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 /* Returns how long it waited in ms */
  40 long (*panic_blink)(long time);
  41 EXPORT_SYMBOL(panic_blink);
  42
  43 static void panic_blink_one_second(void)
  44 {
  45         static long i = 0, end;
  46
  47         if (panic_blink) {
  48                 end = i + MSEC_PER_SEC;
  49
  50                 while (i < end) {
  51                         i += panic_blink(i);
  52                         mdelay(1);
  53                         i++;
  54                 }
  55         } else {
  56                 /*
  57                  * When running under a hypervisor a small mdelay may get
  58                  * rounded up to the hypervisor timeslice. For example, with
  59                  * a 1ms in 10ms hypervisor timeslice we might inflate a
  60                  * mdelay(1) loop by 10x.
  61                  *
  62                  * If we have nothing to blink, spin on 1 second calls to
  63                  * mdelay to avoid this.
  64                  */
  65                 mdelay(MSEC_PER_SEC);
  66         }
  67 }
  68
  69 /**
  70  *      panic - halt the system
  71  *      @fmt: The text string to print
  72  *
  73  *      Display a message, then perform cleanups.
  74  *
  75  *      This function never returns.
  76  */
  77 NORET_TYPE void panic(const char * fmt, ...)
  78 {
  79         static char buf[1024];
  80         va_list args;
  81         long i;
  82
  83         /*
  84          * It's possible to come here directly from a panic-assertion and
  85          * not have preempt disabled. Some functions called from here want
  86          * preempt to be disabled. No point enabling it later though...
  87          */
  88         preempt_disable();
  89
  90         bust_spinlocks(1);
  91         va_start(args, fmt);
  92         vsnprintf(buf, sizeof(buf), fmt, args);
  93         va_end(args);
  94         printk(KERN_EMERG "Kernel panic - not syncing: %s\n",buf);
  95 #ifdef CONFIG_DEBUG_BUGVERBOSE
  96         dump_stack();
  97 #endif
  98
  99         /*
 100          * If we have crashed and we have a crash kernel loaded let it handle
 101          * everything else.
 102          * Do we want to call this before we try to display a message?
 103          */
 104         crash_kexec(NULL);
 105
 106         kmsg_dump(KMSG_DUMP_PANIC);
 107
 108         /*
 109          * Note smp_send_stop is the usual smp shutdown function, which
 110          * unfortunately means it may not be hardened to work in a panic
 111          * situation.
 112          */
 113         smp_send_stop();
 114
 115         atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
 116
 117         bust_spinlocks(0);
 118
 119         if (panic_timeout > 0) {
 120                 /*
 121                  * Delay timeout seconds before rebooting the machine.
 122                  * We can't use the "normal" timers since we just panicked.
 123                  */
 124                 printk(KERN_EMERG "Rebooting in %d seconds..", panic_timeout);
 125
 126                 for (i = 0; i < panic_timeout; i++) {
 127                         touch_nmi_watchdog();
 128                         panic_blink_one_second();
 129                 }
 130                 /*
 131                  * This will not be a clean reboot, with everything
 132                  * shutting down.  But if there is a chance of
 133                  * rebooting the system it will be rebooted.
 134                  */
 135                 emergency_restart();
 136         }
 137 #ifdef __sparc__
 138         {
 139                 extern int stop_a_enabled;
 140                 /* Make sure the user can actually press Stop-A (L1-A) */
 141                 stop_a_enabled = 1;
 142                 printk(KERN_EMERG "Press Stop-A (L1-A) to return to the boot prom\n");
 143         }
 144 #endif
 145 #if defined(CONFIG_S390)
 146         {
 147                 unsigned long caller;
 148
 149                 caller = (unsigned long)__builtin_return_address(0);
 150                 disabled_wait(caller);
 151         }
 152 #endif
 153         local_irq_enable();
 154         while (1) {
 155                 touch_softlockup_watchdog();
 156                 panic_blink_one_second();
 157         }
 158 }
 159
 160 EXPORT_SYMBOL(panic);
 161
 162
 163 struct tnt {
 164         u8      bit;
 165         char    true;
 166         char    false;
 167 };
 168
 169 static const struct tnt tnts[] = {
 170         { TAINT_PROPRIETARY_MODULE,     'P', 'G' },
 171         { TAINT_FORCED_MODULE,          'F', ' ' },
 172         { TAINT_UNSAFE_SMP,             'S', ' ' },
 173         { TAINT_FORCED_RMMOD,           'R', ' ' },
 174         { TAINT_MACHINE_CHECK,          'M', ' ' },
 175         { TAINT_BAD_PAGE,               'B', ' ' },
 176         { TAINT_USER,                   'U', ' ' },
 177         { TAINT_DIE,                    'D', ' ' },
 178         { TAINT_OVERRIDDEN_ACPI_TABLE,  'A', ' ' },
 179         { TAINT_WARN,                   'W', ' ' },
 180         { TAINT_CRAP,                   'C', ' ' },
 181         { TAINT_FIRMWARE_WORKAROUND,    'I', ' ' },
 182 };
 183
 184 /**
 185  *      print_tainted - return a string to represent the kernel taint state.
 186  *
 187  *  'P' - Proprietary module has been loaded.
 188  *  'F' - Module has been forcibly loaded.
 189  *  'S' - SMP with CPUs not designed for SMP.
 190  *  'R' - User forced a module unload.
 191  *  'M' - System experienced a machine check exception.
 192  *  'B' - System has hit bad_page.
 193  *  'U' - Userspace-defined naughtiness.
 194  *  'D' - Kernel has oopsed before
 195  *  'A' - ACPI table overridden.
 196  *  'W' - Taint on warning.
 197  *  'C' - modules from drivers/staging are loaded.
 198  *  'I' - Working around severe firmware bug.
 199  *
 200  *      The string is overwritten by the next call to print_tainted().
 201  */
 202 const char *print_tainted(void)
 203 {
 204         static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
 205
 206         if (tainted_mask) {
 207                 char *s;
 208                 int i;
 209
 210                 s = buf + sprintf(buf, "Tainted: ");
 211                 for (i = 0; i < ARRAY_SIZE(tnts); i++) {
 212                         const struct tnt *t = &tnts[i];
 213                         *s++ = test_bit(t->bit, &tainted_mask) ?
 214                                         t->true : t->false;
 215                 }
 216                 *s = 0;
 217         } else
 218                 snprintf(buf, sizeof(buf), "Not tainted");
 219
 220         return buf;
 221 }
 222
 223 int test_taint(unsigned flag)
 224 {
 225         return test_bit(flag, &tainted_mask);
 226 }
 227 EXPORT_SYMBOL(test_taint);
 228
 229 unsigned long get_taint(void)
 230 {
 231         return tainted_mask;
 232 }
 233
 234 void add_taint(unsigned flag)
 235 {
 236         /*
 237          * Can't trust the integrity of the kernel anymore.
 238          * We don't call directly debug_locks_off() because the issue
 239          * is not necessarily serious enough to set oops_in_progress to 1
 240          * Also we want to keep up lockdep for staging development and
 241          * post-warning case.
 242          */
 243         if (flag != TAINT_CRAP && flag != TAINT_WARN && __debug_locks_off())
 244                 printk(KERN_WARNING "Disabling lock debugging due to kernel taint\n");
 245
 246         set_bit(flag, &tainted_mask);
 247 }
 248 EXPORT_SYMBOL(add_taint);
 249
 250 static void spin_msec(int msecs)
 251 {
 252         int i;
 253
 254         for (i = 0; i < msecs; i++) {
 255                 touch_nmi_watchdog();
 256                 mdelay(1);
 257         }
 258 }
 259
 260 /*
 261  * It just happens that oops_enter() and oops_exit() are identically
 262  * implemented...
 263  */
 264 static void do_oops_enter_exit(void)
 265 {
 266         unsigned long flags;
 267         static int spin_counter;
 268
 269         if (!pause_on_oops)
 270                 return;
 271
 272         spin_lock_irqsave(&pause_on_oops_lock, flags);
 273         if (pause_on_oops_flag == 0) {
 274                 /* This CPU may now print the oops message */
 275                 pause_on_oops_flag = 1;
 276         } else {
 277                 /* We need to stall this CPU */
 278                 if (!spin_counter) {
 279                         /* This CPU gets to do the counting */
 280                         spin_counter = pause_on_oops;
 281                         do {
 282                                 spin_unlock(&pause_on_oops_lock);
 283                                 spin_msec(MSEC_PER_SEC);
 284                                 spin_lock(&pause_on_oops_lock);
 285                         } while (--spin_counter);
 286                         pause_on_oops_flag = 0;
 287                 } else {
 288                         /* This CPU waits for a different one */
 289                         while (spin_counter) {
 290                                 spin_unlock(&pause_on_oops_lock);
 291                                 spin_msec(1);
 292                                 spin_lock(&pause_on_oops_lock);
 293                         }
 294                 }
 295         }
 296         spin_unlock_irqrestore(&pause_on_oops_lock, flags);
 297 }
 298
 299 /*
 300  * Return true if the calling CPU is allowed to print oops-related info.
 301  * This is a bit racy..
 302  */
 303 int oops_may_print(void)
 304 {
 305         return pause_on_oops_flag == 0;
 306 }
 307
 308 /*
 309  * Called when the architecture enters its oops handler, before it prints
 310  * anything.  If this is the first CPU to oops, and it's oopsing the first
 311  * time then let it proceed.
 312  *
 313  * This is all enabled by the pause_on_oops kernel boot option.  We do all
 314  * this to ensure that oopses don't scroll off the screen.  It has the
 315  * side-effect of preventing later-oopsing CPUs from mucking up the display,
 316  * too.
 317  *
 318  * It turns out that the CPU which is allowed to print ends up pausing for
 319  * the right duration, whereas all the other CPUs pause for twice as long:
 320  * once in oops_enter(), once in oops_exit().
 321  */
 322 void oops_enter(void)
 323 {
 324         tracing_off();
 325         /* can't trust the integrity of the kernel anymore: */
 326         debug_locks_off();
 327         do_oops_enter_exit();
 328 }
 329
 330 /*
 331  * 64-bit random ID for oopses:
 332  */
 333 static u64 oops_id;
 334
 335 static int init_oops_id(void)
 336 {
 337         if (!oops_id)
 338                 get_random_bytes(&oops_id, sizeof(oops_id));
 339         else
 340                 oops_id++;
 341
 342         return 0;
 343 }
 344 late_initcall(init_oops_id);
 345
 346 static void print_oops_end_marker(void)
 347 {
 348         init_oops_id();
 349         printk(KERN_WARNING "---[ end trace %016llx ]---\n",
 350                 (unsigned long long)oops_id);
 351 }
 352
 353 /*
 354  * Called when the architecture exits its oops handler, after printing
 355  * everything.
 356  */
 357 void oops_exit(void)
 358 {
 359         do_oops_enter_exit();
 360         print_oops_end_marker();
 361         kmsg_dump(KMSG_DUMP_OOPS);
 362 }
 363
 364 #ifdef WANT_WARN_ON_SLOWPATH
 365 struct slowpath_args {
 366         const char *fmt;
 367         va_list args;
 368 };
 369
 370 static void warn_slowpath_common(const char *file, int line, void *caller,
 371                                  unsigned taint, struct slowpath_args *args)
 372 {
 373         const char *board;
 374
 375         printk(KERN_WARNING "------------[ cut here ]------------\n");
 376         printk(KERN_WARNING "WARNING: at %s:%d %pS()\n", file, line, caller);
 377         board = dmi_get_system_info(DMI_PRODUCT_NAME);
 378         if (board)
 379                 printk(KERN_WARNING "Hardware name: %s\n", board);
 380
 381         if (args)
 382                 vprintk(args->fmt, args->args);
 383
 384         print_modules();
 385         dump_stack();
 386         print_oops_end_marker();
 387         add_taint(taint);
 388 }
 389
 390 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
 391 {
 392         struct slowpath_args args;
 393
 394         args.fmt = fmt;
 395         va_start(args.args, fmt);
 396         warn_slowpath_common(file, line, __builtin_return_address(0),
 397                              TAINT_WARN, &args);
 398         va_end(args.args);
 399 }
 400 EXPORT_SYMBOL(warn_slowpath_fmt);
 401
 402 void warn_slowpath_fmt_taint(const char *file, int line,
 403                              unsigned taint, const char *fmt, ...)
 404 {
 405         struct slowpath_args args;
 406
 407         args.fmt = fmt;
 408         va_start(args.args, fmt);
 409         warn_slowpath_common(file, line, __builtin_return_address(0),
 410                              taint, &args);
 411         va_end(args.args);
 412 }
 413 EXPORT_SYMBOL(warn_slowpath_fmt_taint);
 414
 415 void warn_slowpath_null(const char *file, int line)
 416 {
 417         warn_slowpath_common(file, line, __builtin_return_address(0),
 418                              TAINT_WARN, NULL);
 419 }
 420 EXPORT_SYMBOL(warn_slowpath_null);
 421 #endif
 422
 423 #ifdef CONFIG_CC_STACKPROTECTOR
 424
 425 /*
 426  * Called when gcc's -fstack-protector feature is used, and
 427  * gcc detects corruption of the on-stack canary value
 428  */
 429 void __stack_chk_fail(void)
 430 {
 431         panic("stack-protector: Kernel stack is corrupted in: %p\n",
 432                 __builtin_return_address(0));
 433 }
 434 EXPORT_SYMBOL(__stack_chk_fail);
 435
 436 #endif
 437
 438 core_param(panic, panic_timeout, int, 0644);
 439 core_param(pause_on_oops, pause_on_oops, int, 0644);