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 #include <linux/vt.h>
  26
  27 int panic_on_oops;
  28 static unsigned long tainted_mask;
  29 static int pause_on_oops;
  30 static int pause_on_oops_flag;
  31 static DEFINE_SPINLOCK(pause_on_oops_lock);
  32
  33 int panic_timeout;
  34
  35 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
  36
  37 EXPORT_SYMBOL(panic_notifier_list);
  38
  39 static long no_blink(long time)
  40 {
  41         return 0;
  42 }
  43
  44 /* Returns how long it waited in ms */
  45 long (*panic_blink)(long time);
  46 EXPORT_SYMBOL(panic_blink);
  47
  48 /**
  49  *      panic - halt the system
  50  *      @fmt: The text string to print
  51  *
  52  *      Display a message, then perform cleanups.
  53  *
  54  *      This function never returns.
  55  */
  56 NORET_TYPE void panic(const char * fmt, ...)
  57 {
  58         static char buf[1024];
  59         va_list args;
  60         long i;
  61
  62         /*
  63          * It's possible to come here directly from a panic-assertion and
  64          * not have preempt disabled. Some functions called from here want
  65          * preempt to be disabled. No point enabling it later though...
  66          */
  67         preempt_disable();
  68
  69         /* don't redirect the panic message to some hidden console */
  70         vt_kmsg_redirect(0);
  71
  72         bust_spinlocks(1);
  73         va_start(args, fmt);
  74         vsnprintf(buf, sizeof(buf), fmt, args);
  75         va_end(args);
  76         printk(KERN_EMERG "Kernel panic - not syncing: %s\n",buf);
  77 #ifdef CONFIG_DEBUG_BUGVERBOSE
  78         dump_stack();
  79 #endif
  80
  81         /*
  82          * If we have crashed and we have a crash kernel loaded let it handle
  83          * everything else.
  84          * Do we want to call this before we try to display a message?
  85          */
  86         crash_kexec(NULL);
  87
  88         /*
  89          * Note smp_send_stop is the usual smp shutdown function, which
  90          * unfortunately means it may not be hardened to work in a panic
  91          * situation.
  92          */
  93         smp_send_stop();
  94
  95         atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
  96
  97         bust_spinlocks(0);
  98
  99         if (!panic_blink)
 100                 panic_blink = no_blink;
 101
 102         if (panic_timeout > 0) {
 103                 /*
 104                  * Delay timeout seconds before rebooting the machine.
 105                  * We can't use the "normal" timers since we just panicked.
 106                  */
 107                 printk(KERN_EMERG "Rebooting in %d seconds..", panic_timeout);
 108
 109                 for (i = 0; i < panic_timeout*1000; ) {
 110                         touch_nmi_watchdog();
 111                         i += panic_blink(i);
 112                         mdelay(1);
 113                         i++;
 114                 }
 115                 /*
 116                  * This will not be a clean reboot, with everything
 117                  * shutting down.  But if there is a chance of
 118                  * rebooting the system it will be rebooted.
 119                  */
 120                 emergency_restart();
 121         }
 122 #ifdef __sparc__
 123         {
 124                 extern int stop_a_enabled;
 125                 /* Make sure the user can actually press Stop-A (L1-A) */
 126                 stop_a_enabled = 1;
 127                 printk(KERN_EMERG "Press Stop-A (L1-A) to return to the boot prom\n");
 128         }
 129 #endif
 130 #if defined(CONFIG_S390)
 131         {
 132                 unsigned long caller;
 133
 134                 caller = (unsigned long)__builtin_return_address(0);
 135                 disabled_wait(caller);
 136         }
 137 #endif
 138         local_irq_enable();
 139         for (i = 0; ; ) {
 140                 touch_softlockup_watchdog();
 141                 i += panic_blink(i);
 142                 mdelay(1);
 143                 i++;
 144         }
 145 }
 146
 147 EXPORT_SYMBOL(panic);
 148
 149
 150 struct tnt {
 151         u8      bit;
 152         char    true;
 153         char    false;
 154 };
 155
 156 static const struct tnt tnts[] = {
 157         { TAINT_PROPRIETARY_MODULE,     'P', 'G' },
 158         { TAINT_FORCED_MODULE,          'F', ' ' },
 159         { TAINT_UNSAFE_SMP,             'S', ' ' },
 160         { TAINT_FORCED_RMMOD,           'R', ' ' },
 161         { TAINT_MACHINE_CHECK,          'M', ' ' },
 162         { TAINT_BAD_PAGE,               'B', ' ' },
 163         { TAINT_USER,                   'U', ' ' },
 164         { TAINT_DIE,                    'D', ' ' },
 165         { TAINT_OVERRIDDEN_ACPI_TABLE,  'A', ' ' },
 166         { TAINT_WARN,                   'W', ' ' },
 167         { TAINT_CRAP,                   'C', ' ' },
 168 };
 169
 170 /**
 171  *      print_tainted - return a string to represent the kernel taint state.
 172  *
 173  *  'P' - Proprietary module has been loaded.
 174  *  'F' - Module has been forcibly loaded.
 175  *  'S' - SMP with CPUs not designed for SMP.
 176  *  'R' - User forced a module unload.
 177  *  'M' - System experienced a machine check exception.
 178  *  'B' - System has hit bad_page.
 179  *  'U' - Userspace-defined naughtiness.
 180  *  'D' - Kernel has oopsed before
 181  *  'A' - ACPI table overridden.
 182  *  'W' - Taint on warning.
 183  *  'C' - modules from drivers/staging are loaded.
 184  *
 185  *      The string is overwritten by the next call to print_tainted().
 186  */
 187 const char *print_tainted(void)
 188 {
 189         static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
 190
 191         if (tainted_mask) {
 192                 char *s;
 193                 int i;
 194
 195                 s = buf + sprintf(buf, "Tainted: ");
 196                 for (i = 0; i < ARRAY_SIZE(tnts); i++) {
 197                         const struct tnt *t = &tnts[i];
 198                         *s++ = test_bit(t->bit, &tainted_mask) ?
 199                                         t->true : t->false;
 200                 }
 201                 *s = 0;
 202         } else
 203                 snprintf(buf, sizeof(buf), "Not tainted");
 204
 205         return buf;
 206 }
 207
 208 int test_taint(unsigned flag)
 209 {
 210         return test_bit(flag, &tainted_mask);
 211 }
 212 EXPORT_SYMBOL(test_taint);
 213
 214 unsigned long get_taint(void)
 215 {
 216         return tainted_mask;
 217 }
 218
 219 void add_taint(unsigned flag)
 220 {
 221         /*
 222          * Can't trust the integrity of the kernel anymore.
 223          * We don't call directly debug_locks_off() because the issue
 224          * is not necessarily serious enough to set oops_in_progress to 1
 225          * Also we want to keep up lockdep for staging development and
 226          * post-warning case.
 227          */
 228         if (flag != TAINT_CRAP && flag != TAINT_WARN && __debug_locks_off())
 229                 printk(KERN_WARNING "Disabling lock debugging due to kernel taint\n");
 230
 231         set_bit(flag, &tainted_mask);
 232 }
 233 EXPORT_SYMBOL(add_taint);
 234
 235 static void spin_msec(int msecs)
 236 {
 237         int i;
 238
 239         for (i = 0; i < msecs; i++) {
 240                 touch_nmi_watchdog();
 241                 mdelay(1);
 242         }
 243 }
 244
 245 /*
 246  * It just happens that oops_enter() and oops_exit() are identically
 247  * implemented...
 248  */
 249 static void do_oops_enter_exit(void)
 250 {
 251         unsigned long flags;
 252         static int spin_counter;
 253
 254         if (!pause_on_oops)
 255                 return;
 256
 257         spin_lock_irqsave(&pause_on_oops_lock, flags);
 258         if (pause_on_oops_flag == 0) {
 259                 /* This CPU may now print the oops message */
 260                 pause_on_oops_flag = 1;
 261         } else {
 262                 /* We need to stall this CPU */
 263                 if (!spin_counter) {
 264                         /* This CPU gets to do the counting */
 265                         spin_counter = pause_on_oops;
 266                         do {
 267                                 spin_unlock(&pause_on_oops_lock);
 268                                 spin_msec(MSEC_PER_SEC);
 269                                 spin_lock(&pause_on_oops_lock);
 270                         } while (--spin_counter);
 271                         pause_on_oops_flag = 0;
 272                 } else {
 273                         /* This CPU waits for a different one */
 274                         while (spin_counter) {
 275                                 spin_unlock(&pause_on_oops_lock);
 276                                 spin_msec(1);
 277                                 spin_lock(&pause_on_oops_lock);
 278                         }
 279                 }
 280         }
 281         spin_unlock_irqrestore(&pause_on_oops_lock, flags);
 282 }
 283
 284 /*
 285  * Return true if the calling CPU is allowed to print oops-related info.
 286  * This is a bit racy..
 287  */
 288 int oops_may_print(void)
 289 {
 290         return pause_on_oops_flag == 0;
 291 }
 292
 293 /*
 294  * Called when the architecture enters its oops handler, before it prints
 295  * anything.  If this is the first CPU to oops, and it's oopsing the first
 296  * time then let it proceed.
 297  *
 298  * This is all enabled by the pause_on_oops kernel boot option.  We do all
 299  * this to ensure that oopses don't scroll off the screen.  It has the
 300  * side-effect of preventing later-oopsing CPUs from mucking up the display,
 301  * too.
 302  *
 303  * It turns out that the CPU which is allowed to print ends up pausing for
 304  * the right duration, whereas all the other CPUs pause for twice as long:
 305  * once in oops_enter(), once in oops_exit().
 306  */
 307 void oops_enter(void)
 308 {
 309         tracing_off();
 310         /* can't trust the integrity of the kernel anymore: */
 311         debug_locks_off();
 312         do_oops_enter_exit();
 313 }
 314
 315 /*
 316  * 64-bit random ID for oopses:
 317  */
 318 static u64 oops_id;
 319
 320 static int init_oops_id(void)
 321 {
 322         if (!oops_id)
 323                 get_random_bytes(&oops_id, sizeof(oops_id));
 324         else
 325                 oops_id++;
 326
 327         return 0;
 328 }
 329 late_initcall(init_oops_id);
 330
 331 static void print_oops_end_marker(void)
 332 {
 333         init_oops_id();
 334         printk(KERN_WARNING "---[ end trace %016llx ]---\n",
 335                 (unsigned long long)oops_id);
 336 }
 337
 338 /*
 339  * Called when the architecture exits its oops handler, after printing
 340  * everything.
 341  */
 342 void oops_exit(void)
 343 {
 344         do_oops_enter_exit();
 345         print_oops_end_marker();
 346 }
 347
 348 #ifdef WANT_WARN_ON_SLOWPATH
 349 struct slowpath_args {
 350         const char *fmt;
 351         va_list args;
 352 };
 353
 354 static void warn_slowpath_common(const char *file, int line, void *caller, struct slowpath_args *args)
 355 {
 356         const char *board;
 357
 358         printk(KERN_WARNING "------------[ cut here ]------------\n");
 359         printk(KERN_WARNING "WARNING: at %s:%d %pS()\n", file, line, caller);
 360         board = dmi_get_system_info(DMI_PRODUCT_NAME);
 361         if (board)
 362                 printk(KERN_WARNING "Hardware name: %s\n", board);
 363
 364         if (args)
 365                 vprintk(args->fmt, args->args);
 366
 367         print_modules();
 368         dump_stack();
 369         print_oops_end_marker();
 370         add_taint(TAINT_WARN);
 371 }
 372
 373 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
 374 {
 375         struct slowpath_args args;
 376
 377         args.fmt = fmt;
 378         va_start(args.args, fmt);
 379         warn_slowpath_common(file, line, __builtin_return_address(0), &args);
 380         va_end(args.args);
 381 }
 382 EXPORT_SYMBOL(warn_slowpath_fmt);
 383
 384 void warn_slowpath_null(const char *file, int line)
 385 {
 386         warn_slowpath_common(file, line, __builtin_return_address(0), NULL);
 387 }
 388 EXPORT_SYMBOL(warn_slowpath_null);
 389 #endif
 390
 391 #ifdef CONFIG_CC_STACKPROTECTOR
 392
 393 /*
 394  * Called when gcc's -fstack-protector feature is used, and
 395  * gcc detects corruption of the on-stack canary value
 396  */
 397 void __stack_chk_fail(void)
 398 {
 399         panic("stack-protector: Kernel stack is corrupted in: %p\n",
 400                 __builtin_return_address(0));
 401 }
 402 EXPORT_SYMBOL(__stack_chk_fail);
 403
 404 #endif
 405
 406 core_param(panic, panic_timeout, int, 0644);
 407 core_param(pause_on_oops, pause_on_oops, int, 0644);