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/kernel_stat.h>
  35 #include <linux/security.h>
  36 #include <linux/workqueue.h>
  37 #include <linux/profile.h>
  38 #include <linux/rcupdate.h>
  39 #include <linux/moduleparam.h>
  40 #include <linux/kallsyms.h>
  41 #include <linux/writeback.h>
  42 #include <linux/cpu.h>
  43 #include <linux/efi.h>
  44 #include <linux/unistd.h>
  45 #include <linux/rmap.h>
  46
  47 #include <asm/io.h>
  48 #include <asm/bugs.h>
  49
  50 /*
  51  * This is one of the first .c files built. Error out early
  52  * if we have compiler trouble..
  53  */
  54 #if __GNUC__ == 2 && __GNUC_MINOR__ == 96
  55 #ifdef CONFIG_FRAME_POINTER
  56 #error This compiler cannot compile correctly with frame pointers enabled
  57 #endif
  58 #endif
  59
  60 #ifdef CONFIG_X86_LOCAL_APIC
  61 #include <asm/smp.h>
  62 #endif
  63
  64 /*
  65  * Versions of gcc older than that listed below may actually compile
  66  * and link okay, but the end product can have subtle run time bugs.
  67  * To avoid associated bogus bug reports, we flatly refuse to compile
  68  * with a gcc that is known to be too old from the very beginning.
  69  */
  70 #if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 95)
  71 #error Sorry, your GCC is too old. It builds incorrect kernels.
  72 #endif
  73
  74 extern char *linux_banner;
  75
  76 static int init(void *);
  77
  78 extern void init_IRQ(void);
  79 extern void sock_init(void);
  80 extern void fork_init(unsigned long);
  81 extern void mca_init(void);
  82 extern void sbus_init(void);
  83 extern void sysctl_init(void);
  84 extern void signals_init(void);
  85 extern void buffer_init(void);
  86 extern void pidhash_init(void);
  87 extern void pidmap_init(void);
  88 extern void prio_tree_init(void);
  89 extern void radix_tree_init(void);
  90 extern void free_initmem(void);
  91 extern void populate_rootfs(void);
  92 extern void driver_init(void);
  93 extern void prepare_namespace(void);
  94
  95 #ifdef CONFIG_TC
  96 extern void tc_init(void);
  97 #endif
  98
  99 enum system_states system_state;
 100 EXPORT_SYMBOL(system_state);
 101
 102 /*
 103  * Boot command-line arguments
 104  */
 105 #define MAX_INIT_ARGS 8
 106 #define MAX_INIT_ENVS 8
 107
 108 extern void time_init(void);
 109 /* Default late time init is NULL. archs can override this later. */
 110 void (*late_time_init)(void);
 111 extern void softirq_init(void);
 112
 113 static char *execute_command;
 114
 115 /* Setup configured maximum number of CPUs to activate */
 116 static unsigned int max_cpus = NR_CPUS;
 117
 118 /*
 119  * Setup routine for controlling SMP activation
 120  *
 121  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 122  * activation entirely (the MPS table probe still happens, though).
 123  *
 124  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 125  * greater than 0, limits the maximum number of CPUs activated in
 126  * SMP mode to <NUM>.
 127  */
 128 static int __init nosmp(char *str)
 129 {
 130         max_cpus = 0;
 131         return 1;
 132 }
 133
 134 __setup("nosmp", nosmp);
 135
 136 static int __init maxcpus(char *str)
 137 {
 138         get_option(&str, &max_cpus);
 139         return 1;
 140 }
 141
 142 __setup("maxcpus=", maxcpus);
 143
 144 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 145 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 146 static const char *panic_later, *panic_param;
 147
 148 __setup("profile=", profile_setup);
 149
 150 static int __init obsolete_checksetup(char *line)
 151 {
 152         struct obs_kernel_param *p;
 153         extern struct obs_kernel_param __setup_start, __setup_end;
 154
 155         p = &__setup_start;
 156         do {
 157                 int n = strlen(p->str);
 158                 if (!strncmp(line, p->str, n)) {
 159                         if (!p->setup_func) {
 160                                 printk(KERN_WARNING "Parameter %s is obsolete, ignored\n", p->str);
 161                                 return 1;
 162                         } else if (p->setup_func(line + n))
 163                                 return 1;
 164                 }
 165                 p++;
 166         } while (p < &__setup_end);
 167         return 0;
 168 }
 169
 170 /* this should be approx 2 Bo*oMips to start (note initial shift), and will
 171    still work even if initially too large, it will just take slightly longer */
 172 unsigned long loops_per_jiffy = (1<<12);
 173
 174 EXPORT_SYMBOL(loops_per_jiffy);
 175
 176 /* This is the number of bits of precision for the loops_per_jiffy.  Each
 177    bit takes on average 1.5/HZ seconds.  This (like the original) is a little
 178    better than 1% */
 179 #define LPS_PREC 8
 180
 181 void __devinit calibrate_delay(void)
 182 {
 183         unsigned long ticks, loopbit;
 184         int lps_precision = LPS_PREC;
 185
 186         loops_per_jiffy = (1<<12);
 187
 188         printk("Calibrating delay loop... ");
 189         while ((loops_per_jiffy <<= 1) != 0) {
 190                 /* wait for "start of" clock tick */
 191                 ticks = jiffies;
 192                 while (ticks == jiffies)
 193                         /* nothing */;
 194                 /* Go .. */
 195                 ticks = jiffies;
 196                 __delay(loops_per_jiffy);
 197                 ticks = jiffies - ticks;
 198                 if (ticks)
 199                         break;
 200         }
 201
 202 /* Do a binary approximation to get loops_per_jiffy set to equal one clock
 203    (up to lps_precision bits) */
 204         loops_per_jiffy >>= 1;
 205         loopbit = loops_per_jiffy;
 206         while ( lps_precision-- && (loopbit >>= 1) ) {
 207                 loops_per_jiffy |= loopbit;
 208                 ticks = jiffies;
 209                 while (ticks == jiffies);
 210                 ticks = jiffies;
 211                 __delay(loops_per_jiffy);
 212                 if (jiffies != ticks)   /* longer than 1 tick */
 213                         loops_per_jiffy &= ~loopbit;
 214         }
 215
 216 /* Round the value and print it */
 217         printk("%lu.%02lu BogoMIPS\n",
 218                 loops_per_jiffy/(500000/HZ),
 219                 (loops_per_jiffy/(5000/HZ)) % 100);
 220 }
 221
 222 static int __init debug_kernel(char *str)
 223 {
 224         if (*str)
 225                 return 0;
 226         console_loglevel = 10;
 227         return 1;
 228 }
 229
 230 static int __init quiet_kernel(char *str)
 231 {
 232         if (*str)
 233                 return 0;
 234         console_loglevel = 4;
 235         return 1;
 236 }
 237
 238 __setup("debug", debug_kernel);
 239 __setup("quiet", quiet_kernel);
 240
 241 /* Unknown boot options get handed to init, unless they look like
 242    failed parameters */
 243 static int __init unknown_bootoption(char *param, char *val)
 244 {
 245         /* Change NUL term back to "=", to make "param" the whole string. */
 246         if (val)
 247                 val[-1] = '=';
 248
 249         /* Handle obsolete-style parameters */
 250         if (obsolete_checksetup(param))
 251                 return 0;
 252
 253         /* Preemptive maintenance for "why didn't my mispelled command
 254            line work?" */
 255         if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
 256                 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
 257                 return 0;
 258         }
 259
 260         if (panic_later)
 261                 return 0;
 262
 263         if (val) {
 264                 /* Environment option */
 265                 unsigned int i;
 266                 for (i = 0; envp_init[i]; i++) {
 267                         if (i == MAX_INIT_ENVS) {
 268                                 panic_later = "Too many boot env vars at `%s'";
 269                                 panic_param = param;
 270                         }
 271                 }
 272                 envp_init[i] = param;
 273         } else {
 274                 /* Command line option */
 275                 unsigned int i;
 276                 for (i = 0; argv_init[i]; i++) {
 277                         if (i == MAX_INIT_ARGS) {
 278                                 panic_later = "Too many boot init vars at `%s'";
 279                                 panic_param = param;
 280                         }
 281                 }
 282                 argv_init[i] = param;
 283         }
 284         return 0;
 285 }
 286
 287 static int __init init_setup(char *str)
 288 {
 289         unsigned int i;
 290
 291         execute_command = str;
 292         /* In case LILO is going to boot us with default command line,
 293          * it prepends "auto" before the whole cmdline which makes
 294          * the shell think it should execute a script with such name.
 295          * So we ignore all arguments entered _before_ init=... [MJ]
 296          */
 297         for (i = 1; i < MAX_INIT_ARGS; i++)
 298                 argv_init[i] = NULL;
 299         return 1;
 300 }
 301 __setup("init=", init_setup);
 302
 303 extern void setup_arch(char **);
 304 extern void cpu_idle(void);
 305
 306 #ifndef CONFIG_SMP
 307
 308 #ifdef CONFIG_X86_LOCAL_APIC
 309 static void __init smp_init(void)
 310 {
 311         APIC_init_uniprocessor();
 312 }
 313 #else
 314 #define smp_init()      do { } while (0)
 315 #endif
 316
 317 static inline void setup_per_cpu_areas(void) { }
 318 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 319
 320 #else
 321
 322 #ifdef __GENERIC_PER_CPU
 323 unsigned long __per_cpu_offset[NR_CPUS];
 324
 325 EXPORT_SYMBOL(__per_cpu_offset);
 326
 327 static void __init setup_per_cpu_areas(void)
 328 {
 329         unsigned long size, i;
 330         char *ptr;
 331         /* Created by linker magic */
 332         extern char __per_cpu_start[], __per_cpu_end[];
 333
 334         /* Copy section for each CPU (we discard the original) */
 335         size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
 336 #ifdef CONFIG_MODULES
 337         if (size < PERCPU_ENOUGH_ROOM)
 338                 size = PERCPU_ENOUGH_ROOM;
 339 #endif
 340
 341         ptr = alloc_bootmem(size * NR_CPUS);
 342
 343         for (i = 0; i < NR_CPUS; i++, ptr += size) {
 344                 __per_cpu_offset[i] = ptr - __per_cpu_start;
 345                 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 346         }
 347 }
 348 #endif /* !__GENERIC_PER_CPU */
 349
 350 /* Called by boot processor to activate the rest. */
 351 static void __init smp_init(void)
 352 {
 353         unsigned int i;
 354
 355         /* FIXME: This should be done in userspace --RR */
 356         for_each_present_cpu(i) {
 357                 if (num_online_cpus() >= max_cpus)
 358                         break;
 359                 if (!cpu_online(i))
 360                         cpu_up(i);
 361         }
 362
 363         /* Any cleanup work */
 364         printk("Brought up %ld CPUs\n", (long)num_online_cpus());
 365         smp_cpus_done(max_cpus);
 366 #if 0
 367         /* Get other processors into their bootup holding patterns. */
 368
 369         smp_threads_ready=1;
 370         smp_commence();
 371 #endif
 372 }
 373
 374 #endif
 375
 376 /*
 377  * We need to finalize in a non-__init function or else race conditions
 378  * between the root thread and the init thread may cause start_kernel to
 379  * be reaped by free_initmem before the root thread has proceeded to
 380  * cpu_idle.
 381  *
 382  * gcc-3.4 accidentally inlines this function, so use noinline.
 383  */
 384
 385 static void noinline rest_init(void)
 386 {
 387         kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
 388         unlock_kernel();
 389         cpu_idle();
 390 }
 391
 392 /*
 393  *      Activate the first processor.
 394  */
 395
 396 asmlinkage void __init start_kernel(void)
 397 {
 398         char * command_line;
 399         extern char saved_command_line[];
 400         extern struct kernel_param __start___param[], __stop___param[];
 401 /*
 402  * Interrupts are still disabled. Do necessary setups, then
 403  * enable them
 404  */
 405         lock_kernel();
 406         page_address_init();
 407         printk(linux_banner);
 408         setup_arch(&command_line);
 409         setup_per_cpu_areas();
 410
 411         /*
 412          * Mark the boot cpu "online" so that it can call console drivers in
 413          * printk() and can access its per-cpu storage.
 414          */
 415         smp_prepare_boot_cpu();
 416
 417         /*
 418          * Set up the scheduler prior starting any interrupts (such as the
 419          * timer interrupt). Full topology setup happens at smp_init()
 420          * time - but meanwhile we still have a functioning scheduler.
 421          */
 422         sched_init();
 423
 424         build_all_zonelists();
 425         page_alloc_init();
 426         printk("Kernel command line: %s\n", saved_command_line);
 427         parse_args("Booting kernel", command_line, __start___param,
 428                    __stop___param - __start___param,
 429                    &unknown_bootoption);
 430         sort_main_extable();
 431         trap_init();
 432         rcu_init();
 433         init_IRQ();
 434         pidhash_init();
 435         init_timers();
 436         softirq_init();
 437         time_init();
 438
 439         /*
 440          * HACK ALERT! This is early. We're enabling the console before
 441          * we've done PCI setups etc, and console_init() must be aware of
 442          * this. But we do want output early, in case something goes wrong.
 443          */
 444         console_init();
 445         if (panic_later)
 446                 panic(panic_later, panic_param);
 447         profile_init();
 448         local_irq_enable();
 449 #ifdef CONFIG_BLK_DEV_INITRD
 450         if (initrd_start && !initrd_below_start_ok &&
 451                         initrd_start < min_low_pfn << PAGE_SHIFT) {
 452                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 453                     "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
 454                 initrd_start = 0;
 455         }
 456 #endif
 457         mem_init();
 458         kmem_cache_init();
 459         if (late_time_init)
 460                 late_time_init();
 461         calibrate_delay();
 462         pidmap_init();
 463         pgtable_cache_init();
 464         prio_tree_init();
 465         anon_vma_init();
 466 #ifdef CONFIG_X86
 467         if (efi_enabled)
 468                 efi_enter_virtual_mode();
 469 #endif
 470         fork_init(num_physpages);
 471         proc_caches_init();
 472         buffer_init();
 473         unnamed_dev_init();
 474         security_scaffolding_startup();
 475         vfs_caches_init(num_physpages);
 476         radix_tree_init();
 477         signals_init();
 478         /* rootfs populating might need page-writeback */
 479         page_writeback_init();
 480 #ifdef CONFIG_PROC_FS
 481         proc_root_init();
 482 #endif
 483         check_bugs();
 484
 485         /*
 486          *      We count on the initial thread going ok
 487          *      Like idlers init is an unlocked kernel thread, which will
 488          *      make syscalls (and thus be locked).
 489          */
 490         init_idle(current, smp_processor_id());
 491
 492         /* Do the rest non-__init'ed, we're now alive */
 493         rest_init();
 494 }
 495
 496 static int __initdata initcall_debug;
 497
 498 static int __init initcall_debug_setup(char *str)
 499 {
 500         initcall_debug = 1;
 501         return 1;
 502 }
 503 __setup("initcall_debug", initcall_debug_setup);
 504
 505 struct task_struct *child_reaper = &init_task;
 506
 507 extern initcall_t __initcall_start, __initcall_end;
 508
 509 static void __init do_initcalls(void)
 510 {
 511         initcall_t *call;
 512         int count = preempt_count();
 513
 514         for (call = &__initcall_start; call < &__initcall_end; call++) {
 515                 char *msg;
 516
 517                 if (initcall_debug) {
 518                         printk(KERN_DEBUG "Calling initcall 0x%p", *call);
 519                         print_symbol(": %s()", (unsigned long) *call);
 520                         printk("\n");
 521                 }
 522
 523                 (*call)();
 524
 525                 msg = NULL;
 526                 if (preempt_count() != count) {
 527                         msg = "preemption imbalance";
 528                         preempt_count() = count;
 529                 }
 530                 if (irqs_disabled()) {
 531                         msg = "disabled interrupts";
 532                         local_irq_enable();
 533                 }
 534                 if (msg) {
 535                         printk("error in initcall at 0x%p: "
 536                                 "returned with %s\n", *call, msg);
 537                 }
 538         }
 539
 540         /* Make sure there is no pending stuff from the initcall sequence */
 541         flush_scheduled_work();
 542 }
 543
 544 /*
 545  * Ok, the machine is now initialized. None of the devices
 546  * have been touched yet, but the CPU subsystem is up and
 547  * running, and memory and process management works.
 548  *
 549  * Now we can finally start doing some real work..
 550  */
 551 static void __init do_basic_setup(void)
 552 {
 553         driver_init();
 554
 555 #ifdef CONFIG_SYSCTL
 556         sysctl_init();
 557 #endif
 558
 559         /* Networking initialization needs a process context */
 560         sock_init();
 561
 562         init_workqueues();
 563         do_initcalls();
 564 }
 565
 566 static void do_pre_smp_initcalls(void)
 567 {
 568         extern int spawn_ksoftirqd(void);
 569 #ifdef CONFIG_SMP
 570         extern int migration_init(void);
 571
 572         migration_init();
 573 #endif
 574         spawn_ksoftirqd();
 575 }
 576
 577 static void run_init_process(char *init_filename)
 578 {
 579         argv_init[0] = init_filename;
 580         execve(init_filename, argv_init, envp_init);
 581 }
 582
 583 static inline void fixup_cpu_present_map(void)
 584 {
 585 #ifdef CONFIG_SMP
 586         int i;
 587
 588         /*
 589          * If arch is not hotplug ready and did not populate
 590          * cpu_present_map, just make cpu_present_map same as cpu_possible_map
 591          * for other cpu bringup code to function as normal. e.g smp_init() etc.
 592          */
 593         if (cpus_empty(cpu_present_map)) {
 594                 for_each_cpu(i) {
 595                         cpu_set(i, cpu_present_map);
 596                 }
 597         }
 598 #endif
 599 }
 600
 601 static int init(void * unused)
 602 {
 603         lock_kernel();
 604         /*
 605          * Tell the world that we're going to be the grim
 606          * reaper of innocent orphaned children.
 607          *
 608          * We don't want people to have to make incorrect
 609          * assumptions about where in the task array this
 610          * can be found.
 611          */
 612         child_reaper = current;
 613
 614         /* Sets up cpus_possible() */
 615         smp_prepare_cpus(max_cpus);
 616
 617         do_pre_smp_initcalls();
 618
 619         fixup_cpu_present_map();
 620         smp_init();
 621         sched_init_smp();
 622
 623         /*
 624          * Do this before initcalls, because some drivers want to access
 625          * firmware files.
 626          */
 627         populate_rootfs();
 628
 629         do_basic_setup();
 630
 631         /*
 632          * check if there is an early userspace init.  If yes, let it do all
 633          * the work
 634          */
 635         if (sys_access((const char __user *) "/init", 0) == 0)
 636                 execute_command = "/init";
 637         else
 638                 prepare_namespace();
 639
 640         /*
 641          * Ok, we have completed the initial bootup, and
 642          * we're essentially up and running. Get rid of the
 643          * initmem segments and start the user-mode stuff..
 644          */
 645         free_initmem();
 646         unlock_kernel();
 647         system_state = SYSTEM_RUNNING;
 648
 649         if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
 650                 printk("Warning: unable to open an initial console.\n");
 651
 652         (void) sys_dup(0);
 653         (void) sys_dup(0);
 654
 655         /*
 656          * We try each of these until one succeeds.
 657          *
 658          * The Bourne shell can be used instead of init if we are
 659          * trying to recover a really broken machine.
 660          */
 661
 662         if (execute_command)
 663                 run_init_process(execute_command);
 664
 665         run_init_process("/sbin/init");
 666         run_init_process("/etc/init");
 667         run_init_process("/bin/init");
 668         run_init_process("/bin/sh");
 669
 670         panic("No init found.  Try passing init= option to kernel.");
 671 }