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