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