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