init/main.c

   1 /*
   2  *  linux/init/main.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  *
   6  *  GK 2/5/95  -  Changed to support mounting root fs via NFS
   7  *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
   8  *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
   9  *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
  10  */
  11
  12 #include <linux/types.h>
  13 #include <linux/module.h>
  14 #include <linux/proc_fs.h>
  15 #include <linux/kernel.h>
  16 #include <linux/syscalls.h>
  17 #include <linux/stackprotector.h>
  18 #include <linux/string.h>
  19 #include <linux/ctype.h>
  20 #include <linux/delay.h>
  21 #include <linux/ioport.h>
  22 #include <linux/init.h>
  23 #include <linux/smp_lock.h>
  24 #include <linux/initrd.h>
  25 #include <linux/bootmem.h>
  26 #include <linux/acpi.h>
  27 #include <linux/tty.h>
  28 #include <linux/percpu.h>
  29 #include <linux/kmod.h>
  30 #include <linux/vmalloc.h>
  31 #include <linux/kernel_stat.h>
  32 #include <linux/start_kernel.h>
  33 #include <linux/security.h>
  34 #include <linux/smp.h>
  35 #include <linux/workqueue.h>
  36 #include <linux/profile.h>
  37 #include <linux/rcupdate.h>
  38 #include <linux/moduleparam.h>
  39 #include <linux/kallsyms.h>
  40 #include <linux/writeback.h>
  41 #include <linux/cpu.h>
  42 #include <linux/cpuset.h>
  43 #include <linux/cgroup.h>
  44 #include <linux/efi.h>
  45 #include <linux/tick.h>
  46 #include <linux/interrupt.h>
  47 #include <linux/taskstats_kern.h>
  48 #include <linux/delayacct.h>
  49 #include <linux/unistd.h>
  50 #include <linux/rmap.h>
  51 #include <linux/mempolicy.h>
  52 #include <linux/key.h>
  53 #include <linux/buffer_head.h>
  54 #include <linux/page_cgroup.h>
  55 #include <linux/debug_locks.h>
  56 #include <linux/debugobjects.h>
  57 #include <linux/lockdep.h>
  58 #include <linux/kmemleak.h>
  59 #include <linux/pid_namespace.h>
  60 #include <linux/device.h>
  61 #include <linux/kthread.h>
  62 #include <linux/sched.h>
  63 #include <linux/signal.h>
  64 #include <linux/idr.h>
  65 #include <linux/kgdb.h>
  66 #include <linux/ftrace.h>
  67 #include <linux/async.h>
  68 #include <linux/kmemcheck.h>
  69 #include <linux/kmemtrace.h>
  70 #include <linux/sfi.h>
  71 #include <linux/shmem_fs.h>
  72 #include <linux/slab.h>
  73 #include <trace/boot.h>
  74
  75 #include <asm/io.h>
  76 #include <asm/bugs.h>
  77 #include <asm/setup.h>
  78 #include <asm/sections.h>
  79 #include <asm/cacheflush.h>
  80
  81 #ifdef CONFIG_X86_LOCAL_APIC
  82 #include <asm/smp.h>
  83 #endif
  84
  85 static int kernel_init(void *);
  86
  87 extern void init_IRQ(void);
  88 extern void fork_init(unsigned long);
  89 extern void mca_init(void);
  90 extern void sbus_init(void);
  91 extern void prio_tree_init(void);
  92 extern void radix_tree_init(void);
  93 extern void free_initmem(void);
  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 setup_max_cpus = NR_CPUS;
 129 EXPORT_SYMBOL(setup_max_cpus);
 130
 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 /* this is hard limit */
 156 static int __init nrcpus(char *str)
 157 {
 158         int nr_cpus;
 159
 160         get_option(&str, &nr_cpus);
 161         if (nr_cpus > 0 && nr_cpus < nr_cpu_ids)
 162                 nr_cpu_ids = nr_cpus;
 163
 164         return 0;
 165 }
 166
 167 early_param("nr_cpus", nrcpus);
 168
 169 static int __init maxcpus(char *str)
 170 {
 171         get_option(&str, &setup_max_cpus);
 172         if (setup_max_cpus == 0)
 173                 arch_disable_smp_support();
 174
 175         return 0;
 176 }
 177
 178 early_param("maxcpus", maxcpus);
 179 #else
 180 static const unsigned int setup_max_cpus = NR_CPUS;
 181 #endif
 182
 183 /*
 184  * If set, this is an indication to the drivers that reset the underlying
 185  * device before going ahead with the initialization otherwise driver might
 186  * rely on the BIOS and skip the reset operation.
 187  *
 188  * This is useful if kernel is booting in an unreliable environment.
 189  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 190  * skipped and devices will be in unknown state.
 191  */
 192 unsigned int reset_devices;
 193 EXPORT_SYMBOL(reset_devices);
 194
 195 static int __init set_reset_devices(char *str)
 196 {
 197         reset_devices = 1;
 198         return 1;
 199 }
 200
 201 __setup("reset_devices", set_reset_devices);
 202
 203 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 204 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 205 static const char *panic_later, *panic_param;
 206
 207 extern struct obs_kernel_param __setup_start[], __setup_end[];
 208
 209 static int __init obsolete_checksetup(char *line)
 210 {
 211         struct obs_kernel_param *p;
 212         int had_early_param = 0;
 213
 214         p = __setup_start;
 215         do {
 216                 int n = strlen(p->str);
 217                 if (!strncmp(line, p->str, n)) {
 218                         if (p->early) {
 219                                 /* Already done in parse_early_param?
 220                                  * (Needs exact match on param part).
 221                                  * Keep iterating, as we can have early
 222                                  * params and __setups of same names 8( */
 223                                 if (line[n] == '\0' || line[n] == '=')
 224                                         had_early_param = 1;
 225                         } else if (!p->setup_func) {
 226                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 227                                        " ignored\n", p->str);
 228                                 return 1;
 229                         } else if (p->setup_func(line + n))
 230                                 return 1;
 231                 }
 232                 p++;
 233         } while (p < __setup_end);
 234
 235         return had_early_param;
 236 }
 237
 238 /*
 239  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 240  * still work even if initially too large, it will just take slightly longer
 241  */
 242 unsigned long loops_per_jiffy = (1<<12);
 243
 244 EXPORT_SYMBOL(loops_per_jiffy);
 245
 246 static int __init debug_kernel(char *str)
 247 {
 248         console_loglevel = 10;
 249         return 0;
 250 }
 251
 252 static int __init quiet_kernel(char *str)
 253 {
 254         console_loglevel = 4;
 255         return 0;
 256 }
 257
 258 early_param("debug", debug_kernel);
 259 early_param("quiet", quiet_kernel);
 260
 261 static int __init loglevel(char *str)
 262 {
 263         get_option(&str, &console_loglevel);
 264         return 0;
 265 }
 266
 267 early_param("loglevel", loglevel);
 268
 269 /*
 270  * Unknown boot options get handed to init, unless they look like
 271  * unused parameters (modprobe will find them in /proc/cmdline).
 272  */
 273 static int __init unknown_bootoption(char *param, char *val)
 274 {
 275         /* Change NUL term back to "=", to make "param" the whole string. */
 276         if (val) {
 277                 /* param=val or param="val"? */
 278                 if (val == param+strlen(param)+1)
 279                         val[-1] = '=';
 280                 else if (val == param+strlen(param)+2) {
 281                         val[-2] = '=';
 282                         memmove(val-1, val, strlen(val)+1);
 283                         val--;
 284                 } else
 285                         BUG();
 286         }
 287
 288         /* Handle obsolete-style parameters */
 289         if (obsolete_checksetup(param))
 290                 return 0;
 291
 292         /* Unused module parameter. */
 293         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 294                 return 0;
 295
 296         if (panic_later)
 297                 return 0;
 298
 299         if (val) {
 300                 /* Environment option */
 301                 unsigned int i;
 302                 for (i = 0; envp_init[i]; i++) {
 303                         if (i == MAX_INIT_ENVS) {
 304                                 panic_later = "Too many boot env vars at `%s'";
 305                                 panic_param = param;
 306                         }
 307                         if (!strncmp(param, envp_init[i], val - param))
 308                                 break;
 309                 }
 310                 envp_init[i] = param;
 311         } else {
 312                 /* Command line option */
 313                 unsigned int i;
 314                 for (i = 0; argv_init[i]; i++) {
 315                         if (i == MAX_INIT_ARGS) {
 316                                 panic_later = "Too many boot init vars at `%s'";
 317                                 panic_param = param;
 318                         }
 319                 }
 320                 argv_init[i] = param;
 321         }
 322         return 0;
 323 }
 324
 325 #ifdef CONFIG_DEBUG_PAGEALLOC
 326 int __read_mostly debug_pagealloc_enabled = 0;
 327 #endif
 328
 329 static int __init init_setup(char *str)
 330 {
 331         unsigned int i;
 332
 333         execute_command = str;
 334         /*
 335          * In case LILO is going to boot us with default command line,
 336          * it prepends "auto" before the whole cmdline which makes
 337          * the shell think it should execute a script with such name.
 338          * So we ignore all arguments entered _before_ init=... [MJ]
 339          */
 340         for (i = 1; i < MAX_INIT_ARGS; i++)
 341                 argv_init[i] = NULL;
 342         return 1;
 343 }
 344 __setup("init=", init_setup);
 345
 346 static int __init rdinit_setup(char *str)
 347 {
 348         unsigned int i;
 349
 350         ramdisk_execute_command = str;
 351         /* See "auto" comment in init_setup */
 352         for (i = 1; i < MAX_INIT_ARGS; i++)
 353                 argv_init[i] = NULL;
 354         return 1;
 355 }
 356 __setup("rdinit=", rdinit_setup);
 357
 358 #ifndef CONFIG_SMP
 359
 360 #ifdef CONFIG_X86_LOCAL_APIC
 361 static void __init smp_init(void)
 362 {
 363         APIC_init_uniprocessor();
 364 }
 365 #else
 366 #define smp_init()      do { } while (0)
 367 #endif
 368
 369 static inline void setup_nr_cpu_ids(void) { }
 370 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 371
 372 #else
 373
 374 /* Setup number of possible processor ids */
 375 int nr_cpu_ids __read_mostly = NR_CPUS;
 376 EXPORT_SYMBOL(nr_cpu_ids);
 377
 378 /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
 379 static void __init setup_nr_cpu_ids(void)
 380 {
 381         nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
 382 }
 383
 384 /* Called by boot processor to activate the rest. */
 385 static void __init smp_init(void)
 386 {
 387         unsigned int cpu;
 388
 389         /* FIXME: This should be done in userspace --RR */
 390         for_each_present_cpu(cpu) {
 391                 if (num_online_cpus() >= setup_max_cpus)
 392                         break;
 393                 if (!cpu_online(cpu))
 394                         cpu_up(cpu);
 395         }
 396
 397         /* Any cleanup work */
 398         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 399         smp_cpus_done(setup_max_cpus);
 400 }
 401
 402 #endif
 403
 404 /*
 405  * We need to store the untouched command line for future reference.
 406  * We also need to store the touched command line since the parameter
 407  * parsing is performed in place, and we should allow a component to
 408  * store reference of name/value for future reference.
 409  */
 410 static void __init setup_command_line(char *command_line)
 411 {
 412         saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
 413         static_command_line = alloc_bootmem(strlen (command_line)+1);
 414         strcpy (saved_command_line, boot_command_line);
 415         strcpy (static_command_line, command_line);
 416 }
 417
 418 /*
 419  * We need to finalize in a non-__init function or else race conditions
 420  * between the root thread and the init thread may cause start_kernel to
 421  * be reaped by free_initmem before the root thread has proceeded to
 422  * cpu_idle.
 423  *
 424  * gcc-3.4 accidentally inlines this function, so use noinline.
 425  */
 426
 427 static noinline void __init_refok rest_init(void)
 428         __releases(kernel_lock)
 429 {
 430         int pid;
 431
 432         rcu_scheduler_starting();
 433         kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 434         numa_default_policy();
 435         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 436         rcu_read_lock();
 437         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 438         rcu_read_unlock();
 439         unlock_kernel();
 440
 441         /*
 442          * The boot idle thread must execute schedule()
 443          * at least once to get things moving:
 444          */
 445         init_idle_bootup_task(current);
 446         preempt_enable_no_resched();
 447         schedule();
 448         preempt_disable();
 449
 450         /* Call into cpu_idle with preempt disabled */
 451         cpu_idle();
 452 }
 453
 454 /* Check for early params. */
 455 static int __init do_early_param(char *param, char *val)
 456 {
 457         struct obs_kernel_param *p;
 458
 459         for (p = __setup_start; p < __setup_end; p++) {
 460                 if ((p->early && strcmp(param, p->str) == 0) ||
 461                     (strcmp(param, "console") == 0 &&
 462                      strcmp(p->str, "earlycon") == 0)
 463                 ) {
 464                         if (p->setup_func(val) != 0)
 465                                 printk(KERN_WARNING
 466                                        "Malformed early option '%s'\n", param);
 467                 }
 468         }
 469         /* We accept everything at this stage. */
 470         return 0;
 471 }
 472
 473 void __init parse_early_options(char *cmdline)
 474 {
 475         parse_args("early options", cmdline, NULL, 0, do_early_param);
 476 }
 477
 478 /* Arch code calls this early on, or if not, just before other parsing. */
 479 void __init parse_early_param(void)
 480 {
 481         static __initdata int done = 0;
 482         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 483
 484         if (done)
 485                 return;
 486
 487         /* All fall through to do_early_param. */
 488         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 489         parse_early_options(tmp_cmdline);
 490         done = 1;
 491 }
 492
 493 /*
 494  *      Activate the first processor.
 495  */
 496
 497 static void __init boot_cpu_init(void)
 498 {
 499         int cpu = smp_processor_id();
 500         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 501         set_cpu_online(cpu, true);
 502         set_cpu_active(cpu, true);
 503         set_cpu_present(cpu, true);
 504         set_cpu_possible(cpu, true);
 505 }
 506
 507 void __init __weak smp_setup_processor_id(void)
 508 {
 509 }
 510
 511 void __init __weak thread_info_cache_init(void)
 512 {
 513 }
 514
 515 /*
 516  * Set up kernel memory allocators
 517  */
 518 static void __init mm_init(void)
 519 {
 520         /*
 521          * page_cgroup requires countinous pages as memmap
 522          * and it's bigger than MAX_ORDER unless SPARSEMEM.
 523          */
 524         page_cgroup_init_flatmem();
 525         mem_init();
 526         kmem_cache_init();
 527         pgtable_cache_init();
 528         vmalloc_init();
 529 }
 530
 531 asmlinkage void __init start_kernel(void)
 532 {
 533         char * command_line;
 534         extern struct kernel_param __start___param[], __stop___param[];
 535
 536         smp_setup_processor_id();
 537
 538         /*
 539          * Need to run as early as possible, to initialize the
 540          * lockdep hash:
 541          */
 542         lockdep_init();
 543         debug_objects_early_init();
 544
 545         /*
 546          * Set up the the initial canary ASAP:
 547          */
 548         boot_init_stack_canary();
 549
 550         cgroup_init_early();
 551
 552         local_irq_disable();
 553         early_boot_irqs_off();
 554         early_init_irq_lock_class();
 555
 556 /*
 557  * Interrupts are still disabled. Do necessary setups, then
 558  * enable them
 559  */
 560         lock_kernel();
 561         tick_init();
 562         boot_cpu_init();
 563         page_address_init();
 564         printk(KERN_NOTICE "%s", linux_banner);
 565         setup_arch(&command_line);
 566         mm_init_owner(&init_mm, &init_task);
 567         setup_command_line(command_line);
 568         setup_nr_cpu_ids();
 569         setup_per_cpu_areas();
 570         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 571
 572         build_all_zonelists(NULL);
 573         page_alloc_init();
 574
 575         printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
 576         parse_early_param();
 577         parse_args("Booting kernel", static_command_line, __start___param,
 578                    __stop___param - __start___param,
 579                    &unknown_bootoption);
 580         /*
 581          * These use large bootmem allocations and must precede
 582          * kmem_cache_init()
 583          */
 584         pidhash_init();
 585         vfs_caches_init_early();
 586         sort_main_extable();
 587         trap_init();
 588         mm_init();
 589         /*
 590          * Set up the scheduler prior starting any interrupts (such as the
 591          * timer interrupt). Full topology setup happens at smp_init()
 592          * time - but meanwhile we still have a functioning scheduler.
 593          */
 594         sched_init();
 595         /*
 596          * Disable preemption - early bootup scheduling is extremely
 597          * fragile until we cpu_idle() for the first time.
 598          */
 599         preempt_disable();
 600         if (!irqs_disabled()) {
 601                 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
 602                                 "enabled *very* early, fixing it\n");
 603                 local_irq_disable();
 604         }
 605         rcu_init();
 606         radix_tree_init();
 607         /* init some links before init_ISA_irqs() */
 608         early_irq_init();
 609         init_IRQ();
 610         prio_tree_init();
 611         init_timers();
 612         hrtimers_init();
 613         softirq_init();
 614         timekeeping_init();
 615         time_init();
 616         profile_init();
 617         if (!irqs_disabled())
 618                 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
 619                                  "enabled early\n");
 620         early_boot_irqs_on();
 621         local_irq_enable();
 622
 623         /* Interrupts are enabled now so all GFP allocations are safe. */
 624         gfp_allowed_mask = __GFP_BITS_MASK;
 625
 626         kmem_cache_init_late();
 627
 628         /*
 629          * HACK ALERT! This is early. We're enabling the console before
 630          * we've done PCI setups etc, and console_init() must be aware of
 631          * this. But we do want output early, in case something goes wrong.
 632          */
 633         console_init();
 634         if (panic_later)
 635                 panic(panic_later, panic_param);
 636
 637         lockdep_info();
 638
 639         /*
 640          * Need to run this when irqs are enabled, because it wants
 641          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 642          * too:
 643          */
 644         locking_selftest();
 645
 646 #ifdef CONFIG_BLK_DEV_INITRD
 647         if (initrd_start && !initrd_below_start_ok &&
 648             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 649                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 650                     "disabling it.\n",
 651                     page_to_pfn(virt_to_page((void *)initrd_start)),
 652                     min_low_pfn);
 653                 initrd_start = 0;
 654         }
 655 #endif
 656         page_cgroup_init();
 657         enable_debug_pagealloc();
 658         kmemtrace_init();
 659         kmemleak_init();
 660         debug_objects_mem_init();
 661         idr_init_cache();
 662         setup_per_cpu_pageset();
 663         numa_policy_init();
 664         if (late_time_init)
 665                 late_time_init();
 666         sched_clock_init();
 667         calibrate_delay();
 668         pidmap_init();
 669         anon_vma_init();
 670 #ifdef CONFIG_X86
 671         if (efi_enabled)
 672                 efi_enter_virtual_mode();
 673 #endif
 674         thread_info_cache_init();
 675         cred_init();
 676         fork_init(totalram_pages);
 677         proc_caches_init();
 678         buffer_init();
 679         key_init();
 680         security_init();
 681         dbg_late_init();
 682         vfs_caches_init(totalram_pages);
 683         signals_init();
 684         /* rootfs populating might need page-writeback */
 685         page_writeback_init();
 686 #ifdef CONFIG_PROC_FS
 687         proc_root_init();
 688 #endif
 689         cgroup_init();
 690         cpuset_init();
 691         taskstats_init_early();
 692         delayacct_init();
 693
 694         check_bugs();
 695
 696         acpi_early_init(); /* before LAPIC and SMP init */
 697         sfi_init_late();
 698
 699         ftrace_init();
 700
 701         /* Do the rest non-__init'ed, we're now alive */
 702         rest_init();
 703 }
 704
 705 /* Call all constructor functions linked into the kernel. */
 706 static void __init do_ctors(void)
 707 {
 708 #ifdef CONFIG_CONSTRUCTORS
 709         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 710
 711         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 712                 (*fn)();
 713 #endif
 714 }
 715
 716 int initcall_debug;
 717 core_param(initcall_debug, initcall_debug, bool, 0644);
 718
 719 static char msgbuf[64];
 720 static struct boot_trace_call call;
 721 static struct boot_trace_ret ret;
 722
 723 int do_one_initcall(initcall_t fn)
 724 {
 725         int count = preempt_count();
 726         ktime_t calltime, delta, rettime;
 727
 728         if (initcall_debug) {
 729                 call.caller = task_pid_nr(current);
 730                 printk("calling  %pF @ %i\n", fn, call.caller);
 731                 calltime = ktime_get();
 732                 trace_boot_call(&call, fn);
 733                 enable_boot_trace();
 734         }
 735
 736         ret.result = fn();
 737
 738         if (initcall_debug) {
 739                 disable_boot_trace();
 740                 rettime = ktime_get();
 741                 delta = ktime_sub(rettime, calltime);
 742                 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 743                 trace_boot_ret(&ret, fn);
 744                 printk("initcall %pF returned %d after %Ld usecs\n", fn,
 745                         ret.result, ret.duration);
 746         }
 747
 748         msgbuf[0] = 0;
 749
 750         if (ret.result && ret.result != -ENODEV && initcall_debug)
 751                 sprintf(msgbuf, "error code %d ", ret.result);
 752
 753         if (preempt_count() != count) {
 754                 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
 755                 preempt_count() = count;
 756         }
 757         if (irqs_disabled()) {
 758                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 759                 local_irq_enable();
 760         }
 761         if (msgbuf[0]) {
 762                 printk("initcall %pF returned with %s\n", fn, msgbuf);
 763         }
 764
 765         return ret.result;
 766 }
 767
 768
 769 extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
 770
 771 static void __init do_initcalls(void)
 772 {
 773         initcall_t *fn;
 774
 775         for (fn = __early_initcall_end; fn < __initcall_end; fn++)
 776                 do_one_initcall(*fn);
 777
 778         /* Make sure there is no pending stuff from the initcall sequence */
 779         flush_scheduled_work();
 780 }
 781
 782 /*
 783  * Ok, the machine is now initialized. None of the devices
 784  * have been touched yet, but the CPU subsystem is up and
 785  * running, and memory and process management works.
 786  *
 787  * Now we can finally start doing some real work..
 788  */
 789 static void __init do_basic_setup(void)
 790 {
 791         init_workqueues();
 792         cpuset_init_smp();
 793         usermodehelper_init();
 794         init_tmpfs();
 795         driver_init();
 796         init_irq_proc();
 797         do_ctors();
 798         do_initcalls();
 799 }
 800
 801 static void __init do_pre_smp_initcalls(void)
 802 {
 803         initcall_t *fn;
 804
 805         for (fn = __initcall_start; fn < __early_initcall_end; fn++)
 806                 do_one_initcall(*fn);
 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
 830         current->signal->flags |= SIGNAL_UNKILLABLE;
 831
 832         if (ramdisk_execute_command) {
 833                 run_init_process(ramdisk_execute_command);
 834                 printk(KERN_WARNING "Failed to execute %s\n",
 835                                 ramdisk_execute_command);
 836         }
 837
 838         /*
 839          * We try each of these until one succeeds.
 840          *
 841          * The Bourne shell can be used instead of init if we are
 842          * trying to recover a really broken machine.
 843          */
 844         if (execute_command) {
 845                 run_init_process(execute_command);
 846                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 847                                         "defaults...\n", execute_command);
 848         }
 849         run_init_process("/sbin/init");
 850         run_init_process("/etc/init");
 851         run_init_process("/bin/init");
 852         run_init_process("/bin/sh");
 853
 854         panic("No init found.  Try passing init= option to kernel. "
 855               "See Linux Documentation/init.txt for guidance.");
 856 }
 857
 858 static int __init kernel_init(void * unused)
 859 {
 860         lock_kernel();
 861
 862         /*
 863          * init can allocate pages on any node
 864          */
 865         set_mems_allowed(node_states[N_HIGH_MEMORY]);
 866         /*
 867          * init can run on any cpu.
 868          */
 869         set_cpus_allowed_ptr(current, cpu_all_mask);
 870         /*
 871          * Tell the world that we're going to be the grim
 872          * reaper of innocent orphaned children.
 873          *
 874          * We don't want people to have to make incorrect
 875          * assumptions about where in the task array this
 876          * can be found.
 877          */
 878         init_pid_ns.child_reaper = current;
 879
 880         cad_pid = task_pid(current);
 881
 882         smp_prepare_cpus(setup_max_cpus);
 883
 884         do_pre_smp_initcalls();
 885         start_boot_trace();
 886
 887         smp_init();
 888         sched_init_smp();
 889
 890         do_basic_setup();
 891
 892         /* Open the /dev/console on the rootfs, this should never fail */
 893         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 894                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 895
 896         (void) sys_dup(0);
 897         (void) sys_dup(0);
 898         /*
 899          * check if there is an early userspace init.  If yes, let it do all
 900          * the work
 901          */
 902
 903         if (!ramdisk_execute_command)
 904                 ramdisk_execute_command = "/init";
 905
 906         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 907                 ramdisk_execute_command = NULL;
 908                 prepare_namespace();
 909         }
 910
 911         /*
 912          * Ok, we have completed the initial bootup, and
 913          * we're essentially up and running. Get rid of the
 914          * initmem segments and start the user-mode stuff..
 915          */
 916
 917         init_post();
 918         return 0;
 919 }