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_ext.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 #include <linux/integrity.h>
  82 #include <linux/proc_ns.h>
  83 #include <linux/io.h>
  84
  85 #include <asm/io.h>
  86 #include <asm/bugs.h>
  87 #include <asm/setup.h>
  88 #include <asm/sections.h>
  89 #include <asm/cacheflush.h>
  90
  91 static int kernel_init(void *);
  92
  93 extern void init_IRQ(void);
  94 extern void fork_init(void);
  95 extern void radix_tree_init(void);
  96
  97 /*
  98  * Debug helper: via this flag we know that we are in 'early bootup code'
  99  * where only the boot processor is running with IRQ disabled.  This means
 100  * two things - IRQ must not be enabled before the flag is cleared and some
 101  * operations which are not allowed with IRQ disabled are allowed while the
 102  * flag is set.
 103  */
 104 bool early_boot_irqs_disabled __read_mostly;
 105
 106 enum system_states system_state __read_mostly;
 107 EXPORT_SYMBOL(system_state);
 108
 109 /*
 110  * Boot command-line arguments
 111  */
 112 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 113 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 114
 115 extern void time_init(void);
 116 /* Default late time init is NULL. archs can override this later. */
 117 void (*__initdata late_time_init)(void);
 118
 119 /* Untouched command line saved by arch-specific code. */
 120 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 121 /* Untouched saved command line (eg. for /proc) */
 122 char *saved_command_line;
 123 /* Command line for parameter parsing */
 124 static char *static_command_line;
 125 /* Command line for per-initcall parameter parsing */
 126 static char *initcall_command_line;
 127
 128 static char *execute_command;
 129 static char *ramdisk_execute_command;
 130
 131 /*
 132  * Used to generate warnings if static_key manipulation functions are used
 133  * before jump_label_init is called.
 134  */
 135 bool static_key_initialized __read_mostly;
 136 EXPORT_SYMBOL_GPL(static_key_initialized);
 137
 138 /*
 139  * If set, this is an indication to the drivers that reset the underlying
 140  * device before going ahead with the initialization otherwise driver might
 141  * rely on the BIOS and skip the reset operation.
 142  *
 143  * This is useful if kernel is booting in an unreliable environment.
 144  * For ex. kdump situation where previous kernel has crashed, BIOS has been
 145  * skipped and devices will be in unknown state.
 146  */
 147 unsigned int reset_devices;
 148 EXPORT_SYMBOL(reset_devices);
 149
 150 static int __init set_reset_devices(char *str)
 151 {
 152         reset_devices = 1;
 153         return 1;
 154 }
 155
 156 __setup("reset_devices", set_reset_devices);
 157
 158 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 159 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 160 static const char *panic_later, *panic_param;
 161
 162 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 163
 164 static bool __init obsolete_checksetup(char *line)
 165 {
 166         const struct obs_kernel_param *p;
 167         bool had_early_param = false;
 168
 169         p = __setup_start;
 170         do {
 171                 int n = strlen(p->str);
 172                 if (parameqn(line, p->str, n)) {
 173                         if (p->early) {
 174                                 /* Already done in parse_early_param?
 175                                  * (Needs exact match on param part).
 176                                  * Keep iterating, as we can have early
 177                                  * params and __setups of same names 8( */
 178                                 if (line[n] == '\0' || line[n] == '=')
 179                                         had_early_param = true;
 180                         } else if (!p->setup_func) {
 181                                 pr_warn("Parameter %s is obsolete, ignored\n",
 182                                         p->str);
 183                                 return true;
 184                         } else if (p->setup_func(line + n))
 185                                 return true;
 186                 }
 187                 p++;
 188         } while (p < __setup_end);
 189
 190         return had_early_param;
 191 }
 192
 193 /*
 194  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 195  * still work even if initially too large, it will just take slightly longer
 196  */
 197 unsigned long loops_per_jiffy = (1<<12);
 198 EXPORT_SYMBOL(loops_per_jiffy);
 199
 200 static int __init debug_kernel(char *str)
 201 {
 202         console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
 203         return 0;
 204 }
 205
 206 static int __init quiet_kernel(char *str)
 207 {
 208         console_loglevel = CONSOLE_LOGLEVEL_QUIET;
 209         return 0;
 210 }
 211
 212 early_param("debug", debug_kernel);
 213 early_param("quiet", quiet_kernel);
 214
 215 static int __init loglevel(char *str)
 216 {
 217         int newlevel;
 218
 219         /*
 220          * Only update loglevel value when a correct setting was passed,
 221          * to prevent blind crashes (when loglevel being set to 0) that
 222          * are quite hard to debug
 223          */
 224         if (get_option(&str, &newlevel)) {
 225                 console_loglevel = newlevel;
 226                 return 0;
 227         }
 228
 229         return -EINVAL;
 230 }
 231
 232 early_param("loglevel", loglevel);
 233
 234 /* Change NUL term back to "=", to make "param" the whole string. */
 235 static int __init repair_env_string(char *param, char *val,
 236                                     const char *unused, void *arg)
 237 {
 238         if (val) {
 239                 /* param=val or param="val"? */
 240                 if (val == param+strlen(param)+1)
 241                         val[-1] = '=';
 242                 else if (val == param+strlen(param)+2) {
 243                         val[-2] = '=';
 244                         memmove(val-1, val, strlen(val)+1);
 245                         val--;
 246                 } else
 247                         BUG();
 248         }
 249         return 0;
 250 }
 251
 252 /* Anything after -- gets handed straight to init. */
 253 static int __init set_init_arg(char *param, char *val,
 254                                const char *unused, void *arg)
 255 {
 256         unsigned int i;
 257
 258         if (panic_later)
 259                 return 0;
 260
 261         repair_env_string(param, val, unused, NULL);
 262
 263         for (i = 0; argv_init[i]; i++) {
 264                 if (i == MAX_INIT_ARGS) {
 265                         panic_later = "init";
 266                         panic_param = param;
 267                         return 0;
 268                 }
 269         }
 270         argv_init[i] = param;
 271         return 0;
 272 }
 273
 274 /*
 275  * Unknown boot options get handed to init, unless they look like
 276  * unused parameters (modprobe will find them in /proc/cmdline).
 277  */
 278 static int __init unknown_bootoption(char *param, char *val,
 279                                      const char *unused, void *arg)
 280 {
 281         repair_env_string(param, val, unused, NULL);
 282
 283         /* Handle obsolete-style parameters */
 284         if (obsolete_checksetup(param))
 285                 return 0;
 286
 287         /* Unused module parameter. */
 288         if (strchr(param, '.') && (!val || strchr(param, '.') < val))
 289                 return 0;
 290
 291         if (panic_later)
 292                 return 0;
 293
 294         if (val) {
 295                 /* Environment option */
 296                 unsigned int i;
 297                 for (i = 0; envp_init[i]; i++) {
 298                         if (i == MAX_INIT_ENVS) {
 299                                 panic_later = "env";
 300                                 panic_param = param;
 301                         }
 302                         if (!strncmp(param, envp_init[i], val - param))
 303                                 break;
 304                 }
 305                 envp_init[i] = param;
 306         } else {
 307                 /* Command line option */
 308                 unsigned int i;
 309                 for (i = 0; argv_init[i]; i++) {
 310                         if (i == MAX_INIT_ARGS) {
 311                                 panic_later = "init";
 312                                 panic_param = param;
 313                         }
 314                 }
 315                 argv_init[i] = param;
 316         }
 317         return 0;
 318 }
 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 static const unsigned int setup_max_cpus = NR_CPUS;
 351 static inline void setup_nr_cpu_ids(void) { }
 352 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 353 #endif
 354
 355 /*
 356  * We need to store the untouched command line for future reference.
 357  * We also need to store the touched command line since the parameter
 358  * parsing is performed in place, and we should allow a component to
 359  * store reference of name/value for future reference.
 360  */
 361 static void __init setup_command_line(char *command_line)
 362 {
 363         saved_command_line =
 364                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 365         initcall_command_line =
 366                 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
 367         static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
 368         strcpy(saved_command_line, boot_command_line);
 369         strcpy(static_command_line, command_line);
 370 }
 371
 372 /*
 373  * We need to finalize in a non-__init function or else race conditions
 374  * between the root thread and the init thread may cause start_kernel to
 375  * be reaped by free_initmem before the root thread has proceeded to
 376  * cpu_idle.
 377  *
 378  * gcc-3.4 accidentally inlines this function, so use noinline.
 379  */
 380
 381 static __initdata DECLARE_COMPLETION(kthreadd_done);
 382
 383 static noinline void __init_refok rest_init(void)
 384 {
 385         int pid;
 386
 387         rcu_scheduler_starting();
 388         smpboot_thread_init();
 389         /*
 390          * We need to spawn init first so that it obtains pid 1, however
 391          * the init task will end up wanting to create kthreads, which, if
 392          * we schedule it before we create kthreadd, will OOPS.
 393          */
 394         kernel_thread(kernel_init, NULL, CLONE_FS);
 395         numa_default_policy();
 396         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 397         rcu_read_lock();
 398         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 399         rcu_read_unlock();
 400         complete(&kthreadd_done);
 401
 402         /*
 403          * The boot idle thread must execute schedule()
 404          * at least once to get things moving:
 405          */
 406         init_idle_bootup_task(current);
 407         schedule_preempt_disabled();
 408         /* Call into cpu_idle with preempt disabled */
 409         cpu_startup_entry(CPUHP_ONLINE);
 410 }
 411
 412 /* Check for early params. */
 413 static int __init do_early_param(char *param, char *val,
 414                                  const char *unused, void *arg)
 415 {
 416         const struct obs_kernel_param *p;
 417
 418         for (p = __setup_start; p < __setup_end; p++) {
 419                 if ((p->early && parameq(param, p->str)) ||
 420                     (strcmp(param, "console") == 0 &&
 421                      strcmp(p->str, "earlycon") == 0)
 422                 ) {
 423                         if (p->setup_func(val) != 0)
 424                                 pr_warn("Malformed early option '%s'\n", param);
 425                 }
 426         }
 427         /* We accept everything at this stage. */
 428         return 0;
 429 }
 430
 431 void __init parse_early_options(char *cmdline)
 432 {
 433         parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
 434                    do_early_param);
 435 }
 436
 437 /* Arch code calls this early on, or if not, just before other parsing. */
 438 void __init parse_early_param(void)
 439 {
 440         static int done __initdata;
 441         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 442
 443         if (done)
 444                 return;
 445
 446         /* All fall through to do_early_param. */
 447         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 448         parse_early_options(tmp_cmdline);
 449         done = 1;
 450 }
 451
 452 /*
 453  *      Activate the first processor.
 454  */
 455
 456 static void __init boot_cpu_init(void)
 457 {
 458         int cpu = smp_processor_id();
 459         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 460         set_cpu_online(cpu, true);
 461         set_cpu_active(cpu, true);
 462         set_cpu_present(cpu, true);
 463         set_cpu_possible(cpu, true);
 464 }
 465
 466 void __init __weak smp_setup_processor_id(void)
 467 {
 468 }
 469
 470 # if THREAD_SIZE >= PAGE_SIZE
 471 void __init __weak thread_info_cache_init(void)
 472 {
 473 }
 474 #endif
 475
 476 /*
 477  * Set up kernel memory allocators
 478  */
 479 static void __init mm_init(void)
 480 {
 481         /*
 482          * page_ext requires contiguous pages,
 483          * bigger than MAX_ORDER unless SPARSEMEM.
 484          */
 485         page_ext_init_flatmem();
 486         mem_init();
 487         kmem_cache_init();
 488         percpu_init_late();
 489         pgtable_init();
 490         vmalloc_init();
 491         ioremap_huge_init();
 492 }
 493
 494 asmlinkage __visible void __init start_kernel(void)
 495 {
 496         char *command_line;
 497         char *after_dashes;
 498
 499         set_task_stack_end_magic(&init_task);
 500         smp_setup_processor_id();
 501         debug_objects_early_init();
 502
 503         /*
 504          * Set up the the initial canary ASAP:
 505          */
 506         boot_init_stack_canary();
 507
 508         cgroup_init_early();
 509
 510         local_irq_disable();
 511         early_boot_irqs_disabled = true;
 512
 513 /*
 514  * Interrupts are still disabled. Do necessary setups, then
 515  * enable them
 516  */
 517         boot_cpu_init();
 518         page_address_init();
 519         pr_notice("%s", linux_banner);
 520         setup_arch(&command_line);
 521         mm_init_cpumask(&init_mm);
 522         setup_command_line(command_line);
 523         setup_nr_cpu_ids();
 524         setup_per_cpu_areas();
 525         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 526
 527         build_all_zonelists(NULL, NULL);
 528         page_alloc_init();
 529
 530         pr_notice("Kernel command line: %s\n", boot_command_line);
 531         parse_early_param();
 532         after_dashes = parse_args("Booting kernel",
 533                                   static_command_line, __start___param,
 534                                   __stop___param - __start___param,
 535                                   -1, -1, NULL, &unknown_bootoption);
 536         if (!IS_ERR_OR_NULL(after_dashes))
 537                 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
 538                            NULL, set_init_arg);
 539
 540         jump_label_init();
 541
 542         /*
 543          * These use large bootmem allocations and must precede
 544          * kmem_cache_init()
 545          */
 546         setup_log_buf(0);
 547         pidhash_init();
 548         vfs_caches_init_early();
 549         sort_main_extable();
 550         trap_init();
 551         mm_init();
 552
 553         /*
 554          * Set up the scheduler prior starting any interrupts (such as the
 555          * timer interrupt). Full topology setup happens at smp_init()
 556          * time - but meanwhile we still have a functioning scheduler.
 557          */
 558         sched_init();
 559         /*
 560          * Disable preemption - early bootup scheduling is extremely
 561          * fragile until we cpu_idle() for the first time.
 562          */
 563         preempt_disable();
 564         if (WARN(!irqs_disabled(),
 565                  "Interrupts were enabled *very* early, fixing it\n"))
 566                 local_irq_disable();
 567         idr_init_cache();
 568         rcu_init();
 569
 570         /* trace_printk() and trace points may be used after this */
 571         trace_init();
 572
 573         context_tracking_init();
 574         radix_tree_init();
 575         /* init some links before init_ISA_irqs() */
 576         early_irq_init();
 577         init_IRQ();
 578         tick_init();
 579         rcu_init_nohz();
 580         init_timers();
 581         hrtimers_init();
 582         softirq_init();
 583         timekeeping_init();
 584         time_init();
 585         sched_clock_postinit();
 586         perf_event_init();
 587         profile_init();
 588         call_function_init();
 589         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 590         early_boot_irqs_disabled = false;
 591         local_irq_enable();
 592
 593         kmem_cache_init_late();
 594
 595         /*
 596          * HACK ALERT! This is early. We're enabling the console before
 597          * we've done PCI setups etc, and console_init() must be aware of
 598          * this. But we do want output early, in case something goes wrong.
 599          */
 600         console_init();
 601         if (panic_later)
 602                 panic("Too many boot %s vars at `%s'", panic_later,
 603                       panic_param);
 604
 605         lockdep_info();
 606
 607         /*
 608          * Need to run this when irqs are enabled, because it wants
 609          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 610          * too:
 611          */
 612         locking_selftest();
 613
 614 #ifdef CONFIG_BLK_DEV_INITRD
 615         if (initrd_start && !initrd_below_start_ok &&
 616             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 617                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
 618                     page_to_pfn(virt_to_page((void *)initrd_start)),
 619                     min_low_pfn);
 620                 initrd_start = 0;
 621         }
 622 #endif
 623         page_ext_init();
 624         debug_objects_mem_init();
 625         kmemleak_init();
 626         setup_per_cpu_pageset();
 627         numa_policy_init();
 628         if (late_time_init)
 629                 late_time_init();
 630         sched_clock_init();
 631         calibrate_delay();
 632         pidmap_init();
 633         anon_vma_init();
 634         acpi_early_init();
 635 #ifdef CONFIG_X86
 636         if (efi_enabled(EFI_RUNTIME_SERVICES))
 637                 efi_enter_virtual_mode();
 638 #endif
 639 #ifdef CONFIG_X86_ESPFIX64
 640         /* Should be run before the first non-init thread is created */
 641         init_espfix_bsp();
 642 #endif
 643         thread_info_cache_init();
 644         cred_init();
 645         fork_init();
 646         proc_caches_init();
 647         buffer_init();
 648         key_init();
 649         security_init();
 650         dbg_late_init();
 651         vfs_caches_init();
 652         signals_init();
 653         /* rootfs populating might need page-writeback */
 654         page_writeback_init();
 655         proc_root_init();
 656         nsfs_init();
 657         cpuset_init();
 658         cgroup_init();
 659         taskstats_init_early();
 660         delayacct_init();
 661
 662         check_bugs();
 663
 664         acpi_subsystem_init();
 665         sfi_init_late();
 666
 667         if (efi_enabled(EFI_RUNTIME_SERVICES)) {
 668                 efi_late_init();
 669                 efi_free_boot_services();
 670         }
 671
 672         ftrace_init();
 673
 674         /* Do the rest non-__init'ed, we're now alive */
 675         rest_init();
 676 }
 677
 678 /* Call all constructor functions linked into the kernel. */
 679 static void __init do_ctors(void)
 680 {
 681 #ifdef CONFIG_CONSTRUCTORS
 682         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 683
 684         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 685                 (*fn)();
 686 #endif
 687 }
 688
 689 bool initcall_debug;
 690 core_param(initcall_debug, initcall_debug, bool, 0644);
 691
 692 #ifdef CONFIG_KALLSYMS
 693 struct blacklist_entry {
 694         struct list_head next;
 695         char *buf;
 696 };
 697
 698 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
 699
 700 static int __init initcall_blacklist(char *str)
 701 {
 702         char *str_entry;
 703         struct blacklist_entry *entry;
 704
 705         /* str argument is a comma-separated list of functions */
 706         do {
 707                 str_entry = strsep(&str, ",");
 708                 if (str_entry) {
 709                         pr_debug("blacklisting initcall %s\n", str_entry);
 710                         entry = alloc_bootmem(sizeof(*entry));
 711                         entry->buf = alloc_bootmem(strlen(str_entry) + 1);
 712                         strcpy(entry->buf, str_entry);
 713                         list_add(&entry->next, &blacklisted_initcalls);
 714                 }
 715         } while (str_entry);
 716
 717         return 0;
 718 }
 719
 720 static bool __init_or_module initcall_blacklisted(initcall_t fn)
 721 {
 722         struct blacklist_entry *entry;
 723         char *fn_name;
 724
 725         fn_name = kasprintf(GFP_KERNEL, "%pf", fn);
 726         if (!fn_name)
 727                 return false;
 728
 729         list_for_each_entry(entry, &blacklisted_initcalls, next) {
 730                 if (!strcmp(fn_name, entry->buf)) {
 731                         pr_debug("initcall %s blacklisted\n", fn_name);
 732                         kfree(fn_name);
 733                         return true;
 734                 }
 735         }
 736
 737         kfree(fn_name);
 738         return false;
 739 }
 740 #else
 741 static int __init initcall_blacklist(char *str)
 742 {
 743         pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
 744         return 0;
 745 }
 746
 747 static bool __init_or_module initcall_blacklisted(initcall_t fn)
 748 {
 749         return false;
 750 }
 751 #endif
 752 __setup("initcall_blacklist=", initcall_blacklist);
 753
 754 static int __init_or_module do_one_initcall_debug(initcall_t fn)
 755 {
 756         ktime_t calltime, delta, rettime;
 757         unsigned long long duration;
 758         int ret;
 759
 760         printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
 761         calltime = ktime_get();
 762         ret = fn();
 763         rettime = ktime_get();
 764         delta = ktime_sub(rettime, calltime);
 765         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
 766         printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
 767                  fn, ret, duration);
 768
 769         return ret;
 770 }
 771
 772 int __init_or_module do_one_initcall(initcall_t fn)
 773 {
 774         int count = preempt_count();
 775         int ret;
 776         char msgbuf[64];
 777
 778         if (initcall_blacklisted(fn))
 779                 return -EPERM;
 780
 781         if (initcall_debug)
 782                 ret = do_one_initcall_debug(fn);
 783         else
 784                 ret = fn();
 785
 786         msgbuf[0] = 0;
 787
 788         if (preempt_count() != count) {
 789                 sprintf(msgbuf, "preemption imbalance ");
 790                 preempt_count_set(count);
 791         }
 792         if (irqs_disabled()) {
 793                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
 794                 local_irq_enable();
 795         }
 796         WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
 797
 798         return ret;
 799 }
 800
 801
 802 extern initcall_t __initcall_start[];
 803 extern initcall_t __initcall0_start[];
 804 extern initcall_t __initcall1_start[];
 805 extern initcall_t __initcall2_start[];
 806 extern initcall_t __initcall3_start[];
 807 extern initcall_t __initcall4_start[];
 808 extern initcall_t __initcall5_start[];
 809 extern initcall_t __initcall6_start[];
 810 extern initcall_t __initcall7_start[];
 811 extern initcall_t __initcall_end[];
 812
 813 static initcall_t *initcall_levels[] __initdata = {
 814         __initcall0_start,
 815         __initcall1_start,
 816         __initcall2_start,
 817         __initcall3_start,
 818         __initcall4_start,
 819         __initcall5_start,
 820         __initcall6_start,
 821         __initcall7_start,
 822         __initcall_end,
 823 };
 824
 825 /* Keep these in sync with initcalls in include/linux/init.h */
 826 static char *initcall_level_names[] __initdata = {
 827         "early",
 828         "core",
 829         "postcore",
 830         "arch",
 831         "subsys",
 832         "fs",
 833         "device",
 834         "late",
 835 };
 836
 837 static void __init do_initcall_level(int level)
 838 {
 839         initcall_t *fn;
 840
 841         strcpy(initcall_command_line, saved_command_line);
 842         parse_args(initcall_level_names[level],
 843                    initcall_command_line, __start___param,
 844                    __stop___param - __start___param,
 845                    level, level,
 846                    NULL, &repair_env_string);
 847
 848         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
 849                 do_one_initcall(*fn);
 850 }
 851
 852 static void __init do_initcalls(void)
 853 {
 854         int level;
 855
 856         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
 857                 do_initcall_level(level);
 858 }
 859
 860 /*
 861  * Ok, the machine is now initialized. None of the devices
 862  * have been touched yet, but the CPU subsystem is up and
 863  * running, and memory and process management works.
 864  *
 865  * Now we can finally start doing some real work..
 866  */
 867 static void __init do_basic_setup(void)
 868 {
 869         cpuset_init_smp();
 870         shmem_init();
 871         driver_init();
 872         init_irq_proc();
 873         do_ctors();
 874         usermodehelper_enable();
 875         do_initcalls();
 876         random_int_secret_init();
 877 }
 878
 879 static void __init do_pre_smp_initcalls(void)
 880 {
 881         initcall_t *fn;
 882
 883         for (fn = __initcall_start; fn < __initcall0_start; fn++)
 884                 do_one_initcall(*fn);
 885 }
 886
 887 /*
 888  * This function requests modules which should be loaded by default and is
 889  * called twice right after initrd is mounted and right before init is
 890  * exec'd.  If such modules are on either initrd or rootfs, they will be
 891  * loaded before control is passed to userland.
 892  */
 893 void __init load_default_modules(void)
 894 {
 895         load_default_elevator_module();
 896 }
 897
 898 static int run_init_process(const char *init_filename)
 899 {
 900         argv_init[0] = init_filename;
 901         return do_execve(getname_kernel(init_filename),
 902                 (const char __user *const __user *)argv_init,
 903                 (const char __user *const __user *)envp_init);
 904 }
 905
 906 static int try_to_run_init_process(const char *init_filename)
 907 {
 908         int ret;
 909
 910         ret = run_init_process(init_filename);
 911
 912         if (ret && ret != -ENOENT) {
 913                 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
 914                        init_filename, ret);
 915         }
 916
 917         return ret;
 918 }
 919
 920 static noinline void __init kernel_init_freeable(void);
 921
 922 #ifdef CONFIG_DEBUG_RODATA
 923 static bool rodata_enabled = true;
 924 static int __init set_debug_rodata(char *str)
 925 {
 926         return strtobool(str, &rodata_enabled);
 927 }
 928 __setup("rodata=", set_debug_rodata);
 929
 930 static void mark_readonly(void)
 931 {
 932         if (rodata_enabled)
 933                 mark_rodata_ro();
 934         else
 935                 pr_info("Kernel memory protection disabled.\n");
 936 }
 937 #else
 938 static inline void mark_readonly(void)
 939 {
 940         pr_warn("This architecture does not have kernel memory protection.\n");
 941 }
 942 #endif
 943
 944 static int __ref kernel_init(void *unused)
 945 {
 946         int ret;
 947
 948         kernel_init_freeable();
 949         /* need to finish all async __init code before freeing the memory */
 950         async_synchronize_full();
 951         free_initmem();
 952         mark_readonly();
 953         system_state = SYSTEM_RUNNING;
 954         numa_default_policy();
 955
 956         flush_delayed_fput();
 957
 958         rcu_end_inkernel_boot();
 959
 960         if (ramdisk_execute_command) {
 961                 ret = run_init_process(ramdisk_execute_command);
 962                 if (!ret)
 963                         return 0;
 964                 pr_err("Failed to execute %s (error %d)\n",
 965                        ramdisk_execute_command, ret);
 966         }
 967
 968         /*
 969          * We try each of these until one succeeds.
 970          *
 971          * The Bourne shell can be used instead of init if we are
 972          * trying to recover a really broken machine.
 973          */
 974         if (execute_command) {
 975                 ret = run_init_process(execute_command);
 976                 if (!ret)
 977                         return 0;
 978                 panic("Requested init %s failed (error %d).",
 979                       execute_command, ret);
 980         }
 981         if (!try_to_run_init_process("/sbin/init") ||
 982             !try_to_run_init_process("/etc/init") ||
 983             !try_to_run_init_process("/bin/init") ||
 984             !try_to_run_init_process("/bin/sh"))
 985                 return 0;
 986
 987         panic("No working init found.  Try passing init= option to kernel. "
 988               "See Linux Documentation/init.txt for guidance.");
 989 }
 990
 991 static noinline void __init kernel_init_freeable(void)
 992 {
 993         /*
 994          * Wait until kthreadd is all set-up.
 995          */
 996         wait_for_completion(&kthreadd_done);
 997
 998         /* Now the scheduler is fully set up and can do blocking allocations */
 999         gfp_allowed_mask = __GFP_BITS_MASK;
1000
1001         /*
1002          * init can allocate pages on any node
1003          */
1004         set_mems_allowed(node_states[N_MEMORY]);
1005         /*
1006          * init can run on any cpu.
1007          */
1008         set_cpus_allowed_ptr(current, cpu_all_mask);
1009
1010         cad_pid = task_pid(current);
1011
1012         smp_prepare_cpus(setup_max_cpus);
1013
1014         do_pre_smp_initcalls();
1015         lockup_detector_init();
1016
1017         smp_init();
1018         sched_init_smp();
1019
1020         page_alloc_init_late();
1021
1022         do_basic_setup();
1023
1024         /* Open the /dev/console on the rootfs, this should never fail */
1025         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1026                 pr_err("Warning: unable to open an initial console.\n");
1027
1028         (void) sys_dup(0);
1029         (void) sys_dup(0);
1030         /*
1031          * check if there is an early userspace init.  If yes, let it do all
1032          * the work
1033          */
1034
1035         if (!ramdisk_execute_command)
1036                 ramdisk_execute_command = "/init";
1037
1038         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1039                 ramdisk_execute_command = NULL;
1040                 prepare_namespace();
1041         }
1042
1043         /*
1044          * Ok, we have completed the initial bootup, and
1045          * we're essentially up and running. Get rid of the
1046          * initmem segments and start the user-mode stuff..
1047          *
1048          * rootfs is available now, try loading the public keys
1049          * and default modules
1050          */
1051
1052         integrity_load_keys();
1053         load_default_modules();
1054 }