  74 #include <asm/bugs.h>
  75 #include <asm/setup.h>
  76 #include <asm/sections.h>
  77 #include <asm/cacheflush.h>
  78
  79 #ifdef CONFIG_X86_LOCAL_APIC
  80 #include <asm/smp.h>
  81 #endif
  82
  83 static int kernel_init(void *);
  84
  85 extern void init_IRQ(void);
  86 extern void fork_init(unsigned long);
  87 extern void mca_init(void);
  88 extern void sbus_init(void);
  89 extern void prio_tree_init(void);
  90 extern void radix_tree_init(void);
  91 extern void free_initmem(void);
  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 setup_max_cpus = NR_CPUS;
 127 EXPORT_SYMBOL(setup_max_cpus);
 128
 129
 130 /*
 131  * Setup routine for controlling SMP activation
 132  *
 133  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 134  * activation entirely (the MPS table probe still happens, though).
 135  *
 136  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 137  * greater than 0, limits the maximum number of CPUs activated in
 138  * SMP mode to <NUM>.
 139  */
 140
 141 void __weak arch_disable_smp_support(void) { }
 142
 143 static int __init nosmp(char *str)
 144 {
 145         setup_max_cpus = 0;
 146         arch_disable_smp_support();
 147
 148         return 0;
 149 }
 150
 151 early_param("nosmp", nosmp);
 152
 153 /* this is hard limit */
 154 static int __init nrcpus(char *str)
 155 {
 156         int nr_cpus;
 157
 158         get_option(&str, &nr_cpus);
 159         if (nr_cpus > 0 && nr_cpus < nr_cpu_ids)
 160                 nr_cpu_ids = nr_cpus;
 161
 162         return 0;
 163 }
 164
 165 early_param("nr_cpus", nrcpus);
 166
 167 static int __init maxcpus(char *str)
 168 {
 169         get_option(&str, &setup_max_cpus);
 170         if (setup_max_cpus == 0)
 171                 arch_disable_smp_support();
 172
 173         return 0;
 174 }
 175
 176 early_param("maxcpus", maxcpus);
 177 #else
 178 static const unsigned int setup_max_cpus = NR_CPUS;
 179 #endif
 180
 181 /*
 182  * If set, this is an indication to the drivers that reset the underlying
 183  * device before going ahead with the initialization otherwise driver might
 184  * rely on the BIOS and skip the reset operation.
 185  *
 186  * This is useful if kernel is booting in an unreliable environment.
 187  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 188  * skipped and devices will be in unknown state.
 189  */
 190 unsigned int reset_devices;
 191 EXPORT_SYMBOL(reset_devices);
 192
 193 static int __init set_reset_devices(char *str)
 194 {
 195         reset_devices = 1;
 196         return 1;
 197 }
 198
 199 __setup("reset_devices", set_reset_devices);
 200
 201 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 202 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 203 static const char *panic_later, *panic_param;
 204
 205 extern struct obs_kernel_param __setup_start[], __setup_end[];
 206
 207 static int __init obsolete_checksetup(char *line)
 208 {
 209         struct obs_kernel_param *p;
 210         int had_early_param = 0;
 211
 212         p = __setup_start;
 213         do {
 214                 int n = strlen(p->str);
 215                 if (!strncmp(line, p->str, n)) {
 216                         if (p->early) {
 217                                 /* Already done in parse_early_param?
 218                                  * (Needs exact match on param part).
 219                                  * Keep iterating, as we can have early
 220                                  * params and __setups of same names 8( */
 221                                 if (line[n] == '\0' || line[n] == '=')
 222                                         had_early_param = 1;
 223                         } else if (!p->setup_func) {
 224                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 225                                        " ignored\n", p->str);
 226                                 return 1;
 227                         } else if (p->setup_func(line + n))
 228                                 return 1;
 229                 }
 230                 p++;
 231         } while (p < __setup_end);
 232
 233         return had_early_param;
 234 }
 235
 236 /*
 237  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 238  * still work even if initially too large, it will just take slightly longer
 239  */
 240 unsigned long loops_per_jiffy = (1<<12);
 241
 242 EXPORT_SYMBOL(loops_per_jiffy);
 243
 244 static int __init debug_kernel(char *str)
 245 {
 246         console_loglevel = 10;
 247         return 0;
 248 }
 249
 250 static int __init quiet_kernel(char *str)
 251 {
 252         console_loglevel = 4;
 253         return 0;
 254 }
 255
 256 early_param("debug", debug_kernel);
 257 early_param("quiet", quiet_kernel);
 258
 259 static int __init loglevel(char *str)
 260 {
 261         get_option(&str, &console_loglevel);
 262         return 0;
 263 }
 264
 265 early_param("loglevel", loglevel);
 266
 267 /*
 268  * Unknown boot options get handed to init, unless they look like
 269  * unused parameters (modprobe will find them in /proc/cmdline).
 270  */
 271 static int __init unknown_bootoption(char *param, char *val)
 272 {
 273         /* Change NUL term back to "=", to make "param" the whole string. */
 274         if (val) {
 275                 /* param=val or param="val"? */
 276                 if (val == param+strlen(param)+1)
 277                         val[-1] = '=';
 278                 else if (val == param+strlen(param)+2) {
 279                         val[-2] = '=';
 280                         memmove(val-1, val, strlen(val)+1);
 281                         val--;
 282                 } else
 283                         BUG();
 284         }
 285
 286         /* Handle obsolete-style parameters */
 287         if (obsolete_checksetup(param))
 288                 return 0;
 289
 290         /* Unused module parameter. */
 291         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 292                 return 0;
 293
 294         if (panic_later)
 295                 return 0;
 296
 297         if (val) {
 298                 /* Environment option */
 299                 unsigned int i;
 300                 for (i = 0; envp_init[i]; i++) {
 301                         if (i == MAX_INIT_ENVS) {
 302                                 panic_later = "Too many boot env vars at `%s'";
 303                                 panic_param = param;
 304                         }
 305                         if (!strncmp(param, envp_init[i], val - param))
 306                                 break;
 307                 }
 308                 envp_init[i] = param;
 309         } else {
 310                 /* Command line option */
 311                 unsigned int i;
 312                 for (i = 0; argv_init[i]; i++) {
 313                         if (i == MAX_INIT_ARGS) {
 314                                 panic_later = "Too many boot init vars at `%s'";
 315                                 panic_param = param;
 316                         }
 317                 }
 318                 argv_init[i] = param;
 319         }
 320         return 0;
 321 }
 322
 323 #ifdef CONFIG_DEBUG_PAGEALLOC
 324 int __read_mostly debug_pagealloc_enabled = 0;
 325 #endif
 326
 327 static int __init init_setup(char *str)
 328 {
 329         unsigned int i;
 330
 331         execute_command = str;
 332         /*
 333          * In case LILO is going to boot us with default command line,
 334          * it prepends "auto" before the whole cmdline which makes
 335          * the shell think it should execute a script with such name.
 336          * So we ignore all arguments entered _before_ init=... [MJ]
 337          */
 338         for (i = 1; i < MAX_INIT_ARGS; i++)
 339                 argv_init[i] = NULL;
 340         return 1;
 341 }
 342 __setup("init=", init_setup);
 343
 344 static int __init rdinit_setup(char *str)
 345 {
 346         unsigned int i;
 347
 348         ramdisk_execute_command = str;
 349         /* See "auto" comment in init_setup */
 350         for (i = 1; i < MAX_INIT_ARGS; i++)
 351                 argv_init[i] = NULL;
 352         return 1;
 353 }
 354 __setup("rdinit=", rdinit_setup);
 355
 356 #ifndef CONFIG_SMP
 357
 358 #ifdef CONFIG_X86_LOCAL_APIC
 359 static void __init smp_init(void)
 360 {
 361         APIC_init_uniprocessor();
 362 }
 363 #else
 364 #define smp_init()      do { } while (0)
 365 #endif
 366
 367 static inline void setup_nr_cpu_ids(void) { }
 368 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 369
 370 #else
 371
 372 /* Setup number of possible processor ids */
 373 int nr_cpu_ids __read_mostly = NR_CPUS;
 374 EXPORT_SYMBOL(nr_cpu_ids);
 375
 376 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
 377 static void __init setup_nr_cpu_ids(void)
 378 {
 379         nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
 380 }
 381
 382 /* Called by boot processor to activate the rest. */
 383 static void __init smp_init(void)
 384 {
 385         unsigned int cpu;
 386
 387         /* FIXME: This should be done in userspace --RR */
 388         for_each_present_cpu(cpu) {
 389                 if (num_online_cpus() >= setup_max_cpus)
 390                         break;
 391                 if (!cpu_online(cpu))
 392                         cpu_up(cpu);
 393         }
 394
 395         /* Any cleanup work */
 396         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 397         smp_cpus_done(setup_max_cpus);
 398 }
 399
 400 #endif
 401
 402 /*
 403  * We need to store the untouched command line for future reference.
 404  * We also need to store the touched command line since the parameter
 405  * parsing is performed in place, and we should allow a component to
 406  * store reference of name/value for future reference.
 407  */
 408 static void __init setup_command_line(char *command_line)
 409 {
 410         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 411         static_command_line = alloc_bootmem(strlen (command_line)+1);
 412         strcpy (saved_command_line, boot_command_line);
 413         strcpy (static_command_line, command_line);
 414 }
 415
 416 /*
 417  * We need to finalize in a non-__init function or else race conditions
 418  * between the root thread and the init thread may cause start_kernel to
 419  * be reaped by free_initmem before the root thread has proceeded to
 420  * cpu_idle.
 421  *
 422  * gcc-3.4 accidentally inlines this function, so use noinline.
 423  */
 424
 425 static __initdata DECLARE_COMPLETION(kthreadd_done);
 426
 427 static noinline void __init_refok rest_init(void)
 428         __releases(kernel_lock)
 429 {
 430         int pid;
 431
 432         rcu_scheduler_starting();
 433         /*
 434          * We need to spawn init first so that it obtains pid 1, however
 435          * the init task will end up wanting to create kthreads, which, if
 436          * we schedule it before we create kthreadd, will OOPS.
 437          */
 438         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 439         numa_default_policy();
 440         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 441         rcu_read_lock();
 442         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 443         rcu_read_unlock();
 444         complete(&kthreadd_done);
 445         unlock_kernel();
 446
 447         /*
 448          * The boot idle thread must execute schedule()
 449          * at least once to get things moving:
 450          */
 451         init_idle_bootup_task(current);
 452         preempt_enable_no_resched();
 453         schedule();
 454         preempt_disable();
 455
 456         /* Call into cpu_idle with preempt disabled */
 457         cpu_idle();
 458 }
 459
 460 /* Check for early params. */
 461 static int __init do_early_param(char *param, char *val)
 462 {
 463         struct obs_kernel_param *p;
 464
 465         for (p = __setup_start; p < __setup_end; p++) {
 466                 if ((p->early && strcmp(param, p->str) == 0) ||
 467                     (strcmp(param, "console") == 0 &&
 468                      strcmp(p->str, "earlycon") == 0)
 469                 ) {
 470                         if (p->setup_func(val) != 0)
 471                                 printk(KERN_WARNING
 472                                        "Malformed early option '%s'\n", param);
 473                 }
 474         }
 475         /* We accept everything at this stage. */
 476         return 0;
 477 }
 478
 479 void __init parse_early_options(char *cmdline)
 480 {
 481         parse_args("early options", cmdline, NULL, 0, do_early_param);
 482 }
 483
 484 /* Arch code calls this early on, or if not, just before other parsing. */
 485 void __init parse_early_param(void)
 486 {
 487         static __initdata int done = 0;
 488         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 489
 490         if (done)
 491                 return;
 492
 493         /* All fall through to do_early_param. */
 494         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 495         parse_early_options(tmp_cmdline);
 496         done = 1;
 497 }
 498
 499 /*
 500  *      Activate the first processor.
 501  */
 502
 503 static void __init boot_cpu_init(void)
 504 {
 505         int cpu = smp_processor_id();
 506         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 507         set_cpu_online(cpu, true);
 508         set_cpu_active(cpu, true);
 509         set_cpu_present(cpu, true);
 510         set_cpu_possible(cpu, true);
 511 }
 512
 513 void __init __weak smp_setup_processor_id(void)
 514 {
 515 }
 516
 517 void __init __weak thread_info_cache_init(void)
 518 {
 519 }
 520
 521 /*
 522  * Set up kernel memory allocators
 523  */
 524 static void __init mm_init(void)
 525 {
 526         /*
 527          * page_cgroup requires countinous pages as memmap
 528          * and it's bigger than MAX_ORDER unless SPARSEMEM.
 529          */
 530         page_cgroup_init_flatmem();
 531         mem_init();
 532         kmem_cache_init();
 533         percpu_init_late();
 534         pgtable_cache_init();
 535         vmalloc_init();
 536 }
 537
 538 asmlinkage void __init start_kernel(void)
 539 {
 540         char * command_line;
 541         extern struct kernel_param __start___param[], __stop___param[];
 542
 543         smp_setup_processor_id();
 544
 545         /*
 546          * Need to run as early as possible, to initialize the
 547          * lockdep hash:
 548          */
 549         lockdep_init();
 550         debug_objects_early_init();
 551
 552         /*
 553          * Set up the the initial canary ASAP:
 554          */
 555         boot_init_stack_canary();
 556
 557         cgroup_init_early();
 558
 559         local_irq_disable();
 560         early_boot_irqs_off();
 561         early_init_irq_lock_class();
 562
 563 /*
 564  * Interrupts are still disabled. Do necessary setups, then
 565  * enable them
 566  */
 567         lock_kernel();
 568         tick_init();
 569         boot_cpu_init();
 570         page_address_init();
 571         printk(KERN_NOTICE "%s", linux_banner);
 572         setup_arch(&command_line);
 573         mm_init_owner(&init_mm, &init_task);
 574         setup_command_line(command_line);
 575         setup_nr_cpu_ids();
 576         setup_per_cpu_areas();
 577         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 578
 579         build_all_zonelists(NULL);
 580         page_alloc_init();
 581
 582         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 583         parse_early_param();
 584         parse_args("Booting kernel", static_command_line, __start___param,
 585                    __stop___param - __start___param,
 586                    &unknown_bootoption);
 587         /*
 588          * These use large bootmem allocations and must precede
 589          * kmem_cache_init()
 590          */
 591         pidhash_init();
 592         vfs_caches_init_early();
 593         sort_main_extable();
 594         trap_init();
 595         mm_init();
 596         /*
 597          * Set up the scheduler prior starting any interrupts (such as the
 598          * timer interrupt). Full topology setup happens at smp_init()
 599          * time - but meanwhile we still have a functioning scheduler.
 600          */
 601         sched_init();
 602         /*
 603          * Disable preemption - early bootup scheduling is extremely
 604          * fragile until we cpu_idle() for the first time.
 605          */
 606         preempt_disable();
 607         if (!irqs_disabled()) {
 608                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 609                                 "enabled *very* early, fixing it\n");
 610                 local_irq_disable();
 611         }
 612         rcu_init();
 613         radix_tree_init();
 614         /* init some links before init_ISA_irqs() */
 615         early_irq_init();
 616         init_IRQ();
 617         prio_tree_init();
 618         init_timers();
 619         hrtimers_init();
 620         softirq_init();
 621         timekeeping_init();
 622         time_init();
 623         profile_init();
 624         if (!irqs_disabled())
 625                 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
 626                                  "enabled early\n");
 627         early_boot_irqs_on();
 628         local_irq_enable();
 629
 630         /* Interrupts are enabled now so all GFP allocations are safe. */
 631         gfp_allowed_mask = __GFP_BITS_MASK;
 632
 633         kmem_cache_init_late();
 634
 635         /*
 636          * HACK ALERT! This is early. We're enabling the console before
 637          * we've done PCI setups etc, and console_init() must be aware of
 638          * this. But we do want output early, in case something goes wrong.
 639          */
 640         console_init();
 641         if (panic_later)
 642                 panic(panic_later, panic_param);
 643
 644         lockdep_info();
 645
 646         /*
 647          * Need to run this when irqs are enabled, because it wants
 648          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 649          * too:
 650          */
 651         locking_selftest();
 652
 653 #ifdef CONFIG_BLK_DEV_INITRD
 654         if (initrd_start && !initrd_below_start_ok &&
 655             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 656                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 657                     "disabling it.\n",
 658                     page_to_pfn(virt_to_page((void *)initrd_start)),
 659                     min_low_pfn);
 660                 initrd_start = 0;
 661         }
 662 #endif
 663         page_cgroup_init();
 664         enable_debug_pagealloc();
 665         kmemleak_init();
 666         debug_objects_mem_init();
 667         idr_init_cache();
 668         setup_per_cpu_pageset();
 669         numa_policy_init();
 670         if (late_time_init)
 671                 late_time_init();
 672         sched_clock_init();
 673         calibrate_delay();
 674         pidmap_init();
 675         anon_vma_init();
 676 #ifdef CONFIG_X86
 677         if (efi_enabled)
 678                 efi_enter_virtual_mode();
 679 #endif
 680         thread_info_cache_init();
 681         cred_init();
 682         fork_init(totalram_pages);
 683         proc_caches_init();
 684         buffer_init();
 685         key_init();
 686         security_init();
 687         dbg_late_init();
 688         vfs_caches_init(totalram_pages);
 689         signals_init();
 690         /* rootfs populating might need page-writeback */
 691         page_writeback_init();
 692 #ifdef CONFIG_PROC_FS
 693         proc_root_init();
 694 #endif
 695         cgroup_init();
 696         cpuset_init();
 697         taskstats_init_early();
 698         delayacct_init();
 699
 700         check_bugs();
 701
 702         acpi_early_init(); /* before LAPIC and SMP init */
 703         sfi_init_late();
 704
 705         ftrace_init();
 706
 707         /* Do the rest non-__init'ed, we're now alive */
 708         rest_init();
 709 }
 710
 711 /* Call all constructor functions linked into the kernel. */
 712 static void __init do_ctors(void)
 713 {
 714 #ifdef CONFIG_CONSTRUCTORS
 715         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 716
 717         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 718                 (*fn)();
 719 #endif
 720 }
 721
 722 int initcall_debug;
 723 core_param(initcall_debug, initcall_debug, bool, 0644);
 724
 725 static char msgbuf[64];
 726
 727 int do_one_initcall(initcall_t fn)
 728 {
 729         int count = preempt_count();
 730         ktime_t calltime, delta, rettime;
 731         unsigned long long duration;
 732         int ret;
 733
 734         if (initcall_debug) {
 735                 printk("calling  %pF @ %i\n", fn, task_pid_nr(current));
 736                 calltime = ktime_get();
 737         }
 738
 739         ret = fn();
 740
 741         if (initcall_debug) {
 742                 rettime = ktime_get();
 743                 delta = ktime_sub(rettime, calltime);
 744                 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 745                 printk("initcall %pF returned %d after %lld usecs\n", fn,
 746                         ret, duration);
 747         }
 748
 749         msgbuf[0] = 0;
 750
 751         if (ret && ret != -ENODEV && initcall_debug)
 752                 sprintf(msgbuf, "error code %d ", ret);
 753
 754         if (preempt_count() != count) {
 755                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 756                 preempt_count() = count;
 757         }
 758         if (irqs_disabled()) {
 759                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 760                 local_irq_enable();
 761         }
 762         if (msgbuf[0]) {
 763                 printk("initcall %pF returned with %s\n", fn, msgbuf);
 764         }
 765
 766         return ret;
 767 }
 768
 769
 770 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
 771
 772 static void __init do_initcalls(void)
 773 {
 774         initcall_t *fn;
 775
 776         for (fn = __early_initcall_end; fn < __initcall_end; fn++)
 777                 do_one_initcall(*fn);
 778
 779         /* Make sure there is no pending stuff from the initcall sequence */
 780         flush_scheduled_work();
 781 }
 782
 783 /*
 784  * Ok, the machine is now initialized. None of the devices
 785  * have been touched yet, but the CPU subsystem is up and
 786  * running, and memory and process management works.
 787  *
 788  * Now we can finally start doing some real work..
 789  */
 790 static void __init do_basic_setup(void)
 791 {
 792         init_workqueues();
 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(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         __releases(kernel_lock)
 821 {
 822         /* need to finish all async __init code before freeing the memory */
 823         async_synchronize_full();
 824         free_initmem();
 825         unlock_kernel();
 826         mark_rodata_ro();
 827         system_state = SYSTEM_RUNNING;
 828         numa_default_policy();
 829
 830
 831         current->signal->flags |= SIGNAL_UNKILLABLE;
 832
 833         if (ramdisk_execute_command) {
 834                 run_init_process(ramdisk_execute_command);
 835                 printk(KERN_WARNING "Failed to execute %s\n",
 836                                 ramdisk_execute_command);
 837         }
 838
 839         /*
 840          * We try each of these until one succeeds.
 841          *
 842          * The Bourne shell can be used instead of init if we are
 843          * trying to recover a really broken machine.
 844          */
 845         if (execute_command) {
 846                 run_init_process(execute_command);
 847                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 848                                         "defaults...\n", execute_command);
 849         }
 850         run_init_process("/sbin/init");
 851         run_init_process("/etc/init");
 852         run_init_process("/bin/init");
 853         run_init_process("/bin/sh");
 854
 855         panic("No init found.  Try passing init= option to kernel. "
 856               "See Linux Documentation/init.txt for guidance.");
 857 }
 858
 859 static int __init kernel_init(void * unused)
 860 {
 861         /*
 862          * Wait until kthreadd is all set-up.
 863          */
 864         wait_for_completion(&kthreadd_done);
 865         lock_kernel();
 866
 867         /*
 868          * init can allocate pages on any node
 869          */
 870         set_mems_allowed(node_states[N_HIGH_MEMORY]);
 871         /*
 872          * init can run on any cpu.
 873          */
 874         set_cpus_allowed_ptr(current, cpu_all_mask);
 875         /*
 876          * Tell the world that we're going to be the grim
 877          * reaper of innocent orphaned children.
 878          *
 879          * We don't want people to have to make incorrect
 880          * assumptions about where in the task array this
 881          * can be found.
 882          */
 883         init_pid_ns.child_reaper = current;
 884
 885         cad_pid = task_pid(current);
 886
 887         smp_prepare_cpus(setup_max_cpus);
 888
 889         do_pre_smp_initcalls();
 890
 891         smp_init();
 892         sched_init_smp();
 893
 894         do_basic_setup();
 895
 896         /* Open the /dev/console on the rootfs, this should never fail */
 897         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 898                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 899
 900         (void) sys_dup(0);
 901         (void) sys_dup(0);
 902         /*
 903          * check if there is an early userspace init.  If yes, let it do all
 904          * the work
 905          */
 906
 907         if (!ramdisk_execute_command)
 908                 ramdisk_execute_command = "/init";
 909
 910         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 911                 ramdisk_execute_command = NULL;
 912                 prepare_namespace();
 913         }
 914
 915         /*
 916          * Ok, we have completed the initial bootup, and
 917          * we're essentially up and running. Get rid of the
 918          * initmem segments and start the user-mode stuff..
 919          */
 920
 921         init_post();
 922         return 0;
 923 }