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/kallsyms.h>
  14 #include <linux/notifier.h>
  15 #include <linux/module.h>
  16 #include <linux/random.h>
  17 #include <linux/reboot.h>
  18 #include <linux/delay.h>
  19 #include <linux/kexec.h>
  20 #include <linux/sched.h>
  21 #include <linux/sysrq.h>
  22 #include <linux/init.h>
  23 #include <linux/nmi.h>
  24 #include <linux/dmi.h>
  25
  26 int panic_on_oops;
  27 static unsigned long tainted_mask;
  28 static int pause_on_oops;
  29 static int pause_on_oops_flag;
  30 static DEFINE_SPINLOCK(pause_on_oops_lock);
  31
  32 int panic_timeout;
  33
  34 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
  35
  36 EXPORT_SYMBOL(panic_notifier_list);
  37
  38 static long no_blink(long time)
  39 {
  40         return 0;
  41 }
  42
  43 /* Returns how long it waited in ms */
  44 long (*panic_blink)(long time);
  45 EXPORT_SYMBOL(panic_blink);
  46
  47 /**
  48  *      panic - halt the system
  49  *      @fmt: The text string to print
  50  *
  51  *      Display a message, then perform cleanups.
  52  *
  53  *      This function never returns.
  54  */
  55 NORET_TYPE void panic(const char * fmt, ...)
  56 {
  57         static char buf[1024];
  58         va_list args;
  59         long i;
  60
  61         /*
  62          * It's possible to come here directly from a panic-assertion and
  63          * not have preempt disabled. Some functions called from here want
  64          * preempt to be disabled. No point enabling it later though...
  65          */
  66         preempt_disable();
  67
  68         bust_spinlocks(1);
  69         va_start(args, fmt);
  70         vsnprintf(buf, sizeof(buf), fmt, args);
  71         va_end(args);
  72         printk(KERN_EMERG "Kernel panic - not syncing: %s\n",buf);
  73 #ifdef CONFIG_DEBUG_BUGVERBOSE
  74         dump_stack();
  75 #endif
  76
  77         /*
  78          * If we have crashed and we have a crash kernel loaded let it handle
  79          * everything else.
  80          * Do we want to call this before we try to display a message?
  81          */
  82         crash_kexec(NULL);
  83
  84         /*
  85          * Note smp_send_stop is the usual smp shutdown function, which
  86          * unfortunately means it may not be hardened to work in a panic
  87          * situation.
  88          */
  89         smp_send_stop();
  90
  91         atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
  92
  93         bust_spinlocks(0);
  94
  95         if (!panic_blink)
  96                 panic_blink = no_blink;
  97
  98         if (panic_timeout > 0) {
  99                 /*
 100                  * Delay timeout seconds before rebooting the machine.
 101                  * We can't use the "normal" timers since we just panicked.
 102                  */
 103                 printk(KERN_EMERG "Rebooting in %d seconds..", panic_timeout);
 104
 105                 for (i = 0; i < panic_timeout*1000; ) {
 106                         touch_nmi_watchdog();
 107                         i += panic_blink(i);
 108                         mdelay(1);
 109                         i++;
 110                 }
 111                 /*
 112                  * This will not be a clean reboot, with everything
 113                  * shutting down.  But if there is a chance of
 114                  * rebooting the system it will be rebooted.
 115                  */
 116                 emergency_restart();
 117         }
 118 #ifdef __sparc__
 119         {
 120                 extern int stop_a_enabled;
 121                 /* Make sure the user can actually press Stop-A (L1-A) */
 122                 stop_a_enabled = 1;
 123                 printk(KERN_EMERG "Press Stop-A (L1-A) to return to the boot prom\n");
 124         }
 125 #endif
 126 #if defined(CONFIG_S390)
 127         {
 128                 unsigned long caller;
 129
 130                 caller = (unsigned long)__builtin_return_address(0);
 131                 disabled_wait(caller);
 132         }
 133 #endif
 134         local_irq_enable();
 135         for (i = 0; ; ) {
 136                 touch_softlockup_watchdog();
 137                 i += panic_blink(i);
 138                 mdelay(1);
 139                 i++;
 140         }
 141 }
 142
 143 EXPORT_SYMBOL(panic);
 144
 145
 146 struct tnt {
 147         u8      bit;
 148         char    true;
 149         char    false;
 150 };
 151
 152 static const struct tnt tnts[] = {
 153         { TAINT_PROPRIETARY_MODULE,     'P', 'G' },
 154         { TAINT_FORCED_MODULE,          'F', ' ' },
 155         { TAINT_UNSAFE_SMP,             'S', ' ' },
 156         { TAINT_FORCED_RMMOD,           'R', ' ' },
 157         { TAINT_MACHINE_CHECK,          'M', ' ' },
 158         { TAINT_BAD_PAGE,               'B', ' ' },
 159         { TAINT_USER,                   'U', ' ' },
 160         { TAINT_DIE,                    'D', ' ' },
 161         { TAINT_OVERRIDDEN_ACPI_TABLE,  'A', ' ' },
 162         { TAINT_WARN,                   'W', ' ' },
 163         { TAINT_CRAP,                   'C', ' ' },
 164 };
 165
 166 /**
 167  *      print_tainted - return a string to represent the kernel taint state.
 168  *
 169  *  'P' - Proprietary module has been loaded.
 170  *  'F' - Module has been forcibly loaded.
 171  *  'S' - SMP with CPUs not designed for SMP.
 172  *  'R' - User forced a module unload.
 173  *  'M' - System experienced a machine check exception.
 174  *  'B' - System has hit bad_page.
 175  *  'U' - Userspace-defined naughtiness.
 176  *  'D' - Kernel has oopsed before
 177  *  'A' - ACPI table overridden.
 178  *  'W' - Taint on warning.
 179  *  'C' - modules from drivers/staging are loaded.
 180  *
 181  *      The string is overwritten by the next call to print_tainted().
 182  */
 183 const char *print_tainted(void)
 184 {
 185         static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
 186
 187         if (tainted_mask) {
 188                 char *s;
 189                 int i;
 190
 191                 s = buf + sprintf(buf, "Tainted: ");
 192                 for (i = 0; i < ARRAY_SIZE(tnts); i++) {
 193                         const struct tnt *t = &tnts[i];
 194                         *s++ = test_bit(t->bit, &tainted_mask) ?
 195                                         t->true : t->false;
 196                 }
 197                 *s = 0;
 198         } else
 199                 snprintf(buf, sizeof(buf), "Not tainted");
 200
 201         return buf;
 202 }
 203
 204 int test_taint(unsigned flag)
 205 {
 206         return test_bit(flag, &tainted_mask);
 207 }
 208 EXPORT_SYMBOL(test_taint);
 209
 210 unsigned long get_taint(void)
 211 {
 212         return tainted_mask;
 213 }
 214
 215 void add_taint(unsigned flag)
 216 {
 217         /*
 218          * Can't trust the integrity of the kernel anymore.
 219          * We don't call directly debug_locks_off() because the issue
 220          * is not necessarily serious enough to set oops_in_progress to 1
 221          * Also we want to keep up lockdep for staging development and
 222          * post-warning case.
 223          */
 224         if (flag != TAINT_CRAP && flag != TAINT_WARN && __debug_locks_off())
 225                 printk(KERN_WARNING "Disabling lock debugging due to kernel taint\n");
 226
 227         set_bit(flag, &tainted_mask);
 228 }
 229 EXPORT_SYMBOL(add_taint);
 230
 231 static void spin_msec(int msecs)
 232 {
 233         int i;
 234
 235         for (i = 0; i < msecs; i++) {
 236                 touch_nmi_watchdog();
 237                 mdelay(1);
 238         }
 239 }
 240
 241 /*
 242  * It just happens that oops_enter() and oops_exit() are identically
 243  * implemented...
 244  */
 245 static void do_oops_enter_exit(void)
 246 {
 247         unsigned long flags;
 248         static int spin_counter;
 249
 250         if (!pause_on_oops)
 251                 return;
 252
 253         spin_lock_irqsave(&pause_on_oops_lock, flags);
 254         if (pause_on_oops_flag == 0) {
 255                 /* This CPU may now print the oops message */
 256                 pause_on_oops_flag = 1;
 257         } else {
 258                 /* We need to stall this CPU */
 259                 if (!spin_counter) {
 260                         /* This CPU gets to do the counting */
 261                         spin_counter = pause_on_oops;
 262                         do {
 263                                 spin_unlock(&pause_on_oops_lock);
 264                                 spin_msec(MSEC_PER_SEC);
 265                                 spin_lock(&pause_on_oops_lock);
 266                         } while (--spin_counter);
 267                         pause_on_oops_flag = 0;
 268                 } else {
 269                         /* This CPU waits for a different one */
 270                         while (spin_counter) {
 271                                 spin_unlock(&pause_on_oops_lock);
 272                                 spin_msec(1);
 273                                 spin_lock(&pause_on_oops_lock);
 274                         }
 275                 }
 276         }
 277         spin_unlock_irqrestore(&pause_on_oops_lock, flags);
 278 }
 279
 280 /*
 281  * Return true if the calling CPU is allowed to print oops-related info.
 282  * This is a bit racy..
 283  */
 284 int oops_may_print(void)
 285 {
 286         return pause_on_oops_flag == 0;
 287 }
 288
 289 /*
 290  * Called when the architecture enters its oops handler, before it prints
 291  * anything.  If this is the first CPU to oops, and it's oopsing the first
 292  * time then let it proceed.
 293  *
 294  * This is all enabled by the pause_on_oops kernel boot option.  We do all
 295  * this to ensure that oopses don't scroll off the screen.  It has the
 296  * side-effect of preventing later-oopsing CPUs from mucking up the display,
 297  * too.
 298  *
 299  * It turns out that the CPU which is allowed to print ends up pausing for
 300  * the right duration, whereas all the other CPUs pause for twice as long:
 301  * once in oops_enter(), once in oops_exit().
 302  */
 303 void oops_enter(void)
 304 {
 305         tracing_off();
 306         /* can't trust the integrity of the kernel anymore: */
 307         debug_locks_off();
 308         do_oops_enter_exit();
 309 }
 310
 311 /*
 312  * 64-bit random ID for oopses:
 313  */
 314 static u64 oops_id;
 315
 316 static int init_oops_id(void)
 317 {
 318         if (!oops_id)
 319                 get_random_bytes(&oops_id, sizeof(oops_id));
 320         else
 321                 oops_id++;
 322
 323         return 0;
 324 }
 325 late_initcall(init_oops_id);
 326
 327 static void print_oops_end_marker(void)
 328 {
 329         init_oops_id();
 330         printk(KERN_WARNING "---[ end trace %016llx ]---\n",
 331                 (unsigned long long)oops_id);
 332 }
 333
 334 /*
 335  * Called when the architecture exits its oops handler, after printing
 336  * everything.
 337  */
 338 void oops_exit(void)
 339 {
 340         do_oops_enter_exit();
 341         print_oops_end_marker();
 342 }
 343
 344 #ifdef WANT_WARN_ON_SLOWPATH
 345 struct slowpath_args {
 346         const char *fmt;
 347         va_list args;
 348 };
 349
 350 static void warn_slowpath_common(const char *file, int line, void *caller, struct slowpath_args *args)
 351 {
 352         const char *board;
 353
 354         printk(KERN_WARNING "------------[ cut here ]------------\n");
 355         printk(KERN_WARNING "WARNING: at %s:%d %pS()\n", file, line, caller);
 356         board = dmi_get_system_info(DMI_PRODUCT_NAME);
 357         if (board)
 358                 printk(KERN_WARNING "Hardware name: %s\n", board);
 359
 360         if (args)
 361                 vprintk(args->fmt, args->args);
 362
 363         print_modules();
 364         dump_stack();
 365         print_oops_end_marker();
 366         add_taint(TAINT_WARN);
 367 }
 368
 369 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
 370 {
 371         struct slowpath_args args;
 372
 373         args.fmt = fmt;
 374         va_start(args.args, fmt);
 375         warn_slowpath_common(file, line, __builtin_return_address(0), &args);
 376         va_end(args.args);
 377 }
 378 EXPORT_SYMBOL(warn_slowpath_fmt);
 379
 380 void warn_slowpath_null(const char *file, int line)
 381 {
 382         warn_slowpath_common(file, line, __builtin_return_address(0), NULL);
 383 }
 384 EXPORT_SYMBOL(warn_slowpath_null);
 385 #endif
 386
 387 #ifdef CONFIG_CC_STACKPROTECTOR
 388
 389 /*
 390  * Called when gcc's -fstack-protector feature is used, and
 391  * gcc detects corruption of the on-stack canary value
 392  */
 393 void __stack_chk_fail(void)
 394 {
 395         panic("stack-protector: Kernel stack is corrupted in: %p\n",
 396                 __builtin_return_address(0));
 397 }
 398 EXPORT_SYMBOL(__stack_chk_fail);
 399
 400 #endif
 401
 402 core_param(panic, panic_timeout, int, 0644);
 403 core_param(pause_on_oops, pause_on_oops, int, 0644);