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 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         bust_spinlocks(1);
  70         va_start(args, fmt);
  71         vsnprintf(buf, sizeof(buf), fmt, args);
  72         va_end(args);
  73         printk(KERN_EMERG "Kernel panic - not syncing: %s\n",buf);
  74 #ifdef CONFIG_DEBUG_BUGVERBOSE
  75         dump_stack();
  76 #endif
  77
  78         /*
  79          * If we have crashed and we have a crash kernel loaded let it handle
  80          * everything else.
  81          * Do we want to call this before we try to display a message?
  82          */
  83         crash_kexec(NULL);
  84
  85         kmsg_dump(KMSG_DUMP_PANIC);
  86
  87         /*
  88          * Note smp_send_stop is the usual smp shutdown function, which
  89          * unfortunately means it may not be hardened to work in a panic
  90          * situation.
  91          */
  92         smp_send_stop();
  93
  94         atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
  95
  96         bust_spinlocks(0);
  97
  98         if (!panic_blink)
  99                 panic_blink = no_blink;
 100
 101         if (panic_timeout > 0) {
 102                 /*
 103                  * Delay timeout seconds before rebooting the machine.
 104                  * We can't use the "normal" timers since we just panicked.
 105                  */
 106                 printk(KERN_EMERG "Rebooting in %d seconds..", panic_timeout);
 107
 108                 for (i = 0; i < panic_timeout*1000; ) {
 109                         touch_nmi_watchdog();
 110                         i += panic_blink(i);
 111                         mdelay(1);
 112                         i++;
 113                 }
 114                 /*
 115                  * This will not be a clean reboot, with everything
 116                  * shutting down.  But if there is a chance of
 117                  * rebooting the system it will be rebooted.
 118                  */
 119                 emergency_restart();
 120         }
 121 #ifdef __sparc__
 122         {
 123                 extern int stop_a_enabled;
 124                 /* Make sure the user can actually press Stop-A (L1-A) */
 125                 stop_a_enabled = 1;
 126                 printk(KERN_EMERG "Press Stop-A (L1-A) to return to the boot prom\n");
 127         }
 128 #endif
 129 #if defined(CONFIG_S390)
 130         {
 131                 unsigned long caller;
 132
 133                 caller = (unsigned long)__builtin_return_address(0);
 134                 disabled_wait(caller);
 135         }
 136 #endif
 137         local_irq_enable();
 138         for (i = 0; ; ) {
 139                 touch_softlockup_watchdog();
 140                 i += panic_blink(i);
 141                 mdelay(1);
 142                 i++;
 143         }
 144 }
 145
 146 EXPORT_SYMBOL(panic);
 147
 148
 149 struct tnt {
 150         u8      bit;
 151         char    true;
 152         char    false;
 153 };
 154
 155 static const struct tnt tnts[] = {
 156         { TAINT_PROPRIETARY_MODULE,     'P', 'G' },
 157         { TAINT_FORCED_MODULE,          'F', ' ' },
 158         { TAINT_UNSAFE_SMP,             'S', ' ' },
 159         { TAINT_FORCED_RMMOD,           'R', ' ' },
 160         { TAINT_MACHINE_CHECK,          'M', ' ' },
 161         { TAINT_BAD_PAGE,               'B', ' ' },
 162         { TAINT_USER,                   'U', ' ' },
 163         { TAINT_DIE,                    'D', ' ' },
 164         { TAINT_OVERRIDDEN_ACPI_TABLE,  'A', ' ' },
 165         { TAINT_WARN,                   'W', ' ' },
 166         { TAINT_CRAP,                   'C', ' ' },
 167 };
 168
 169 /**
 170  *      print_tainted - return a string to represent the kernel taint state.
 171  *
 172  *  'P' - Proprietary module has been loaded.
 173  *  'F' - Module has been forcibly loaded.
 174  *  'S' - SMP with CPUs not designed for SMP.
 175  *  'R' - User forced a module unload.
 176  *  'M' - System experienced a machine check exception.
 177  *  'B' - System has hit bad_page.
 178  *  'U' - Userspace-defined naughtiness.
 179  *  'D' - Kernel has oopsed before
 180  *  'A' - ACPI table overridden.
 181  *  'W' - Taint on warning.
 182  *  'C' - modules from drivers/staging are loaded.
 183  *
 184  *      The string is overwritten by the next call to print_tainted().
 185  */
 186 const char *print_tainted(void)
 187 {
 188         static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
 189
 190         if (tainted_mask) {
 191                 char *s;
 192                 int i;
 193
 194                 s = buf + sprintf(buf, "Tainted: ");
 195                 for (i = 0; i < ARRAY_SIZE(tnts); i++) {
 196                         const struct tnt *t = &tnts[i];
 197                         *s++ = test_bit(t->bit, &tainted_mask) ?
 198                                         t->true : t->false;
 199                 }
 200                 *s = 0;
 201         } else
 202                 snprintf(buf, sizeof(buf), "Not tainted");
 203
 204         return buf;
 205 }
 206
 207 int test_taint(unsigned flag)
 208 {
 209         return test_bit(flag, &tainted_mask);
 210 }
 211 EXPORT_SYMBOL(test_taint);
 212
 213 unsigned long get_taint(void)
 214 {
 215         return tainted_mask;
 216 }
 217
 218 void add_taint(unsigned flag)
 219 {
 220         /*
 221          * Can't trust the integrity of the kernel anymore.
 222          * We don't call directly debug_locks_off() because the issue
 223          * is not necessarily serious enough to set oops_in_progress to 1
 224          * Also we want to keep up lockdep for staging development and
 225          * post-warning case.
 226          */
 227         if (flag != TAINT_CRAP && flag != TAINT_WARN && __debug_locks_off())
 228                 printk(KERN_WARNING "Disabling lock debugging due to kernel taint\n");
 229
 230         set_bit(flag, &tainted_mask);
 231 }
 232 EXPORT_SYMBOL(add_taint);
 233
 234 static void spin_msec(int msecs)
 235 {
 236         int i;
 237
 238         for (i = 0; i < msecs; i++) {
 239                 touch_nmi_watchdog();
 240                 mdelay(1);
 241         }
 242 }
 243
 244 /*
 245  * It just happens that oops_enter() and oops_exit() are identically
 246  * implemented...
 247  */
 248 static void do_oops_enter_exit(void)
 249 {
 250         unsigned long flags;
 251         static int spin_counter;
 252
 253         if (!pause_on_oops)
 254                 return;
 255
 256         spin_lock_irqsave(&pause_on_oops_lock, flags);
 257         if (pause_on_oops_flag == 0) {
 258                 /* This CPU may now print the oops message */
 259                 pause_on_oops_flag = 1;
 260         } else {
 261                 /* We need to stall this CPU */
 262                 if (!spin_counter) {
 263                         /* This CPU gets to do the counting */
 264                         spin_counter = pause_on_oops;
 265                         do {
 266                                 spin_unlock(&pause_on_oops_lock);
 267                                 spin_msec(MSEC_PER_SEC);
 268                                 spin_lock(&pause_on_oops_lock);
 269                         } while (--spin_counter);
 270                         pause_on_oops_flag = 0;
 271                 } else {
 272                         /* This CPU waits for a different one */
 273                         while (spin_counter) {
 274                                 spin_unlock(&pause_on_oops_lock);
 275                                 spin_msec(1);
 276                                 spin_lock(&pause_on_oops_lock);
 277                         }
 278                 }
 279         }
 280         spin_unlock_irqrestore(&pause_on_oops_lock, flags);
 281 }
 282
 283 /*
 284  * Return true if the calling CPU is allowed to print oops-related info.
 285  * This is a bit racy..
 286  */
 287 int oops_may_print(void)
 288 {
 289         return pause_on_oops_flag == 0;
 290 }
 291
 292 /*
 293  * Called when the architecture enters its oops handler, before it prints
 294  * anything.  If this is the first CPU to oops, and it's oopsing the first
 295  * time then let it proceed.
 296  *
 297  * This is all enabled by the pause_on_oops kernel boot option.  We do all
 298  * this to ensure that oopses don't scroll off the screen.  It has the
 299  * side-effect of preventing later-oopsing CPUs from mucking up the display,
 300  * too.
 301  *
 302  * It turns out that the CPU which is allowed to print ends up pausing for
 303  * the right duration, whereas all the other CPUs pause for twice as long:
 304  * once in oops_enter(), once in oops_exit().
 305  */
 306 void oops_enter(void)
 307 {
 308         tracing_off();
 309         /* can't trust the integrity of the kernel anymore: */
 310         debug_locks_off();
 311         do_oops_enter_exit();
 312 }
 313
 314 /*
 315  * 64-bit random ID for oopses:
 316  */
 317 static u64 oops_id;
 318
 319 static int init_oops_id(void)
 320 {
 321         if (!oops_id)
 322                 get_random_bytes(&oops_id, sizeof(oops_id));
 323         else
 324                 oops_id++;
 325
 326         return 0;
 327 }
 328 late_initcall(init_oops_id);
 329
 330 static void print_oops_end_marker(void)
 331 {
 332         init_oops_id();
 333         printk(KERN_WARNING "---[ end trace %016llx ]---\n",
 334                 (unsigned long long)oops_id);
 335 }
 336
 337 /*
 338  * Called when the architecture exits its oops handler, after printing
 339  * everything.
 340  */
 341 void oops_exit(void)
 342 {
 343         do_oops_enter_exit();
 344         print_oops_end_marker();
 345         kmsg_dump(KMSG_DUMP_OOPS);
 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);