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 #include <linux/console.h>
  26
  27 #define PANIC_TIMER_STEP 100
  28 #define PANIC_BLINK_SPD 18
  29
  30 /* Machine specific panic information string */
  31 char *mach_panic_string;
  32
  33 int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE;
  34 static unsigned long tainted_mask;
  35 static int pause_on_oops;
  36 static int pause_on_oops_flag;
  37 static DEFINE_SPINLOCK(pause_on_oops_lock);
  38
  39 #ifndef CONFIG_PANIC_TIMEOUT
  40 #define CONFIG_PANIC_TIMEOUT 0
  41 #endif
  42 int panic_timeout = CONFIG_PANIC_TIMEOUT;
  43 EXPORT_SYMBOL_GPL(panic_timeout);
  44
  45 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
  46
  47 EXPORT_SYMBOL(panic_notifier_list);
  48
  49 static long no_blink(int state)
  50 {
  51         return 0;
  52 }
  53
  54 /* Returns how long it waited in ms */
  55 long (*panic_blink)(int state);
  56 EXPORT_SYMBOL(panic_blink);
  57
  58 /*
  59  * Stop ourself in panic -- architecture code may override this
  60  */
  61 void __weak panic_smp_self_stop(void)
  62 {
  63         while (1)
  64                 cpu_relax();
  65 }
  66
  67 /**
  68  *      panic - halt the system
  69  *      @fmt: The text string to print
  70  *
  71  *      Display a message, then perform cleanups.
  72  *
  73  *      This function never returns.
  74  */
  75 void panic(const char *fmt, ...)
  76 {
  77         static DEFINE_SPINLOCK(panic_lock);
  78         static char buf[1024];
  79         va_list args;
  80         long i, i_next = 0;
  81         int state = 0;
  82
  83         /*
  84          * Disable local interrupts. This will prevent panic_smp_self_stop
  85          * from deadlocking the first cpu that invokes the panic, since
  86          * there is nothing to prevent an interrupt handler (that runs
  87          * after the panic_lock is acquired) from invoking panic again.
  88          */
  89         local_irq_disable();
  90
  91         /*
  92          * It's possible to come here directly from a panic-assertion and
  93          * not have preempt disabled. Some functions called from here want
  94          * preempt to be disabled. No point enabling it later though...
  95          *
  96          * Only one CPU is allowed to execute the panic code from here. For
  97          * multiple parallel invocations of panic, all other CPUs either
  98          * stop themself or will wait until they are stopped by the 1st CPU
  99          * with smp_send_stop().
 100          */
 101         if (!spin_trylock(&panic_lock))
 102                 panic_smp_self_stop();
 103
 104         console_verbose();
 105         bust_spinlocks(1);
 106         va_start(args, fmt);
 107         vsnprintf(buf, sizeof(buf), fmt, args);
 108         va_end(args);
 109         printk(KERN_EMERG "Kernel panic - not syncing: %s\n",buf);
 110 #ifdef CONFIG_DEBUG_BUGVERBOSE
 111         /*
 112          * Avoid nested stack-dumping if a panic occurs during oops processing
 113          */
 114         if (!test_taint(TAINT_DIE) && oops_in_progress <= 1)
 115                 dump_stack();
 116 #endif
 117
 118         /*
 119          * If we have crashed and we have a crash kernel loaded let it handle
 120          * everything else.
 121          * Do we want to call this before we try to display a message?
 122          */
 123         crash_kexec(NULL);
 124
 125         /*
 126          * Note smp_send_stop is the usual smp shutdown function, which
 127          * unfortunately means it may not be hardened to work in a panic
 128          * situation.
 129          */
 130         smp_send_stop();
 131
 132         kmsg_dump(KMSG_DUMP_PANIC);
 133
 134         atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
 135
 136         bust_spinlocks(0);
 137
 138         console_flush_on_panic();
 139
 140         if (!panic_blink)
 141                 panic_blink = no_blink;
 142
 143         if (panic_timeout > 0) {
 144                 /*
 145                  * Delay timeout seconds before rebooting the machine.
 146                  * We can't use the "normal" timers since we just panicked.
 147                  */
 148                 printk(KERN_EMERG "Rebooting in %d seconds..", panic_timeout);
 149
 150                 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
 151                         touch_nmi_watchdog();
 152                         if (i >= i_next) {
 153                                 i += panic_blink(state ^= 1);
 154                                 i_next = i + 3600 / PANIC_BLINK_SPD;
 155                         }
 156                         mdelay(PANIC_TIMER_STEP);
 157                 }
 158         }
 159         if (panic_timeout != 0) {
 160                 /*
 161                  * This will not be a clean reboot, with everything
 162                  * shutting down.  But if there is a chance of
 163                  * rebooting the system it will be rebooted.
 164                  */
 165                 emergency_restart();
 166         }
 167 #ifdef __sparc__
 168         {
 169                 extern int stop_a_enabled;
 170                 /* Make sure the user can actually press Stop-A (L1-A) */
 171                 stop_a_enabled = 1;
 172                 printk(KERN_EMERG "Press Stop-A (L1-A) to return to the boot prom\n");
 173         }
 174 #endif
 175 #if defined(CONFIG_S390)
 176         {
 177                 unsigned long caller;
 178
 179                 caller = (unsigned long)__builtin_return_address(0);
 180                 disabled_wait(caller);
 181         }
 182 #endif
 183         local_irq_enable();
 184         for (i = 0; ; i += PANIC_TIMER_STEP) {
 185                 touch_softlockup_watchdog();
 186                 if (i >= i_next) {
 187                         i += panic_blink(state ^= 1);
 188                         i_next = i + 3600 / PANIC_BLINK_SPD;
 189                 }
 190                 mdelay(PANIC_TIMER_STEP);
 191         }
 192 }
 193
 194 EXPORT_SYMBOL(panic);
 195
 196
 197 struct tnt {
 198         u8      bit;
 199         char    true;
 200         char    false;
 201 };
 202
 203 static const struct tnt tnts[] = {
 204         { TAINT_PROPRIETARY_MODULE,     'P', 'G' },
 205         { TAINT_FORCED_MODULE,          'F', ' ' },
 206         { TAINT_UNSAFE_SMP,             'S', ' ' },
 207         { TAINT_FORCED_RMMOD,           'R', ' ' },
 208         { TAINT_MACHINE_CHECK,          'M', ' ' },
 209         { TAINT_BAD_PAGE,               'B', ' ' },
 210         { TAINT_USER,                   'U', ' ' },
 211         { TAINT_DIE,                    'D', ' ' },
 212         { TAINT_OVERRIDDEN_ACPI_TABLE,  'A', ' ' },
 213         { TAINT_WARN,                   'W', ' ' },
 214         { TAINT_CRAP,                   'C', ' ' },
 215         { TAINT_FIRMWARE_WORKAROUND,    'I', ' ' },
 216         { TAINT_OOT_MODULE,             'O', ' ' },
 217 };
 218
 219 /**
 220  *      print_tainted - return a string to represent the kernel taint state.
 221  *
 222  *  'P' - Proprietary module has been loaded.
 223  *  'F' - Module has been forcibly loaded.
 224  *  'S' - SMP with CPUs not designed for SMP.
 225  *  'R' - User forced a module unload.
 226  *  'M' - System experienced a machine check exception.
 227  *  'B' - System has hit bad_page.
 228  *  'U' - Userspace-defined naughtiness.
 229  *  'D' - Kernel has oopsed before
 230  *  'A' - ACPI table overridden.
 231  *  'W' - Taint on warning.
 232  *  'C' - modules from drivers/staging are loaded.
 233  *  'I' - Working around severe firmware bug.
 234  *  'O' - Out-of-tree module has been loaded.
 235  *
 236  *      The string is overwritten by the next call to print_tainted().
 237  */
 238 const char *print_tainted(void)
 239 {
 240         static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
 241
 242         if (tainted_mask) {
 243                 char *s;
 244                 int i;
 245
 246                 s = buf + sprintf(buf, "Tainted: ");
 247                 for (i = 0; i < ARRAY_SIZE(tnts); i++) {
 248                         const struct tnt *t = &tnts[i];
 249                         *s++ = test_bit(t->bit, &tainted_mask) ?
 250                                         t->true : t->false;
 251                 }
 252                 *s = 0;
 253         } else
 254                 snprintf(buf, sizeof(buf), "Not tainted");
 255
 256         return buf;
 257 }
 258
 259 int test_taint(unsigned flag)
 260 {
 261         return test_bit(flag, &tainted_mask);
 262 }
 263 EXPORT_SYMBOL(test_taint);
 264
 265 unsigned long get_taint(void)
 266 {
 267         return tainted_mask;
 268 }
 269
 270 /**
 271  * add_taint: add a taint flag if not already set.
 272  * @flag: one of the TAINT_* constants.
 273  * @lockdep_ok: whether lock debugging is still OK.
 274  *
 275  * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
 276  * some notewortht-but-not-corrupting cases, it can be set to true.
 277  */
 278 void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
 279 {
 280         if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
 281                 printk(KERN_WARNING
 282                        "Disabling lock debugging due to kernel taint\n");
 283
 284         set_bit(flag, &tainted_mask);
 285 }
 286 EXPORT_SYMBOL(add_taint);
 287
 288 static void spin_msec(int msecs)
 289 {
 290         int i;
 291
 292         for (i = 0; i < msecs; i++) {
 293                 touch_nmi_watchdog();
 294                 mdelay(1);
 295         }
 296 }
 297
 298 /*
 299  * It just happens that oops_enter() and oops_exit() are identically
 300  * implemented...
 301  */
 302 static void do_oops_enter_exit(void)
 303 {
 304         unsigned long flags;
 305         static int spin_counter;
 306
 307         if (!pause_on_oops)
 308                 return;
 309
 310         spin_lock_irqsave(&pause_on_oops_lock, flags);
 311         if (pause_on_oops_flag == 0) {
 312                 /* This CPU may now print the oops message */
 313                 pause_on_oops_flag = 1;
 314         } else {
 315                 /* We need to stall this CPU */
 316                 if (!spin_counter) {
 317                         /* This CPU gets to do the counting */
 318                         spin_counter = pause_on_oops;
 319                         do {
 320                                 spin_unlock(&pause_on_oops_lock);
 321                                 spin_msec(MSEC_PER_SEC);
 322                                 spin_lock(&pause_on_oops_lock);
 323                         } while (--spin_counter);
 324                         pause_on_oops_flag = 0;
 325                 } else {
 326                         /* This CPU waits for a different one */
 327                         while (spin_counter) {
 328                                 spin_unlock(&pause_on_oops_lock);
 329                                 spin_msec(1);
 330                                 spin_lock(&pause_on_oops_lock);
 331                         }
 332                 }
 333         }
 334         spin_unlock_irqrestore(&pause_on_oops_lock, flags);
 335 }
 336
 337 /*
 338  * Return true if the calling CPU is allowed to print oops-related info.
 339  * This is a bit racy..
 340  */
 341 int oops_may_print(void)
 342 {
 343         return pause_on_oops_flag == 0;
 344 }
 345
 346 /*
 347  * Called when the architecture enters its oops handler, before it prints
 348  * anything.  If this is the first CPU to oops, and it's oopsing the first
 349  * time then let it proceed.
 350  *
 351  * This is all enabled by the pause_on_oops kernel boot option.  We do all
 352  * this to ensure that oopses don't scroll off the screen.  It has the
 353  * side-effect of preventing later-oopsing CPUs from mucking up the display,
 354  * too.
 355  *
 356  * It turns out that the CPU which is allowed to print ends up pausing for
 357  * the right duration, whereas all the other CPUs pause for twice as long:
 358  * once in oops_enter(), once in oops_exit().
 359  */
 360 void oops_enter(void)
 361 {
 362         tracing_off();
 363         /* can't trust the integrity of the kernel anymore: */
 364         debug_locks_off();
 365         do_oops_enter_exit();
 366 }
 367
 368 /*
 369  * 64-bit random ID for oopses:
 370  */
 371 static u64 oops_id;
 372
 373 static int init_oops_id(void)
 374 {
 375         if (!oops_id)
 376                 get_random_bytes(&oops_id, sizeof(oops_id));
 377         else
 378                 oops_id++;
 379
 380         return 0;
 381 }
 382 late_initcall(init_oops_id);
 383
 384 void print_oops_end_marker(void)
 385 {
 386         init_oops_id();
 387
 388         if (mach_panic_string)
 389                 printk(KERN_WARNING "Board Information: %s\n",
 390                        mach_panic_string);
 391
 392         printk(KERN_WARNING "---[ end trace %016llx ]---\n",
 393                 (unsigned long long)oops_id);
 394 }
 395
 396 /*
 397  * Called when the architecture exits its oops handler, after printing
 398  * everything.
 399  */
 400 void oops_exit(void)
 401 {
 402         do_oops_enter_exit();
 403         print_oops_end_marker();
 404         kmsg_dump(KMSG_DUMP_OOPS);
 405 }
 406
 407 #ifdef WANT_WARN_ON_SLOWPATH
 408 struct slowpath_args {
 409         const char *fmt;
 410         va_list args;
 411 };
 412
 413 static void warn_slowpath_common(const char *file, int line, void *caller,
 414                                  unsigned taint, struct slowpath_args *args)
 415 {
 416         printk(KERN_WARNING "------------[ cut here ]------------\n");
 417         printk(KERN_WARNING "WARNING: at %s:%d %pS()\n", file, line, caller);
 418
 419         if (args)
 420                 vprintk(args->fmt, args->args);
 421
 422         print_modules();
 423         dump_stack();
 424         print_oops_end_marker();
 425         /* Just a warning, don't kill lockdep. */
 426         add_taint(taint, LOCKDEP_STILL_OK);
 427 }
 428
 429 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
 430 {
 431         struct slowpath_args args;
 432
 433         args.fmt = fmt;
 434         va_start(args.args, fmt);
 435         warn_slowpath_common(file, line, __builtin_return_address(0),
 436                              TAINT_WARN, &args);
 437         va_end(args.args);
 438 }
 439 EXPORT_SYMBOL(warn_slowpath_fmt);
 440
 441 void warn_slowpath_fmt_taint(const char *file, int line,
 442                              unsigned taint, const char *fmt, ...)
 443 {
 444         struct slowpath_args args;
 445
 446         args.fmt = fmt;
 447         va_start(args.args, fmt);
 448         warn_slowpath_common(file, line, __builtin_return_address(0),
 449                              taint, &args);
 450         va_end(args.args);
 451 }
 452 EXPORT_SYMBOL(warn_slowpath_fmt_taint);
 453
 454 void warn_slowpath_null(const char *file, int line)
 455 {
 456         warn_slowpath_common(file, line, __builtin_return_address(0),
 457                              TAINT_WARN, NULL);
 458 }
 459 EXPORT_SYMBOL(warn_slowpath_null);
 460 #endif
 461
 462 #ifdef CONFIG_CC_STACKPROTECTOR
 463
 464 /*
 465  * Called when gcc's -fstack-protector feature is used, and
 466  * gcc detects corruption of the on-stack canary value
 467  */
 468 void __stack_chk_fail(void)
 469 {
 470 /*
 471         panic("stack-protector: Kernel stack is corrupted in: %p\n",
 472                 __builtin_return_address(0));
 473 */
 474     BUG();
 475     printk(KERN_ERR "stack-protector: Kernel stack is corrupted in: %p\n",
 476                 __builtin_return_address(0));
 477 }
 478 EXPORT_SYMBOL(__stack_chk_fail);
 479
 480 #endif
 481
 482 core_param(panic, panic_timeout, int, 0644);
 483 core_param(pause_on_oops, pause_on_oops, int, 0644);
 484
 485 static int __init oops_setup(char *s)
 486 {
 487         if (!s)
 488                 return -EINVAL;
 489         if (!strcmp(s, "panic"))
 490                 panic_on_oops = 1;
 491         return 0;
 492 }
 493 early_param("oops", oops_setup);