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/kmemleak.h>
  60 #include <linux/pid_namespace.h>
  61 #include <linux/device.h>
  62 #include <linux/kthread.h>
  63 #include <linux/sched.h>
  64 #include <linux/signal.h>
  65 #include <linux/idr.h>
  66 #include <linux/ftrace.h>
  67 #include <linux/async.h>
  68 #include <linux/kmemcheck.h>
  69 #include <linux/kmemtrace.h>
  70 #include <trace/boot.h>
  71
  72 #include <asm/io.h>
  73 #include <asm/bugs.h>
  74 #include <asm/setup.h>
  75 #include <asm/sections.h>
  76 #include <asm/cacheflush.h>
  77
  78 #ifdef CONFIG_X86_LOCAL_APIC
  79 #include <asm/smp.h>
  80 #endif
  81
  82 static int kernel_init(void *);
  83
  84 extern void init_IRQ(void);
  85 extern void fork_init(unsigned long);
  86 extern void mca_init(void);
  87 extern void sbus_init(void);
  88 extern void prio_tree_init(void);
  89 extern void radix_tree_init(void);
  90 extern void free_initmem(void);
  91 #ifdef  CONFIG_ACPI
  92 extern void acpi_early_init(void);
  93 #else
  94 static inline void acpi_early_init(void) { }
  95 #endif
  96 #ifndef CONFIG_DEBUG_RODATA
  97 static inline void mark_rodata_ro(void) { }
  98 #endif
  99
 100 #ifdef CONFIG_TC
 101 extern void tc_init(void);
 102 #endif
 103
 104 enum system_states system_state __read_mostly;
 105 EXPORT_SYMBOL(system_state);
 106
 107 /*
 108  * Boot command-line arguments
 109  */
 110 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 111 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 112
 113 extern void time_init(void);
 114 /* Default late time init is NULL. archs can override this later. */
 115 void (*__initdata late_time_init)(void);
 116 extern void softirq_init(void);
 117
 118 /* Untouched command line saved by arch-specific code. */
 119 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 120 /* Untouched saved command line (eg. for /proc) */
 121 char *saved_command_line;
 122 /* Command line for parameter parsing */
 123 static char *static_command_line;
 124
 125 static char *execute_command;
 126 static char *ramdisk_execute_command;
 127
 128 #ifdef CONFIG_SMP
 129 /* Setup configured maximum number of CPUs to activate */
 130 unsigned int __initdata setup_max_cpus = NR_CPUS;
 131
 132 /*
 133  * Setup routine for controlling SMP activation
 134  *
 135  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 136  * activation entirely (the MPS table probe still happens, though).
 137  *
 138  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 139  * greater than 0, limits the maximum number of CPUs activated in
 140  * SMP mode to <NUM>.
 141  */
 142
 143 void __weak arch_disable_smp_support(void) { }
 144
 145 static int __init nosmp(char *str)
 146 {
 147         setup_max_cpus = 0;
 148         arch_disable_smp_support();
 149
 150         return 0;
 151 }
 152
 153 early_param("nosmp", nosmp);
 154
 155 static int __init maxcpus(char *str)
 156 {
 157         get_option(&str, &setup_max_cpus);
 158         if (setup_max_cpus == 0)
 159                 arch_disable_smp_support();
 160
 161         return 0;
 162 }
 163
 164 early_param("maxcpus", maxcpus);
 165 #else
 166 const unsigned int setup_max_cpus = NR_CPUS;
 167 #endif
 168
 169 /*
 170  * If set, this is an indication to the drivers that reset the underlying
 171  * device before going ahead with the initialization otherwise driver might
 172  * rely on the BIOS and skip the reset operation.
 173  *
 174  * This is useful if kernel is booting in an unreliable environment.
 175  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 176  * skipped and devices will be in unknown state.
 177  */
 178 unsigned int reset_devices;
 179 EXPORT_SYMBOL(reset_devices);
 180
 181 static int __init set_reset_devices(char *str)
 182 {
 183         reset_devices = 1;
 184         return 1;
 185 }
 186
 187 __setup("reset_devices", set_reset_devices);
 188
 189 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 190 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 191 static const char *panic_later, *panic_param;
 192
 193 extern struct obs_kernel_param __setup_start[], __setup_end[];
 194
 195 static int __init obsolete_checksetup(char *line)
 196 {
 197         struct obs_kernel_param *p;
 198         int had_early_param = 0;
 199
 200         p = __setup_start;
 201         do {
 202                 int n = strlen(p->str);
 203                 if (!strncmp(line, p->str, n)) {
 204                         if (p->early) {
 205                                 /* Already done in parse_early_param?
 206                                  * (Needs exact match on param part).
 207                                  * Keep iterating, as we can have early
 208                                  * params and __setups of same names 8( */
 209                                 if (line[n] == '\0' || line[n] == '=')
 210                                         had_early_param = 1;
 211                         } else if (!p->setup_func) {
 212                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 213                                        " ignored\n", p->str);
 214                                 return 1;
 215                         } else if (p->setup_func(line + n))
 216                                 return 1;
 217                 }
 218                 p++;
 219         } while (p < __setup_end);
 220
 221         return had_early_param;
 222 }
 223
 224 /*
 225  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 226  * still work even if initially too large, it will just take slightly longer
 227  */
 228 unsigned long loops_per_jiffy = (1<<12);
 229
 230 EXPORT_SYMBOL(loops_per_jiffy);
 231
 232 static int __init debug_kernel(char *str)
 233 {
 234         console_loglevel = 10;
 235         return 0;
 236 }
 237
 238 static int __init quiet_kernel(char *str)
 239 {
 240         console_loglevel = 4;
 241         return 0;
 242 }
 243
 244 early_param("debug", debug_kernel);
 245 early_param("quiet", quiet_kernel);
 246
 247 static int __init loglevel(char *str)
 248 {
 249         get_option(&str, &console_loglevel);
 250         return 0;
 251 }
 252
 253 early_param("loglevel", loglevel);
 254
 255 /*
 256  * Unknown boot options get handed to init, unless they look like
 257  * failed parameters
 258  */
 259 static int __init unknown_bootoption(char *param, char *val)
 260 {
 261         /* Change NUL term back to "=", to make "param" the whole string. */
 262         if (val) {
 263                 /* param=val or param="val"? */
 264                 if (val == param+strlen(param)+1)
 265                         val[-1] = '=';
 266                 else if (val == param+strlen(param)+2) {
 267                         val[-2] = '=';
 268                         memmove(val-1, val, strlen(val)+1);
 269                         val--;
 270                 } else
 271                         BUG();
 272         }
 273
 274         /* Handle obsolete-style parameters */
 275         if (obsolete_checksetup(param))
 276                 return 0;
 277
 278         /*
 279          * Preemptive maintenance for "why didn't my misspelled command
 280          * line work?"
 281          */
 282         if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
 283                 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
 284                 return 0;
 285         }
 286
 287         if (panic_later)
 288                 return 0;
 289
 290         if (val) {
 291                 /* Environment option */
 292                 unsigned int i;
 293                 for (i = 0; envp_init[i]; i++) {
 294                         if (i == MAX_INIT_ENVS) {
 295                                 panic_later = "Too many boot env vars at `%s'";
 296                                 panic_param = param;
 297                         }
 298                         if (!strncmp(param, envp_init[i], val - param))
 299                                 break;
 300                 }
 301                 envp_init[i] = param;
 302         } else {
 303                 /* Command line option */
 304                 unsigned int i;
 305                 for (i = 0; argv_init[i]; i++) {
 306                         if (i == MAX_INIT_ARGS) {
 307                                 panic_later = "Too many boot init vars at `%s'";
 308                                 panic_param = param;
 309                         }
 310                 }
 311                 argv_init[i] = param;
 312         }
 313         return 0;
 314 }
 315
 316 #ifdef CONFIG_DEBUG_PAGEALLOC
 317 int __read_mostly debug_pagealloc_enabled = 0;
 318 #endif
 319
 320 static int __init init_setup(char *str)
 321 {
 322         unsigned int i;
 323
 324         execute_command = str;
 325         /*
 326          * In case LILO is going to boot us with default command line,
 327          * it prepends "auto" before the whole cmdline which makes
 328          * the shell think it should execute a script with such name.
 329          * So we ignore all arguments entered _before_ init=... [MJ]
 330          */
 331         for (i = 1; i < MAX_INIT_ARGS; i++)
 332                 argv_init[i] = NULL;
 333         return 1;
 334 }
 335 __setup("init=", init_setup);
 336
 337 static int __init rdinit_setup(char *str)
 338 {
 339         unsigned int i;
 340
 341         ramdisk_execute_command = str;
 342         /* See "auto" comment in init_setup */
 343         for (i = 1; i < MAX_INIT_ARGS; i++)
 344                 argv_init[i] = NULL;
 345         return 1;
 346 }
 347 __setup("rdinit=", rdinit_setup);
 348
 349 #ifndef CONFIG_SMP
 350
 351 #ifdef CONFIG_X86_LOCAL_APIC
 352 static void __init smp_init(void)
 353 {
 354         APIC_init_uniprocessor();
 355 }
 356 #else
 357 #define smp_init()      do { } while (0)
 358 #endif
 359
 360 static inline void setup_per_cpu_areas(void) { }
 361 static inline void setup_nr_cpu_ids(void) { }
 362 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 363
 364 #else
 365
 366 #if NR_CPUS > BITS_PER_LONG
 367 cpumask_t cpu_mask_all __read_mostly = CPU_MASK_ALL;
 368 EXPORT_SYMBOL(cpu_mask_all);
 369 #endif
 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 #ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
 382 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
 383
 384 EXPORT_SYMBOL(__per_cpu_offset);
 385
 386 static void __init setup_per_cpu_areas(void)
 387 {
 388         unsigned long size, i;
 389         char *ptr;
 390         unsigned long nr_possible_cpus = num_possible_cpus();
 391
 392         /* Copy section for each CPU (we discard the original) */
 393         size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
 394         ptr = alloc_bootmem_pages(size * nr_possible_cpus);
 395
 396         for_each_possible_cpu(i) {
 397                 __per_cpu_offset[i] = ptr - __per_cpu_start;
 398                 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 399                 ptr += size;
 400         }
 401 }
 402 #endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
 403
 404 /* Called by boot processor to activate the rest. */
 405 static void __init smp_init(void)
 406 {
 407         unsigned int cpu;
 408
 409         /*
 410          * Set up the current CPU as possible to migrate to.
 411          * The other ones will be done by cpu_up/cpu_down()
 412          */
 413         set_cpu_active(smp_processor_id(), true);
 414
 415         /* FIXME: This should be done in userspace --RR */
 416         for_each_present_cpu(cpu) {
 417                 if (num_online_cpus() >= setup_max_cpus)
 418                         break;
 419                 if (!cpu_online(cpu))
 420                         cpu_up(cpu);
 421         }
 422
 423         /* Any cleanup work */
 424         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 425         smp_cpus_done(setup_max_cpus);
 426 }
 427
 428 #endif
 429
 430 /*
 431  * We need to store the untouched command line for future reference.
 432  * We also need to store the touched command line since the parameter
 433  * parsing is performed in place, and we should allow a component to
 434  * store reference of name/value for future reference.
 435  */
 436 static void __init setup_command_line(char *command_line)
 437 {
 438         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 439         static_command_line = alloc_bootmem(strlen (command_line)+1);
 440         strcpy (saved_command_line, boot_command_line);
 441         strcpy (static_command_line, command_line);
 442 }
 443
 444 /*
 445  * We need to finalize in a non-__init function or else race conditions
 446  * between the root thread and the init thread may cause start_kernel to
 447  * be reaped by free_initmem before the root thread has proceeded to
 448  * cpu_idle.
 449  *
 450  * gcc-3.4 accidentally inlines this function, so use noinline.
 451  */
 452
 453 static noinline void __init_refok rest_init(void)
 454         __releases(kernel_lock)
 455 {
 456         int pid;
 457
 458         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 459         numa_default_policy();
 460         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 461         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 462         unlock_kernel();
 463
 464         /*
 465          * The boot idle thread must execute schedule()
 466          * at least once to get things moving:
 467          */
 468         init_idle_bootup_task(current);
 469         rcu_scheduler_starting();
 470         preempt_enable_no_resched();
 471         schedule();
 472         preempt_disable();
 473
 474         /* Call into cpu_idle with preempt disabled */
 475         cpu_idle();
 476 }
 477
 478 /* Check for early params. */
 479 static int __init do_early_param(char *param, char *val)
 480 {
 481         struct obs_kernel_param *p;
 482
 483         for (p = __setup_start; p < __setup_end; p++) {
 484                 if ((p->early && strcmp(param, p->str) == 0) ||
 485                     (strcmp(param, "console") == 0 &&
 486                      strcmp(p->str, "earlycon") == 0)
 487                 ) {
 488                         if (p->setup_func(val) != 0)
 489                                 printk(KERN_WARNING
 490                                        "Malformed early option '%s'\n", param);
 491                 }
 492         }
 493         /* We accept everything at this stage. */
 494         return 0;
 495 }
 496
 497 void __init parse_early_options(char *cmdline)
 498 {
 499         parse_args("early options", cmdline, NULL, 0, do_early_param);
 500 }
 501
 502 /* Arch code calls this early on, or if not, just before other parsing. */
 503 void __init parse_early_param(void)
 504 {
 505         static __initdata int done = 0;
 506         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 507
 508         if (done)
 509                 return;
 510
 511         /* All fall through to do_early_param. */
 512         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 513         parse_early_options(tmp_cmdline);
 514         done = 1;
 515 }
 516
 517 /*
 518  *      Activate the first processor.
 519  */
 520
 521 static void __init boot_cpu_init(void)
 522 {
 523         int cpu = smp_processor_id();
 524         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 525         set_cpu_online(cpu, true);
 526         set_cpu_present(cpu, true);
 527         set_cpu_possible(cpu, true);
 528 }
 529
 530 void __init __weak smp_setup_processor_id(void)
 531 {
 532 }
 533
 534 void __init __weak thread_info_cache_init(void)
 535 {
 536 }
 537
 538 /*
 539  * Set up kernel memory allocators
 540  */
 541 static void __init mm_init(void)
 542 {
 543         mem_init();
 544         kmem_cache_init();
 545         vmalloc_init();
 546 }
 547
 548 asmlinkage void __init start_kernel(void)
 549 {
 550         char * command_line;
 551         extern struct kernel_param __start___param[], __stop___param[];
 552
 553         smp_setup_processor_id();
 554
 555         /*
 556          * Need to run as early as possible, to initialize the
 557          * lockdep hash:
 558          */
 559         lockdep_init();
 560         debug_objects_early_init();
 561
 562         /*
 563          * Set up the the initial canary ASAP:
 564          */
 565         boot_init_stack_canary();
 566
 567         cgroup_init_early();
 568
 569         local_irq_disable();
 570         early_boot_irqs_off();
 571         early_init_irq_lock_class();
 572
 573 /*
 574  * Interrupts are still disabled. Do necessary setups, then
 575  * enable them
 576  */
 577         lock_kernel();
 578         tick_init();
 579         boot_cpu_init();
 580         page_address_init();
 581         printk(KERN_NOTICE "%s", linux_banner);
 582         setup_arch(&command_line);
 583         mm_init_owner(&init_mm, &init_task);
 584         setup_command_line(command_line);
 585         setup_per_cpu_areas();
 586         setup_nr_cpu_ids();
 587         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 588
 589         build_all_zonelists();
 590         page_alloc_init();
 591
 592         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 593         parse_early_param();
 594         parse_args("Booting kernel", static_command_line, __start___param,
 595                    __stop___param - __start___param,
 596                    &unknown_bootoption);
 597         /*
 598          * These use large bootmem allocations and must precede
 599          * kmem_cache_init()
 600          */
 601         pidhash_init();
 602         vfs_caches_init_early();
 603         sort_main_extable();
 604         trap_init();
 605         mm_init();
 606         /*
 607          * Set up the scheduler prior starting any interrupts (such as the
 608          * timer interrupt). Full topology setup happens at smp_init()
 609          * time - but meanwhile we still have a functioning scheduler.
 610          */
 611         sched_init();
 612         /*
 613          * Disable preemption - early bootup scheduling is extremely
 614          * fragile until we cpu_idle() for the first time.
 615          */
 616         preempt_disable();
 617         if (!irqs_disabled()) {
 618                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 619                                 "enabled *very* early, fixing it\n");
 620                 local_irq_disable();
 621         }
 622         rcu_init();
 623         /* init some links before init_ISA_irqs() */
 624         early_irq_init();
 625         init_IRQ();
 626         prio_tree_init();
 627         init_timers();
 628         hrtimers_init();
 629         softirq_init();
 630         timekeeping_init();
 631         time_init();
 632         sched_clock_init();
 633         profile_init();
 634         if (!irqs_disabled())
 635                 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
 636                                  "enabled early\n");
 637         early_boot_irqs_on();
 638         local_irq_enable();
 639
 640         /*
 641          * HACK ALERT! This is early. We're enabling the console before
 642          * we've done PCI setups etc, and console_init() must be aware of
 643          * this. But we do want output early, in case something goes wrong.
 644          */
 645         console_init();
 646         if (panic_later)
 647                 panic(panic_later, panic_param);
 648
 649         lockdep_info();
 650
 651         /*
 652          * Need to run this when irqs are enabled, because it wants
 653          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 654          * too:
 655          */
 656         locking_selftest();
 657
 658 #ifdef CONFIG_BLK_DEV_INITRD
 659         if (initrd_start && !initrd_below_start_ok &&
 660             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 661                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 662                     "disabling it.\n",
 663                     page_to_pfn(virt_to_page((void *)initrd_start)),
 664                     min_low_pfn);
 665                 initrd_start = 0;
 666         }
 667 #endif
 668         cpuset_init_early();
 669         page_cgroup_init();
 670         enable_debug_pagealloc();
 671         cpu_hotplug_init();
 672         kmemtrace_init();
 673         kmemleak_init();
 674         debug_objects_mem_init();
 675         idr_init_cache();
 676         setup_per_cpu_pageset();
 677         numa_policy_init();
 678         if (late_time_init)
 679                 late_time_init();
 680         calibrate_delay();
 681         pidmap_init();
 682         pgtable_cache_init();
 683         anon_vma_init();
 684 #ifdef CONFIG_X86
 685         if (efi_enabled)
 686                 efi_enter_virtual_mode();
 687 #endif
 688         thread_info_cache_init();
 689         cred_init();
 690         fork_init(num_physpages);
 691         proc_caches_init();
 692         buffer_init();
 693         key_init();
 694         security_init();
 695         vfs_caches_init(num_physpages);
 696         radix_tree_init();
 697         signals_init();
 698         /* rootfs populating might need page-writeback */
 699         page_writeback_init();
 700 #ifdef CONFIG_PROC_FS
 701         proc_root_init();
 702 #endif
 703         cgroup_init();
 704         cpuset_init();
 705         taskstats_init_early();
 706         delayacct_init();
 707
 708         check_bugs();
 709
 710         acpi_early_init(); /* before LAPIC and SMP init */
 711
 712         ftrace_init();
 713
 714         /* Do the rest non-__init'ed, we're now alive */
 715         rest_init();
 716 }
 717
 718 int initcall_debug;
 719 core_param(initcall_debug, initcall_debug, bool, 0644);
 720
 721 int do_one_initcall(initcall_t fn)
 722 {
 723         int count = preempt_count();
 724         ktime_t calltime, delta, rettime;
 725         char msgbuf[64];
 726         struct boot_trace_call call;
 727         struct boot_trace_ret ret;
 728
 729         if (initcall_debug) {
 730                 call.caller = task_pid_nr(current);
 731                 printk("calling  %pF @ %i\n", fn, call.caller);
 732                 calltime = ktime_get();
 733                 trace_boot_call(&call, fn);
 734                 enable_boot_trace();
 735         }
 736
 737         ret.result = fn();
 738
 739         if (initcall_debug) {
 740                 disable_boot_trace();
 741                 rettime = ktime_get();
 742                 delta = ktime_sub(rettime, calltime);
 743                 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 744                 trace_boot_ret(&ret, fn);
 745                 printk("initcall %pF returned %d after %Ld usecs\n", fn,
 746                         ret.result, ret.duration);
 747         }
 748
 749         msgbuf[0] = 0;
 750
 751         if (ret.result && ret.result != -ENODEV && initcall_debug)
 752                 sprintf(msgbuf, "error code %d ", ret.result);
 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.result;
 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 *call;
 775
 776         for (call = __early_initcall_end; call < __initcall_end; call++)
 777                 do_one_initcall(*call);
 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         rcu_init_sched(); /* needed by module_init stage. */
 793         init_workqueues();
 794         cpuset_init_smp();
 795         usermodehelper_init();
 796         driver_init();
 797         init_irq_proc();
 798         do_initcalls();
 799 }
 800
 801 static void __init do_pre_smp_initcalls(void)
 802 {
 803         initcall_t *call;
 804
 805         for (call = __initcall_start; call < __early_initcall_end; call++)
 806                 do_one_initcall(*call);
 807 }
 808
 809 static void run_init_process(char *init_filename)
 810 {
 811         argv_init[0] = init_filename;
 812         kernel_execve(init_filename, argv_init, envp_init);
 813 }
 814
 815 /* This is a non __init function. Force it to be noinline otherwise gcc
 816  * makes it inline to init() and it becomes part of init.text section
 817  */
 818 static noinline int init_post(void)
 819         __releases(kernel_lock)
 820 {
 821         /* need to finish all async __init code before freeing the memory */
 822         async_synchronize_full();
 823         free_initmem();
 824         unlock_kernel();
 825         mark_rodata_ro();
 826         system_state = SYSTEM_RUNNING;
 827         numa_default_policy();
 828
 829         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 830                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 831
 832         (void) sys_dup(0);
 833         (void) sys_dup(0);
 834
 835         current->signal->flags |= SIGNAL_UNKILLABLE;
 836
 837         if (ramdisk_execute_command) {
 838                 run_init_process(ramdisk_execute_command);
 839                 printk(KERN_WARNING "Failed to execute %s\n",
 840                                 ramdisk_execute_command);
 841         }
 842
 843         /*
 844          * We try each of these until one succeeds.
 845          *
 846          * The Bourne shell can be used instead of init if we are
 847          * trying to recover a really broken machine.
 848          */
 849         if (execute_command) {
 850                 run_init_process(execute_command);
 851                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 852                                         "defaults...\n", execute_command);
 853         }
 854         run_init_process("/sbin/init");
 855         run_init_process("/etc/init");
 856         run_init_process("/bin/init");
 857         run_init_process("/bin/sh");
 858
 859         panic("No init found.  Try passing init= option to kernel.");
 860 }
 861
 862 static int __init kernel_init(void * unused)
 863 {
 864         lock_kernel();
 865         /*
 866          * init can run on any cpu.
 867          */
 868         set_cpus_allowed_ptr(current, cpu_all_mask);
 869         /*
 870          * Tell the world that we're going to be the grim
 871          * reaper of innocent orphaned children.
 872          *
 873          * We don't want people to have to make incorrect
 874          * assumptions about where in the task array this
 875          * can be found.
 876          */
 877         init_pid_ns.child_reaper = current;
 878
 879         cad_pid = task_pid(current);
 880
 881         smp_prepare_cpus(setup_max_cpus);
 882
 883         do_pre_smp_initcalls();
 884         start_boot_trace();
 885
 886         smp_init();
 887         sched_init_smp();
 888
 889         do_basic_setup();
 890
 891         /*
 892          * check if there is an early userspace init.  If yes, let it do all
 893          * the work
 894          */
 895
 896         if (!ramdisk_execute_command)
 897                 ramdisk_execute_command = "/init";
 898
 899         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 900                 ramdisk_execute_command = NULL;
 901                 prepare_namespace();
 902         }
 903
 904         /*
 905          * Ok, we have completed the initial bootup, and
 906          * we're essentially up and running. Get rid of the
 907          * initmem segments and start the user-mode stuff..
 908          */
 909
 910         init_post();
 911         return 0;
 912 }