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/module.h>
  12 #include <linux/sched.h>
  13 #include <linux/delay.h>
  14 #include <linux/reboot.h>
  15 #include <linux/notifier.h>
  16 #include <linux/init.h>
  17 #include <linux/sysrq.h>
  18 #include <linux/interrupt.h>
  19 #include <linux/nmi.h>
  20 #include <linux/kexec.h>
  21 #include <linux/debug_locks.h>
  22 #include <linux/random.h>
  23 #include <linux/kallsyms.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
  56 NORET_TYPE void panic(const char * fmt, ...)
  57 {
  58         long i;
  59         static char buf[1024];
  60         va_list args;
  61 #if defined(CONFIG_S390)
  62         unsigned long caller = (unsigned long) __builtin_return_address(0);
  63 #endif
  64
  65         /*
  66          * It's possible to come here directly from a panic-assertion and not
  67          * have preempt disabled. Some functions called from here want
  68          * preempt to be disabled. No point enabling it later though...
  69          */
  70         preempt_disable();
  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         bust_spinlocks(0);
  81
  82         /*
  83          * If we have crashed and we have a crash kernel loaded let it handle
  84          * everything else.
  85          * Do we want to call this before we try to display a message?
  86          */
  87         crash_kexec(NULL);
  88
  89 #ifdef CONFIG_SMP
  90         /*
  91          * Note smp_send_stop is the usual smp shutdown function, which
  92          * unfortunately means it may not be hardened to work in a panic
  93          * situation.
  94          */
  95         smp_send_stop();
  96 #endif
  97
  98         atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
  99
 100         if (!panic_blink)
 101                 panic_blink = no_blink;
 102
 103         if (panic_timeout > 0) {
 104                 /*
 105                  * Delay timeout seconds before rebooting the machine.
 106                  * We can't use the "normal" timers since we just panicked..
 107                  */
 108                 printk(KERN_EMERG "Rebooting in %d seconds..",panic_timeout);
 109                 for (i = 0; i < panic_timeout*1000; ) {
 110                         touch_nmi_watchdog();
 111                         i += panic_blink(i);
 112                         mdelay(1);
 113                         i++;
 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         disabled_wait(caller);
 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 }
 140
 141 EXPORT_SYMBOL(panic);
 142
 143
 144 struct tnt {
 145         u8 bit;
 146         char true;
 147         char false;
 148 };
 149
 150 static const struct tnt tnts[] = {
 151         { TAINT_PROPRIETARY_MODULE, 'P', 'G' },
 152         { TAINT_FORCED_MODULE, 'F', ' ' },
 153         { TAINT_UNSAFE_SMP, 'S', ' ' },
 154         { TAINT_FORCED_RMMOD, 'R', ' ' },
 155         { TAINT_MACHINE_CHECK, 'M', ' ' },
 156         { TAINT_BAD_PAGE, 'B', ' ' },
 157         { TAINT_USER, 'U', ' ' },
 158         { TAINT_DIE, 'D', ' ' },
 159         { TAINT_OVERRIDDEN_ACPI_TABLE, 'A', ' ' },
 160         { TAINT_WARN, 'W', ' ' },
 161         { TAINT_CRAP, 'C', ' ' },
 162 };
 163
 164 /**
 165  *      print_tainted - return a string to represent the kernel taint state.
 166  *
 167  *  'P' - Proprietary module has been loaded.
 168  *  'F' - Module has been forcibly loaded.
 169  *  'S' - SMP with CPUs not designed for SMP.
 170  *  'R' - User forced a module unload.
 171  *  'M' - System experienced a machine check exception.
 172  *  'B' - System has hit bad_page.
 173  *  'U' - Userspace-defined naughtiness.
 174  *  'D' - Kernel has oopsed before
 175  *  'A' - ACPI table overridden.
 176  *  'W' - Taint on warning.
 177  *  'C' - modules from drivers/staging are loaded.
 178  *
 179  *      The string is overwritten by the next call to print_taint().
 180  */
 181 const char *print_tainted(void)
 182 {
 183         static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
 184
 185         if (tainted_mask) {
 186                 char *s;
 187                 int i;
 188
 189                 s = buf + sprintf(buf, "Tainted: ");
 190                 for (i = 0; i < ARRAY_SIZE(tnts); i++) {
 191                         const struct tnt *t = &tnts[i];
 192                         *s++ = test_bit(t->bit, &tainted_mask) ?
 193                                         t->true : t->false;
 194                 }
 195                 *s = 0;
 196         } else
 197                 snprintf(buf, sizeof(buf), "Not tainted");
 198         return(buf);
 199 }
 200
 201 int test_taint(unsigned flag)
 202 {
 203         return test_bit(flag, &tainted_mask);
 204 }
 205 EXPORT_SYMBOL(test_taint);
 206
 207 unsigned long get_taint(void)
 208 {
 209         return tainted_mask;
 210 }
 211
 212 void add_taint(unsigned flag)
 213 {
 214         debug_locks = 0; /* can't trust the integrity of the kernel anymore */
 215         set_bit(flag, &tainted_mask);
 216 }
 217 EXPORT_SYMBOL(add_taint);
 218
 219 static void spin_msec(int msecs)
 220 {
 221         int i;
 222
 223         for (i = 0; i < msecs; i++) {
 224                 touch_nmi_watchdog();
 225                 mdelay(1);
 226         }
 227 }
 228
 229 /*
 230  * It just happens that oops_enter() and oops_exit() are identically
 231  * implemented...
 232  */
 233 static void do_oops_enter_exit(void)
 234 {
 235         unsigned long flags;
 236         static int spin_counter;
 237
 238         if (!pause_on_oops)
 239                 return;
 240
 241         spin_lock_irqsave(&pause_on_oops_lock, flags);
 242         if (pause_on_oops_flag == 0) {
 243                 /* This CPU may now print the oops message */
 244                 pause_on_oops_flag = 1;
 245         } else {
 246                 /* We need to stall this CPU */
 247                 if (!spin_counter) {
 248                         /* This CPU gets to do the counting */
 249                         spin_counter = pause_on_oops;
 250                         do {
 251                                 spin_unlock(&pause_on_oops_lock);
 252                                 spin_msec(MSEC_PER_SEC);
 253                                 spin_lock(&pause_on_oops_lock);
 254                         } while (--spin_counter);
 255                         pause_on_oops_flag = 0;
 256                 } else {
 257                         /* This CPU waits for a different one */
 258                         while (spin_counter) {
 259                                 spin_unlock(&pause_on_oops_lock);
 260                                 spin_msec(1);
 261                                 spin_lock(&pause_on_oops_lock);
 262                         }
 263                 }
 264         }
 265         spin_unlock_irqrestore(&pause_on_oops_lock, flags);
 266 }
 267
 268 /*
 269  * Return true if the calling CPU is allowed to print oops-related info.  This
 270  * is a bit racy..
 271  */
 272 int oops_may_print(void)
 273 {
 274         return pause_on_oops_flag == 0;
 275 }
 276
 277 /*
 278  * Called when the architecture enters its oops handler, before it prints
 279  * anything.  If this is the first CPU to oops, and it's oopsing the first time
 280  * then let it proceed.
 281  *
 282  * This is all enabled by the pause_on_oops kernel boot option.  We do all this
 283  * to ensure that oopses don't scroll off the screen.  It has the side-effect
 284  * of preventing later-oopsing CPUs from mucking up the display, too.
 285  *
 286  * It turns out that the CPU which is allowed to print ends up pausing for the
 287  * right duration, whereas all the other CPUs pause for twice as long: once in
 288  * oops_enter(), once in oops_exit().
 289  */
 290 void oops_enter(void)
 291 {
 292         debug_locks_off(); /* can't trust the integrity of the kernel anymore */
 293         do_oops_enter_exit();
 294 }
 295
 296 /*
 297  * 64-bit random ID for oopses:
 298  */
 299 static u64 oops_id;
 300
 301 static int init_oops_id(void)
 302 {
 303         if (!oops_id)
 304                 get_random_bytes(&oops_id, sizeof(oops_id));
 305         else
 306                 oops_id++;
 307
 308         return 0;
 309 }
 310 late_initcall(init_oops_id);
 311
 312 static void print_oops_end_marker(void)
 313 {
 314         init_oops_id();
 315         printk(KERN_WARNING "---[ end trace %016llx ]---\n",
 316                 (unsigned long long)oops_id);
 317 }
 318
 319 /*
 320  * Called when the architecture exits its oops handler, after printing
 321  * everything.
 322  */
 323 void oops_exit(void)
 324 {
 325         do_oops_enter_exit();
 326         print_oops_end_marker();
 327 }
 328
 329 #ifdef WANT_WARN_ON_SLOWPATH
 330 void warn_slowpath(const char *file, int line, const char *fmt, ...)
 331 {
 332         va_list args;
 333         char function[KSYM_SYMBOL_LEN];
 334         unsigned long caller = (unsigned long)__builtin_return_address(0);
 335         const char *board;
 336
 337         sprint_symbol(function, caller);
 338
 339         printk(KERN_WARNING "------------[ cut here ]------------\n");
 340         printk(KERN_WARNING "WARNING: at %s:%d %s()\n", file,
 341                 line, function);
 342         board = dmi_get_system_info(DMI_PRODUCT_NAME);
 343         if (board)
 344                 printk(KERN_WARNING "Hardware name: %s\n", board);
 345
 346         if (fmt) {
 347                 va_start(args, fmt);
 348                 vprintk(fmt, args);
 349                 va_end(args);
 350         }
 351
 352         print_modules();
 353         dump_stack();
 354         print_oops_end_marker();
 355         add_taint(TAINT_WARN);
 356 }
 357 EXPORT_SYMBOL(warn_slowpath);
 358 #endif
 359
 360 #ifdef CONFIG_CC_STACKPROTECTOR
 361
 362 /*
 363  * Called when gcc's -fstack-protector feature is used, and
 364  * gcc detects corruption of the on-stack canary value
 365  */
 366 void __stack_chk_fail(void)
 367 {
 368         panic("stack-protector: Kernel stack is corrupted in: %p\n",
 369                 __builtin_return_address(0));
 370 }
 371 EXPORT_SYMBOL(__stack_chk_fail);
 372
 373 #endif
 374
 375 core_param(panic, panic_timeout, int, 0644);
 376 core_param(pause_on_oops, pause_on_oops, int, 0644);