linux/kernel/panic.c
<<
>>
Prefs
   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/ftrace.h>
  19#include <linux/reboot.h>
  20#include <linux/delay.h>
  21#include <linux/kexec.h>
  22#include <linux/sched.h>
  23#include <linux/sysrq.h>
  24#include <linux/init.h>
  25#include <linux/nmi.h>
  26
  27#define PANIC_TIMER_STEP 100
  28#define PANIC_BLINK_SPD 18
  29
  30int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE;
  31static unsigned long tainted_mask;
  32static int pause_on_oops;
  33static int pause_on_oops_flag;
  34static DEFINE_SPINLOCK(pause_on_oops_lock);
  35
  36int panic_timeout;
  37EXPORT_SYMBOL_GPL(panic_timeout);
  38
  39ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
  40
  41EXPORT_SYMBOL(panic_notifier_list);
  42
  43static long no_blink(int state)
  44{
  45        return 0;
  46}
  47
  48/* Returns how long it waited in ms */
  49long (*panic_blink)(int state);
  50EXPORT_SYMBOL(panic_blink);
  51
  52/*
  53 * Stop ourself in panic -- architecture code may override this
  54 */
  55void __weak panic_smp_self_stop(void)
  56{
  57        while (1)
  58                cpu_relax();
  59}
  60
  61/**
  62 *      panic - halt the system
  63 *      @fmt: The text string to print
  64 *
  65 *      Display a message, then perform cleanups.
  66 *
  67 *      This function never returns.
  68 */
  69void panic(const char *fmt, ...)
  70{
  71        static DEFINE_SPINLOCK(panic_lock);
  72        static char buf[1024];
  73        va_list args;
  74        long i, i_next = 0;
  75        int state = 0;
  76
  77        /*
  78         * Disable local interrupts. This will prevent panic_smp_self_stop
  79         * from deadlocking the first cpu that invokes the panic, since
  80         * there is nothing to prevent an interrupt handler (that runs
  81         * after the panic_lock is acquired) from invoking panic again.
  82         */
  83        local_irq_disable();
  84
  85        /*
  86         * It's possible to come here directly from a panic-assertion and
  87         * not have preempt disabled. Some functions called from here want
  88         * preempt to be disabled. No point enabling it later though...
  89         *
  90         * Only one CPU is allowed to execute the panic code from here. For
  91         * multiple parallel invocations of panic, all other CPUs either
  92         * stop themself or will wait until they are stopped by the 1st CPU
  93         * with smp_send_stop().
  94         */
  95        if (!spin_trylock(&panic_lock))
  96                panic_smp_self_stop();
  97
  98        console_verbose();
  99        bust_spinlocks(1);
 100        va_start(args, fmt);
 101        vsnprintf(buf, sizeof(buf), fmt, args);
 102        va_end(args);
 103        printk(KERN_EMERG "Kernel panic - not syncing: %s\n",buf);
 104#ifdef CONFIG_DEBUG_BUGVERBOSE
 105        /*
 106         * Avoid nested stack-dumping if a panic occurs during oops processing
 107         */
 108        if (!test_taint(TAINT_DIE) && oops_in_progress <= 1)
 109                dump_stack();
 110#endif
 111
 112        /*
 113         * If we have crashed and we have a crash kernel loaded let it handle
 114         * everything else.
 115         * Do we want to call this before we try to display a message?
 116         */
 117        crash_kexec(NULL);
 118
 119        /*
 120         * Note smp_send_stop is the usual smp shutdown function, which
 121         * unfortunately means it may not be hardened to work in a panic
 122         * situation.
 123         */
 124        smp_send_stop();
 125
 126        kmsg_dump(KMSG_DUMP_PANIC);
 127
 128        atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
 129
 130        bust_spinlocks(0);
 131
 132        if (!panic_blink)
 133                panic_blink = no_blink;
 134
 135        if (panic_timeout > 0) {
 136                /*
 137                 * Delay timeout seconds before rebooting the machine.
 138                 * We can't use the "normal" timers since we just panicked.
 139                 */
 140                printk(KERN_EMERG "Rebooting in %d seconds..", panic_timeout);
 141
 142                for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
 143                        touch_nmi_watchdog();
 144                        if (i >= i_next) {
 145                                i += panic_blink(state ^= 1);
 146                                i_next = i + 3600 / PANIC_BLINK_SPD;
 147                        }
 148                        mdelay(PANIC_TIMER_STEP);
 149                }
 150        }
 151        if (panic_timeout != 0) {
 152                /*
 153                 * This will not be a clean reboot, with everything
 154                 * shutting down.  But if there is a chance of
 155                 * rebooting the system it will be rebooted.
 156                 */
 157                emergency_restart();
 158        }
 159#ifdef __sparc__
 160        {
 161                extern int stop_a_enabled;
 162                /* Make sure the user can actually press Stop-A (L1-A) */
 163                stop_a_enabled = 1;
 164                printk(KERN_EMERG "Press Stop-A (L1-A) to return to the boot prom\n");
 165        }
 166#endif
 167#if defined(CONFIG_S390)
 168        {
 169                unsigned long caller;
 170
 171                caller = (unsigned long)__builtin_return_address(0);
 172                disabled_wait(caller);
 173        }
 174#endif
 175        local_irq_enable();
 176        for (i = 0; ; i += PANIC_TIMER_STEP) {
 177                touch_softlockup_watchdog();
 178                if (i >= i_next) {
 179                        i += panic_blink(state ^= 1);
 180                        i_next = i + 3600 / PANIC_BLINK_SPD;
 181                }
 182                mdelay(PANIC_TIMER_STEP);
 183        }
 184}
 185
 186EXPORT_SYMBOL(panic);
 187
 188
 189struct tnt {
 190        u8      bit;
 191        char    true;
 192        char    false;
 193};
 194
 195static const struct tnt tnts[] = {
 196        { TAINT_PROPRIETARY_MODULE,     'P', 'G' },
 197        { TAINT_FORCED_MODULE,          'F', ' ' },
 198        { TAINT_UNSAFE_SMP,             'S', ' ' },
 199        { TAINT_FORCED_RMMOD,           'R', ' ' },
 200        { TAINT_MACHINE_CHECK,          'M', ' ' },
 201        { TAINT_BAD_PAGE,               'B', ' ' },
 202        { TAINT_USER,                   'U', ' ' },
 203        { TAINT_DIE,                    'D', ' ' },
 204        { TAINT_OVERRIDDEN_ACPI_TABLE,  'A', ' ' },
 205        { TAINT_WARN,                   'W', ' ' },
 206        { TAINT_CRAP,                   'C', ' ' },
 207        { TAINT_FIRMWARE_WORKAROUND,    'I', ' ' },
 208        { TAINT_OOT_MODULE,             'O', ' ' },
 209};
 210
 211/**
 212 *      print_tainted - return a string to represent the kernel taint state.
 213 *
 214 *  'P' - Proprietary module has been loaded.
 215 *  'F' - Module has been forcibly loaded.
 216 *  'S' - SMP with CPUs not designed for SMP.
 217 *  'R' - User forced a module unload.
 218 *  'M' - System experienced a machine check exception.
 219 *  'B' - System has hit bad_page.
 220 *  'U' - Userspace-defined naughtiness.
 221 *  'D' - Kernel has oopsed before
 222 *  'A' - ACPI table overridden.
 223 *  'W' - Taint on warning.
 224 *  'C' - modules from drivers/staging are loaded.
 225 *  'I' - Working around severe firmware bug.
 226 *  'O' - Out-of-tree module has been loaded.
 227 *
 228 *      The string is overwritten by the next call to print_tainted().
 229 */
 230const char *print_tainted(void)
 231{
 232        static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
 233
 234        if (tainted_mask) {
 235                char *s;
 236                int i;
 237
 238                s = buf + sprintf(buf, "Tainted: ");
 239                for (i = 0; i < ARRAY_SIZE(tnts); i++) {
 240                        const struct tnt *t = &tnts[i];
 241                        *s++ = test_bit(t->bit, &tainted_mask) ?
 242                                        t->true : t->false;
 243                }
 244                *s = 0;
 245        } else
 246                snprintf(buf, sizeof(buf), "Not tainted");
 247
 248        return buf;
 249}
 250
 251int test_taint(unsigned flag)
 252{
 253        return test_bit(flag, &tainted_mask);
 254}
 255EXPORT_SYMBOL(test_taint);
 256
 257unsigned long get_taint(void)
 258{
 259        return tainted_mask;
 260}
 261
 262/**
 263 * add_taint: add a taint flag if not already set.
 264 * @flag: one of the TAINT_* constants.
 265 * @lockdep_ok: whether lock debugging is still OK.
 266 *
 267 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
 268 * some notewortht-but-not-corrupting cases, it can be set to true.
 269 */
 270void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
 271{
 272        if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
 273                printk(KERN_WARNING
 274                       "Disabling lock debugging due to kernel taint\n");
 275
 276        set_bit(flag, &tainted_mask);
 277}
 278EXPORT_SYMBOL(add_taint);
 279
 280static void spin_msec(int msecs)
 281{
 282        int i;
 283
 284        for (i = 0; i < msecs; i++) {
 285                touch_nmi_watchdog();
 286                mdelay(1);
 287        }
 288}
 289
 290/*
 291 * It just happens that oops_enter() and oops_exit() are identically
 292 * implemented...
 293 */
 294static void do_oops_enter_exit(void)
 295{
 296        unsigned long flags;
 297        static int spin_counter;
 298
 299        if (!pause_on_oops)
 300                return;
 301
 302        spin_lock_irqsave(&pause_on_oops_lock, flags);
 303        if (pause_on_oops_flag == 0) {
 304                /* This CPU may now print the oops message */
 305                pause_on_oops_flag = 1;
 306        } else {
 307                /* We need to stall this CPU */
 308                if (!spin_counter) {
 309                        /* This CPU gets to do the counting */
 310                        spin_counter = pause_on_oops;
 311                        do {
 312                                spin_unlock(&pause_on_oops_lock);
 313                                spin_msec(MSEC_PER_SEC);
 314                                spin_lock(&pause_on_oops_lock);
 315                        } while (--spin_counter);
 316                        pause_on_oops_flag = 0;
 317                } else {
 318                        /* This CPU waits for a different one */
 319                        while (spin_counter) {
 320                                spin_unlock(&pause_on_oops_lock);
 321                                spin_msec(1);
 322                                spin_lock(&pause_on_oops_lock);
 323                        }
 324                }
 325        }
 326        spin_unlock_irqrestore(&pause_on_oops_lock, flags);
 327}
 328
 329/*
 330 * Return true if the calling CPU is allowed to print oops-related info.
 331 * This is a bit racy..
 332 */
 333int oops_may_print(void)
 334{
 335        return pause_on_oops_flag == 0;
 336}
 337
 338/*
 339 * Called when the architecture enters its oops handler, before it prints
 340 * anything.  If this is the first CPU to oops, and it's oopsing the first
 341 * time then let it proceed.
 342 *
 343 * This is all enabled by the pause_on_oops kernel boot option.  We do all
 344 * this to ensure that oopses don't scroll off the screen.  It has the
 345 * side-effect of preventing later-oopsing CPUs from mucking up the display,
 346 * too.
 347 *
 348 * It turns out that the CPU which is allowed to print ends up pausing for
 349 * the right duration, whereas all the other CPUs pause for twice as long:
 350 * once in oops_enter(), once in oops_exit().
 351 */
 352void oops_enter(void)
 353{
 354        tracing_off();
 355        /* can't trust the integrity of the kernel anymore: */
 356        debug_locks_off();
 357        do_oops_enter_exit();
 358}
 359
 360/*
 361 * 64-bit random ID for oopses:
 362 */
 363static u64 oops_id;
 364
 365static int init_oops_id(void)
 366{
 367        if (!oops_id)
 368                get_random_bytes(&oops_id, sizeof(oops_id));
 369        else
 370                oops_id++;
 371
 372        return 0;
 373}
 374late_initcall(init_oops_id);
 375
 376void print_oops_end_marker(void)
 377{
 378        init_oops_id();
 379        printk(KERN_WARNING "---[ end trace %016llx ]---\n",
 380                (unsigned long long)oops_id);
 381}
 382
 383/*
 384 * Called when the architecture exits its oops handler, after printing
 385 * everything.
 386 */
 387void oops_exit(void)
 388{
 389        do_oops_enter_exit();
 390        print_oops_end_marker();
 391        kmsg_dump(KMSG_DUMP_OOPS);
 392}
 393
 394#ifdef WANT_WARN_ON_SLOWPATH
 395struct slowpath_args {
 396        const char *fmt;
 397        va_list args;
 398};
 399
 400static void warn_slowpath_common(const char *file, int line, void *caller,
 401                                 unsigned taint, struct slowpath_args *args)
 402{
 403        disable_trace_on_warning();
 404
 405        pr_warn("------------[ cut here ]------------\n");
 406        pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS()\n",
 407                raw_smp_processor_id(), current->pid, file, line, caller);
 408
 409        if (args)
 410                vprintk(args->fmt, args->args);
 411
 412        print_modules();
 413        dump_stack();
 414        print_oops_end_marker();
 415        /* Just a warning, don't kill lockdep. */
 416        add_taint(taint, LOCKDEP_STILL_OK);
 417}
 418
 419void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
 420{
 421        struct slowpath_args args;
 422
 423        args.fmt = fmt;
 424        va_start(args.args, fmt);
 425        warn_slowpath_common(file, line, __builtin_return_address(0),
 426                             TAINT_WARN, &args);
 427        va_end(args.args);
 428}
 429EXPORT_SYMBOL(warn_slowpath_fmt);
 430
 431void warn_slowpath_fmt_taint(const char *file, int line,
 432                             unsigned taint, const char *fmt, ...)
 433{
 434        struct slowpath_args args;
 435
 436        args.fmt = fmt;
 437        va_start(args.args, fmt);
 438        warn_slowpath_common(file, line, __builtin_return_address(0),
 439                             taint, &args);
 440        va_end(args.args);
 441}
 442EXPORT_SYMBOL(warn_slowpath_fmt_taint);
 443
 444void warn_slowpath_null(const char *file, int line)
 445{
 446        warn_slowpath_common(file, line, __builtin_return_address(0),
 447                             TAINT_WARN, NULL);
 448}
 449EXPORT_SYMBOL(warn_slowpath_null);
 450#endif
 451
 452#ifdef CONFIG_CC_STACKPROTECTOR
 453
 454/*
 455 * Called when gcc's -fstack-protector feature is used, and
 456 * gcc detects corruption of the on-stack canary value
 457 */
 458void __stack_chk_fail(void)
 459{
 460        panic("stack-protector: Kernel stack is corrupted in: %p\n",
 461                __builtin_return_address(0));
 462}
 463EXPORT_SYMBOL(__stack_chk_fail);
 464
 465#endif
 466
 467core_param(panic, panic_timeout, int, 0644);
 468core_param(pause_on_oops, pause_on_oops, int, 0644);
 469
 470static int __init oops_setup(char *s)
 471{
 472        if (!s)
 473                return -EINVAL;
 474        if (!strcmp(s, "panic"))
 475                panic_on_oops = 1;
 476        return 0;
 477}
 478early_param("oops", oops_setup);
 479