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 __KERNEL_SYSCALLS__
  13
  14 #include <linux/config.h>
  15 #include <linux/types.h>
  16 #include <linux/module.h>
  17 #include <linux/proc_fs.h>
  18 #include <linux/devfs_fs_kernel.h>
  19 #include <linux/kernel.h>
  20 #include <linux/syscalls.h>
  21 #include <linux/string.h>
  22 #include <linux/ctype.h>
  23 #include <linux/delay.h>
  24 #include <linux/utsname.h>
  25 #include <linux/ioport.h>
  26 #include <linux/init.h>
  27 #include <linux/smp_lock.h>
  28 #include <linux/initrd.h>
  29 #include <linux/hdreg.h>
  30 #include <linux/bootmem.h>
  31 #include <linux/tty.h>
  32 #include <linux/gfp.h>
  33 #include <linux/percpu.h>
  34 #include <linux/kmod.h>
  35 #include <linux/kernel_stat.h>
  36 #include <linux/security.h>
  37 #include <linux/workqueue.h>
  38 #include <linux/profile.h>
  39 #include <linux/rcupdate.h>
  40 #include <linux/moduleparam.h>
  41 #include <linux/kallsyms.h>
  42 #include <linux/writeback.h>
  43 #include <linux/cpu.h>
  44 #include <linux/cpuset.h>
  45 #include <linux/efi.h>
  46 #include <linux/unistd.h>
  47 #include <linux/rmap.h>
  48 #include <linux/mempolicy.h>
  49 #include <linux/key.h>
  50 #include <linux/unwind.h>
  51
  52 #include <asm/io.h>
  53 #include <asm/bugs.h>
  54 #include <asm/setup.h>
  55 #include <asm/sections.h>
  56 #include <asm/cacheflush.h>
  57
  58 #ifdef CONFIG_X86_LOCAL_APIC
  59 #include <asm/smp.h>
  60 #endif
  61
  62 /*
  63  * This is one of the first .c files built. Error out early if we have compiler
  64  * trouble.
  65  *
  66  * Versions of gcc older than that listed below may actually compile and link
  67  * okay, but the end product can have subtle run time bugs.  To avoid associated
  68  * bogus bug reports, we flatly refuse to compile with a gcc that is known to be
  69  * too old from the very beginning.
  70  */
  71 #if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
  72 #error Sorry, your GCC is too old. It builds incorrect kernels.
  73 #endif
  74
  75 static int init(void *);
  76
  77 extern void init_IRQ(void);
  78 extern void fork_init(unsigned long);
  79 extern void mca_init(void);
  80 extern void sbus_init(void);
  81 extern void sysctl_init(void);
  82 extern void signals_init(void);
  83 extern void buffer_init(void);
  84 extern void pidhash_init(void);
  85 extern void pidmap_init(void);
  86 extern void prio_tree_init(void);
  87 extern void radix_tree_init(void);
  88 extern void free_initmem(void);
  89 extern void populate_rootfs(void);
  90 extern void driver_init(void);
  91 extern void prepare_namespace(void);
  92 #ifdef  CONFIG_ACPI
  93 extern void acpi_early_init(void);
  94 #else
  95 static inline void acpi_early_init(void) { }
  96 #endif
  97 #ifndef CONFIG_DEBUG_RODATA
  98 static inline void mark_rodata_ro(void) { }
  99 #endif
 100
 101 #ifdef CONFIG_TC
 102 extern void tc_init(void);
 103 #endif
 104
 105 enum system_states system_state;
 106 EXPORT_SYMBOL(system_state);
 107
 108 /*
 109  * Boot command-line arguments
 110  */
 111 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 112 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 113
 114 extern void time_init(void);
 115 /* Default late time init is NULL. archs can override this later. */
 116 void (*late_time_init)(void);
 117 extern void softirq_init(void);
 118
 119 /* Untouched command line (eg. for /proc) saved by arch-specific code. */
 120 char saved_command_line[COMMAND_LINE_SIZE];
 121
 122 static char *execute_command;
 123 static char *ramdisk_execute_command;
 124
 125 /* Setup configured maximum number of CPUs to activate */
 126 static unsigned int max_cpus = NR_CPUS;
 127
 128 /*
 129  * Setup routine for controlling SMP activation
 130  *
 131  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 132  * activation entirely (the MPS table probe still happens, though).
 133  *
 134  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 135  * greater than 0, limits the maximum number of CPUs activated in
 136  * SMP mode to <NUM>.
 137  */
 138 static int __init nosmp(char *str)
 139 {
 140         max_cpus = 0;
 141         return 1;
 142 }
 143
 144 __setup("nosmp", nosmp);
 145
 146 static int __init maxcpus(char *str)
 147 {
 148         get_option(&str, &max_cpus);
 149         return 1;
 150 }
 151
 152 __setup("maxcpus=", maxcpus);
 153
 154 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 155 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 156 static const char *panic_later, *panic_param;
 157
 158 extern struct obs_kernel_param __setup_start[], __setup_end[];
 159
 160 static int __init obsolete_checksetup(char *line)
 161 {
 162         struct obs_kernel_param *p;
 163
 164         p = __setup_start;
 165         do {
 166                 int n = strlen(p->str);
 167                 if (!strncmp(line, p->str, n)) {
 168                         if (p->early) {
 169                                 /* Already done in parse_early_param?  (Needs
 170                                  * exact match on param part) */
 171                                 if (line[n] == '\0' || line[n] == '=')
 172                                         return 1;
 173                         } else if (!p->setup_func) {
 174                                 printk(KERN_WARNING "Parameter %s is obsolete,"
 175                                        " ignored\n", p->str);
 176                                 return 1;
 177                         } else if (p->setup_func(line + n))
 178                                 return 1;
 179                 }
 180                 p++;
 181         } while (p < __setup_end);
 182         return 0;
 183 }
 184
 185 /*
 186  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 187  * still work even if initially too large, it will just take slightly longer
 188  */
 189 unsigned long loops_per_jiffy = (1<<12);
 190
 191 EXPORT_SYMBOL(loops_per_jiffy);
 192
 193 static int __init debug_kernel(char *str)
 194 {
 195         if (*str)
 196                 return 0;
 197         console_loglevel = 10;
 198         return 1;
 199 }
 200
 201 static int __init quiet_kernel(char *str)
 202 {
 203         if (*str)
 204                 return 0;
 205         console_loglevel = 4;
 206         return 1;
 207 }
 208
 209 __setup("debug", debug_kernel);
 210 __setup("quiet", quiet_kernel);
 211
 212 static int __init loglevel(char *str)
 213 {
 214         get_option(&str, &console_loglevel);
 215         return 1;
 216 }
 217
 218 __setup("loglevel=", loglevel);
 219
 220 /*
 221  * Unknown boot options get handed to init, unless they look like
 222  * failed parameters
 223  */
 224 static int __init unknown_bootoption(char *param, char *val)
 225 {
 226         /* Change NUL term back to "=", to make "param" the whole string. */
 227         if (val) {
 228                 /* param=val or param="val"? */
 229                 if (val == param+strlen(param)+1)
 230                         val[-1] = '=';
 231                 else if (val == param+strlen(param)+2) {
 232                         val[-2] = '=';
 233                         memmove(val-1, val, strlen(val)+1);
 234                         val--;
 235                 } else
 236                         BUG();
 237         }
 238
 239         /* Handle obsolete-style parameters */
 240         if (obsolete_checksetup(param))
 241                 return 0;
 242
 243         /*
 244          * Preemptive maintenance for "why didn't my mispelled command
 245          * line work?"
 246          */
 247         if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
 248                 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
 249                 return 0;
 250         }
 251
 252         if (panic_later)
 253                 return 0;
 254
 255         if (val) {
 256                 /* Environment option */
 257                 unsigned int i;
 258                 for (i = 0; envp_init[i]; i++) {
 259                         if (i == MAX_INIT_ENVS) {
 260                                 panic_later = "Too many boot env vars at `%s'";
 261                                 panic_param = param;
 262                         }
 263                         if (!strncmp(param, envp_init[i], val - param))
 264                                 break;
 265                 }
 266                 envp_init[i] = param;
 267         } else {
 268                 /* Command line option */
 269                 unsigned int i;
 270                 for (i = 0; argv_init[i]; i++) {
 271                         if (i == MAX_INIT_ARGS) {
 272                                 panic_later = "Too many boot init vars at `%s'";
 273                                 panic_param = param;
 274                         }
 275                 }
 276                 argv_init[i] = param;
 277         }
 278         return 0;
 279 }
 280
 281 static int __init init_setup(char *str)
 282 {
 283         unsigned int i;
 284
 285         execute_command = str;
 286         /*
 287          * In case LILO is going to boot us with default command line,
 288          * it prepends "auto" before the whole cmdline which makes
 289          * the shell think it should execute a script with such name.
 290          * So we ignore all arguments entered _before_ init=... [MJ]
 291          */
 292         for (i = 1; i < MAX_INIT_ARGS; i++)
 293                 argv_init[i] = NULL;
 294         return 1;
 295 }
 296 __setup("init=", init_setup);
 297
 298 static int __init rdinit_setup(char *str)
 299 {
 300         unsigned int i;
 301
 302         ramdisk_execute_command = str;
 303         /* See "auto" comment in init_setup */
 304         for (i = 1; i < MAX_INIT_ARGS; i++)
 305                 argv_init[i] = NULL;
 306         return 1;
 307 }
 308 __setup("rdinit=", rdinit_setup);
 309
 310 #ifndef CONFIG_SMP
 311
 312 #ifdef CONFIG_X86_LOCAL_APIC
 313 static void __init smp_init(void)
 314 {
 315         APIC_init_uniprocessor();
 316 }
 317 #else
 318 #define smp_init()      do { } while (0)
 319 #endif
 320
 321 static inline void setup_per_cpu_areas(void) { }
 322 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 323
 324 #else
 325
 326 #ifdef __GENERIC_PER_CPU
 327 unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
 328
 329 EXPORT_SYMBOL(__per_cpu_offset);
 330
 331 static void __init setup_per_cpu_areas(void)
 332 {
 333         unsigned long size, i;
 334         char *ptr;
 335         unsigned long nr_possible_cpus = num_possible_cpus();
 336
 337         /* Copy section for each CPU (we discard the original) */
 338         size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
 339 #ifdef CONFIG_MODULES
 340         if (size < PERCPU_ENOUGH_ROOM)
 341                 size = PERCPU_ENOUGH_ROOM;
 342 #endif
 343         ptr = alloc_bootmem(size * nr_possible_cpus);
 344
 345         for_each_possible_cpu(i) {
 346                 __per_cpu_offset[i] = ptr - __per_cpu_start;
 347                 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 348                 ptr += size;
 349         }
 350 }
 351 #endif /* !__GENERIC_PER_CPU */
 352
 353 /* Called by boot processor to activate the rest. */
 354 static void __init smp_init(void)
 355 {
 356         unsigned int i;
 357
 358         /* FIXME: This should be done in userspace --RR */
 359         for_each_present_cpu(i) {
 360                 if (num_online_cpus() >= max_cpus)
 361                         break;
 362                 if (!cpu_online(i))
 363                         cpu_up(i);
 364         }
 365
 366         /* Any cleanup work */
 367         printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
 368         smp_cpus_done(max_cpus);
 369 #if 0
 370         /* Get other processors into their bootup holding patterns. */
 371
 372         smp_commence();
 373 #endif
 374 }
 375
 376 #endif
 377
 378 /*
 379  * We need to finalize in a non-__init function or else race conditions
 380  * between the root thread and the init thread may cause start_kernel to
 381  * be reaped by free_initmem before the root thread has proceeded to
 382  * cpu_idle.
 383  *
 384  * gcc-3.4 accidentally inlines this function, so use noinline.
 385  */
 386
 387 static void noinline rest_init(void)
 388         __releases(kernel_lock)
 389 {
 390         kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
 391         numa_default_policy();
 392         unlock_kernel();
 393
 394         /*
 395          * The boot idle thread must execute schedule()
 396          * at least one to get things moving:
 397          */
 398         preempt_enable_no_resched();
 399         schedule();
 400         preempt_disable();
 401
 402         /* Call into cpu_idle with preempt disabled */
 403         cpu_idle();
 404 }
 405
 406 /* Check for early params. */
 407 static int __init do_early_param(char *param, char *val)
 408 {
 409         struct obs_kernel_param *p;
 410
 411         for (p = __setup_start; p < __setup_end; p++) {
 412                 if (p->early && strcmp(param, p->str) == 0) {
 413                         if (p->setup_func(val) != 0)
 414                                 printk(KERN_WARNING
 415                                        "Malformed early option '%s'\n", param);
 416                 }
 417         }
 418         /* We accept everything at this stage. */
 419         return 0;
 420 }
 421
 422 /* Arch code calls this early on, or if not, just before other parsing. */
 423 void __init parse_early_param(void)
 424 {
 425         static __initdata int done = 0;
 426         static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
 427
 428         if (done)
 429                 return;
 430
 431         /* All fall through to do_early_param. */
 432         strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
 433         parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
 434         done = 1;
 435 }
 436
 437 /*
 438  *      Activate the first processor.
 439  */
 440
 441 static void __init boot_cpu_init(void)
 442 {
 443         int cpu = smp_processor_id();
 444         /* Mark the boot cpu "present", "online" etc for SMP and UP case */
 445         cpu_set(cpu, cpu_online_map);
 446         cpu_set(cpu, cpu_present_map);
 447         cpu_set(cpu, cpu_possible_map);
 448 }
 449
 450 asmlinkage void __init start_kernel(void)
 451 {
 452         char * command_line;
 453         extern struct kernel_param __start___param[], __stop___param[];
 454 /*
 455  * Interrupts are still disabled. Do necessary setups, then
 456  * enable them
 457  */
 458         lock_kernel();
 459         boot_cpu_init();
 460         page_address_init();
 461         printk(KERN_NOTICE);
 462         printk(linux_banner);
 463         setup_arch(&command_line);
 464         setup_per_cpu_areas();
 465         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 466
 467         /*
 468          * Set up the scheduler prior starting any interrupts (such as the
 469          * timer interrupt). Full topology setup happens at smp_init()
 470          * time - but meanwhile we still have a functioning scheduler.
 471          */
 472         sched_init();
 473         /*
 474          * Disable preemption - early bootup scheduling is extremely
 475          * fragile until we cpu_idle() for the first time.
 476          */
 477         preempt_disable();
 478         build_all_zonelists();
 479         page_alloc_init();
 480         printk(KERN_NOTICE "Kernel command line: %s\n", saved_command_line);
 481         parse_early_param();
 482         parse_args("Booting kernel", command_line, __start___param,
 483                    __stop___param - __start___param,
 484                    &unknown_bootoption);
 485         sort_main_extable();
 486         unwind_init();
 487         trap_init();
 488         rcu_init();
 489         init_IRQ();
 490         pidhash_init();
 491         init_timers();
 492         hrtimers_init();
 493         softirq_init();
 494         time_init();
 495
 496         /*
 497          * HACK ALERT! This is early. We're enabling the console before
 498          * we've done PCI setups etc, and console_init() must be aware of
 499          * this. But we do want output early, in case something goes wrong.
 500          */
 501         console_init();
 502         if (panic_later)
 503                 panic(panic_later, panic_param);
 504         profile_init();
 505         local_irq_enable();
 506 #ifdef CONFIG_BLK_DEV_INITRD
 507         if (initrd_start && !initrd_below_start_ok &&
 508                         initrd_start < min_low_pfn << PAGE_SHIFT) {
 509                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 510                     "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
 511                 initrd_start = 0;
 512         }
 513 #endif
 514         vfs_caches_init_early();
 515         cpuset_init_early();
 516         mem_init();
 517         kmem_cache_init();
 518         setup_per_cpu_pageset();
 519         numa_policy_init();
 520         if (late_time_init)
 521                 late_time_init();
 522         calibrate_delay();
 523         pidmap_init();
 524         pgtable_cache_init();
 525         prio_tree_init();
 526         anon_vma_init();
 527 #ifdef CONFIG_X86
 528         if (efi_enabled)
 529                 efi_enter_virtual_mode();
 530 #endif
 531         fork_init(num_physpages);
 532         proc_caches_init();
 533         buffer_init();
 534         unnamed_dev_init();
 535         key_init();
 536         security_init();
 537         vfs_caches_init(num_physpages);
 538         radix_tree_init();
 539         signals_init();
 540         /* rootfs populating might need page-writeback */
 541         page_writeback_init();
 542 #ifdef CONFIG_PROC_FS
 543         proc_root_init();
 544 #endif
 545         cpuset_init();
 546
 547         check_bugs();
 548
 549         acpi_early_init(); /* before LAPIC and SMP init */
 550
 551         /* Do the rest non-__init'ed, we're now alive */
 552         rest_init();
 553 }
 554
 555 static int __initdata initcall_debug;
 556
 557 static int __init initcall_debug_setup(char *str)
 558 {
 559         initcall_debug = 1;
 560         return 1;
 561 }
 562 __setup("initcall_debug", initcall_debug_setup);
 563
 564 struct task_struct *child_reaper = &init_task;
 565
 566 extern initcall_t __initcall_start[], __initcall_end[];
 567
 568 static void __init do_initcalls(void)
 569 {
 570         initcall_t *call;
 571         int count = preempt_count();
 572
 573         for (call = __initcall_start; call < __initcall_end; call++) {
 574                 char *msg = NULL;
 575                 char msgbuf[40];
 576                 int result;
 577
 578                 if (initcall_debug) {
 579                         printk("Calling initcall 0x%p", *call);
 580                         print_fn_descriptor_symbol(": %s()",
 581                                         (unsigned long) *call);
 582                         printk("\n");
 583                 }
 584
 585                 result = (*call)();
 586
 587                 if (result && result != -ENODEV && initcall_debug) {
 588                         sprintf(msgbuf, "error code %d", result);
 589                         msg = msgbuf;
 590                 }
 591                 if (preempt_count() != count) {
 592                         msg = "preemption imbalance";
 593                         preempt_count() = count;
 594                 }
 595                 if (irqs_disabled()) {
 596                         msg = "disabled interrupts";
 597                         local_irq_enable();
 598                 }
 599                 if (msg) {
 600                         printk(KERN_WARNING "initcall at 0x%p", *call);
 601                         print_fn_descriptor_symbol(": %s()",
 602                                         (unsigned long) *call);
 603                         printk(": returned with %s\n", msg);
 604                 }
 605         }
 606
 607         /* Make sure there is no pending stuff from the initcall sequence */
 608         flush_scheduled_work();
 609 }
 610
 611 /*
 612  * Ok, the machine is now initialized. None of the devices
 613  * have been touched yet, but the CPU subsystem is up and
 614  * running, and memory and process management works.
 615  *
 616  * Now we can finally start doing some real work..
 617  */
 618 static void __init do_basic_setup(void)
 619 {
 620         /* drivers will send hotplug events */
 621         init_workqueues();
 622         usermodehelper_init();
 623         driver_init();
 624
 625 #ifdef CONFIG_SYSCTL
 626         sysctl_init();
 627 #endif
 628
 629         do_initcalls();
 630 }
 631
 632 static void do_pre_smp_initcalls(void)
 633 {
 634         extern int spawn_ksoftirqd(void);
 635 #ifdef CONFIG_SMP
 636         extern int migration_init(void);
 637
 638         migration_init();
 639 #endif
 640         spawn_ksoftirqd();
 641         spawn_softlockup_task();
 642 }
 643
 644 static void run_init_process(char *init_filename)
 645 {
 646         argv_init[0] = init_filename;
 647         execve(init_filename, argv_init, envp_init);
 648 }
 649
 650 static int init(void * unused)
 651 {
 652         lock_kernel();
 653         /*
 654          * init can run on any cpu.
 655          */
 656         set_cpus_allowed(current, CPU_MASK_ALL);
 657         /*
 658          * Tell the world that we're going to be the grim
 659          * reaper of innocent orphaned children.
 660          *
 661          * We don't want people to have to make incorrect
 662          * assumptions about where in the task array this
 663          * can be found.
 664          */
 665         child_reaper = current;
 666
 667         smp_prepare_cpus(max_cpus);
 668
 669         do_pre_smp_initcalls();
 670
 671         smp_init();
 672         sched_init_smp();
 673
 674         cpuset_init_smp();
 675
 676         /*
 677          * Do this before initcalls, because some drivers want to access
 678          * firmware files.
 679          */
 680         populate_rootfs();
 681
 682         do_basic_setup();
 683
 684         /*
 685          * check if there is an early userspace init.  If yes, let it do all
 686          * the work
 687          */
 688
 689         if (!ramdisk_execute_command)
 690                 ramdisk_execute_command = "/init";
 691
 692         if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
 693                 ramdisk_execute_command = NULL;
 694                 prepare_namespace();
 695         }
 696
 697         /*
 698          * Ok, we have completed the initial bootup, and
 699          * we're essentially up and running. Get rid of the
 700          * initmem segments and start the user-mode stuff..
 701          */
 702         free_initmem();
 703         unlock_kernel();
 704         mark_rodata_ro();
 705         system_state = SYSTEM_RUNNING;
 706         numa_default_policy();
 707
 708         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 709                 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
 710
 711         (void) sys_dup(0);
 712         (void) sys_dup(0);
 713
 714         if (ramdisk_execute_command) {
 715                 run_init_process(ramdisk_execute_command);
 716                 printk(KERN_WARNING "Failed to execute %s\n",
 717                                 ramdisk_execute_command);
 718         }
 719
 720         /*
 721          * We try each of these until one succeeds.
 722          *
 723          * The Bourne shell can be used instead of init if we are
 724          * trying to recover a really broken machine.
 725          */
 726         if (execute_command) {
 727                 run_init_process(execute_command);
 728                 printk(KERN_WARNING "Failed to execute %s.  Attempting "
 729                                         "defaults...\n", execute_command);
 730         }
 731         run_init_process("/sbin/init");
 732         run_init_process("/etc/init");
 733         run_init_process("/bin/init");
 734         run_init_process("/bin/sh");
 735
 736         panic("No init found.  Try passing init= option to kernel.");
 737 }