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