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