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/ftrace.h>
  66 #include <linux/async.h>
  67 #include <linux/kmemcheck.h>
  68 #include <linux/kmemtrace.h>
  69 #include <linux/sfi.h>
  70 #include <linux/shmem_fs.h>
  71 #include <linux/slab.h>
  72 #include <trace/boot.h>
  73
  74 #include <asm/io.h>
  75 #include <asm/bugs.h>
  76 #include <asm/setup.h>
  77 #include <asm/sections.h>
  78 #include <asm/cacheflush.h>
  79
  80 #ifdef CONFIG_X86_LOCAL_APIC
  81 #include <asm/smp.h>
  82 #endif
  83
  84 static int kernel_init(void *);
  85
  86 extern void init_IRQ(void);
  87 extern void fork_init(unsigned long);
  88 extern void mca_init(void);
  89 extern void sbus_init(void);
  90 extern void prio_tree_init(void);
  91 extern void radix_tree_init(void);
  92 extern void free_initmem(void);
  93 #ifndef CONFIG_DEBUG_RODATA
  94 static inline void mark_rodata_ro(void) { }
  95 #endif
  96
  97 #ifdef CONFIG_TC
  98 extern void tc_init(void);
  99 #endif
 100
 101 enum system_states system_state __read_mostly;
 102 EXPORT_SYMBOL(system_state);
 103
 104 /*
 105  * Boot command-line arguments
 106  */
 107 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 108 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 109
 110 extern void time_init(void);
 111 /* Default late time init is NULL. archs can override this later. */
 112 void (*__initdata late_time_init)(void);
 113 extern void softirq_init(void);
 114
 115 /* Untouched command line saved by arch-specific code. */
 116 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 117 /* Untouched saved command line (eg. for /proc) */
 118 char *saved_command_line;
 119 /* Command line for parameter parsing */
 120 static char *static_command_line;
 121
 122 static char *execute_command;
 123 static char *ramdisk_execute_command;
 124
 125 #ifdef CONFIG_SMP
 126 /* Setup configured maximum number of CPUs to activate */
 127 unsigned int __initdata setup_max_cpus = NR_CPUS;
 128
 129 /*
 130  * Setup routine for controlling SMP activation
 131  *
 132  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 133  * activation entirely (the MPS table probe still happens, though).
 134  *
 135  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 136  * greater than 0, limits the maximum number of CPUs activated in
 137  * SMP mode to <NUM>.
 138  */
 139
 140 void __weak arch_disable_smp_support(void) { }
 141
 142 static int __init nosmp(char *str)
 143 {
 144         setup_max_cpus = 0;
 145         arch_disable_smp_support();
 146
 147         return 0;
 148 }
 149
 150 early_param("nosmp", nosmp);
 151
 152 /* 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 char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 201 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 202 static const char *panic_later, *panic_param;
 203
 204 extern struct obs_kernel_param __setup_start[], __setup_end[];
 205
 206 static int __init obsolete_checksetup(char *line)
 207 {
 208         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 noinline void __init_refok rest_init(void)
 425         __releases(kernel_lock)
 426 {
 427         int pid;
 428
 429         rcu_scheduler_starting();
 430         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 431         numa_default_policy();
 432         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 433         rcu_read_lock();
 434         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 435         rcu_read_unlock();
 436         unlock_kernel();
 437
 438         /*
 439          * The boot idle thread must execute schedule()
 440          * at least once to get things moving:
 441          */
 442         init_idle_bootup_task(current);
 443         preempt_enable_no_resched();
 444         schedule();
 445         preempt_disable();
 446
 447         /* Call into cpu_idle with preempt disabled */
 448         cpu_idle();
 449 }
 450
 451 /* Check for early params. */
 452 static int __init do_early_param(char *param, char *val)
 453 {
 454         struct obs_kernel_param *p;
 455
 456         for (p = __setup_start; p < __setup_end; p++) {
 457                 if ((p->early && strcmp(param, p->str) == 0) ||
 458                     (strcmp(param, "console") == 0 &&
 459                      strcmp(p->str, "earlycon") == 0)
 460                 ) {
 461                         if (p->setup_func(val) != 0)
 462                                 printk(KERN_WARNING
 463                                        "Malformed early option '%s'\n", param);
 464                 }
 465         }
 466         /* We accept everything at this stage. */
 467         return 0;
 468 }
 469
 470 void __init parse_early_options(char *cmdline)
 471 {
 472         parse_args("early options", cmdline, NULL, 0, do_early_param);
 473 }
 474
 475 /* Arch code calls this early on, or if not, just before other parsing. */
 476 void __init parse_early_param(void)
 477 {
 478         static __initdata int done = 0;
 479         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 480
 481         if (done)
 482                 return;
 483
 484         /* All fall through to do_early_param. */
 485         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 486         parse_early_options(tmp_cmdline);
 487         done = 1;
 488 }
 489
 490 /*
 491  *      Activate the first processor.
 492  */
 493
 494 static void __init boot_cpu_init(void)
 495 {
 496         int cpu = smp_processor_id();
 497         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 498         set_cpu_online(cpu, true);
 499         set_cpu_active(cpu, true);
 500         set_cpu_present(cpu, true);
 501         set_cpu_possible(cpu, true);
 502 }
 503
 504 void __init __weak smp_setup_processor_id(void)
 505 {
 506 }
 507
 508 void __init __weak thread_info_cache_init(void)
 509 {
 510 }
 511
 512 /*
 513  * Set up kernel memory allocators
 514  */
 515 static void __init mm_init(void)
 516 {
 517         /*
 518          * page_cgroup requires countinous pages as memmap
 519          * and it's bigger than MAX_ORDER unless SPARSEMEM.
 520          */
 521         page_cgroup_init_flatmem();
 522         mem_init();
 523         kmem_cache_init();
 524         pgtable_cache_init();
 525         vmalloc_init();
 526 }
 527
 528 asmlinkage void __init start_kernel(void)
 529 {
 530         char * command_line;
 531         extern struct kernel_param __start___param[], __stop___param[];
 532
 533         smp_setup_processor_id();
 534
 535         /*
 536          * Need to run as early as possible, to initialize the
 537          * lockdep hash:
 538          */
 539         lockdep_init();
 540         debug_objects_early_init();
 541
 542         /*
 543          * Set up the the initial canary ASAP:
 544          */
 545         boot_init_stack_canary();
 546
 547         cgroup_init_early();
 548
 549         local_irq_disable();
 550         early_boot_irqs_off();
 551         early_init_irq_lock_class();
 552
 553 /*
 554  * Interrupts are still disabled. Do necessary setups, then
 555  * enable them
 556  */
 557         lock_kernel();
 558         tick_init();
 559         boot_cpu_init();
 560         page_address_init();
 561         printk(KERN_NOTICE "%s", linux_banner);
 562         setup_arch(&command_line);
 563         mm_init_owner(&init_mm, &init_task);
 564         setup_command_line(command_line);
 565         setup_nr_cpu_ids();
 566         setup_per_cpu_areas();
 567         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 568
 569         build_all_zonelists();
 570         page_alloc_init();
 571
 572         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 573         parse_early_param();
 574         parse_args("Booting kernel", static_command_line, __start___param,
 575                    __stop___param - __start___param,
 576                    &unknown_bootoption);
 577         /*
 578          * These use large bootmem allocations and must precede
 579          * kmem_cache_init()
 580          */
 581         pidhash_init();
 582         vfs_caches_init_early();
 583         sort_main_extable();
 584         trap_init();
 585         mm_init();
 586         /*
 587          * Set up the scheduler prior starting any interrupts (such as the
 588          * timer interrupt). Full topology setup happens at smp_init()
 589          * time - but meanwhile we still have a functioning scheduler.
 590          */
 591         sched_init();
 592         /*
 593          * Disable preemption - early bootup scheduling is extremely
 594          * fragile until we cpu_idle() for the first time.
 595          */
 596         preempt_disable();
 597         if (!irqs_disabled()) {
 598                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 599                                 "enabled *very* early, fixing it\n");
 600                 local_irq_disable();
 601         }
 602         rcu_init();
 603         radix_tree_init();
 604         /* init some links before init_ISA_irqs() */
 605         early_irq_init();
 606         init_IRQ();
 607         prio_tree_init();
 608         init_timers();
 609         hrtimers_init();
 610         softirq_init();
 611         timekeeping_init();
 612         time_init();
 613         profile_init();
 614         if (!irqs_disabled())
 615                 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
 616                                  "enabled early\n");
 617         early_boot_irqs_on();
 618         local_irq_enable();
 619
 620         /* Interrupts are enabled now so all GFP allocations are safe. */
 621         gfp_allowed_mask = __GFP_BITS_MASK;
 622
 623         kmem_cache_init_late();
 624
 625         /*
 626          * HACK ALERT! This is early. We're enabling the console before
 627          * we've done PCI setups etc, and console_init() must be aware of
 628          * this. But we do want output early, in case something goes wrong.
 629          */
 630         console_init();
 631         if (panic_later)
 632                 panic(panic_later, panic_param);
 633
 634         lockdep_info();
 635
 636         /*
 637          * Need to run this when irqs are enabled, because it wants
 638          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 639          * too:
 640          */
 641         locking_selftest();
 642
 643 #ifdef CONFIG_BLK_DEV_INITRD
 644         if (initrd_start && !initrd_below_start_ok &&
 645             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 646                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 647                     "disabling it.\n",
 648                     page_to_pfn(virt_to_page((void *)initrd_start)),
 649                     min_low_pfn);
 650                 initrd_start = 0;
 651         }
 652 #endif
 653         page_cgroup_init();
 654         enable_debug_pagealloc();
 655         kmemtrace_init();
 656         kmemleak_init();
 657         debug_objects_mem_init();
 658         idr_init_cache();
 659         setup_per_cpu_pageset();
 660         numa_policy_init();
 661         if (late_time_init)
 662                 late_time_init();
 663         sched_clock_init();
 664         calibrate_delay();
 665         pidmap_init();
 666         anon_vma_init();
 667 #ifdef CONFIG_X86
 668         if (efi_enabled)
 669                 efi_enter_virtual_mode();
 670 #endif
 671         thread_info_cache_init();
 672         cred_init();
 673         fork_init(totalram_pages);
 674         proc_caches_init();
 675         buffer_init();
 676         key_init();
 677         security_init();
 678         vfs_caches_init(totalram_pages);
 679         signals_init();
 680         /* rootfs populating might need page-writeback */
 681         page_writeback_init();
 682 #ifdef CONFIG_PROC_FS
 683         proc_root_init();
 684 #endif
 685         cgroup_init();
 686         cpuset_init();
 687         taskstats_init_early();
 688         delayacct_init();
 689
 690         check_bugs();
 691
 692         acpi_early_init(); /* before LAPIC and SMP init */
 693         sfi_init_late();
 694
 695         ftrace_init();
 696
 697         /* Do the rest non-__init'ed, we're now alive */
 698         rest_init();
 699 }
 700
 701 /* Call all constructor functions linked into the kernel. */
 702 static void __init do_ctors(void)
 703 {
 704 #ifdef CONFIG_CONSTRUCTORS
 705         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 706
 707         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 708                 (*fn)();
 709 #endif
 710 }
 711
 712 int initcall_debug;
 713 core_param(initcall_debug, initcall_debug, bool, 0644);
 714
 715 static char msgbuf[64];
 716 static struct boot_trace_call call;
 717 static struct boot_trace_ret ret;
 718
 719 int do_one_initcall(initcall_t fn)
 720 {
 721         int count = preempt_count();
 722         ktime_t calltime, delta, rettime;
 723
 724         if (initcall_debug) {
 725                 call.caller = task_pid_nr(current);
 726                 printk("calling  %pF @ %i\n", fn, call.caller);
 727                 calltime = ktime_get();
 728                 trace_boot_call(&call, fn);
 729                 enable_boot_trace();
 730         }
 731
 732         ret.result = fn();
 733
 734         if (initcall_debug) {
 735                 disable_boot_trace();
 736                 rettime = ktime_get();
 737                 delta = ktime_sub(rettime, calltime);
 738                 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 739                 trace_boot_ret(&ret, fn);
 740                 printk("initcall %pF returned %d after %Ld usecs\n", fn,
 741                         ret.result, ret.duration);
 742         }
 743
 744         msgbuf[0] = 0;
 745
 746         if (ret.result && ret.result != -ENODEV && initcall_debug)
 747                 sprintf(msgbuf, "error code %d ", ret.result);
 748
 749         if (preempt_count() != count) {
 750                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 751                 preempt_count() = count;
 752         }
 753         if (irqs_disabled()) {
 754                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 755                 local_irq_enable();
 756         }
 757         if (msgbuf[0]) {
 758                 printk("initcall %pF returned with %s\n", fn, msgbuf);
 759         }
 760
 761         return ret.result;
 762 }
 763
 764
 765 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
 766
 767 static void __init do_initcalls(void)
 768 {
 769         initcall_t *fn;
 770
 771         for (fn = __early_initcall_end; fn < __initcall_end; fn++)
 772                 do_one_initcall(*fn);
 773
 774         /* Make sure there is no pending stuff from the initcall sequence */
 775         flush_scheduled_work();
 776 }
 777
 778 /*
 779  * Ok, the machine is now initialized. None of the devices
 780  * have been touched yet, but the CPU subsystem is up and
 781  * running, and memory and process management works.
 782  *
 783  * Now we can finally start doing some real work..
 784  */
 785 static void __init do_basic_setup(void)
 786 {
 787         init_workqueues();
 788         cpuset_init_smp();
 789         usermodehelper_init();
 790         init_tmpfs();
 791         driver_init();
 792         init_irq_proc();
 793         do_ctors();
 794         do_initcalls();
 795 }
 796
 797 static void __init do_pre_smp_initcalls(void)
 798 {
 799         initcall_t *fn;
 800
 801         for (fn = __initcall_start; fn < __early_initcall_end; fn++)
 802                 do_one_initcall(*fn);
 803 }
 804
 805 static void run_init_process(char *init_filename)
 806 {
 807         argv_init[0] = init_filename;
 808         kernel_execve(init_filename, argv_init, envp_init);
 809 }
 810
 811 /* This is a non __init function. Force it to be noinline otherwise gcc
 812  * makes it inline to init() and it becomes part of init.text section
 813  */
 814 static noinline int init_post(void)
 815         __releases(kernel_lock)
 816 {
 817         /* need to finish all async __init code before freeing the memory */
 818         async_synchronize_full();
 819         free_initmem();
 820         unlock_kernel();
 821         mark_rodata_ro();
 822         system_state = SYSTEM_RUNNING;
 823         numa_default_policy();
 824
 825
 826         current->signal->flags |= SIGNAL_UNKILLABLE;
 827
 828         if (ramdisk_execute_command) {
 829                 run_init_process(ramdisk_execute_command);
 830                 printk(KERN_WARNING "Failed to execute %s\n",
 831                                 ramdisk_execute_command);
 832         }
 833
 834         /*
 835          * We try each of these until one succeeds.
 836          *
 837          * The Bourne shell can be used instead of init if we are
 838          * trying to recover a really broken machine.
 839          */
 840         if (execute_command) {
 841                 run_init_process(execute_command);
 842                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 843                                         "defaults...\n", execute_command);
 844         }
 845         run_init_process("/sbin/init");
 846         run_init_process("/etc/init");
 847         run_init_process("/bin/init");
 848         run_init_process("/bin/sh");
 849
 850         panic("No init found.  Try passing init= option to kernel. "
 851               "See Linux Documentation/init.txt for guidance.");
 852 }
 853
 854 static int __init kernel_init(void * unused)
 855 {
 856         lock_kernel();
 857
 858         /*
 859          * init can allocate pages on any node
 860          */
 861         set_mems_allowed(node_states[N_HIGH_MEMORY]);
 862         /*
 863          * init can run on any cpu.
 864          */
 865         set_cpus_allowed_ptr(current, cpu_all_mask);
 866         /*
 867          * Tell the world that we're going to be the grim
 868          * reaper of innocent orphaned children.
 869          *
 870          * We don't want people to have to make incorrect
 871          * assumptions about where in the task array this
 872          * can be found.
 873          */
 874         init_pid_ns.child_reaper = current;
 875
 876         cad_pid = task_pid(current);
 877
 878         smp_prepare_cpus(setup_max_cpus);
 879
 880         do_pre_smp_initcalls();
 881         start_boot_trace();
 882
 883         smp_init();
 884         sched_init_smp();
 885
 886         do_basic_setup();
 887
 888         /* Open the /dev/console on the rootfs, this should never fail */
 889         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 890                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 891
 892         (void) sys_dup(0);
 893         (void) sys_dup(0);
 894         /*
 895          * check if there is an early userspace init.  If yes, let it do all
 896          * the work
 897          */
 898
 899         if (!ramdisk_execute_command)
 900                 ramdisk_execute_command = "/init";
 901
 902         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 903                 ramdisk_execute_command = NULL;
 904                 prepare_namespace();
 905         }
 906
 907         /*
 908          * Ok, we have completed the initial bootup, and
 909          * we're essentially up and running. Get rid of the
 910          * initmem segments and start the user-mode stuff..
 911          */
 912
 913         init_post();
 914         return 0;
 915 }