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
  46 #include <asm/io.h>
  47 #include <asm/bugs.h>
  48
  49 /*
  50  * This is one of the first .c files built. Error out early
  51  * if we have compiler trouble..
  52  */
  53 #if __GNUC__ == 2 && __GNUC_MINOR__ == 96
  54 #ifdef CONFIG_FRAME_POINTER
  55 #error This compiler cannot compile correctly with frame pointers enabled
  56 #endif
  57 #endif
  58
  59 #ifdef CONFIG_X86_LOCAL_APIC
  60 #include <asm/smp.h>
  61 #endif
  62
  63 /*
  64  * Versions of gcc older than that listed below may actually compile
  65  * and link okay, but the end product can have subtle run time bugs.
  66  * To avoid associated bogus bug reports, we flatly refuse to compile
  67  * with a gcc that is known to be too old from the very beginning.
  68  */
  69 #if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 95)
  70 #error Sorry, your GCC is too old. It builds incorrect kernels.
  71 #endif
  72
  73 extern char *linux_banner;
  74
  75 static int init(void *);
  76
  77 extern void init_IRQ(void);
  78 extern void sock_init(void);
  79 extern void fork_init(unsigned long);
  80 extern void mca_init(void);
  81 extern void sbus_init(void);
  82 extern void sysctl_init(void);
  83 extern void signals_init(void);
  84 extern void buffer_init(void);
  85 extern void pidhash_init(void);
  86 extern void pidmap_init(void);
  87 extern void pte_chain_init(void);
  88 extern void radix_tree_init(void);
  89 extern void free_initmem(void);
  90 extern void populate_rootfs(void);
  91 extern void driver_init(void);
  92
  93 #ifdef CONFIG_TC
  94 extern void tc_init(void);
  95 #endif
  96
  97 /*
  98  * Are we up and running (ie do we have all the infrastructure
  99  * set up)
 100  */
 101 int system_running;
 102
 103 /*
 104  * Boot command-line arguments
 105  */
 106 #define MAX_INIT_ARGS 8
 107 #define MAX_INIT_ENVS 8
 108
 109 extern void time_init(void);
 110 /* Default late time init is NULL. archs can override this later. */
 111 void (*late_time_init)(void);
 112 extern void softirq_init(void);
 113
 114 static char *execute_command;
 115
 116 /* Setup configured maximum number of CPUs to activate */
 117 static unsigned int max_cpus = NR_CPUS;
 118
 119 /*
 120  * Setup routine for controlling SMP activation
 121  *
 122  * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
 123  * activation entirely (the MPS table probe still happens, though).
 124  *
 125  * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
 126  * greater than 0, limits the maximum number of CPUs activated in
 127  * SMP mode to <NUM>.
 128  */
 129 static int __init nosmp(char *str)
 130 {
 131         max_cpus = 0;
 132         return 1;
 133 }
 134
 135 __setup("nosmp", nosmp);
 136
 137 static int __init maxcpus(char *str)
 138 {
 139         get_option(&str, &max_cpus);
 140         return 1;
 141 }
 142
 143 __setup("maxcpus=", maxcpus);
 144
 145 static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 146 char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 147 static const char *panic_later, *panic_param;
 148
 149 __setup("profile=", profile_setup);
 150
 151 static int __init obsolete_checksetup(char *line)
 152 {
 153         struct obs_kernel_param *p;
 154         extern struct obs_kernel_param __setup_start, __setup_end;
 155
 156         p = &__setup_start;
 157         do {
 158                 int n = strlen(p->str);
 159                 if (!strncmp(line, p->str, n)) {
 160                         if (!p->setup_func) {
 161                                 printk(KERN_WARNING "Parameter %s is obsolete, ignored\n", p->str);
 162                                 return 1;
 163                         } else if (p->setup_func(line + n))
 164                                 return 1;
 165                 }
 166                 p++;
 167         } while (p < &__setup_end);
 168         return 0;
 169 }
 170
 171 /* this should be approx 2 Bo*oMips to start (note initial shift), and will
 172    still work even if initially too large, it will just take slightly longer */
 173 unsigned long loops_per_jiffy = (1<<12);
 174
 175 #ifndef __ia64__
 176 EXPORT_SYMBOL(loops_per_jiffy);
 177 #endif
 178
 179 /* This is the number of bits of precision for the loops_per_jiffy.  Each
 180    bit takes on average 1.5/HZ seconds.  This (like the original) is a little
 181    better than 1% */
 182 #define LPS_PREC 8
 183
 184 void __init calibrate_delay(void)
 185 {
 186         unsigned long ticks, loopbit;
 187         int lps_precision = LPS_PREC;
 188
 189         loops_per_jiffy = (1<<12);
 190
 191         printk("Calibrating delay loop... ");
 192         while (loops_per_jiffy <<= 1) {
 193                 /* wait for "start of" clock tick */
 194                 ticks = jiffies;
 195                 while (ticks == jiffies)
 196                         /* nothing */;
 197                 /* Go .. */
 198                 ticks = jiffies;
 199                 __delay(loops_per_jiffy);
 200                 ticks = jiffies - ticks;
 201                 if (ticks)
 202                         break;
 203         }
 204
 205 /* Do a binary approximation to get loops_per_jiffy set to equal one clock
 206    (up to lps_precision bits) */
 207         loops_per_jiffy >>= 1;
 208         loopbit = loops_per_jiffy;
 209         while ( lps_precision-- && (loopbit >>= 1) ) {
 210                 loops_per_jiffy |= loopbit;
 211                 ticks = jiffies;
 212                 while (ticks == jiffies);
 213                 ticks = jiffies;
 214                 __delay(loops_per_jiffy);
 215                 if (jiffies != ticks)   /* longer than 1 tick */
 216                         loops_per_jiffy &= ~loopbit;
 217         }
 218
 219 /* Round the value and print it */
 220         printk("%lu.%02lu BogoMIPS\n",
 221                 loops_per_jiffy/(500000/HZ),
 222                 (loops_per_jiffy/(5000/HZ)) % 100);
 223 }
 224
 225 static int __init debug_kernel(char *str)
 226 {
 227         if (*str)
 228                 return 0;
 229         console_loglevel = 10;
 230         return 1;
 231 }
 232
 233 static int __init quiet_kernel(char *str)
 234 {
 235         if (*str)
 236                 return 0;
 237         console_loglevel = 4;
 238         return 1;
 239 }
 240
 241 __setup("debug", debug_kernel);
 242 __setup("quiet", quiet_kernel);
 243
 244 /* Unknown boot options get handed to init, unless they look like
 245    failed parameters */
 246 static int __init unknown_bootoption(char *param, char *val)
 247 {
 248         /* Change NUL term back to "=", to make "param" the whole string. */
 249         if (val)
 250                 val[-1] = '=';
 251
 252         /* Handle obsolete-style parameters */
 253         if (obsolete_checksetup(param))
 254                 return 0;
 255
 256         /* Preemptive maintenance for "why didn't my mispelled command
 257            line work?" */
 258         if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
 259                 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
 260                 return 0;
 261         }
 262
 263         if (panic_later)
 264                 return 0;
 265
 266         if (val) {
 267                 /* Environment option */
 268                 unsigned int i;
 269                 for (i = 0; envp_init[i]; i++) {
 270                         if (i == MAX_INIT_ENVS) {
 271                                 panic_later = "Too many boot env vars at `%s'";
 272                                 panic_param = param;
 273                         }
 274                 }
 275                 envp_init[i] = param;
 276         } else {
 277                 /* Command line option */
 278                 unsigned int i;
 279                 for (i = 0; argv_init[i]; i++) {
 280                         if (i == MAX_INIT_ARGS) {
 281                                 panic_later = "Too many boot init vars at `%s'";
 282                                 panic_param = param;
 283                         }
 284                 }
 285                 argv_init[i] = param;
 286         }
 287         return 0;
 288 }
 289
 290 static int __init init_setup(char *str)
 291 {
 292         unsigned int i;
 293
 294         execute_command = str;
 295         /* In case LILO is going to boot us with default command line,
 296          * it prepends "auto" before the whole cmdline which makes
 297          * the shell think it should execute a script with such name.
 298          * So we ignore all arguments entered _before_ init=... [MJ]
 299          */
 300         for (i = 1; i < MAX_INIT_ARGS; i++)
 301                 argv_init[i] = NULL;
 302         return 1;
 303 }
 304 __setup("init=", init_setup);
 305
 306 extern void setup_arch(char **);
 307 extern void cpu_idle(void);
 308
 309 #ifndef CONFIG_SMP
 310
 311 #ifdef CONFIG_X86_LOCAL_APIC
 312 static void __init smp_init(void)
 313 {
 314         APIC_init_uniprocessor();
 315 }
 316 #else
 317 #define smp_init()      do { } while (0)
 318 #endif
 319
 320 static inline void setup_per_cpu_areas(void) { }
 321 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 322
 323 #else
 324
 325 #ifdef __GENERIC_PER_CPU
 326 unsigned long __per_cpu_offset[NR_CPUS];
 327
 328 EXPORT_SYMBOL(__per_cpu_offset);
 329
 330 static void __init setup_per_cpu_areas(void)
 331 {
 332         unsigned long size, i;
 333         char *ptr;
 334         /* Created by linker magic */
 335         extern char __per_cpu_start[], __per_cpu_end[];
 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
 344         ptr = alloc_bootmem(size * NR_CPUS);
 345
 346         for (i = 0; i < NR_CPUS; i++, ptr += size) {
 347                 __per_cpu_offset[i] = ptr - __per_cpu_start;
 348                 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 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         unsigned j = 1;
 358
 359         /* FIXME: This should be done in userspace --RR */
 360         for (i = 0; i < NR_CPUS; i++) {
 361                 if (num_online_cpus() >= max_cpus)
 362                         break;
 363                 if (cpu_possible(i) && !cpu_online(i)) {
 364                         cpu_up(i);
 365                         j++;
 366                 }
 367         }
 368
 369         /* Any cleanup work */
 370         printk("Brought up %u CPUs\n", j);
 371         smp_cpus_done(max_cpus);
 372 #if 0
 373         /* Get other processors into their bootup holding patterns. */
 374
 375         smp_threads_ready=1;
 376         smp_commence();
 377 #endif
 378 }
 379
 380 #endif
 381
 382 /*
 383  * We need to finalize in a non-__init function or else race conditions
 384  * between the root thread and the init thread may cause start_kernel to
 385  * be reaped by free_initmem before the root thread has proceeded to
 386  * cpu_idle.
 387  *
 388  * gcc-3.4 accidentally inlines this function, so use noinline.
 389  */
 390
 391 static void noinline rest_init(void)
 392 {
 393         kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
 394         unlock_kernel();
 395         cpu_idle();
 396 }
 397
 398 /*
 399  *      Activate the first processor.
 400  */
 401
 402 asmlinkage void __init start_kernel(void)
 403 {
 404         char * command_line;
 405         extern char saved_command_line[];
 406         extern struct kernel_param __start___param[], __stop___param[];
 407 /*
 408  * Interrupts are still disabled. Do necessary setups, then
 409  * enable them
 410  */
 411         lock_kernel();
 412         page_address_init();
 413         printk(linux_banner);
 414         setup_arch(&command_line);
 415         setup_per_cpu_areas();
 416
 417         /*
 418          * Mark the boot cpu "online" so that it can call console drivers in
 419          * printk() and can access its per-cpu storage.
 420          */
 421         smp_prepare_boot_cpu();
 422
 423         build_all_zonelists();
 424         page_alloc_init();
 425         printk("Kernel command line: %s\n", saved_command_line);
 426         parse_args("Booting kernel", command_line, __start___param,
 427                    __stop___param - __start___param,
 428                    &unknown_bootoption);
 429         sort_main_extable();
 430         trap_init();
 431         rcu_init();
 432         init_IRQ();
 433         pidhash_init();
 434         sched_init();
 435         softirq_init();
 436         time_init();
 437
 438         /*
 439          * HACK ALERT! This is early. We're enabling the console before
 440          * we've done PCI setups etc, and console_init() must be aware of
 441          * this. But we do want output early, in case something goes wrong.
 442          */
 443         console_init();
 444         if (panic_later)
 445                 panic(panic_later, panic_param);
 446         profile_init();
 447         local_irq_enable();
 448 #ifdef CONFIG_BLK_DEV_INITRD
 449         if (initrd_start && !initrd_below_start_ok &&
 450                         initrd_start < min_low_pfn << PAGE_SHIFT) {
 451                 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
 452                     "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
 453                 initrd_start = 0;
 454         }
 455 #endif
 456         mem_init();
 457         kmem_cache_init();
 458         if (late_time_init)
 459                 late_time_init();
 460         calibrate_delay();
 461         pidmap_init();
 462         pgtable_cache_init();
 463         pte_chain_init();
 464 #ifdef CONFIG_X86
 465         if (efi_enabled)
 466                 efi_enter_virtual_mode();
 467 #endif
 468         fork_init(num_physpages);
 469         proc_caches_init();
 470         buffer_init();
 471         unnamed_dev_init();
 472         security_scaffolding_startup();
 473         vfs_caches_init(num_physpages);
 474         radix_tree_init();
 475         signals_init();
 476         /* rootfs populating might need page-writeback */
 477         page_writeback_init();
 478         populate_rootfs();
 479 #ifdef CONFIG_PROC_FS
 480         proc_root_init();
 481 #endif
 482         check_bugs();
 483         printk("POSIX conformance testing by UNIFIX\n");
 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         node_nr_running_init();
 575         spawn_ksoftirqd();
 576 }
 577
 578 static void run_init_process(char *init_filename)
 579 {
 580         argv_init[0] = init_filename;
 581         execve(init_filename, argv_init, envp_init);
 582 }
 583
 584 extern void prepare_namespace(void);
 585
 586 static int init(void * unused)
 587 {
 588         lock_kernel();
 589         /*
 590          * Tell the world that we're going to be the grim
 591          * reaper of innocent orphaned children.
 592          *
 593          * We don't want people to have to make incorrect
 594          * assumptions about where in the task array this
 595          * can be found.
 596          */
 597         child_reaper = current;
 598
 599         /* Sets up cpus_possible() */
 600         smp_prepare_cpus(max_cpus);
 601
 602         do_pre_smp_initcalls();
 603
 604         smp_init();
 605         do_basic_setup();
 606
 607         prepare_namespace();
 608
 609         /*
 610          * Ok, we have completed the initial bootup, and
 611          * we're essentially up and running. Get rid of the
 612          * initmem segments and start the user-mode stuff..
 613          */
 614         free_initmem();
 615         unlock_kernel();
 616         system_running = 1;
 617
 618         if (sys_open("/dev/console", O_RDWR, 0) < 0)
 619                 printk("Warning: unable to open an initial console.\n");
 620
 621         (void) sys_dup(0);
 622         (void) sys_dup(0);
 623
 624         /*
 625          * We try each of these until one succeeds.
 626          *
 627          * The Bourne shell can be used instead of init if we are
 628          * trying to recover a really broken machine.
 629          */
 630
 631         if (execute_command)
 632                 run_init_process(execute_command);
 633
 634         run_init_process("/sbin/init");
 635         run_init_process("/etc/init");
 636         run_init_process("/bin/init");
 637         run_init_process("/bin/sh");
 638
 639         panic("No init found.  Try passing init= option to kernel.");
 640 }