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 #define DEBUG           /* Enable initcall_debug */
  13
  14 #include <linux/types.h>
  15 #include <linux/module.h>
  16 #include <linux/proc_fs.h>
  17 #include <linux/kernel.h>
  18 #include <linux/syscalls.h>
  19 #include <linux/stackprotector.h>
  20 #include <linux/string.h>
  21 #include <linux/ctype.h>
  22 #include <linux/delay.h>
  23 #include <linux/ioport.h>
  24 #include <linux/init.h>
  25 #include <linux/initrd.h>
  26 #include <linux/bootmem.h>
  27 #include <linux/acpi.h>
  28 #include <linux/tty.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/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/sfi.h>
  70 #include <linux/shmem_fs.h>
  71 #include <linux/slab.h>
  72 #include <linux/perf_event.h>
  73 #include <linux/file.h>
  74 #include <linux/ptrace.h>
  75 #include <linux/blkdev.h>
  76 #include <linux/elevator.h>
  77 #include <linux/sched_clock.h>
  78 #include <linux/context_tracking.h>
  79 #include <linux/random.h>
  80 #include <linux/list.h>
  81
  82 #include <asm/io.h>
  83 #include <asm/bugs.h>
  84 #include <asm/setup.h>
  85 #include <asm/sections.h>
  86 #include <asm/cacheflush.h>
  87
  88 #ifdef CONFIG_X86_LOCAL_APIC
  89 #include <asm/smp.h>
  90 #endif
  91
  92 static int kernel_init(void *);
  93
  94 extern void init_IRQ(void);
  95 extern void fork_init(unsigned long);
  96 extern void radix_tree_init(void);
  97 #ifndef CONFIG_DEBUG_RODATA
  98 static inline void mark_rodata_ro(void) { }
  99 #endif
 100
 101 /*
 102  * Debug helper: via this flag we know that we are in 'early bootup code'
 103  * where only the boot processor is running with IRQ disabled.  This means
 104  * two things - IRQ must not be enabled before the flag is cleared and some
 105  * operations which are not allowed with IRQ disabled are allowed while the
 106  * flag is set.
 107  */
 108 bool early_boot_irqs_disabled __read_mostly;
 109
 110 enum system_states system_state __read_mostly;
 111 EXPORT_SYMBOL(system_state);
 112
 113 /*
 114  * Boot command-line arguments
 115  */
 116 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 117 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 118
 119 extern void time_init(void);
 120 /* Default late time init is NULL. archs can override this later. */
 121 void (*__initdata late_time_init)(void);
 122
 123 /* Untouched command line saved by arch-specific code. */
 124 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 125 /* Untouched saved command line (eg. for /proc) */
 126 char *saved_command_line;
 127 /* Command line for parameter parsing */
 128 static char *static_command_line;
 129 /* Command line for per-initcall parameter parsing */
 130 static char *initcall_command_line;
 131
 132 static char *execute_command;
 133 static char *ramdisk_execute_command;
 134
 135 /*
 136  * Used to generate warnings if static_key manipulation functions are used
 137  * before jump_label_init is called.
 138  */
 139 bool static_key_initialized __read_mostly = false;
 140 EXPORT_SYMBOL_GPL(static_key_initialized);
 141
 142 /*
 143  * If set, this is an indication to the drivers that reset the underlying
 144  * device before going ahead with the initialization otherwise driver might
 145  * rely on the BIOS and skip the reset operation.
 146  *
 147  * This is useful if kernel is booting in an unreliable environment.
 148  * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 149  * skipped and devices will be in unknown state.
 150  */
 151 unsigned int reset_devices;
 152 EXPORT_SYMBOL(reset_devices);
 153
 154 static int __init set_reset_devices(char *str)
 155 {
 156         reset_devices = 1;
 157         return 1;
 158 }
 159
 160 __setup("reset_devices", set_reset_devices);
 161
 162 static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 163 const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 164 static const char *panic_later, *panic_param;
 165
 166 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 167
 168 static int __init obsolete_checksetup(char *line)
 169 {
 170         const struct obs_kernel_param *p;
 171         int had_early_param = 0;
 172
 173         p = __setup_start;
 174         do {
 175                 int n = strlen(p->str);
 176                 if (parameqn(line, p->str, n)) {
 177                         if (p->early) {
 178                                 /* Already done in parse_early_param?
 179                                  * (Needs exact match on param part).
 180                                  * Keep iterating, as we can have early
 181                                  * params and __setups of same names 8( */
 182                                 if (line[n] == '\0' || line[n] == '=')
 183                                         had_early_param = 1;
 184                         } else if (!p->setup_func) {
 185                                 pr_warn("Parameter %s is obsolete, ignored\n",
 186                                         p->str);
 187                                 return 1;
 188                         } else if (p->setup_func(line + n))
 189                                 return 1;
 190                 }
 191                 p++;
 192         } while (p < __setup_end);
 193
 194         return had_early_param;
 195 }
 196
 197 /*
 198  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 199  * still work even if initially too large, it will just take slightly longer
 200  */
 201 unsigned long loops_per_jiffy = (1<<12);
 202
 203 EXPORT_SYMBOL(loops_per_jiffy);
 204
 205 static int __init debug_kernel(char *str)
 206 {
 207         console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
 208         return 0;
 209 }
 210
 211 static int __init quiet_kernel(char *str)
 212 {
 213         console_loglevel = CONSOLE_LOGLEVEL_QUIET;
 214         return 0;
 215 }
 216
 217 early_param("debug", debug_kernel);
 218 early_param("quiet", quiet_kernel);
 219
 220 static int __init loglevel(char *str)
 221 {
 222         int newlevel;
 223
 224         /*
 225          * Only update loglevel value when a correct setting was passed,
 226          * to prevent blind crashes (when loglevel being set to 0) that
 227          * are quite hard to debug
 228          */
 229         if (get_option(&str, &newlevel)) {
 230                 console_loglevel = newlevel;
 231                 return 0;
 232         }
 233
 234         return -EINVAL;
 235 }
 236
 237 early_param("loglevel", loglevel);
 238
 239 /* Change NUL term back to "=", to make "param" the whole string. */
 240 static int __init repair_env_string(char *param, char *val, const char *unused)
 241 {
 242         if (val) {
 243                 /* param=val or param="val"? */
 244                 if (val == param+strlen(param)+1)
 245                         val[-1] = '=';
 246                 else if (val == param+strlen(param)+2) {
 247                         val[-2] = '=';
 248                         memmove(val-1, val, strlen(val)+1);
 249                         val--;
 250                 } else
 251                         BUG();
 252         }
 253         return 0;
 254 }
 255
 256 /* Anything after -- gets handed straight to init. */
 257 static int __init set_init_arg(char *param, char *val, const char *unused)
 258 {
 259         unsigned int i;
 260
 261         if (panic_later)
 262                 return 0;
 263
 264         repair_env_string(param, val, unused);
 265
 266         for (i = 0; argv_init[i]; i++) {
 267                 if (i == MAX_INIT_ARGS) {
 268                         panic_later = "init";
 269                         panic_param = param;
 270                         return 0;
 271                 }
 272         }
 273         argv_init[i] = param;
 274         return 0;
 275 }
 276
 277 /*
 278  * Unknown boot options get handed to init, unless they look like
 279  * unused parameters (modprobe will find them in /proc/cmdline).
 280  */
 281 static int __init unknown_bootoption(char *param, char *val, const char *unused)
 282 {
 283         repair_env_string(param, val, unused);
 284
 285         /* Handle obsolete-style parameters */
 286         if (obsolete_checksetup(param))
 287                 return 0;
 288
 289         /* Unused module parameter. */
 290         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 291                 return 0;
 292
 293         if (panic_later)
 294                 return 0;
 295
 296         if (val) {
 297                 /* Environment option */
 298                 unsigned int i;
 299                 for (i = 0; envp_init[i]; i++) {
 300                         if (i == MAX_INIT_ENVS) {
 301                                 panic_later = "env";
 302                                 panic_param = param;
 303                         }
 304                         if (!strncmp(param, envp_init[i], val - param))
 305                                 break;
 306                 }
 307                 envp_init[i] = param;
 308         } else {
 309                 /* Command line option */
 310                 unsigned int i;
 311                 for (i = 0; argv_init[i]; i++) {
 312                         if (i == MAX_INIT_ARGS) {
 313                                 panic_later = "init";
 314                                 panic_param = param;
 315                         }
 316                 }
 317                 argv_init[i] = param;
 318         }
 319         return 0;
 320 }
 321
 322 static int __init init_setup(char *str)
 323 {
 324         unsigned int i;
 325
 326         execute_command = str;
 327         /*
 328          * In case LILO is going to boot us with default command line,
 329          * it prepends "auto" before the whole cmdline which makes
 330          * the shell think it should execute a script with such name.
 331          * So we ignore all arguments entered _before_ init=... [MJ]
 332          */
 333         for (i = 1; i < MAX_INIT_ARGS; i++)
 334                 argv_init[i] = NULL;
 335         return 1;
 336 }
 337 __setup("init=", init_setup);
 338
 339 static int __init rdinit_setup(char *str)
 340 {
 341         unsigned int i;
 342
 343         ramdisk_execute_command = str;
 344         /* See "auto" comment in init_setup */
 345         for (i = 1; i < MAX_INIT_ARGS; i++)
 346                 argv_init[i] = NULL;
 347         return 1;
 348 }
 349 __setup("rdinit=", rdinit_setup);
 350
 351 #ifndef CONFIG_SMP
 352 static const unsigned int setup_max_cpus = NR_CPUS;
 353 #ifdef CONFIG_X86_LOCAL_APIC
 354 static void __init smp_init(void)
 355 {
 356         APIC_init_uniprocessor();
 357 }
 358 #else
 359 #define smp_init()      do { } while (0)
 360 #endif
 361
 362 static inline void setup_nr_cpu_ids(void) { }
 363 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 364 #endif
 365
 366 /*
 367  * We need to store the untouched command line for future reference.
 368  * We also need to store the touched command line since the parameter
 369  * parsing is performed in place, and we should allow a component to
 370  * store reference of name/value for future reference.
 371  */
 372 static void __init setup_command_line(char *command_line)
 373 {
 374         saved_command_line =
 375                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 376         initcall_command_line =
 377                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 378         static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
 379         strcpy (saved_command_line, boot_command_line);
 380         strcpy (static_command_line, command_line);
 381 }
 382
 383 /*
 384  * We need to finalize in a non-__init function or else race conditions
 385  * between the root thread and the init thread may cause start_kernel to
 386  * be reaped by free_initmem before the root thread has proceeded to
 387  * cpu_idle.
 388  *
 389  * gcc-3.4 accidentally inlines this function, so use noinline.
 390  */
 391
 392 static __initdata DECLARE_COMPLETION(kthreadd_done);
 393
 394 static noinline void __init_refok rest_init(void)
 395 {
 396         int pid;
 397
 398         rcu_scheduler_starting();
 399         smpboot_thread_init();
 400         /*
 401          * We need to spawn init first so that it obtains pid 1, however
 402          * the init task will end up wanting to create kthreads, which, if
 403          * we schedule it before we create kthreadd, will OOPS.
 404          */
 405         kernel_thread(kernel_init, NULL, CLONE_FS);
 406         numa_default_policy();
 407         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 408         rcu_read_lock();
 409         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 410         rcu_read_unlock();
 411         complete(&kthreadd_done);
 412
 413         /*
 414          * The boot idle thread must execute schedule()
 415          * at least once to get things moving:
 416          */
 417         init_idle_bootup_task(current);
 418         schedule_preempt_disabled();
 419         /* Call into cpu_idle with preempt disabled */
 420         cpu_startup_entry(CPUHP_ONLINE);
 421 }
 422
 423 /* Check for early params. */
 424 static int __init do_early_param(char *param, char *val, const char *unused)
 425 {
 426         const struct obs_kernel_param *p;
 427
 428         for (p = __setup_start; p < __setup_end; p++) {
 429                 if ((p->early && parameq(param, p->str)) ||
 430                     (strcmp(param, "console") == 0 &&
 431                      strcmp(p->str, "earlycon") == 0)
 432                 ) {
 433                         if (p->setup_func(val) != 0)
 434                                 pr_warn("Malformed early option '%s'\n", param);
 435                 }
 436         }
 437         /* We accept everything at this stage. */
 438         return 0;
 439 }
 440
 441 void __init parse_early_options(char *cmdline)
 442 {
 443         parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
 444 }
 445
 446 /* Arch code calls this early on, or if not, just before other parsing. */
 447 void __init parse_early_param(void)
 448 {
 449         static __initdata int done = 0;
 450         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 451
 452         if (done)
 453                 return;
 454
 455         /* All fall through to do_early_param. */
 456         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 457         parse_early_options(tmp_cmdline);
 458         done = 1;
 459 }
 460
 461 /*
 462  *      Activate the first processor.
 463  */
 464
 465 static void __init boot_cpu_init(void)
 466 {
 467         int cpu = smp_processor_id();
 468         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 469         set_cpu_online(cpu, true);
 470         set_cpu_active(cpu, true);
 471         set_cpu_present(cpu, true);
 472         set_cpu_possible(cpu, true);
 473 }
 474
 475 void __init __weak smp_setup_processor_id(void)
 476 {
 477 }
 478
 479 # if THREAD_SIZE >= PAGE_SIZE
 480 void __init __weak thread_info_cache_init(void)
 481 {
 482 }
 483 #endif
 484
 485 /*
 486  * Set up kernel memory allocators
 487  */
 488 static void __init mm_init(void)
 489 {
 490         /*
 491          * page_cgroup requires contiguous pages,
 492          * bigger than MAX_ORDER unless SPARSEMEM.
 493          */
 494         page_cgroup_init_flatmem();
 495         mem_init();
 496         kmem_cache_init();
 497         percpu_init_late();
 498         pgtable_init();
 499         vmalloc_init();
 500 }
 501
 502 asmlinkage __visible void __init start_kernel(void)
 503 {
 504         char * command_line, *after_dashes;
 505         extern const struct kernel_param __start___param[], __stop___param[];
 506
 507         /*
 508          * Need to run as early as possible, to initialize the
 509          * lockdep hash:
 510          */
 511         lockdep_init();
 512         smp_setup_processor_id();
 513         debug_objects_early_init();
 514
 515         /*
 516          * Set up the the initial canary ASAP:
 517          */
 518         boot_init_stack_canary();
 519
 520         cgroup_init_early();
 521
 522         local_irq_disable();
 523         early_boot_irqs_disabled = true;
 524
 525 /*
 526  * Interrupts are still disabled. Do necessary setups, then
 527  * enable them
 528  */
 529         boot_cpu_init();
 530         page_address_init();
 531         pr_notice("%s", linux_banner);
 532         setup_arch(&command_line);
 533         mm_init_cpumask(&init_mm);
 534         setup_command_line(command_line);
 535         setup_nr_cpu_ids();
 536         setup_per_cpu_areas();
 537         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 538
 539         build_all_zonelists(NULL, NULL);
 540         page_alloc_init();
 541
 542         pr_notice("Kernel command line: %s\n", boot_command_line);
 543         parse_early_param();
 544         after_dashes = parse_args("Booting kernel",
 545                                   static_command_line, __start___param,
 546                                   __stop___param - __start___param,
 547                                   -1, -1, &unknown_bootoption);
 548         if (!IS_ERR_OR_NULL(after_dashes))
 549                 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
 550                            set_init_arg);
 551
 552         jump_label_init();
 553
 554         /*
 555          * These use large bootmem allocations and must precede
 556          * kmem_cache_init()
 557          */
 558         setup_log_buf(0);
 559         pidhash_init();
 560         vfs_caches_init_early();
 561         sort_main_extable();
 562         trap_init();
 563         mm_init();
 564
 565         /*
 566          * Set up the scheduler prior starting any interrupts (such as the
 567          * timer interrupt). Full topology setup happens at smp_init()
 568          * time - but meanwhile we still have a functioning scheduler.
 569          */
 570         sched_init();
 571         /*
 572          * Disable preemption - early bootup scheduling is extremely
 573          * fragile until we cpu_idle() for the first time.
 574          */
 575         preempt_disable();
 576         if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
 577                 local_irq_disable();
 578         idr_init_cache();
 579         rcu_init();
 580         tick_nohz_init();
 581         context_tracking_init();
 582         radix_tree_init();
 583         /* init some links before init_ISA_irqs() */
 584         early_irq_init();
 585         init_IRQ();
 586         tick_init();
 587         init_timers();
 588         hrtimers_init();
 589         softirq_init();
 590         timekeeping_init();
 591         time_init();
 592         sched_clock_postinit();
 593         perf_event_init();
 594         profile_init();
 595         call_function_init();
 596         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 597         early_boot_irqs_disabled = false;
 598         local_irq_enable();
 599
 600         kmem_cache_init_late();
 601
 602         /*
 603          * HACK ALERT! This is early. We're enabling the console before
 604          * we've done PCI setups etc, and console_init() must be aware of
 605          * this. But we do want output early, in case something goes wrong.
 606          */
 607         console_init();
 608         if (panic_later)
 609                 panic("Too many boot %s vars at `%s'", panic_later,
 610                       panic_param);
 611
 612         lockdep_info();
 613
 614         /*
 615          * Need to run this when irqs are enabled, because it wants
 616          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 617          * too:
 618          */
 619         locking_selftest();
 620
 621 #ifdef CONFIG_BLK_DEV_INITRD
 622         if (initrd_start && !initrd_below_start_ok &&
 623             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 624                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
 625                     page_to_pfn(virt_to_page((void *)initrd_start)),
 626                     min_low_pfn);
 627                 initrd_start = 0;
 628         }
 629 #endif
 630         page_cgroup_init();
 631         debug_objects_mem_init();
 632         kmemleak_init();
 633         setup_per_cpu_pageset();
 634         numa_policy_init();
 635         if (late_time_init)
 636                 late_time_init();
 637         sched_clock_init();
 638         calibrate_delay();
 639         pidmap_init();
 640         anon_vma_init();
 641         acpi_early_init();
 642 #ifdef CONFIG_X86
 643         if (efi_enabled(EFI_RUNTIME_SERVICES))
 644                 efi_enter_virtual_mode();
 645 #endif
 646 #ifdef CONFIG_X86_ESPFIX64
 647         /* Should be run before the first non-init thread is created */
 648         init_espfix_bsp();
 649 #endif
 650         thread_info_cache_init();
 651         cred_init();
 652         fork_init(totalram_pages);
 653         proc_caches_init();
 654         buffer_init();
 655         key_init();
 656         security_init();
 657         dbg_late_init();
 658         vfs_caches_init(totalram_pages);
 659         signals_init();
 660         /* rootfs populating might need page-writeback */
 661         page_writeback_init();
 662         proc_root_init();
 663         cgroup_init();
 664         cpuset_init();
 665         taskstats_init_early();
 666         delayacct_init();
 667
 668         check_bugs();
 669
 670         acpi_subsystem_init();
 671         sfi_init_late();
 672
 673         if (efi_enabled(EFI_RUNTIME_SERVICES)) {
 674                 efi_late_init();
 675                 efi_free_boot_services();
 676         }
 677
 678         ftrace_init();
 679
 680         /* Do the rest non-__init'ed, we're now alive */
 681         rest_init();
 682 }
 683
 684 /* Call all constructor functions linked into the kernel. */
 685 static void __init do_ctors(void)
 686 {
 687 #ifdef CONFIG_CONSTRUCTORS
 688         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 689
 690         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 691                 (*fn)();
 692 #endif
 693 }
 694
 695 bool initcall_debug;
 696 core_param(initcall_debug, initcall_debug, bool, 0644);
 697
 698 #ifdef CONFIG_KALLSYMS
 699 struct blacklist_entry {
 700         struct list_head next;
 701         char *buf;
 702 };
 703
 704 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
 705
 706 static int __init initcall_blacklist(char *str)
 707 {
 708         char *str_entry;
 709         struct blacklist_entry *entry;
 710
 711         /* str argument is a comma-separated list of functions */
 712         do {
 713                 str_entry = strsep(&str, ",");
 714                 if (str_entry) {
 715                         pr_debug("blacklisting initcall %s\n", str_entry);
 716                         entry = alloc_bootmem(sizeof(*entry));
 717                         entry->buf = alloc_bootmem(strlen(str_entry) + 1);
 718                         strcpy(entry->buf, str_entry);
 719                         list_add(&entry->next, &blacklisted_initcalls);
 720                 }
 721         } while (str_entry);
 722
 723         return 0;
 724 }
 725
 726 static bool __init_or_module initcall_blacklisted(initcall_t fn)
 727 {
 728         struct list_head *tmp;
 729         struct blacklist_entry *entry;
 730         char *fn_name;
 731
 732         fn_name = kasprintf(GFP_KERNEL, "%pf", fn);
 733         if (!fn_name)
 734                 return false;
 735
 736         list_for_each(tmp, &blacklisted_initcalls) {
 737                 entry = list_entry(tmp, struct blacklist_entry, next);
 738                 if (!strcmp(fn_name, entry->buf)) {
 739                         pr_debug("initcall %s blacklisted\n", fn_name);
 740                         kfree(fn_name);
 741                         return true;
 742                 }
 743         }
 744
 745         kfree(fn_name);
 746         return false;
 747 }
 748 #else
 749 static int __init initcall_blacklist(char *str)
 750 {
 751         pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
 752         return 0;
 753 }
 754
 755 static bool __init_or_module initcall_blacklisted(initcall_t fn)
 756 {
 757         return false;
 758 }
 759 #endif
 760 __setup("initcall_blacklist=", initcall_blacklist);
 761
 762 static int __init_or_module do_one_initcall_debug(initcall_t fn)
 763 {
 764         ktime_t calltime, delta, rettime;
 765         unsigned long long duration;
 766         int ret;
 767
 768         printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
 769         calltime = ktime_get();
 770         ret = fn();
 771         rettime = ktime_get();
 772         delta = ktime_sub(rettime, calltime);
 773         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 774         printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
 775                  fn, ret, duration);
 776
 777         return ret;
 778 }
 779
 780 int __init_or_module do_one_initcall(initcall_t fn)
 781 {
 782         int count = preempt_count();
 783         int ret;
 784         char msgbuf[64];
 785
 786         if (initcall_blacklisted(fn))
 787                 return -EPERM;
 788
 789         if (initcall_debug)
 790                 ret = do_one_initcall_debug(fn);
 791         else
 792                 ret = fn();
 793
 794         msgbuf[0] = 0;
 795
 796         if (preempt_count() != count) {
 797                 sprintf(msgbuf, "preemption imbalance ");
 798                 preempt_count_set(count);
 799         }
 800         if (irqs_disabled()) {
 801                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 802                 local_irq_enable();
 803         }
 804         WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
 805
 806         return ret;
 807 }
 808
 809
 810 extern initcall_t __initcall_start[];
 811 extern initcall_t __initcall0_start[];
 812 extern initcall_t __initcall1_start[];
 813 extern initcall_t __initcall2_start[];
 814 extern initcall_t __initcall3_start[];
 815 extern initcall_t __initcall4_start[];
 816 extern initcall_t __initcall5_start[];
 817 extern initcall_t __initcall6_start[];
 818 extern initcall_t __initcall7_start[];
 819 extern initcall_t __initcall_end[];
 820
 821 static initcall_t *initcall_levels[] __initdata = {
 822         __initcall0_start,
 823         __initcall1_start,
 824         __initcall2_start,
 825         __initcall3_start,
 826         __initcall4_start,
 827         __initcall5_start,
 828         __initcall6_start,
 829         __initcall7_start,
 830         __initcall_end,
 831 };
 832
 833 /* Keep these in sync with initcalls in include/linux/init.h */
 834 static char *initcall_level_names[] __initdata = {
 835         "early",
 836         "core",
 837         "postcore",
 838         "arch",
 839         "subsys",
 840         "fs",
 841         "device",
 842         "late",
 843 };
 844
 845 static void __init do_initcall_level(int level)
 846 {
 847         extern const struct kernel_param __start___param[], __stop___param[];
 848         initcall_t *fn;
 849
 850         strcpy(initcall_command_line, saved_command_line);
 851         parse_args(initcall_level_names[level],
 852                    initcall_command_line, __start___param,
 853                    __stop___param - __start___param,
 854                    level, level,
 855                    &repair_env_string);
 856
 857         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
 858                 do_one_initcall(*fn);
 859 }
 860
 861 static void __init do_initcalls(void)
 862 {
 863         int level;
 864
 865         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
 866                 do_initcall_level(level);
 867 }
 868
 869 /*
 870  * Ok, the machine is now initialized. None of the devices
 871  * have been touched yet, but the CPU subsystem is up and
 872  * running, and memory and process management works.
 873  *
 874  * Now we can finally start doing some real work..
 875  */
 876 static void __init do_basic_setup(void)
 877 {
 878         cpuset_init_smp();
 879         usermodehelper_init();
 880         shmem_init();
 881         driver_init();
 882         init_irq_proc();
 883         do_ctors();
 884         usermodehelper_enable();
 885         do_initcalls();
 886         random_int_secret_init();
 887 }
 888
 889 static void __init do_pre_smp_initcalls(void)
 890 {
 891         initcall_t *fn;
 892
 893         for (fn = __initcall_start; fn < __initcall0_start; fn++)
 894                 do_one_initcall(*fn);
 895 }
 896
 897 /*
 898  * This function requests modules which should be loaded by default and is
 899  * called twice right after initrd is mounted and right before init is
 900  * exec'd.  If such modules are on either initrd or rootfs, they will be
 901  * loaded before control is passed to userland.
 902  */
 903 void __init load_default_modules(void)
 904 {
 905         load_default_elevator_module();
 906 }
 907
 908 static int run_init_process(const char *init_filename)
 909 {
 910         argv_init[0] = init_filename;
 911         return do_execve(getname_kernel(init_filename),
 912                 (const char __user *const __user *)argv_init,
 913                 (const char __user *const __user *)envp_init);
 914 }
 915
 916 static int try_to_run_init_process(const char *init_filename)
 917 {
 918         int ret;
 919
 920         ret = run_init_process(init_filename);
 921
 922         if (ret && ret != -ENOENT) {
 923                 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
 924                        init_filename, ret);
 925         }
 926
 927         return ret;
 928 }
 929
 930 static noinline void __init kernel_init_freeable(void);
 931
 932 static int __ref kernel_init(void *unused)
 933 {
 934         int ret;
 935
 936         kernel_init_freeable();
 937         /* need to finish all async __init code before freeing the memory */
 938         async_synchronize_full();
 939         free_initmem();
 940         mark_rodata_ro();
 941         system_state = SYSTEM_RUNNING;
 942         numa_default_policy();
 943
 944         flush_delayed_fput();
 945
 946         if (ramdisk_execute_command) {
 947                 ret = run_init_process(ramdisk_execute_command);
 948                 if (!ret)
 949                         return 0;
 950                 pr_err("Failed to execute %s (error %d)\n",
 951                        ramdisk_execute_command, ret);
 952         }
 953
 954         /*
 955          * We try each of these until one succeeds.
 956          *
 957          * The Bourne shell can be used instead of init if we are
 958          * trying to recover a really broken machine.
 959          */
 960         if (execute_command) {
 961                 ret = run_init_process(execute_command);
 962                 if (!ret)
 963                         return 0;
 964                 pr_err("Failed to execute %s (error %d).  Attempting defaults...\n",
 965                         execute_command, ret);
 966         }
 967         if (!try_to_run_init_process("/sbin/init") ||
 968             !try_to_run_init_process("/etc/init") ||
 969             !try_to_run_init_process("/bin/init") ||
 970             !try_to_run_init_process("/bin/sh"))
 971                 return 0;
 972
 973         panic("No working init found.  Try passing init= option to kernel. "
 974               "See Linux Documentation/init.txt for guidance.");
 975 }
 976
 977 static noinline void __init kernel_init_freeable(void)
 978 {
 979         /*
 980          * Wait until kthreadd is all set-up.
 981          */
 982         wait_for_completion(&kthreadd_done);
 983
 984         /* Now the scheduler is fully set up and can do blocking allocations */
 985         gfp_allowed_mask = __GFP_BITS_MASK;
 986
 987         /*
 988          * init can allocate pages on any node
 989          */
 990         set_mems_allowed(node_states[N_MEMORY]);
 991         /*
 992          * init can run on any cpu.
 993          */
 994         set_cpus_allowed_ptr(current, cpu_all_mask);
 995
 996         cad_pid = task_pid(current);
 997
 998         smp_prepare_cpus(setup_max_cpus);
 999
1000         do_pre_smp_initcalls();
1001         lockup_detector_init();
1002
1003         smp_init();
1004         sched_init_smp();
1005
1006         do_basic_setup();
1007
1008         /* Open the /dev/console on the rootfs, this should never fail */
1009         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1010                 pr_err("Warning: unable to open an initial console.\n");
1011
1012         (void) sys_dup(0);
1013         (void) sys_dup(0);
1014         /*
1015          * check if there is an early userspace init.  If yes, let it do all
1016          * the work
1017          */
1018
1019         if (!ramdisk_execute_command)
1020                 ramdisk_execute_command = "/init";
1021
1022         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1023                 ramdisk_execute_command = NULL;
1024                 prepare_namespace();
1025         }
1026
1027         /*
1028          * Ok, we have completed the initial bootup, and
1029          * we're essentially up and running. Get rid of the
1030          * initmem segments and start the user-mode stuff..
1031          */
1032
1033         /* rootfs is available now, try loading default modules */
1034         load_default_modules();
1035 }