init/main.c

   1 /*
   2  *  linux/init/main.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  *
   6  *  GK 2/5/95  -  Changed to support mounting root fs via NFS
   7  *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
   8  *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
   9  *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
  10  */
  11
  12 #include <linux/types.h>
  13 #include <linux/module.h>
  14 #include <linux/proc_fs.h>
  15 #include <linux/kernel.h>
  16 #include <linux/syscalls.h>
  17 #include <linux/string.h>
  18 #include <linux/ctype.h>
  19 #include <linux/delay.h>
  20 #include <linux/utsname.h>
  21 #include <linux/ioport.h>
  22 #include <linux/init.h>
  23 #include <linux/smp_lock.h>
  24 #include <linux/initrd.h>
  25 #include <linux/bootmem.h>
  26 #include <linux/tty.h>
  27 #include <linux/gfp.h>
  28 #include <linux/percpu.h>
  29 #include <linux/kmod.h>
  30 #include <linux/vmalloc.h>
  31 #include <linux/kernel_stat.h>
  32 #include <linux/start_kernel.h>
  33 #include <linux/security.h>
  34 #include <linux/smp.h>
  35 #include <linux/workqueue.h>
  36 #include <linux/profile.h>
  37 #include <linux/rcupdate.h>
  38 #include <linux/moduleparam.h>
  39 #include <linux/kallsyms.h>
  40 #include <linux/writeback.h>
  41 #include <linux/cpu.h>
  42 #include <linux/cpuset.h>
  43 #include <linux/cgroup.h>
  44 #include <linux/efi.h>
  45 #include <linux/tick.h>
  46 #include <linux/interrupt.h>
  47 #include <linux/taskstats_kern.h>
  48 #include <linux/delayacct.h>
  49 #include <linux/unistd.h>
  50 #include <linux/rmap.h>
  51 #include <linux/mempolicy.h>
  52 #include <linux/key.h>
  53 #include <linux/unwind.h>
  54 #include <linux/buffer_head.h>
  55 #include <linux/page_cgroup.h>
  56 #include <linux/debug_locks.h>
  57 #include <linux/debugobjects.h>
  58 #include <linux/lockdep.h>
  59 #include <linux/pid_namespace.h>
  60 #include <linux/device.h>
  61 #include <linux/kthread.h>
  62 #include <linux/sched.h>
  63 #include <linux/signal.h>
  64 #include <linux/idr.h>
  65 #include <linux/ftrace.h>
  66 #include <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 /*
  79  * This is one of the first .c files built. Error out early if we have compiler
  80  * trouble.
  81  */
  82
  83 #if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
  84 #warning gcc-4.1.0 is known to miscompile the kernel.  A different compiler version is recommended.
  85 #endif
  86
  87 static int kernel_init(void *);
  88
  89 extern void init_IRQ(void);
  90 extern void fork_init(unsigned long);
  91 extern void mca_init(void);
  92 extern void sbus_init(void);
  93 extern void prio_tree_init(void);
  94 extern void radix_tree_init(void);
  95 extern void free_initmem(void);
  96 #ifdef  CONFIG_ACPI
  97 extern void acpi_early_init(void);
  98 #else
  99 static inline void acpi_early_init(void) { }
 100 #endif
 101 #ifndef CONFIG_DEBUG_RODATA
 102 static inline void mark_rodata_ro(void) { }
 103 #endif
 104
 105 #ifdef CONFIG_TC
 106 extern void tc_init(void);
 107 #endif
 108
 109 enum system_states system_state;
 110 EXPORT_SYMBOL(system_state);
 111
 112 /*
 113  * Boot command-line arguments
 114  */
 115 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 116 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 117
 118 extern void time_init(void);
 119 /* Default late time init is NULL. archs can override this later. */
 120 void (*late_time_init)(void);
 121 extern void softirq_init(void);
 122
 123 /* Untouched command line saved by arch-specific code. */
 124 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 125 /* Untouched saved command line (eg. for /proc) */
 126 char *saved_command_line;
 127 /* Command line for parameter parsing */
 128 static char *static_command_line;
 129
 130 static char *execute_command;
 131 static char *ramdisk_execute_command;
 132
 133 #ifdef CONFIG_SMP
 134 /* Setup configured maximum number of CPUs to activate */
 135 unsigned int __initdata setup_max_cpus = NR_CPUS;
 136
 137 /*
 138  * Setup routine for controlling SMP activation
 139  *
 140  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 141  * activation entirely (the MPS table probe still happens, though).
 142  *
 143  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 144  * greater than 0, limits the maximum number of CPUs activated in
 145  * SMP mode to <NUM>.
 146  */
 147 #ifndef CONFIG_X86_IO_APIC
 148 static inline void disable_ioapic_setup(void) {};
 149 #endif
 150
 151 static int __init nosmp(char *str)
 152 {
 153         setup_max_cpus = 0;
 154         disable_ioapic_setup();
 155         return 0;
 156 }
 157
 158 early_param("nosmp", nosmp);
 159
 160 static int __init maxcpus(char *str)
 161 {
 162         get_option(&str, &setup_max_cpus);
 163         if (setup_max_cpus == 0)
 164                 disable_ioapic_setup();
 165
 166         return 0;
 167 }
 168
 169 early_param("maxcpus", maxcpus);
 170 #else
 171 #define setup_max_cpus NR_CPUS
 172 #endif
 173
 174 /*
 175  * If set, this is an indication to the drivers that reset the underlying
 176  * device before going ahead with the initialization otherwise driver might
 177  * rely on the BIOS and skip the reset operation.
 178  *
 179  * This is useful if kernel is booting in an unreliable environment.
 180  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 181  * skipped and devices will be in unknown state.
 182  */
 183 unsigned int reset_devices;
 184 EXPORT_SYMBOL(reset_devices);
 185
 186 static int __init set_reset_devices(char *str)
 187 {
 188         reset_devices = 1;
 189         return 1;
 190 }
 191
 192 __setup("reset_devices", set_reset_devices);
 193
 194 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 195 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 196 static const char *panic_later, *panic_param;
 197
 198 extern struct obs_kernel_param __setup_start[], __setup_end[];
 199
 200 static int __init obsolete_checksetup(char *line)
 201 {
 202         struct obs_kernel_param *p;
 203         int had_early_param = 0;
 204
 205         p = __setup_start;
 206         do {
 207                 int n = strlen(p->str);
 208                 if (!strncmp(line, p->str, n)) {
 209                         if (p->early) {
 210                                 /* Already done in parse_early_param?
 211                                  * (Needs exact match on param part).
 212                                  * Keep iterating, as we can have early
 213                                  * params and __setups of same names 8( */
 214                                 if (line[n] == '\0' || line[n] == '=')
 215                                         had_early_param = 1;
 216                         } else if (!p->setup_func) {
 217                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 218                                        " ignored\n", p->str);
 219                                 return 1;
 220                         } else if (p->setup_func(line + n))
 221                                 return 1;
 222                 }
 223                 p++;
 224         } while (p < __setup_end);
 225
 226         return had_early_param;
 227 }
 228
 229 /*
 230  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 231  * still work even if initially too large, it will just take slightly longer
 232  */
 233 unsigned long loops_per_jiffy = (1<<12);
 234
 235 EXPORT_SYMBOL(loops_per_jiffy);
 236
 237 static int __init debug_kernel(char *str)
 238 {
 239         console_loglevel = 10;
 240         return 0;
 241 }
 242
 243 static int __init quiet_kernel(char *str)
 244 {
 245         console_loglevel = 4;
 246         return 0;
 247 }
 248
 249 early_param("debug", debug_kernel);
 250 early_param("quiet", quiet_kernel);
 251
 252 static int __init loglevel(char *str)
 253 {
 254         get_option(&str, &console_loglevel);
 255         return 0;
 256 }
 257
 258 early_param("loglevel", loglevel);
 259
 260 /*
 261  * Unknown boot options get handed to init, unless they look like
 262  * failed parameters
 263  */
 264 static int __init unknown_bootoption(char *param, char *val)
 265 {
 266         /* Change NUL term back to "=", to make "param" the whole string. */
 267         if (val) {
 268                 /* param=val or param="val"? */
 269                 if (val == param+strlen(param)+1)
 270                         val[-1] = '=';
 271                 else if (val == param+strlen(param)+2) {
 272                         val[-2] = '=';
 273                         memmove(val-1, val, strlen(val)+1);
 274                         val--;
 275                 } else
 276                         BUG();
 277         }
 278
 279         /* Handle obsolete-style parameters */
 280         if (obsolete_checksetup(param))
 281                 return 0;
 282
 283         /*
 284          * Preemptive maintenance for "why didn't my misspelled command
 285          * line work?"
 286          */
 287         if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
 288                 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
 289                 return 0;
 290         }
 291
 292         if (panic_later)
 293                 return 0;
 294
 295         if (val) {
 296                 /* Environment option */
 297                 unsigned int i;
 298                 for (i = 0; envp_init[i]; i++) {
 299                         if (i == MAX_INIT_ENVS) {
 300                                 panic_later = "Too many boot env vars at `%s'";
 301                                 panic_param = param;
 302                         }
 303                         if (!strncmp(param, envp_init[i], val - param))
 304                                 break;
 305                 }
 306                 envp_init[i] = param;
 307         } else {
 308                 /* Command line option */
 309                 unsigned int i;
 310                 for (i = 0; argv_init[i]; i++) {
 311                         if (i == MAX_INIT_ARGS) {
 312                                 panic_later = "Too many boot init vars at `%s'";
 313                                 panic_param = param;
 314                         }
 315                 }
 316                 argv_init[i] = param;
 317         }
 318         return 0;
 319 }
 320
 321 #ifdef CONFIG_DEBUG_PAGEALLOC
 322 int __read_mostly debug_pagealloc_enabled = 0;
 323 #endif
 324
 325 static int __init init_setup(char *str)
 326 {
 327         unsigned int i;
 328
 329         execute_command = str;
 330         /*
 331          * In case LILO is going to boot us with default command line,
 332          * it prepends "auto" before the whole cmdline which makes
 333          * the shell think it should execute a script with such name.
 334          * So we ignore all arguments entered _before_ init=... [MJ]
 335          */
 336         for (i = 1; i < MAX_INIT_ARGS; i++)
 337                 argv_init[i] = NULL;
 338         return 1;
 339 }
 340 __setup("init=", init_setup);
 341
 342 static int __init rdinit_setup(char *str)
 343 {
 344         unsigned int i;
 345
 346         ramdisk_execute_command = str;
 347         /* See "auto" comment in init_setup */
 348         for (i = 1; i < MAX_INIT_ARGS; i++)
 349                 argv_init[i] = NULL;
 350         return 1;
 351 }
 352 __setup("rdinit=", rdinit_setup);
 353
 354 #ifndef CONFIG_SMP
 355
 356 #ifdef CONFIG_X86_LOCAL_APIC
 357 static void __init smp_init(void)
 358 {
 359         APIC_init_uniprocessor();
 360 }
 361 #else
 362 #define smp_init()      do { } while (0)
 363 #endif
 364
 365 static inline void setup_per_cpu_areas(void) { }
 366 static inline void setup_nr_cpu_ids(void) { }
 367 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 368
 369 #else
 370
 371 #if NR_CPUS > BITS_PER_LONG
 372 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
 373 EXPORT_SYMBOL(cpu_mask_all);
 374 #endif
 375
 376 /* Setup number of possible processor ids */
 377 int nr_cpu_ids __read_mostly = NR_CPUS;
 378 EXPORT_SYMBOL(nr_cpu_ids);
 379
 380 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
 381 static void __init setup_nr_cpu_ids(void)
 382 {
 383         int cpu, highest_cpu = 0;
 384
 385         for_each_possible_cpu(cpu)
 386                 highest_cpu = cpu;
 387
 388         nr_cpu_ids = highest_cpu + 1;
 389 }
 390
 391 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
 392 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
 393
 394 EXPORT_SYMBOL(__per_cpu_offset);
 395
 396 static void __init setup_per_cpu_areas(void)
 397 {
 398         unsigned long size, i;
 399         char *ptr;
 400         unsigned long nr_possible_cpus = num_possible_cpus();
 401
 402         /* Copy section for each CPU (we discard the original) */
 403         size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
 404         ptr = alloc_bootmem_pages(size * nr_possible_cpus);
 405
 406         for_each_possible_cpu(i) {
 407                 __per_cpu_offset[i] = ptr - __per_cpu_start;
 408                 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 409                 ptr += size;
 410         }
 411 }
 412 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
 413
 414 /* Called by boot processor to activate the rest. */
 415 static void __init smp_init(void)
 416 {
 417         unsigned int cpu;
 418
 419         /*
 420          * Set up the current CPU as possible to migrate to.
 421          * The other ones will be done by cpu_up/cpu_down()
 422          */
 423         cpu = smp_processor_id();
 424         cpu_set(cpu, cpu_active_map);
 425
 426         /* FIXME: This should be done in userspace --RR */
 427         for_each_present_cpu(cpu) {
 428                 if (num_online_cpus() >= setup_max_cpus)
 429                         break;
 430                 if (!cpu_online(cpu))
 431                         cpu_up(cpu);
 432         }
 433
 434         /* Any cleanup work */
 435         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 436         smp_cpus_done(setup_max_cpus);
 437 }
 438
 439 #endif
 440
 441 /*
 442  * We need to store the untouched command line for future reference.
 443  * We also need to store the touched command line since the parameter
 444  * parsing is performed in place, and we should allow a component to
 445  * store reference of name/value for future reference.
 446  */
 447 static void __init setup_command_line(char *command_line)
 448 {
 449         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 450         static_command_line = alloc_bootmem(strlen (command_line)+1);
 451         strcpy (saved_command_line, boot_command_line);
 452         strcpy (static_command_line, command_line);
 453 }
 454
 455 /*
 456  * We need to finalize in a non-__init function or else race conditions
 457  * between the root thread and the init thread may cause start_kernel to
 458  * be reaped by free_initmem before the root thread has proceeded to
 459  * cpu_idle.
 460  *
 461  * gcc-3.4 accidentally inlines this function, so use noinline.
 462  */
 463
 464 static void noinline __init_refok rest_init(void)
 465         __releases(kernel_lock)
 466 {
 467         int pid;
 468
 469         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 470         numa_default_policy();
 471         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 472         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 473         unlock_kernel();
 474
 475         /*
 476          * The boot idle thread must execute schedule()
 477          * at least once to get things moving:
 478          */
 479         init_idle_bootup_task(current);
 480         preempt_enable_no_resched();
 481         schedule();
 482         preempt_disable();
 483
 484         /* Call into cpu_idle with preempt disabled */
 485         cpu_idle();
 486 }
 487
 488 /* Check for early params. */
 489 static int __init do_early_param(char *param, char *val)
 490 {
 491         struct obs_kernel_param *p;
 492
 493         for (p = __setup_start; p < __setup_end; p++) {
 494                 if ((p->early && strcmp(param, p->str) == 0) ||
 495                     (strcmp(param, "console") == 0 &&
 496                      strcmp(p->str, "earlycon") == 0)
 497                 ) {
 498                         if (p->setup_func(val) != 0)
 499                                 printk(KERN_WARNING
 500                                        "Malformed early option '%s'\n", param);
 501                 }
 502         }
 503         /* We accept everything at this stage. */
 504         return 0;
 505 }
 506
 507 /* Arch code calls this early on, or if not, just before other parsing. */
 508 void __init parse_early_param(void)
 509 {
 510         static __initdata int done = 0;
 511         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 512
 513         if (done)
 514                 return;
 515
 516         /* All fall through to do_early_param. */
 517         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 518         parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
 519         done = 1;
 520 }
 521
 522 /*
 523  *      Activate the first processor.
 524  */
 525
 526 static void __init boot_cpu_init(void)
 527 {
 528         int cpu = smp_processor_id();
 529         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 530         cpu_set(cpu, cpu_online_map);
 531         cpu_set(cpu, cpu_present_map);
 532         cpu_set(cpu, cpu_possible_map);
 533 }
 534
 535 void __init __weak smp_setup_processor_id(void)
 536 {
 537 }
 538
 539 void __init __weak thread_info_cache_init(void)
 540 {
 541 }
 542
 543 asmlinkage void __init start_kernel(void)
 544 {
 545         char * command_line;
 546         extern struct kernel_param __start___param[], __stop___param[];
 547
 548         smp_setup_processor_id();
 549
 550         /*
 551          * Need to run as early as possible, to initialize the
 552          * lockdep hash:
 553          */
 554         unwind_init();
 555         lockdep_init();
 556         debug_objects_early_init();
 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);
 572         printk(linux_banner);
 573         setup_arch(&command_line);
 574         mm_init_owner(&init_mm, &init_task);
 575         setup_command_line(command_line);
 576         unwind_setup();
 577         setup_per_cpu_areas();
 578         setup_nr_cpu_ids();
 579         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 580
 581         /*
 582          * Set up the scheduler prior starting any interrupts (such as the
 583          * timer interrupt). Full topology setup happens at smp_init()
 584          * time - but meanwhile we still have a functioning scheduler.
 585          */
 586         sched_init();
 587         /*
 588          * Disable preemption - early bootup scheduling is extremely
 589          * fragile until we cpu_idle() for the first time.
 590          */
 591         preempt_disable();
 592         build_all_zonelists();
 593         page_alloc_init();
 594         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 595         parse_early_param();
 596         parse_args("Booting kernel", static_command_line, __start___param,
 597                    __stop___param - __start___param,
 598                    &unknown_bootoption);
 599         if (!irqs_disabled()) {
 600                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 601                                 "enabled *very* early, fixing it\n");
 602                 local_irq_disable();
 603         }
 604         sort_main_extable();
 605         trap_init();
 606         rcu_init();
 607         /* init some links before init_ISA_irqs() */
 608         early_irq_init();
 609         init_IRQ();
 610         pidhash_init();
 611         init_timers();
 612         hrtimers_init();
 613         softirq_init();
 614         timekeeping_init();
 615         time_init();
 616         sched_clock_init();
 617         profile_init();
 618         if (!irqs_disabled())
 619                 printk("start_kernel(): bug: interrupts were enabled early\n");
 620         early_boot_irqs_on();
 621         local_irq_enable();
 622
 623         /*
 624          * HACK ALERT! This is early. We're enabling the console before
 625          * we've done PCI setups etc, and console_init() must be aware of
 626          * this. But we do want output early, in case something goes wrong.
 627          */
 628         console_init();
 629         if (panic_later)
 630                 panic(panic_later, panic_param);
 631
 632         lockdep_info();
 633
 634         /*
 635          * Need to run this when irqs are enabled, because it wants
 636          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 637          * too:
 638          */
 639         locking_selftest();
 640
 641 #ifdef CONFIG_BLK_DEV_INITRD
 642         if (initrd_start && !initrd_below_start_ok &&
 643             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 644                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 645                     "disabling it.\n",
 646                     page_to_pfn(virt_to_page((void *)initrd_start)),
 647                     min_low_pfn);
 648                 initrd_start = 0;
 649         }
 650 #endif
 651         vmalloc_init();
 652         vfs_caches_init_early();
 653         cpuset_init_early();
 654         page_cgroup_init();
 655         mem_init();
 656         enable_debug_pagealloc();
 657         cpu_hotplug_init();
 658         kmem_cache_init();
 659         debug_objects_mem_init();
 660         idr_init_cache();
 661         setup_per_cpu_pageset();
 662         numa_policy_init();
 663         if (late_time_init)
 664                 late_time_init();
 665         calibrate_delay();
 666         pidmap_init();
 667         pgtable_cache_init();
 668         prio_tree_init();
 669         anon_vma_init();
 670 #ifdef CONFIG_X86
 671         if (efi_enabled)
 672                 efi_enter_virtual_mode();
 673 #endif
 674         thread_info_cache_init();
 675         cred_init();
 676         fork_init(num_physpages);
 677         proc_caches_init();
 678         buffer_init();
 679         key_init();
 680         security_init();
 681         vfs_caches_init(num_physpages);
 682         radix_tree_init();
 683         signals_init();
 684         /* rootfs populating might need page-writeback */
 685         page_writeback_init();
 686 #ifdef CONFIG_PROC_FS
 687         proc_root_init();
 688 #endif
 689         cgroup_init();
 690         cpuset_init();
 691         taskstats_init_early();
 692         delayacct_init();
 693
 694         check_bugs();
 695
 696         acpi_early_init(); /* before LAPIC and SMP init */
 697
 698         ftrace_init();
 699
 700         /* Do the rest non-__init'ed, we're now alive */
 701         rest_init();
 702 }
 703
 704 static int initcall_debug;
 705 core_param(initcall_debug, initcall_debug, bool, 0644);
 706
 707 int do_one_initcall(initcall_t fn)
 708 {
 709         int count = preempt_count();
 710         ktime_t calltime, delta, rettime;
 711         char msgbuf[64];
 712         struct boot_trace_call call;
 713         struct boot_trace_ret ret;
 714
 715         if (initcall_debug) {
 716                 call.caller = task_pid_nr(current);
 717                 printk("calling  %pF @ %i\n", fn, call.caller);
 718                 calltime = ktime_get();
 719                 trace_boot_call(&call, fn);
 720                 enable_boot_trace();
 721         }
 722
 723         ret.result = fn();
 724
 725         if (initcall_debug) {
 726                 disable_boot_trace();
 727                 rettime = ktime_get();
 728                 delta = ktime_sub(rettime, calltime);
 729                 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 730                 trace_boot_ret(&ret, fn);
 731                 printk("initcall %pF returned %d after %Ld usecs\n", fn,
 732                         ret.result, ret.duration);
 733         }
 734
 735         msgbuf[0] = 0;
 736
 737         if (ret.result && ret.result != -ENODEV && initcall_debug)
 738                 sprintf(msgbuf, "error code %d ", ret.result);
 739
 740         if (preempt_count() != count) {
 741                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 742                 preempt_count() = count;
 743         }
 744         if (irqs_disabled()) {
 745                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 746                 local_irq_enable();
 747         }
 748         if (msgbuf[0]) {
 749                 printk("initcall %pF returned with %s\n", fn, msgbuf);
 750         }
 751
 752         return ret.result;
 753 }
 754
 755
 756 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
 757
 758 static void __init do_initcalls(void)
 759 {
 760         initcall_t *call;
 761
 762         for (call = __early_initcall_end; call < __initcall_end; call++)
 763                 do_one_initcall(*call);
 764
 765         /* Make sure there is no pending stuff from the initcall sequence */
 766         flush_scheduled_work();
 767 }
 768
 769 /*
 770  * Ok, the machine is now initialized. None of the devices
 771  * have been touched yet, but the CPU subsystem is up and
 772  * running, and memory and process management works.
 773  *
 774  * Now we can finally start doing some real work..
 775  */
 776 static void __init do_basic_setup(void)
 777 {
 778         rcu_init_sched(); /* needed by module_init stage. */
 779         init_workqueues();
 780         usermodehelper_init();
 781         driver_init();
 782         init_irq_proc();
 783         do_initcalls();
 784 }
 785
 786 static void __init do_pre_smp_initcalls(void)
 787 {
 788         initcall_t *call;
 789
 790         for (call = __initcall_start; call < __early_initcall_end; call++)
 791                 do_one_initcall(*call);
 792 }
 793
 794 static void run_init_process(char *init_filename)
 795 {
 796         argv_init[0] = init_filename;
 797         kernel_execve(init_filename, argv_init, envp_init);
 798 }
 799
 800 /* This is a non __init function. Force it to be noinline otherwise gcc
 801  * makes it inline to init() and it becomes part of init.text section
 802  */
 803 static int noinline init_post(void)
 804 {
 805         free_initmem();
 806         unlock_kernel();
 807         mark_rodata_ro();
 808         system_state = SYSTEM_RUNNING;
 809         numa_default_policy();
 810
 811         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 812                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 813
 814         (void) sys_dup(0);
 815         (void) sys_dup(0);
 816
 817         current->signal->flags |= SIGNAL_UNKILLABLE;
 818
 819         if (ramdisk_execute_command) {
 820                 run_init_process(ramdisk_execute_command);
 821                 printk(KERN_WARNING "Failed to execute %s\n",
 822                                 ramdisk_execute_command);
 823         }
 824
 825         /*
 826          * We try each of these until one succeeds.
 827          *
 828          * The Bourne shell can be used instead of init if we are
 829          * trying to recover a really broken machine.
 830          */
 831         if (execute_command) {
 832                 run_init_process(execute_command);
 833                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 834                                         "defaults...\n", execute_command);
 835         }
 836         run_init_process("/sbin/init");
 837         run_init_process("/etc/init");
 838         run_init_process("/bin/init");
 839         run_init_process("/bin/sh");
 840
 841         panic("No init found.  Try passing init= option to kernel.");
 842 }
 843
 844 static int __init kernel_init(void * unused)
 845 {
 846         lock_kernel();
 847         /*
 848          * init can run on any cpu.
 849          */
 850         set_cpus_allowed_ptr(current, CPU_MASK_ALL_PTR);
 851         /*
 852          * Tell the world that we're going to be the grim
 853          * reaper of innocent orphaned children.
 854          *
 855          * We don't want people to have to make incorrect
 856          * assumptions about where in the task array this
 857          * can be found.
 858          */
 859         init_pid_ns.child_reaper = current;
 860
 861         cad_pid = task_pid(current);
 862
 863         smp_prepare_cpus(setup_max_cpus);
 864
 865         do_pre_smp_initcalls();
 866         start_boot_trace();
 867
 868         smp_init();
 869         sched_init_smp();
 870
 871         cpuset_init_smp();
 872
 873         do_basic_setup();
 874
 875         /*
 876          * check if there is an early userspace init.  If yes, let it do all
 877          * the work
 878          */
 879
 880         if (!ramdisk_execute_command)
 881                 ramdisk_execute_command = "/init";
 882
 883         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 884                 ramdisk_execute_command = NULL;
 885                 prepare_namespace();
 886         }
 887
 888         /*
 889          * Ok, we have completed the initial bootup, and
 890          * we're essentially up and running. Get rid of the
 891          * initmem segments and start the user-mode stuff..
 892          */
 893
 894         init_post();
 895         return 0;
 896 }