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/ftrace.h>
  19 #include <linux/reboot.h>
  20 #include <linux/delay.h>
  21 #include <linux/kexec.h>
  22 #include <linux/sched.h>
  23 #include <linux/sysrq.h>
  24 #include <linux/init.h>
  25 #include <linux/nmi.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 static bool crash_kexec_post_notifiers;
  36
  37 int panic_timeout = CONFIG_PANIC_TIMEOUT;
  38 EXPORT_SYMBOL_GPL(panic_timeout);
  39
  40 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
  41
  42 EXPORT_SYMBOL(panic_notifier_list);
  43
  44 static long no_blink(int state)
  45 {
  46         return 0;
  47 }
  48
  49 /* Returns how long it waited in ms */
  50 long (*panic_blink)(int state);
  51 EXPORT_SYMBOL(panic_blink);
  52
  53 /*
  54  * Stop ourself in panic -- architecture code may override this
  55  */
  56 void __weak panic_smp_self_stop(void)
  57 {
  58         while (1)
  59                 cpu_relax();
  60 }
  61
  62 /**
  63  *      panic - halt the system
  64  *      @fmt: The text string to print
  65  *
  66  *      Display a message, then perform cleanups.
  67  *
  68  *      This function never returns.
  69  */
  70 void panic(const char *fmt, ...)
  71 {
  72         static DEFINE_SPINLOCK(panic_lock);
  73         static char buf[1024];
  74         va_list args;
  75         long i, i_next = 0;
  76         int state = 0;
  77
  78         /*
  79          * Disable local interrupts. This will prevent panic_smp_self_stop
  80          * from deadlocking the first cpu that invokes the panic, since
  81          * there is nothing to prevent an interrupt handler (that runs
  82          * after the panic_lock is acquired) from invoking panic again.
  83          */
  84         local_irq_disable();
  85
  86         /*
  87          * It's possible to come here directly from a panic-assertion and
  88          * not have preempt disabled. Some functions called from here want
  89          * preempt to be disabled. No point enabling it later though...
  90          *
  91          * Only one CPU is allowed to execute the panic code from here. For
  92          * multiple parallel invocations of panic, all other CPUs either
  93          * stop themself or will wait until they are stopped by the 1st CPU
  94          * with smp_send_stop().
  95          */
  96         if (!spin_trylock(&panic_lock))
  97                 panic_smp_self_stop();
  98
  99         console_verbose();
 100         bust_spinlocks(1);
 101         va_start(args, fmt);
 102         vsnprintf(buf, sizeof(buf), fmt, args);
 103         va_end(args);
 104         pr_emerg("Kernel panic - not syncing: %s\n", buf);
 105 #ifdef CONFIG_DEBUG_BUGVERBOSE
 106         /*
 107          * Avoid nested stack-dumping if a panic occurs during oops processing
 108          */
 109         if (!test_taint(TAINT_DIE) && oops_in_progress <= 1)
 110                 dump_stack();
 111 #endif
 112
 113         /*
 114          * If we have crashed and we have a crash kernel loaded let it handle
 115          * everything else.
 116          * If we want to run this after calling panic_notifiers, pass
 117          * the "crash_kexec_post_notifiers" option to the kernel.
 118          */
 119         if (!crash_kexec_post_notifiers)
 120                 crash_kexec(NULL);
 121
 122         /*
 123          * Note smp_send_stop is the usual smp shutdown function, which
 124          * unfortunately means it may not be hardened to work in a panic
 125          * situation.
 126          */
 127         smp_send_stop();
 128
 129         /*
 130          * Run any panic handlers, including those that might need to
 131          * add information to the kmsg dump output.
 132          */
 133         atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
 134
 135         kmsg_dump(KMSG_DUMP_PANIC);
 136
 137         /*
 138          * If you doubt kdump always works fine in any situation,
 139          * "crash_kexec_post_notifiers" offers you a chance to run
 140          * panic_notifiers and dumping kmsg before kdump.
 141          * Note: since some panic_notifiers can make crashed kernel
 142          * more unstable, it can increase risks of the kdump failure too.
 143          */
 144         crash_kexec(NULL);
 145
 146         bust_spinlocks(0);
 147
 148         if (!panic_blink)
 149                 panic_blink = no_blink;
 150
 151         if (panic_timeout > 0) {
 152                 /*
 153                  * Delay timeout seconds before rebooting the machine.
 154                  * We can't use the "normal" timers since we just panicked.
 155                  */
 156                 pr_emerg("Rebooting in %d seconds..", panic_timeout);
 157
 158                 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
 159                         touch_nmi_watchdog();
 160                         if (i >= i_next) {
 161                                 i += panic_blink(state ^= 1);
 162                                 i_next = i + 3600 / PANIC_BLINK_SPD;
 163                         }
 164                         mdelay(PANIC_TIMER_STEP);
 165                 }
 166         }
 167         if (panic_timeout != 0) {
 168                 /*
 169                  * This will not be a clean reboot, with everything
 170                  * shutting down.  But if there is a chance of
 171                  * rebooting the system it will be rebooted.
 172                  */
 173                 emergency_restart();
 174         }
 175 #ifdef __sparc__
 176         {
 177                 extern int stop_a_enabled;
 178                 /* Make sure the user can actually press Stop-A (L1-A) */
 179                 stop_a_enabled = 1;
 180                 pr_emerg("Press Stop-A (L1-A) to return to the boot prom\n");
 181         }
 182 #endif
 183 #if defined(CONFIG_S390)
 184         {
 185                 unsigned long caller;
 186
 187                 caller = (unsigned long)__builtin_return_address(0);
 188                 disabled_wait(caller);
 189         }
 190 #endif
 191         pr_emerg("---[ end Kernel panic - not syncing: %s\n", buf);
 192         local_irq_enable();
 193         for (i = 0; ; i += PANIC_TIMER_STEP) {
 194                 touch_softlockup_watchdog();
 195                 if (i >= i_next) {
 196                         i += panic_blink(state ^= 1);
 197                         i_next = i + 3600 / PANIC_BLINK_SPD;
 198                 }
 199                 mdelay(PANIC_TIMER_STEP);
 200         }
 201 }
 202
 203 EXPORT_SYMBOL(panic);
 204
 205
 206 struct tnt {
 207         u8      bit;
 208         char    true;
 209         char    false;
 210 };
 211
 212 static const struct tnt tnts[] = {
 213         { TAINT_PROPRIETARY_MODULE,     'P', 'G' },
 214         { TAINT_FORCED_MODULE,          'F', ' ' },
 215         { TAINT_CPU_OUT_OF_SPEC,        'S', ' ' },
 216         { TAINT_FORCED_RMMOD,           'R', ' ' },
 217         { TAINT_MACHINE_CHECK,          'M', ' ' },
 218         { TAINT_BAD_PAGE,               'B', ' ' },
 219         { TAINT_USER,                   'U', ' ' },
 220         { TAINT_DIE,                    'D', ' ' },
 221         { TAINT_OVERRIDDEN_ACPI_TABLE,  'A', ' ' },
 222         { TAINT_WARN,                   'W', ' ' },
 223         { TAINT_CRAP,                   'C', ' ' },
 224         { TAINT_FIRMWARE_WORKAROUND,    'I', ' ' },
 225         { TAINT_OOT_MODULE,             'O', ' ' },
 226         { TAINT_UNSIGNED_MODULE,        'E', ' ' },
 227         { TAINT_SOFTLOCKUP,             'L', ' ' },
 228 };
 229
 230 /**
 231  *      print_tainted - return a string to represent the kernel taint state.
 232  *
 233  *  'P' - Proprietary module has been loaded.
 234  *  'F' - Module has been forcibly loaded.
 235  *  'S' - SMP with CPUs not designed for SMP.
 236  *  'R' - User forced a module unload.
 237  *  'M' - System experienced a machine check exception.
 238  *  'B' - System has hit bad_page.
 239  *  'U' - Userspace-defined naughtiness.
 240  *  'D' - Kernel has oopsed before
 241  *  'A' - ACPI table overridden.
 242  *  'W' - Taint on warning.
 243  *  'C' - modules from drivers/staging are loaded.
 244  *  'I' - Working around severe firmware bug.
 245  *  'O' - Out-of-tree module has been loaded.
 246  *  'E' - Unsigned module has been loaded.
 247  *  'L' - A soft lockup has previously occurred.
 248  *
 249  *      The string is overwritten by the next call to print_tainted().
 250  */
 251 const char *print_tainted(void)
 252 {
 253         static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ")];
 254
 255         if (tainted_mask) {
 256                 char *s;
 257                 int i;
 258
 259                 s = buf + sprintf(buf, "Tainted: ");
 260                 for (i = 0; i < ARRAY_SIZE(tnts); i++) {
 261                         const struct tnt *t = &tnts[i];
 262                         *s++ = test_bit(t->bit, &tainted_mask) ?
 263                                         t->true : t->false;
 264                 }
 265                 *s = 0;
 266         } else
 267                 snprintf(buf, sizeof(buf), "Not tainted");
 268
 269         return buf;
 270 }
 271
 272 int test_taint(unsigned flag)
 273 {
 274         return test_bit(flag, &tainted_mask);
 275 }
 276 EXPORT_SYMBOL(test_taint);
 277
 278 unsigned long get_taint(void)
 279 {
 280         return tainted_mask;
 281 }
 282
 283 /**
 284  * add_taint: add a taint flag if not already set.
 285  * @flag: one of the TAINT_* constants.
 286  * @lockdep_ok: whether lock debugging is still OK.
 287  *
 288  * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
 289  * some notewortht-but-not-corrupting cases, it can be set to true.
 290  */
 291 void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
 292 {
 293         if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
 294                 pr_warn("Disabling lock debugging due to kernel taint\n");
 295
 296         set_bit(flag, &tainted_mask);
 297 }
 298 EXPORT_SYMBOL(add_taint);
 299
 300 static void spin_msec(int msecs)
 301 {
 302         int i;
 303
 304         for (i = 0; i < msecs; i++) {
 305                 touch_nmi_watchdog();
 306                 mdelay(1);
 307         }
 308 }
 309
 310 /*
 311  * It just happens that oops_enter() and oops_exit() are identically
 312  * implemented...
 313  */
 314 static void do_oops_enter_exit(void)
 315 {
 316         unsigned long flags;
 317         static int spin_counter;
 318
 319         if (!pause_on_oops)
 320                 return;
 321
 322         spin_lock_irqsave(&pause_on_oops_lock, flags);
 323         if (pause_on_oops_flag == 0) {
 324                 /* This CPU may now print the oops message */
 325                 pause_on_oops_flag = 1;
 326         } else {
 327                 /* We need to stall this CPU */
 328                 if (!spin_counter) {
 329                         /* This CPU gets to do the counting */
 330                         spin_counter = pause_on_oops;
 331                         do {
 332                                 spin_unlock(&pause_on_oops_lock);
 333                                 spin_msec(MSEC_PER_SEC);
 334                                 spin_lock(&pause_on_oops_lock);
 335                         } while (--spin_counter);
 336                         pause_on_oops_flag = 0;
 337                 } else {
 338                         /* This CPU waits for a different one */
 339                         while (spin_counter) {
 340                                 spin_unlock(&pause_on_oops_lock);
 341                                 spin_msec(1);
 342                                 spin_lock(&pause_on_oops_lock);
 343                         }
 344                 }
 345         }
 346         spin_unlock_irqrestore(&pause_on_oops_lock, flags);
 347 }
 348
 349 /*
 350  * Return true if the calling CPU is allowed to print oops-related info.
 351  * This is a bit racy..
 352  */
 353 int oops_may_print(void)
 354 {
 355         return pause_on_oops_flag == 0;
 356 }
 357
 358 /*
 359  * Called when the architecture enters its oops handler, before it prints
 360  * anything.  If this is the first CPU to oops, and it's oopsing the first
 361  * time then let it proceed.
 362  *
 363  * This is all enabled by the pause_on_oops kernel boot option.  We do all
 364  * this to ensure that oopses don't scroll off the screen.  It has the
 365  * side-effect of preventing later-oopsing CPUs from mucking up the display,
 366  * too.
 367  *
 368  * It turns out that the CPU which is allowed to print ends up pausing for
 369  * the right duration, whereas all the other CPUs pause for twice as long:
 370  * once in oops_enter(), once in oops_exit().
 371  */
 372 void oops_enter(void)
 373 {
 374         tracing_off();
 375         /* can't trust the integrity of the kernel anymore: */
 376         debug_locks_off();
 377         do_oops_enter_exit();
 378 }
 379
 380 /*
 381  * 64-bit random ID for oopses:
 382  */
 383 static u64 oops_id;
 384
 385 static int init_oops_id(void)
 386 {
 387         if (!oops_id)
 388                 get_random_bytes(&oops_id, sizeof(oops_id));
 389         else
 390                 oops_id++;
 391
 392         return 0;
 393 }
 394 late_initcall(init_oops_id);
 395
 396 void print_oops_end_marker(void)
 397 {
 398         init_oops_id();
 399         pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id);
 400 }
 401
 402 /*
 403  * Called when the architecture exits its oops handler, after printing
 404  * everything.
 405  */
 406 void oops_exit(void)
 407 {
 408         do_oops_enter_exit();
 409         print_oops_end_marker();
 410         kmsg_dump(KMSG_DUMP_OOPS);
 411 }
 412
 413 #ifdef WANT_WARN_ON_SLOWPATH
 414 struct slowpath_args {
 415         const char *fmt;
 416         va_list args;
 417 };
 418
 419 static void warn_slowpath_common(const char *file, int line, void *caller,
 420                                  unsigned taint, struct slowpath_args *args)
 421 {
 422         disable_trace_on_warning();
 423
 424         pr_warn("------------[ cut here ]------------\n");
 425         pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS()\n",
 426                 raw_smp_processor_id(), current->pid, file, line, caller);
 427
 428         if (args)
 429                 vprintk(args->fmt, args->args);
 430
 431         print_modules();
 432         dump_stack();
 433         print_oops_end_marker();
 434         /* Just a warning, don't kill lockdep. */
 435         add_taint(taint, LOCKDEP_STILL_OK);
 436 }
 437
 438 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
 439 {
 440         struct slowpath_args args;
 441
 442         args.fmt = fmt;
 443         va_start(args.args, fmt);
 444         warn_slowpath_common(file, line, __builtin_return_address(0),
 445                              TAINT_WARN, &args);
 446         va_end(args.args);
 447 }
 448 EXPORT_SYMBOL(warn_slowpath_fmt);
 449
 450 void warn_slowpath_fmt_taint(const char *file, int line,
 451                              unsigned taint, const char *fmt, ...)
 452 {
 453         struct slowpath_args args;
 454
 455         args.fmt = fmt;
 456         va_start(args.args, fmt);
 457         warn_slowpath_common(file, line, __builtin_return_address(0),
 458                              taint, &args);
 459         va_end(args.args);
 460 }
 461 EXPORT_SYMBOL(warn_slowpath_fmt_taint);
 462
 463 void warn_slowpath_null(const char *file, int line)
 464 {
 465         warn_slowpath_common(file, line, __builtin_return_address(0),
 466                              TAINT_WARN, NULL);
 467 }
 468 EXPORT_SYMBOL(warn_slowpath_null);
 469 #endif
 470
 471 #ifdef CONFIG_CC_STACKPROTECTOR
 472
 473 /*
 474  * Called when gcc's -fstack-protector feature is used, and
 475  * gcc detects corruption of the on-stack canary value
 476  */
 477 __visible void __stack_chk_fail(void)
 478 {
 479         panic("stack-protector: Kernel stack is corrupted in: %p\n",
 480                 __builtin_return_address(0));
 481 }
 482 EXPORT_SYMBOL(__stack_chk_fail);
 483
 484 #endif
 485
 486 core_param(panic, panic_timeout, int, 0644);
 487 core_param(pause_on_oops, pause_on_oops, int, 0644);
 488
 489 static int __init setup_crash_kexec_post_notifiers(char *s)
 490 {
 491         crash_kexec_post_notifiers = true;
 492         return 0;
 493 }
 494 early_param("crash_kexec_post_notifiers", setup_crash_kexec_post_notifiers);
 495
 496 static int __init oops_setup(char *s)
 497 {
 498         if (!s)
 499                 return -EINVAL;
 500         if (!strcmp(s, "panic"))
 501                 panic_on_oops = 1;
 502         return 0;
 503 }
 504 early_param("oops", oops_setup);