Root/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/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 = CONFIG_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    pr_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    /*
127     * Run any panic handlers, including those that might need to
128     * add information to the kmsg dump output.
129     */
130    atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
131
132    kmsg_dump(KMSG_DUMP_PANIC);
133
134    bust_spinlocks(0);
135
136    if (!panic_blink)
137        panic_blink = no_blink;
138
139    if (panic_timeout > 0) {
140        /*
141         * Delay timeout seconds before rebooting the machine.
142         * We can't use the "normal" timers since we just panicked.
143         */
144        pr_emerg("Rebooting in %d seconds..", panic_timeout);
145
146        for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
147            touch_nmi_watchdog();
148            if (i >= i_next) {
149                i += panic_blink(state ^= 1);
150                i_next = i + 3600 / PANIC_BLINK_SPD;
151            }
152            mdelay(PANIC_TIMER_STEP);
153        }
154    }
155    if (panic_timeout != 0) {
156        /*
157         * This will not be a clean reboot, with everything
158         * shutting down. But if there is a chance of
159         * rebooting the system it will be rebooted.
160         */
161        emergency_restart();
162    }
163#ifdef __sparc__
164    {
165        extern int stop_a_enabled;
166        /* Make sure the user can actually press Stop-A (L1-A) */
167        stop_a_enabled = 1;
168        pr_emerg("Press Stop-A (L1-A) to return to the boot prom\n");
169    }
170#endif
171#if defined(CONFIG_S390)
172    {
173        unsigned long caller;
174
175        caller = (unsigned long)__builtin_return_address(0);
176        disabled_wait(caller);
177    }
178#endif
179    pr_emerg("---[ end Kernel panic - not syncing: %s\n", buf);
180    local_irq_enable();
181    for (i = 0; ; i += PANIC_TIMER_STEP) {
182        touch_softlockup_watchdog();
183        if (i >= i_next) {
184            i += panic_blink(state ^= 1);
185            i_next = i + 3600 / PANIC_BLINK_SPD;
186        }
187        mdelay(PANIC_TIMER_STEP);
188    }
189}
190
191EXPORT_SYMBOL(panic);
192
193
194struct tnt {
195    u8 bit;
196    char true;
197    char false;
198};
199
200static const struct tnt tnts[] = {
201    { TAINT_PROPRIETARY_MODULE, 'P', 'G' },
202    { TAINT_FORCED_MODULE, 'F', ' ' },
203    { TAINT_CPU_OUT_OF_SPEC, 'S', ' ' },
204    { TAINT_FORCED_RMMOD, 'R', ' ' },
205    { TAINT_MACHINE_CHECK, 'M', ' ' },
206    { TAINT_BAD_PAGE, 'B', ' ' },
207    { TAINT_USER, 'U', ' ' },
208    { TAINT_DIE, 'D', ' ' },
209    { TAINT_OVERRIDDEN_ACPI_TABLE, 'A', ' ' },
210    { TAINT_WARN, 'W', ' ' },
211    { TAINT_CRAP, 'C', ' ' },
212    { TAINT_FIRMWARE_WORKAROUND, 'I', ' ' },
213    { TAINT_OOT_MODULE, 'O', ' ' },
214    { TAINT_UNSIGNED_MODULE, 'E', ' ' },
215};
216
217/**
218 * print_tainted - return a string to represent the kernel taint state.
219 *
220 * 'P' - Proprietary module has been loaded.
221 * 'F' - Module has been forcibly loaded.
222 * 'S' - SMP with CPUs not designed for SMP.
223 * 'R' - User forced a module unload.
224 * 'M' - System experienced a machine check exception.
225 * 'B' - System has hit bad_page.
226 * 'U' - Userspace-defined naughtiness.
227 * 'D' - Kernel has oopsed before
228 * 'A' - ACPI table overridden.
229 * 'W' - Taint on warning.
230 * 'C' - modules from drivers/staging are loaded.
231 * 'I' - Working around severe firmware bug.
232 * 'O' - Out-of-tree module has been loaded.
233 * 'E' - Unsigned module has been loaded.
234 *
235 * The string is overwritten by the next call to print_tainted().
236 */
237const char *print_tainted(void)
238{
239    static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ")];
240
241    if (tainted_mask) {
242        char *s;
243        int i;
244
245        s = buf + sprintf(buf, "Tainted: ");
246        for (i = 0; i < ARRAY_SIZE(tnts); i++) {
247            const struct tnt *t = &tnts[i];
248            *s++ = test_bit(t->bit, &tainted_mask) ?
249                    t->true : t->false;
250        }
251        *s = 0;
252    } else
253        snprintf(buf, sizeof(buf), "Not tainted");
254
255    return buf;
256}
257
258int test_taint(unsigned flag)
259{
260    return test_bit(flag, &tainted_mask);
261}
262EXPORT_SYMBOL(test_taint);
263
264unsigned long get_taint(void)
265{
266    return tainted_mask;
267}
268
269/**
270 * add_taint: add a taint flag if not already set.
271 * @flag: one of the TAINT_* constants.
272 * @lockdep_ok: whether lock debugging is still OK.
273 *
274 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
275 * some notewortht-but-not-corrupting cases, it can be set to true.
276 */
277void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
278{
279    if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
280        pr_warn("Disabling lock debugging due to kernel taint\n");
281
282    set_bit(flag, &tainted_mask);
283}
284EXPORT_SYMBOL(add_taint);
285
286static void spin_msec(int msecs)
287{
288    int i;
289
290    for (i = 0; i < msecs; i++) {
291        touch_nmi_watchdog();
292        mdelay(1);
293    }
294}
295
296/*
297 * It just happens that oops_enter() and oops_exit() are identically
298 * implemented...
299 */
300static void do_oops_enter_exit(void)
301{
302    unsigned long flags;
303    static int spin_counter;
304
305    if (!pause_on_oops)
306        return;
307
308    spin_lock_irqsave(&pause_on_oops_lock, flags);
309    if (pause_on_oops_flag == 0) {
310        /* This CPU may now print the oops message */
311        pause_on_oops_flag = 1;
312    } else {
313        /* We need to stall this CPU */
314        if (!spin_counter) {
315            /* This CPU gets to do the counting */
316            spin_counter = pause_on_oops;
317            do {
318                spin_unlock(&pause_on_oops_lock);
319                spin_msec(MSEC_PER_SEC);
320                spin_lock(&pause_on_oops_lock);
321            } while (--spin_counter);
322            pause_on_oops_flag = 0;
323        } else {
324            /* This CPU waits for a different one */
325            while (spin_counter) {
326                spin_unlock(&pause_on_oops_lock);
327                spin_msec(1);
328                spin_lock(&pause_on_oops_lock);
329            }
330        }
331    }
332    spin_unlock_irqrestore(&pause_on_oops_lock, flags);
333}
334
335/*
336 * Return true if the calling CPU is allowed to print oops-related info.
337 * This is a bit racy..
338 */
339int oops_may_print(void)
340{
341    return pause_on_oops_flag == 0;
342}
343
344/*
345 * Called when the architecture enters its oops handler, before it prints
346 * anything. If this is the first CPU to oops, and it's oopsing the first
347 * time then let it proceed.
348 *
349 * This is all enabled by the pause_on_oops kernel boot option. We do all
350 * this to ensure that oopses don't scroll off the screen. It has the
351 * side-effect of preventing later-oopsing CPUs from mucking up the display,
352 * too.
353 *
354 * It turns out that the CPU which is allowed to print ends up pausing for
355 * the right duration, whereas all the other CPUs pause for twice as long:
356 * once in oops_enter(), once in oops_exit().
357 */
358void oops_enter(void)
359{
360    tracing_off();
361    /* can't trust the integrity of the kernel anymore: */
362    debug_locks_off();
363    do_oops_enter_exit();
364}
365
366/*
367 * 64-bit random ID for oopses:
368 */
369static u64 oops_id;
370
371static int init_oops_id(void)
372{
373    if (!oops_id)
374        get_random_bytes(&oops_id, sizeof(oops_id));
375    else
376        oops_id++;
377
378    return 0;
379}
380late_initcall(init_oops_id);
381
382void print_oops_end_marker(void)
383{
384    init_oops_id();
385    pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id);
386}
387
388/*
389 * Called when the architecture exits its oops handler, after printing
390 * everything.
391 */
392void oops_exit(void)
393{
394    do_oops_enter_exit();
395    print_oops_end_marker();
396    kmsg_dump(KMSG_DUMP_OOPS);
397}
398
399#ifdef WANT_WARN_ON_SLOWPATH
400struct slowpath_args {
401    const char *fmt;
402    va_list args;
403};
404
405static void warn_slowpath_common(const char *file, int line, void *caller,
406                 unsigned taint, struct slowpath_args *args)
407{
408    disable_trace_on_warning();
409
410    pr_warn("------------[ cut here ]------------\n");
411    pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS()\n",
412        raw_smp_processor_id(), current->pid, file, line, caller);
413
414    if (args)
415        vprintk(args->fmt, args->args);
416
417    print_modules();
418    dump_stack();
419    print_oops_end_marker();
420    /* Just a warning, don't kill lockdep. */
421    add_taint(taint, LOCKDEP_STILL_OK);
422}
423
424void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
425{
426    struct slowpath_args args;
427
428    args.fmt = fmt;
429    va_start(args.args, fmt);
430    warn_slowpath_common(file, line, __builtin_return_address(0),
431                 TAINT_WARN, &args);
432    va_end(args.args);
433}
434EXPORT_SYMBOL(warn_slowpath_fmt);
435
436void warn_slowpath_fmt_taint(const char *file, int line,
437                 unsigned taint, const char *fmt, ...)
438{
439    struct slowpath_args args;
440
441    args.fmt = fmt;
442    va_start(args.args, fmt);
443    warn_slowpath_common(file, line, __builtin_return_address(0),
444                 taint, &args);
445    va_end(args.args);
446}
447EXPORT_SYMBOL(warn_slowpath_fmt_taint);
448
449void warn_slowpath_null(const char *file, int line)
450{
451    warn_slowpath_common(file, line, __builtin_return_address(0),
452                 TAINT_WARN, NULL);
453}
454EXPORT_SYMBOL(warn_slowpath_null);
455#endif
456
457#ifdef CONFIG_CC_STACKPROTECTOR
458
459/*
460 * Called when gcc's -fstack-protector feature is used, and
461 * gcc detects corruption of the on-stack canary value
462 */
463__visible void __stack_chk_fail(void)
464{
465    panic("stack-protector: Kernel stack is corrupted in: %p\n",
466        __builtin_return_address(0));
467}
468EXPORT_SYMBOL(__stack_chk_fail);
469
470#endif
471
472core_param(panic, panic_timeout, int, 0644);
473core_param(pause_on_oops, pause_on_oops, int, 0644);
474
475static int __init oops_setup(char *s)
476{
477    if (!s)
478        return -EINVAL;
479    if (!strcmp(s, "panic"))
480        panic_on_oops = 1;
481    return 0;
482}
483early_param("oops", oops_setup);
484

Archive Download this file



interactive