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         kmsg_dump(KMSG_DUMP_PANIC);
  79         /*
  80          * If we have crashed and we have a crash kernel loaded let it handle
  81          * everything else.
  82          * Do we want to call this before we try to display a message?
  83          */
  84         crash_kexec(NULL);
  85
  86         /*
  87          * Note smp_send_stop is the usual smp shutdown function, which
  88          * unfortunately means it may not be hardened to work in a panic
  89          * situation.
  90          */
  91         smp_send_stop();
  92
  93         atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
  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         bust_spinlocks(0);
 142 }
 143
 144 EXPORT_SYMBOL(panic);
 145
 146
 147 struct tnt {
 148         u8      bit;
 149         char    true;
 150         char    false;
 151 };
 152
 153 static const struct tnt tnts[] = {
 154         { TAINT_PROPRIETARY_MODULE,     'P', 'G' },
 155         { TAINT_FORCED_MODULE,          'F', ' ' },
 156         { TAINT_UNSAFE_SMP,             'S', ' ' },
 157         { TAINT_FORCED_RMMOD,           'R', ' ' },
 158         { TAINT_MACHINE_CHECK,          'M', ' ' },
 159         { TAINT_BAD_PAGE,               'B', ' ' },
 160         { TAINT_USER,                   'U', ' ' },
 161         { TAINT_DIE,                    'D', ' ' },
 162         { TAINT_OVERRIDDEN_ACPI_TABLE,  'A', ' ' },
 163         { TAINT_WARN,                   'W', ' ' },
 164         { TAINT_CRAP,                   'C', ' ' },
 165 };
 166
 167 /**
 168  *      print_tainted - return a string to represent the kernel taint state.
 169  *
 170  *  'P' - Proprietary module has been loaded.
 171  *  'F' - Module has been forcibly loaded.
 172  *  'S' - SMP with CPUs not designed for SMP.
 173  *  'R' - User forced a module unload.
 174  *  'M' - System experienced a machine check exception.
 175  *  'B' - System has hit bad_page.
 176  *  'U' - Userspace-defined naughtiness.
 177  *  'D' - Kernel has oopsed before
 178  *  'A' - ACPI table overridden.
 179  *  'W' - Taint on warning.
 180  *  'C' - modules from drivers/staging are loaded.
 181  *
 182  *      The string is overwritten by the next call to print_tainted().
 183  */
 184 const char *print_tainted(void)
 185 {
 186         static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
 187
 188         if (tainted_mask) {
 189                 char *s;
 190                 int i;
 191
 192                 s = buf + sprintf(buf, "Tainted: ");
 193                 for (i = 0; i < ARRAY_SIZE(tnts); i++) {
 194                         const struct tnt *t = &tnts[i];
 195                         *s++ = test_bit(t->bit, &tainted_mask) ?
 196                                         t->true : t->false;
 197                 }
 198                 *s = 0;
 199         } else
 200                 snprintf(buf, sizeof(buf), "Not tainted");
 201
 202         return buf;
 203 }
 204
 205 int test_taint(unsigned flag)
 206 {
 207         return test_bit(flag, &tainted_mask);
 208 }
 209 EXPORT_SYMBOL(test_taint);
 210
 211 unsigned long get_taint(void)
 212 {
 213         return tainted_mask;
 214 }
 215
 216 void add_taint(unsigned flag)
 217 {
 218         /*
 219          * Can't trust the integrity of the kernel anymore.
 220          * We don't call directly debug_locks_off() because the issue
 221          * is not necessarily serious enough to set oops_in_progress to 1
 222          * Also we want to keep up lockdep for staging development and
 223          * post-warning case.
 224          */
 225         if (flag != TAINT_CRAP && flag != TAINT_WARN && __debug_locks_off())
 226                 printk(KERN_WARNING "Disabling lock debugging due to kernel taint\n");
 227
 228         set_bit(flag, &tainted_mask);
 229 }
 230 EXPORT_SYMBOL(add_taint);
 231
 232 static void spin_msec(int msecs)
 233 {
 234         int i;
 235
 236         for (i = 0; i < msecs; i++) {
 237                 touch_nmi_watchdog();
 238                 mdelay(1);
 239         }
 240 }
 241
 242 /*
 243  * It just happens that oops_enter() and oops_exit() are identically
 244  * implemented...
 245  */
 246 static void do_oops_enter_exit(void)
 247 {
 248         unsigned long flags;
 249         static int spin_counter;
 250
 251         if (!pause_on_oops)
 252                 return;
 253
 254         spin_lock_irqsave(&pause_on_oops_lock, flags);
 255         if (pause_on_oops_flag == 0) {
 256                 /* This CPU may now print the oops message */
 257                 pause_on_oops_flag = 1;
 258         } else {
 259                 /* We need to stall this CPU */
 260                 if (!spin_counter) {
 261                         /* This CPU gets to do the counting */
 262                         spin_counter = pause_on_oops;
 263                         do {
 264                                 spin_unlock(&pause_on_oops_lock);
 265                                 spin_msec(MSEC_PER_SEC);
 266                                 spin_lock(&pause_on_oops_lock);
 267                         } while (--spin_counter);
 268                         pause_on_oops_flag = 0;
 269                 } else {
 270                         /* This CPU waits for a different one */
 271                         while (spin_counter) {
 272                                 spin_unlock(&pause_on_oops_lock);
 273                                 spin_msec(1);
 274                                 spin_lock(&pause_on_oops_lock);
 275                         }
 276                 }
 277         }
 278         spin_unlock_irqrestore(&pause_on_oops_lock, flags);
 279 }
 280
 281 /*
 282  * Return true if the calling CPU is allowed to print oops-related info.
 283  * This is a bit racy..
 284  */
 285 int oops_may_print(void)
 286 {
 287         return pause_on_oops_flag == 0;
 288 }
 289
 290 /*
 291  * Called when the architecture enters its oops handler, before it prints
 292  * anything.  If this is the first CPU to oops, and it's oopsing the first
 293  * time then let it proceed.
 294  *
 295  * This is all enabled by the pause_on_oops kernel boot option.  We do all
 296  * this to ensure that oopses don't scroll off the screen.  It has the
 297  * side-effect of preventing later-oopsing CPUs from mucking up the display,
 298  * too.
 299  *
 300  * It turns out that the CPU which is allowed to print ends up pausing for
 301  * the right duration, whereas all the other CPUs pause for twice as long:
 302  * once in oops_enter(), once in oops_exit().
 303  */
 304 void oops_enter(void)
 305 {
 306         tracing_off();
 307         /* can't trust the integrity of the kernel anymore: */
 308         debug_locks_off();
 309         do_oops_enter_exit();
 310 }
 311
 312 /*
 313  * 64-bit random ID for oopses:
 314  */
 315 static u64 oops_id;
 316
 317 static int init_oops_id(void)
 318 {
 319         if (!oops_id)
 320                 get_random_bytes(&oops_id, sizeof(oops_id));
 321         else
 322                 oops_id++;
 323
 324         return 0;
 325 }
 326 late_initcall(init_oops_id);
 327
 328 static void print_oops_end_marker(void)
 329 {
 330         init_oops_id();
 331         printk(KERN_WARNING "---[ end trace %016llx ]---\n",
 332                 (unsigned long long)oops_id);
 333 }
 334
 335 /*
 336  * Called when the architecture exits its oops handler, after printing
 337  * everything.
 338  */
 339 void oops_exit(void)
 340 {
 341         do_oops_enter_exit();
 342         print_oops_end_marker();
 343         kmsg_dump(KMSG_DUMP_OOPS);
 344 }
 345
 346 #ifdef WANT_WARN_ON_SLOWPATH
 347 struct slowpath_args {
 348         const char *fmt;
 349         va_list args;
 350 };
 351
 352 static void warn_slowpath_common(const char *file, int line, void *caller, struct slowpath_args *args)
 353 {
 354         const char *board;
 355
 356         printk(KERN_WARNING "------------[ cut here ]------------\n");
 357         printk(KERN_WARNING "WARNING: at %s:%d %pS()\n", file, line, caller);
 358         board = dmi_get_system_info(DMI_PRODUCT_NAME);
 359         if (board)
 360                 printk(KERN_WARNING "Hardware name: %s\n", board);
 361
 362         if (args)
 363                 vprintk(args->fmt, args->args);
 364
 365         print_modules();
 366         dump_stack();
 367         print_oops_end_marker();
 368         add_taint(TAINT_WARN);
 369 }
 370
 371 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
 372 {
 373         struct slowpath_args args;
 374
 375         args.fmt = fmt;
 376         va_start(args.args, fmt);
 377         warn_slowpath_common(file, line, __builtin_return_address(0), &args);
 378         va_end(args.args);
 379 }
 380 EXPORT_SYMBOL(warn_slowpath_fmt);
 381
 382 void warn_slowpath_null(const char *file, int line)
 383 {
 384         warn_slowpath_common(file, line, __builtin_return_address(0), NULL);
 385 }
 386 EXPORT_SYMBOL(warn_slowpath_null);
 387 #endif
 388
 389 #ifdef CONFIG_CC_STACKPROTECTOR
 390
 391 /*
 392  * Called when gcc's -fstack-protector feature is used, and
 393  * gcc detects corruption of the on-stack canary value
 394  */
 395 void __stack_chk_fail(void)
 396 {
 397         panic("stack-protector: Kernel stack is corrupted in: %p\n",
 398                 __builtin_return_address(0));
 399 }
 400 EXPORT_SYMBOL(__stack_chk_fail);
 401
 402 #endif
 403
 404 core_param(panic, panic_timeout, int, 0644);
 405 core_param(pause_on_oops, pause_on_oops, int, 0644);