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         bust_spinlocks(0);
  78
  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 #ifdef CONFIG_SMP
  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 #endif
  94
  95         atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
  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                 for (i = 0; i < panic_timeout*1000; ) {
 107                         touch_nmi_watchdog();
 108                         i += panic_blink(i);
 109                         mdelay(1);
 110                         i++;
 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         disabled_wait(caller);
 128 #endif
 129         local_irq_enable();
 130         for (i = 0;;) {
 131                 touch_softlockup_watchdog();
 132                 i += panic_blink(i);
 133                 mdelay(1);
 134                 i++;
 135         }
 136 }
 137
 138 EXPORT_SYMBOL(panic);
 139
 140
 141 struct tnt {
 142         u8 bit;
 143         char true;
 144         char false;
 145 };
 146
 147 static const struct tnt tnts[] = {
 148         { TAINT_PROPRIETARY_MODULE, 'P', 'G' },
 149         { TAINT_FORCED_MODULE, 'F', ' ' },
 150         { TAINT_UNSAFE_SMP, 'S', ' ' },
 151         { TAINT_FORCED_RMMOD, 'R', ' ' },
 152         { TAINT_MACHINE_CHECK, 'M', ' ' },
 153         { TAINT_BAD_PAGE, 'B', ' ' },
 154         { TAINT_USER, 'U', ' ' },
 155         { TAINT_DIE, 'D', ' ' },
 156         { TAINT_OVERRIDDEN_ACPI_TABLE, 'A', ' ' },
 157         { TAINT_WARN, 'W', ' ' },
 158         { TAINT_CRAP, 'C', ' ' },
 159 };
 160
 161 /**
 162  *      print_tainted - return a string to represent the kernel taint state.
 163  *
 164  *  'P' - Proprietary module has been loaded.
 165  *  'F' - Module has been forcibly loaded.
 166  *  'S' - SMP with CPUs not designed for SMP.
 167  *  'R' - User forced a module unload.
 168  *  'M' - System experienced a machine check exception.
 169  *  'B' - System has hit bad_page.
 170  *  'U' - Userspace-defined naughtiness.
 171  *  'D' - Kernel has oopsed before
 172  *  'A' - ACPI table overridden.
 173  *  'W' - Taint on warning.
 174  *  'C' - modules from drivers/staging are loaded.
 175  *
 176  *      The string is overwritten by the next call to print_taint().
 177  */
 178 const char *print_tainted(void)
 179 {
 180         static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
 181
 182         if (tainted_mask) {
 183                 char *s;
 184                 int i;
 185
 186                 s = buf + sprintf(buf, "Tainted: ");
 187                 for (i = 0; i < ARRAY_SIZE(tnts); i++) {
 188                         const struct tnt *t = &tnts[i];
 189                         *s++ = test_bit(t->bit, &tainted_mask) ?
 190                                         t->true : t->false;
 191                 }
 192                 *s = 0;
 193         } else
 194                 snprintf(buf, sizeof(buf), "Not tainted");
 195         return(buf);
 196 }
 197
 198 int test_taint(unsigned flag)
 199 {
 200         return test_bit(flag, &tainted_mask);
 201 }
 202 EXPORT_SYMBOL(test_taint);
 203
 204 unsigned long get_taint(void)
 205 {
 206         return tainted_mask;
 207 }
 208
 209 void add_taint(unsigned flag)
 210 {
 211         debug_locks = 0; /* can't trust the integrity of the kernel anymore */
 212         set_bit(flag, &tainted_mask);
 213 }
 214 EXPORT_SYMBOL(add_taint);
 215
 216 static void spin_msec(int msecs)
 217 {
 218         int i;
 219
 220         for (i = 0; i < msecs; i++) {
 221                 touch_nmi_watchdog();
 222                 mdelay(1);
 223         }
 224 }
 225
 226 /*
 227  * It just happens that oops_enter() and oops_exit() are identically
 228  * implemented...
 229  */
 230 static void do_oops_enter_exit(void)
 231 {
 232         unsigned long flags;
 233         static int spin_counter;
 234
 235         if (!pause_on_oops)
 236                 return;
 237
 238         spin_lock_irqsave(&pause_on_oops_lock, flags);
 239         if (pause_on_oops_flag == 0) {
 240                 /* This CPU may now print the oops message */
 241                 pause_on_oops_flag = 1;
 242         } else {
 243                 /* We need to stall this CPU */
 244                 if (!spin_counter) {
 245                         /* This CPU gets to do the counting */
 246                         spin_counter = pause_on_oops;
 247                         do {
 248                                 spin_unlock(&pause_on_oops_lock);
 249                                 spin_msec(MSEC_PER_SEC);
 250                                 spin_lock(&pause_on_oops_lock);
 251                         } while (--spin_counter);
 252                         pause_on_oops_flag = 0;
 253                 } else {
 254                         /* This CPU waits for a different one */
 255                         while (spin_counter) {
 256                                 spin_unlock(&pause_on_oops_lock);
 257                                 spin_msec(1);
 258                                 spin_lock(&pause_on_oops_lock);
 259                         }
 260                 }
 261         }
 262         spin_unlock_irqrestore(&pause_on_oops_lock, flags);
 263 }
 264
 265 /*
 266  * Return true if the calling CPU is allowed to print oops-related info.  This
 267  * is a bit racy..
 268  */
 269 int oops_may_print(void)
 270 {
 271         return pause_on_oops_flag == 0;
 272 }
 273
 274 /*
 275  * Called when the architecture enters its oops handler, before it prints
 276  * anything.  If this is the first CPU to oops, and it's oopsing the first time
 277  * then let it proceed.
 278  *
 279  * This is all enabled by the pause_on_oops kernel boot option.  We do all this
 280  * to ensure that oopses don't scroll off the screen.  It has the side-effect
 281  * of preventing later-oopsing CPUs from mucking up the display, too.
 282  *
 283  * It turns out that the CPU which is allowed to print ends up pausing for the
 284  * right duration, whereas all the other CPUs pause for twice as long: once in
 285  * oops_enter(), once in oops_exit().
 286  */
 287 void oops_enter(void)
 288 {
 289         debug_locks_off(); /* can't trust the integrity of the kernel anymore */
 290         do_oops_enter_exit();
 291 }
 292
 293 /*
 294  * 64-bit random ID for oopses:
 295  */
 296 static u64 oops_id;
 297
 298 static int init_oops_id(void)
 299 {
 300         if (!oops_id)
 301                 get_random_bytes(&oops_id, sizeof(oops_id));
 302
 303         return 0;
 304 }
 305 late_initcall(init_oops_id);
 306
 307 static void print_oops_end_marker(void)
 308 {
 309         init_oops_id();
 310         printk(KERN_WARNING "---[ end trace %016llx ]---\n",
 311                 (unsigned long long)oops_id);
 312 }
 313
 314 /*
 315  * Called when the architecture exits its oops handler, after printing
 316  * everything.
 317  */
 318 void oops_exit(void)
 319 {
 320         do_oops_enter_exit();
 321         print_oops_end_marker();
 322 }
 323
 324 #ifdef WANT_WARN_ON_SLOWPATH
 325 void warn_slowpath(const char *file, int line, const char *fmt, ...)
 326 {
 327         va_list args;
 328         char function[KSYM_SYMBOL_LEN];
 329         unsigned long caller = (unsigned long)__builtin_return_address(0);
 330         const char *board;
 331
 332         sprint_symbol(function, caller);
 333
 334         printk(KERN_WARNING "------------[ cut here ]------------\n");
 335         printk(KERN_WARNING "WARNING: at %s:%d %s()\n", file,
 336                 line, function);
 337         board = dmi_get_system_info(DMI_PRODUCT_NAME);
 338         if (board)
 339                 printk(KERN_WARNING "Hardware name: %s\n", board);
 340
 341         if (fmt) {
 342                 va_start(args, fmt);
 343                 vprintk(fmt, args);
 344                 va_end(args);
 345         }
 346
 347         print_modules();
 348         dump_stack();
 349         print_oops_end_marker();
 350         add_taint(TAINT_WARN);
 351 }
 352 EXPORT_SYMBOL(warn_slowpath);
 353 #endif
 354
 355 #ifdef CONFIG_CC_STACKPROTECTOR
 356 /*
 357  * Called when gcc's -fstack-protector feature is used, and
 358  * gcc detects corruption of the on-stack canary value
 359  */
 360 void __stack_chk_fail(void)
 361 {
 362         panic("stack-protector: Kernel stack is corrupted");
 363 }
 364 EXPORT_SYMBOL(__stack_chk_fail);
 365 #endif
 366
 367 core_param(panic, panic_timeout, int, 0644);
 368 core_param(pause_on_oops, pause_on_oops, int, 0644);