kernel/panic.c

   1 /*
   2  *  linux/kernel/panic.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 /*
   8  * This function is used through-out the kernel (including mm and fs)
   9  * to indicate a major problem.
  10  */
  11 #include <linux/debug_locks.h>
  12 #include <linux/interrupt.h>
  13 #include <linux/kmsg_dump.h>
  14 #include <linux/kallsyms.h>
  15 #include <linux/notifier.h>
  16 #include <linux/module.h>
  17 #include <linux/random.h>
  18 #include <linux/reboot.h>
  19 #include <linux/delay.h>
  20 #include <linux/kexec.h>
  21 #include <linux/sched.h>
  22 #include <linux/sysrq.h>
  23 #include <linux/init.h>
  24 #include <linux/nmi.h>
  25
  26 #define PANIC_TIMER_STEP 100
  27 #define PANIC_BLINK_SPD 18
  28
  29 /* Machine specific panic information string */
  30 char *mach_panic_string;
  31
  32 int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE;
  33 static unsigned long tainted_mask;
  34 static int pause_on_oops;
  35 static int pause_on_oops_flag;
  36 static DEFINE_SPINLOCK(pause_on_oops_lock);
  37
  38 #ifndef CONFIG_PANIC_TIMEOUT
  39 #define CONFIG_PANIC_TIMEOUT 0
  40 #endif
  41 int panic_timeout = CONFIG_PANIC_TIMEOUT;
  42 EXPORT_SYMBOL_GPL(panic_timeout);
  43
  44 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
  45
  46 EXPORT_SYMBOL(panic_notifier_list);
  47
  48 static long no_blink(int state)
  49 {
  50         return 0;
  51 }
  52
  53 /* Returns how long it waited in ms */
  54 long (*panic_blink)(int state);
  55 EXPORT_SYMBOL(panic_blink);
  56
  57 /*
  58  * Stop ourself in panic -- architecture code may override this
  59  */
  60 void __weak panic_smp_self_stop(void)
  61 {
  62         while (1)
  63                 cpu_relax();
  64 }
  65
  66 /**
  67  *      panic - halt the system
  68  *      @fmt: The text string to print
  69  *
  70  *      Display a message, then perform cleanups.
  71  *
  72  *      This function never returns.
  73  */
  74 void panic(const char *fmt, ...)
  75 {
  76         static DEFINE_SPINLOCK(panic_lock);
  77         static char buf[1024];
  78         va_list args;
  79         long i, i_next = 0;
  80         int state = 0;
  81
  82         /*
  83          * Disable local interrupts. This will prevent panic_smp_self_stop
  84          * from deadlocking the first cpu that invokes the panic, since
  85          * there is nothing to prevent an interrupt handler (that runs
  86          * after the panic_lock is acquired) from invoking panic again.
  87          */
  88         local_irq_disable();
  89
  90         /*
  91          * It's possible to come here directly from a panic-assertion and
  92          * not have preempt disabled. Some functions called from here want
  93          * preempt to be disabled. No point enabling it later though...
  94          *
  95          * Only one CPU is allowed to execute the panic code from here. For
  96          * multiple parallel invocations of panic, all other CPUs either
  97          * stop themself or will wait until they are stopped by the 1st CPU
  98          * with smp_send_stop().
  99          */
 100         if (!spin_trylock(&panic_lock))
 101                 panic_smp_self_stop();
 102
 103         console_verbose();
 104         bust_spinlocks(1);
 105         va_start(args, fmt);
 106         vsnprintf(buf, sizeof(buf), fmt, args);
 107         va_end(args);
 108         printk(KERN_EMERG "Kernel panic - not syncing: %s\n",buf);
 109 #ifdef CONFIG_DEBUG_BUGVERBOSE
 110         /*
 111          * Avoid nested stack-dumping if a panic occurs during oops processing
 112          */
 113         if (!test_taint(TAINT_DIE) && oops_in_progress <= 1)
 114                 dump_stack();
 115 #endif
 116
 117         /*
 118          * If we have crashed and we have a crash kernel loaded let it handle
 119          * everything else.
 120          * Do we want to call this before we try to display a message?
 121          */
 122         crash_kexec(NULL);
 123
 124         /*
 125          * Note smp_send_stop is the usual smp shutdown function, which
 126          * unfortunately means it may not be hardened to work in a panic
 127          * situation.
 128          */
 129         smp_send_stop();
 130
 131         kmsg_dump(KMSG_DUMP_PANIC);
 132
 133         atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
 134
 135         bust_spinlocks(0);
 136
 137         if (!panic_blink)
 138                 panic_blink = no_blink;
 139
 140         if (panic_timeout > 0) {
 141                 /*
 142                  * Delay timeout seconds before rebooting the machine.
 143                  * We can't use the "normal" timers since we just panicked.
 144                  */
 145                 printk(KERN_EMERG "Rebooting in %d seconds..", panic_timeout);
 146
 147                 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
 148                         touch_nmi_watchdog();
 149                         if (i >= i_next) {
 150                                 i += panic_blink(state ^= 1);
 151                                 i_next = i + 3600 / PANIC_BLINK_SPD;
 152                         }
 153                         mdelay(PANIC_TIMER_STEP);
 154                 }
 155         }
 156         if (panic_timeout != 0) {
 157                 /*
 158                  * This will not be a clean reboot, with everything
 159                  * shutting down.  But if there is a chance of
 160                  * rebooting the system it will be rebooted.
 161                  */
 162                 emergency_restart();
 163         }
 164 #ifdef __sparc__
 165         {
 166                 extern int stop_a_enabled;
 167                 /* Make sure the user can actually press Stop-A (L1-A) */
 168                 stop_a_enabled = 1;
 169                 printk(KERN_EMERG "Press Stop-A (L1-A) to return to the boot prom\n");
 170         }
 171 #endif
 172 #if defined(CONFIG_S390)
 173         {
 174                 unsigned long caller;
 175
 176                 caller = (unsigned long)__builtin_return_address(0);
 177                 disabled_wait(caller);
 178         }
 179 #endif
 180         local_irq_enable();
 181         for (i = 0; ; i += PANIC_TIMER_STEP) {
 182                 touch_softlockup_watchdog();
 183                 if (i >= i_next) {
 184                         i += panic_blink(state ^= 1);
 185                         i_next = i + 3600 / PANIC_BLINK_SPD;
 186                 }
 187                 mdelay(PANIC_TIMER_STEP);
 188         }
 189 }
 190
 191 EXPORT_SYMBOL(panic);
 192
 193
 194 struct tnt {
 195         u8      bit;
 196         char    true;
 197         char    false;
 198 };
 199
 200 static const struct tnt tnts[] = {
 201         { TAINT_PROPRIETARY_MODULE,     'P', 'G' },
 202         { TAINT_FORCED_MODULE,          'F', ' ' },
 203         { TAINT_UNSAFE_SMP,             'S', ' ' },
 204         { TAINT_FORCED_RMMOD,           'R', ' ' },
 205         { TAINT_MACHINE_CHECK,          'M', ' ' },
 206         { TAINT_BAD_PAGE,               'B', ' ' },
 207         { TAINT_USER,                   'U', ' ' },
 208         { TAINT_DIE,                    'D', ' ' },
 209         { TAINT_OVERRIDDEN_ACPI_TABLE,  'A', ' ' },
 210         { TAINT_WARN,                   'W', ' ' },
 211         { TAINT_CRAP,                   'C', ' ' },
 212         { TAINT_FIRMWARE_WORKAROUND,    'I', ' ' },
 213         { TAINT_OOT_MODULE,             'O', ' ' },
 214 };
 215
 216 /**
 217  *      print_tainted - return a string to represent the kernel taint state.
 218  *
 219  *  'P' - Proprietary module has been loaded.
 220  *  'F' - Module has been forcibly loaded.
 221  *  'S' - SMP with CPUs not designed for SMP.
 222  *  'R' - User forced a module unload.
 223  *  'M' - System experienced a machine check exception.
 224  *  'B' - System has hit bad_page.
 225  *  'U' - Userspace-defined naughtiness.
 226  *  'D' - Kernel has oopsed before
 227  *  'A' - ACPI table overridden.
 228  *  'W' - Taint on warning.
 229  *  'C' - modules from drivers/staging are loaded.
 230  *  'I' - Working around severe firmware bug.
 231  *  'O' - Out-of-tree module has been loaded.
 232  *
 233  *      The string is overwritten by the next call to print_tainted().
 234  */
 235 const char *print_tainted(void)
 236 {
 237         static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
 238
 239         if (tainted_mask) {
 240                 char *s;
 241                 int i;
 242
 243                 s = buf + sprintf(buf, "Tainted: ");
 244                 for (i = 0; i < ARRAY_SIZE(tnts); i++) {
 245                         const struct tnt *t = &tnts[i];
 246                         *s++ = test_bit(t->bit, &tainted_mask) ?
 247                                         t->true : t->false;
 248                 }
 249                 *s = 0;
 250         } else
 251                 snprintf(buf, sizeof(buf), "Not tainted");
 252
 253         return buf;
 254 }
 255
 256 int test_taint(unsigned flag)
 257 {
 258         return test_bit(flag, &tainted_mask);
 259 }
 260 EXPORT_SYMBOL(test_taint);
 261
 262 unsigned long get_taint(void)
 263 {
 264         return tainted_mask;
 265 }
 266
 267 /**
 268  * add_taint: add a taint flag if not already set.
 269  * @flag: one of the TAINT_* constants.
 270  * @lockdep_ok: whether lock debugging is still OK.
 271  *
 272  * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
 273  * some notewortht-but-not-corrupting cases, it can be set to true.
 274  */
 275 void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
 276 {
 277         if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
 278                 printk(KERN_WARNING
 279                        "Disabling lock debugging due to kernel taint\n");
 280
 281         set_bit(flag, &tainted_mask);
 282 }
 283 EXPORT_SYMBOL(add_taint);
 284
 285 static void spin_msec(int msecs)
 286 {
 287         int i;
 288
 289         for (i = 0; i < msecs; i++) {
 290                 touch_nmi_watchdog();
 291                 mdelay(1);
 292         }
 293 }
 294
 295 /*
 296  * It just happens that oops_enter() and oops_exit() are identically
 297  * implemented...
 298  */
 299 static void do_oops_enter_exit(void)
 300 {
 301         unsigned long flags;
 302         static int spin_counter;
 303
 304         if (!pause_on_oops)
 305                 return;
 306
 307         spin_lock_irqsave(&pause_on_oops_lock, flags);
 308         if (pause_on_oops_flag == 0) {
 309                 /* This CPU may now print the oops message */
 310                 pause_on_oops_flag = 1;
 311         } else {
 312                 /* We need to stall this CPU */
 313                 if (!spin_counter) {
 314                         /* This CPU gets to do the counting */
 315                         spin_counter = pause_on_oops;
 316                         do {
 317                                 spin_unlock(&pause_on_oops_lock);
 318                                 spin_msec(MSEC_PER_SEC);
 319                                 spin_lock(&pause_on_oops_lock);
 320                         } while (--spin_counter);
 321                         pause_on_oops_flag = 0;
 322                 } else {
 323                         /* This CPU waits for a different one */
 324                         while (spin_counter) {
 325                                 spin_unlock(&pause_on_oops_lock);
 326                                 spin_msec(1);
 327                                 spin_lock(&pause_on_oops_lock);
 328                         }
 329                 }
 330         }
 331         spin_unlock_irqrestore(&pause_on_oops_lock, flags);
 332 }
 333
 334 /*
 335  * Return true if the calling CPU is allowed to print oops-related info.
 336  * This is a bit racy..
 337  */
 338 int oops_may_print(void)
 339 {
 340         return pause_on_oops_flag == 0;
 341 }
 342
 343 /*
 344  * Called when the architecture enters its oops handler, before it prints
 345  * anything.  If this is the first CPU to oops, and it's oopsing the first
 346  * time then let it proceed.
 347  *
 348  * This is all enabled by the pause_on_oops kernel boot option.  We do all
 349  * this to ensure that oopses don't scroll off the screen.  It has the
 350  * side-effect of preventing later-oopsing CPUs from mucking up the display,
 351  * too.
 352  *
 353  * It turns out that the CPU which is allowed to print ends up pausing for
 354  * the right duration, whereas all the other CPUs pause for twice as long:
 355  * once in oops_enter(), once in oops_exit().
 356  */
 357 void oops_enter(void)
 358 {
 359         tracing_off();
 360         /* can't trust the integrity of the kernel anymore: */
 361         debug_locks_off();
 362         do_oops_enter_exit();
 363 }
 364
 365 /*
 366  * 64-bit random ID for oopses:
 367  */
 368 static u64 oops_id;
 369
 370 static int init_oops_id(void)
 371 {
 372         if (!oops_id)
 373                 get_random_bytes(&oops_id, sizeof(oops_id));
 374         else
 375                 oops_id++;
 376
 377         return 0;
 378 }
 379 late_initcall(init_oops_id);
 380
 381 void print_oops_end_marker(void)
 382 {
 383         init_oops_id();
 384
 385         if (mach_panic_string)
 386                 printk(KERN_WARNING "Board Information: %s\n",
 387                        mach_panic_string);
 388
 389         printk(KERN_WARNING "---[ end trace %016llx ]---\n",
 390                 (unsigned long long)oops_id);
 391 }
 392
 393 /*
 394  * Called when the architecture exits its oops handler, after printing
 395  * everything.
 396  */
 397 void oops_exit(void)
 398 {
 399         do_oops_enter_exit();
 400         print_oops_end_marker();
 401         kmsg_dump(KMSG_DUMP_OOPS);
 402 }
 403
 404 #ifdef WANT_WARN_ON_SLOWPATH
 405 struct slowpath_args {
 406         const char *fmt;
 407         va_list args;
 408 };
 409
 410 static void warn_slowpath_common(const char *file, int line, void *caller,
 411                                  unsigned taint, struct slowpath_args *args)
 412 {
 413         printk(KERN_WARNING "------------[ cut here ]------------\n");
 414         printk(KERN_WARNING "WARNING: at %s:%d %pS()\n", file, line, caller);
 415
 416         if (args)
 417                 vprintk(args->fmt, args->args);
 418
 419         print_modules();
 420         dump_stack();
 421         print_oops_end_marker();
 422         /* Just a warning, don't kill lockdep. */
 423         add_taint(taint, LOCKDEP_STILL_OK);
 424 }
 425
 426 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
 427 {
 428         struct slowpath_args args;
 429
 430         args.fmt = fmt;
 431         va_start(args.args, fmt);
 432         warn_slowpath_common(file, line, __builtin_return_address(0),
 433                              TAINT_WARN, &args);
 434         va_end(args.args);
 435 }
 436 EXPORT_SYMBOL(warn_slowpath_fmt);
 437
 438 void warn_slowpath_fmt_taint(const char *file, int line,
 439                              unsigned taint, const char *fmt, ...)
 440 {
 441         struct slowpath_args args;
 442
 443         args.fmt = fmt;
 444         va_start(args.args, fmt);
 445         warn_slowpath_common(file, line, __builtin_return_address(0),
 446                              taint, &args);
 447         va_end(args.args);
 448 }
 449 EXPORT_SYMBOL(warn_slowpath_fmt_taint);
 450
 451 void warn_slowpath_null(const char *file, int line)
 452 {
 453         warn_slowpath_common(file, line, __builtin_return_address(0),
 454                              TAINT_WARN, NULL);
 455 }
 456 EXPORT_SYMBOL(warn_slowpath_null);
 457 #endif
 458
 459 #ifdef CONFIG_CC_STACKPROTECTOR
 460
 461 /*
 462  * Called when gcc's -fstack-protector feature is used, and
 463  * gcc detects corruption of the on-stack canary value
 464  */
 465 void __stack_chk_fail(void)
 466 {
 467 /*
 468         panic("stack-protector: Kernel stack is corrupted in: %p\n",
 469                 __builtin_return_address(0));
 470 */
 471     BUG();
 472     printk(KERN_ERR "stack-protector: Kernel stack is corrupted in: %p\n",
 473                 __builtin_return_address(0));
 474 }
 475 EXPORT_SYMBOL(__stack_chk_fail);
 476
 477 #endif
 478
 479 core_param(panic, panic_timeout, int, 0644);
 480 core_param(pause_on_oops, pause_on_oops, int, 0644);
 481
 482 static int __init oops_setup(char *s)
 483 {
 484         if (!s)
 485                 return -EINVAL;
 486         if (!strcmp(s, "panic"))
 487                 panic_on_oops = 1;
 488         return 0;
 489 }
 490 early_param("oops", oops_setup);