4 * Copyright (C) 1991, 1992 Linus Torvalds
8 * This function is used through-out the kernel (including mm and fs)
9 * to indicate a major problem.
11 #include <linux/debug_locks.h>
12 #include <linux/sched/debug.h>
13 #include <linux/interrupt.h>
14 #include <linux/kmsg_dump.h>
15 #include <linux/kallsyms.h>
16 #include <linux/notifier.h>
17 #include <linux/vt_kern.h>
18 #include <linux/module.h>
19 #include <linux/random.h>
20 #include <linux/ftrace.h>
21 #include <linux/reboot.h>
22 #include <linux/delay.h>
23 #include <linux/kexec.h>
24 #include <linux/sched.h>
25 #include <linux/sysrq.h>
26 #include <linux/init.h>
27 #include <linux/nmi.h>
28 #include <linux/console.h>
29 #include <linux/bug.h>
30 #include <linux/ratelimit.h>
31 #include <linux/debugfs.h>
32 #include <linux/sysfs.h>
33 #include <asm/sections.h>
35 #define PANIC_TIMER_STEP 100
36 #define PANIC_BLINK_SPD 18
38 int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE;
39 static unsigned long tainted_mask =
40 IS_ENABLED(CONFIG_GCC_PLUGIN_RANDSTRUCT) ? (1 << TAINT_RANDSTRUCT) : 0;
41 static int pause_on_oops;
42 static int pause_on_oops_flag;
43 static DEFINE_SPINLOCK(pause_on_oops_lock);
44 bool crash_kexec_post_notifiers;
45 int panic_on_warn __read_mostly;
46 static unsigned int warn_limit __read_mostly;
48 int panic_timeout = CONFIG_PANIC_TIMEOUT;
49 EXPORT_SYMBOL_GPL(panic_timeout);
51 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
53 EXPORT_SYMBOL(panic_notifier_list);
56 static struct ctl_table kern_panic_table[] = {
58 .procname = "warn_limit",
60 .maxlen = sizeof(warn_limit),
62 .proc_handler = proc_douintvec,
67 static __init int kernel_panic_sysctls_init(void)
69 register_sysctl_init("kernel", kern_panic_table);
72 late_initcall(kernel_panic_sysctls_init);
75 static atomic_t warn_count = ATOMIC_INIT(0);
78 static ssize_t warn_count_show(struct kobject *kobj, struct kobj_attribute *attr,
81 return sysfs_emit(page, "%d\n", atomic_read(&warn_count));
84 static struct kobj_attribute warn_count_attr = __ATTR_RO(warn_count);
86 static __init int kernel_panic_sysfs_init(void)
88 sysfs_add_file_to_group(kernel_kobj, &warn_count_attr.attr, NULL);
91 late_initcall(kernel_panic_sysfs_init);
94 static long no_blink(int state)
99 /* Returns how long it waited in ms */
100 long (*panic_blink)(int state);
101 EXPORT_SYMBOL(panic_blink);
104 * Stop ourself in panic -- architecture code may override this
106 void __weak panic_smp_self_stop(void)
113 * Stop ourselves in NMI context if another CPU has already panicked. Arch code
114 * may override this to prepare for crash dumping, e.g. save regs info.
116 void __weak nmi_panic_self_stop(struct pt_regs *regs)
118 panic_smp_self_stop();
122 * Stop other CPUs in panic. Architecture dependent code may override this
123 * with more suitable version. For example, if the architecture supports
124 * crash dump, it should save registers of each stopped CPU and disable
125 * per-CPU features such as virtualization extensions.
127 void __weak crash_smp_send_stop(void)
129 static int cpus_stopped;
132 * This function can be called twice in panic path, but obviously
133 * we execute this only once.
139 * Note smp_send_stop is the usual smp shutdown function, which
140 * unfortunately means it may not be hardened to work in a panic
147 atomic_t panic_cpu = ATOMIC_INIT(PANIC_CPU_INVALID);
150 * A variant of panic() called from NMI context. We return if we've already
151 * panicked on this CPU. If another CPU already panicked, loop in
152 * nmi_panic_self_stop() which can provide architecture dependent code such
153 * as saving register state for crash dump.
155 void nmi_panic(struct pt_regs *regs, const char *msg)
159 cpu = raw_smp_processor_id();
160 old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, cpu);
162 if (old_cpu == PANIC_CPU_INVALID)
164 else if (old_cpu != cpu)
165 nmi_panic_self_stop(regs);
167 EXPORT_SYMBOL(nmi_panic);
169 void check_panic_on_warn(const char *origin)
174 panic("%s: panic_on_warn set ...\n", origin);
176 limit = READ_ONCE(warn_limit);
177 if (atomic_inc_return(&warn_count) >= limit && limit)
178 panic("%s: system warned too often (kernel.warn_limit is %d)",
183 * panic - halt the system
184 * @fmt: The text string to print
186 * Display a message, then perform cleanups.
188 * This function never returns.
190 void panic(const char *fmt, ...)
192 static char buf[1024];
196 int old_cpu, this_cpu;
197 bool _crash_kexec_post_notifiers = crash_kexec_post_notifiers;
201 * This thread may hit another WARN() in the panic path.
202 * Resetting this prevents additional WARN() from panicking the
203 * system on this thread. Other threads are blocked by the
204 * panic_mutex in panic().
210 * Disable local interrupts. This will prevent panic_smp_self_stop
211 * from deadlocking the first cpu that invokes the panic, since
212 * there is nothing to prevent an interrupt handler (that runs
213 * after setting panic_cpu) from invoking panic() again.
216 preempt_disable_notrace();
219 * It's possible to come here directly from a panic-assertion and
220 * not have preempt disabled. Some functions called from here want
221 * preempt to be disabled. No point enabling it later though...
223 * Only one CPU is allowed to execute the panic code from here. For
224 * multiple parallel invocations of panic, all other CPUs either
225 * stop themself or will wait until they are stopped by the 1st CPU
226 * with smp_send_stop().
228 * `old_cpu == PANIC_CPU_INVALID' means this is the 1st CPU which
229 * comes here, so go ahead.
230 * `old_cpu == this_cpu' means we came from nmi_panic() which sets
231 * panic_cpu to this CPU. In this case, this is also the 1st CPU.
233 this_cpu = raw_smp_processor_id();
234 old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, this_cpu);
236 if (old_cpu != PANIC_CPU_INVALID && old_cpu != this_cpu)
237 panic_smp_self_stop();
242 vsnprintf(buf, sizeof(buf), fmt, args);
244 pr_emerg("Kernel panic - not syncing: %s\n", buf);
245 #ifdef CONFIG_DEBUG_BUGVERBOSE
247 * Avoid nested stack-dumping if a panic occurs during oops processing
249 if (!test_taint(TAINT_DIE) && oops_in_progress <= 1)
254 * If we have crashed and we have a crash kernel loaded let it handle
256 * If we want to run this after calling panic_notifiers, pass
257 * the "crash_kexec_post_notifiers" option to the kernel.
259 * Bypass the panic_cpu check and call __crash_kexec directly.
261 if (!_crash_kexec_post_notifiers) {
262 printk_safe_flush_on_panic();
266 * Note smp_send_stop is the usual smp shutdown function, which
267 * unfortunately means it may not be hardened to work in a
273 * If we want to do crash dump after notifier calls and
274 * kmsg_dump, we will need architecture dependent extra
275 * works in addition to stopping other CPUs.
277 crash_smp_send_stop();
281 * Run any panic handlers, including those that might need to
282 * add information to the kmsg dump output.
284 atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
286 /* Call flush even twice. It tries harder with a single online CPU */
287 printk_safe_flush_on_panic();
288 kmsg_dump(KMSG_DUMP_PANIC);
291 * If you doubt kdump always works fine in any situation,
292 * "crash_kexec_post_notifiers" offers you a chance to run
293 * panic_notifiers and dumping kmsg before kdump.
294 * Note: since some panic_notifiers can make crashed kernel
295 * more unstable, it can increase risks of the kdump failure too.
297 * Bypass the panic_cpu check and call __crash_kexec directly.
299 if (_crash_kexec_post_notifiers)
308 * We may have ended up stopping the CPU holding the lock (in
309 * smp_send_stop()) while still having some valuable data in the console
310 * buffer. Try to acquire the lock then release it regardless of the
311 * result. The release will also print the buffers out. Locks debug
312 * should be disabled to avoid reporting bad unlock balance when
313 * panic() is not being callled from OOPS.
316 console_flush_on_panic();
319 panic_blink = no_blink;
321 if (panic_timeout > 0) {
323 * Delay timeout seconds before rebooting the machine.
324 * We can't use the "normal" timers since we just panicked.
326 pr_emerg("Rebooting in %d seconds..\n", panic_timeout);
328 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
329 touch_nmi_watchdog();
331 i += panic_blink(state ^= 1);
332 i_next = i + 3600 / PANIC_BLINK_SPD;
334 mdelay(PANIC_TIMER_STEP);
337 if (panic_timeout != 0) {
339 * This will not be a clean reboot, with everything
340 * shutting down. But if there is a chance of
341 * rebooting the system it will be rebooted.
347 extern int stop_a_enabled;
348 /* Make sure the user can actually press Stop-A (L1-A) */
350 pr_emerg("Press Stop-A (L1-A) from sun keyboard or send break\n"
351 "twice on console to return to the boot prom\n");
354 #if defined(CONFIG_S390)
356 unsigned long caller;
358 caller = (unsigned long)__builtin_return_address(0);
359 disabled_wait(caller);
362 pr_emerg("---[ end Kernel panic - not syncing: %s ]---\n", buf);
364 for (i = 0; ; i += PANIC_TIMER_STEP) {
365 touch_softlockup_watchdog();
367 i += panic_blink(state ^= 1);
368 i_next = i + 3600 / PANIC_BLINK_SPD;
370 mdelay(PANIC_TIMER_STEP);
374 EXPORT_SYMBOL(panic);
377 * TAINT_FORCED_RMMOD could be a per-module flag but the module
378 * is being removed anyway.
380 const struct taint_flag taint_flags[TAINT_FLAGS_COUNT] = {
381 [ TAINT_PROPRIETARY_MODULE ] = { 'P', 'G', true },
382 [ TAINT_FORCED_MODULE ] = { 'F', ' ', true },
383 [ TAINT_CPU_OUT_OF_SPEC ] = { 'S', ' ', false },
384 [ TAINT_FORCED_RMMOD ] = { 'R', ' ', false },
385 [ TAINT_MACHINE_CHECK ] = { 'M', ' ', false },
386 [ TAINT_BAD_PAGE ] = { 'B', ' ', false },
387 [ TAINT_USER ] = { 'U', ' ', false },
388 [ TAINT_DIE ] = { 'D', ' ', false },
389 [ TAINT_OVERRIDDEN_ACPI_TABLE ] = { 'A', ' ', false },
390 [ TAINT_WARN ] = { 'W', ' ', false },
391 [ TAINT_CRAP ] = { 'C', ' ', true },
392 [ TAINT_FIRMWARE_WORKAROUND ] = { 'I', ' ', false },
393 [ TAINT_OOT_MODULE ] = { 'O', ' ', true },
394 [ TAINT_UNSIGNED_MODULE ] = { 'E', ' ', true },
395 [ TAINT_SOFTLOCKUP ] = { 'L', ' ', false },
396 [ TAINT_LIVEPATCH ] = { 'K', ' ', true },
397 [ TAINT_AUX ] = { 'X', ' ', true },
398 [ TAINT_RANDSTRUCT ] = { 'T', ' ', true },
402 * print_tainted - return a string to represent the kernel taint state.
404 * For individual taint flag meanings, see Documentation/sysctl/kernel.txt
406 * The string is overwritten by the next call to print_tainted(),
407 * but is always NULL terminated.
409 const char *print_tainted(void)
411 static char buf[TAINT_FLAGS_COUNT + sizeof("Tainted: ")];
413 BUILD_BUG_ON(ARRAY_SIZE(taint_flags) != TAINT_FLAGS_COUNT);
419 s = buf + sprintf(buf, "Tainted: ");
420 for (i = 0; i < TAINT_FLAGS_COUNT; i++) {
421 const struct taint_flag *t = &taint_flags[i];
422 *s++ = test_bit(i, &tainted_mask) ?
423 t->c_true : t->c_false;
427 snprintf(buf, sizeof(buf), "Not tainted");
432 int test_taint(unsigned flag)
434 return test_bit(flag, &tainted_mask);
436 EXPORT_SYMBOL(test_taint);
438 unsigned long get_taint(void)
444 * add_taint: add a taint flag if not already set.
445 * @flag: one of the TAINT_* constants.
446 * @lockdep_ok: whether lock debugging is still OK.
448 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
449 * some notewortht-but-not-corrupting cases, it can be set to true.
451 void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
453 if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
454 pr_warn("Disabling lock debugging due to kernel taint\n");
456 set_bit(flag, &tainted_mask);
458 EXPORT_SYMBOL(add_taint);
460 static void spin_msec(int msecs)
464 for (i = 0; i < msecs; i++) {
465 touch_nmi_watchdog();
471 * It just happens that oops_enter() and oops_exit() are identically
474 static void do_oops_enter_exit(void)
477 static int spin_counter;
482 spin_lock_irqsave(&pause_on_oops_lock, flags);
483 if (pause_on_oops_flag == 0) {
484 /* This CPU may now print the oops message */
485 pause_on_oops_flag = 1;
487 /* We need to stall this CPU */
489 /* This CPU gets to do the counting */
490 spin_counter = pause_on_oops;
492 spin_unlock(&pause_on_oops_lock);
493 spin_msec(MSEC_PER_SEC);
494 spin_lock(&pause_on_oops_lock);
495 } while (--spin_counter);
496 pause_on_oops_flag = 0;
498 /* This CPU waits for a different one */
499 while (spin_counter) {
500 spin_unlock(&pause_on_oops_lock);
502 spin_lock(&pause_on_oops_lock);
506 spin_unlock_irqrestore(&pause_on_oops_lock, flags);
510 * Return true if the calling CPU is allowed to print oops-related info.
511 * This is a bit racy..
513 int oops_may_print(void)
515 return pause_on_oops_flag == 0;
519 * Called when the architecture enters its oops handler, before it prints
520 * anything. If this is the first CPU to oops, and it's oopsing the first
521 * time then let it proceed.
523 * This is all enabled by the pause_on_oops kernel boot option. We do all
524 * this to ensure that oopses don't scroll off the screen. It has the
525 * side-effect of preventing later-oopsing CPUs from mucking up the display,
528 * It turns out that the CPU which is allowed to print ends up pausing for
529 * the right duration, whereas all the other CPUs pause for twice as long:
530 * once in oops_enter(), once in oops_exit().
532 void oops_enter(void)
535 /* can't trust the integrity of the kernel anymore: */
537 do_oops_enter_exit();
541 * 64-bit random ID for oopses:
545 static int init_oops_id(void)
548 get_random_bytes(&oops_id, sizeof(oops_id));
554 late_initcall(init_oops_id);
556 void print_oops_end_marker(void)
559 pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id);
563 * Called when the architecture exits its oops handler, after printing
568 do_oops_enter_exit();
569 print_oops_end_marker();
570 kmsg_dump(KMSG_DUMP_OOPS);
578 void __warn(const char *file, int line, void *caller, unsigned taint,
579 struct pt_regs *regs, struct warn_args *args)
581 disable_trace_on_warning();
587 pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS\n",
588 raw_smp_processor_id(), current->pid, file, line,
591 pr_warn("WARNING: CPU: %d PID: %d at %pS\n",
592 raw_smp_processor_id(), current->pid, caller);
595 vprintk(args->fmt, args->args);
597 check_panic_on_warn("kernel");
606 print_irqtrace_events(current);
608 print_oops_end_marker();
610 /* Just a warning, don't kill lockdep. */
611 add_taint(taint, LOCKDEP_STILL_OK);
614 #ifdef WANT_WARN_ON_SLOWPATH
615 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
617 struct warn_args args;
620 va_start(args.args, fmt);
621 __warn(file, line, __builtin_return_address(0), TAINT_WARN, NULL,
625 EXPORT_SYMBOL(warn_slowpath_fmt);
627 void warn_slowpath_fmt_taint(const char *file, int line,
628 unsigned taint, const char *fmt, ...)
630 struct warn_args args;
633 va_start(args.args, fmt);
634 __warn(file, line, __builtin_return_address(0), taint, NULL, &args);
637 EXPORT_SYMBOL(warn_slowpath_fmt_taint);
639 void warn_slowpath_null(const char *file, int line)
642 __warn(file, line, __builtin_return_address(0), TAINT_WARN, NULL, NULL);
644 EXPORT_SYMBOL(warn_slowpath_null);
646 void __warn_printk(const char *fmt, ...)
656 EXPORT_SYMBOL(__warn_printk);
661 /* Support resetting WARN*_ONCE state */
663 static int clear_warn_once_set(void *data, u64 val)
665 generic_bug_clear_once();
666 memset(__start_once, 0, __end_once - __start_once);
670 DEFINE_SIMPLE_ATTRIBUTE(clear_warn_once_fops,
675 static __init int register_warn_debugfs(void)
677 /* Don't care about failure */
678 debugfs_create_file("clear_warn_once", 0200, NULL,
679 NULL, &clear_warn_once_fops);
683 device_initcall(register_warn_debugfs);
686 #ifdef CONFIG_STACKPROTECTOR
689 * Called when gcc's -fstack-protector feature is used, and
690 * gcc detects corruption of the on-stack canary value
692 __visible void __stack_chk_fail(void)
694 panic("stack-protector: Kernel stack is corrupted in: %pB",
695 __builtin_return_address(0));
697 EXPORT_SYMBOL(__stack_chk_fail);
701 #ifdef CONFIG_ARCH_HAS_REFCOUNT
702 void refcount_error_report(struct pt_regs *regs, const char *err)
704 WARN_RATELIMIT(1, "refcount_t %s at %pB in %s[%d], uid/euid: %u/%u\n",
705 err, (void *)instruction_pointer(regs),
706 current->comm, task_pid_nr(current),
707 from_kuid_munged(&init_user_ns, current_uid()),
708 from_kuid_munged(&init_user_ns, current_euid()));
712 core_param(panic, panic_timeout, int, 0644);
713 core_param(pause_on_oops, pause_on_oops, int, 0644);
714 core_param(panic_on_warn, panic_on_warn, int, 0644);
715 core_param(crash_kexec_post_notifiers, crash_kexec_post_notifiers, bool, 0644);
717 static int __init oops_setup(char *s)
721 if (!strcmp(s, "panic"))
725 early_param("oops", oops_setup);