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