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>
32 #define PANIC_TIMER_STEP 100
33 #define PANIC_BLINK_SPD 18
35 int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE;
36 static unsigned long tainted_mask;
37 static int pause_on_oops;
38 static int pause_on_oops_flag;
39 static DEFINE_SPINLOCK(pause_on_oops_lock);
40 bool crash_kexec_post_notifiers;
41 int panic_on_warn __read_mostly;
43 int panic_timeout = CONFIG_PANIC_TIMEOUT;
44 EXPORT_SYMBOL_GPL(panic_timeout);
46 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
48 EXPORT_SYMBOL(panic_notifier_list);
50 static long no_blink(int state)
55 /* Returns how long it waited in ms */
56 long (*panic_blink)(int state);
57 EXPORT_SYMBOL(panic_blink);
60 * Stop ourself in panic -- architecture code may override this
62 void __weak panic_smp_self_stop(void)
69 * Stop ourselves in NMI context if another CPU has already panicked. Arch code
70 * may override this to prepare for crash dumping, e.g. save regs info.
72 void __weak nmi_panic_self_stop(struct pt_regs *regs)
74 panic_smp_self_stop();
78 * Stop other CPUs in panic. Architecture dependent code may override this
79 * with more suitable version. For example, if the architecture supports
80 * crash dump, it should save registers of each stopped CPU and disable
81 * per-CPU features such as virtualization extensions.
83 void __weak crash_smp_send_stop(void)
85 static int cpus_stopped;
88 * This function can be called twice in panic path, but obviously
89 * we execute this only once.
95 * Note smp_send_stop is the usual smp shutdown function, which
96 * unfortunately means it may not be hardened to work in a panic
103 atomic_t panic_cpu = ATOMIC_INIT(PANIC_CPU_INVALID);
106 * A variant of panic() called from NMI context. We return if we've already
107 * panicked on this CPU. If another CPU already panicked, loop in
108 * nmi_panic_self_stop() which can provide architecture dependent code such
109 * as saving register state for crash dump.
111 void nmi_panic(struct pt_regs *regs, const char *msg)
115 cpu = raw_smp_processor_id();
116 old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, cpu);
118 if (old_cpu == PANIC_CPU_INVALID)
120 else if (old_cpu != cpu)
121 nmi_panic_self_stop(regs);
123 EXPORT_SYMBOL(nmi_panic);
126 * panic - halt the system
127 * @fmt: The text string to print
129 * Display a message, then perform cleanups.
131 * This function never returns.
133 void panic(const char *fmt, ...)
135 static char buf[1024];
139 int old_cpu, this_cpu;
140 bool _crash_kexec_post_notifiers = crash_kexec_post_notifiers;
143 * Disable local interrupts. This will prevent panic_smp_self_stop
144 * from deadlocking the first cpu that invokes the panic, since
145 * there is nothing to prevent an interrupt handler (that runs
146 * after setting panic_cpu) from invoking panic() again.
149 preempt_disable_notrace();
152 * It's possible to come here directly from a panic-assertion and
153 * not have preempt disabled. Some functions called from here want
154 * preempt to be disabled. No point enabling it later though...
156 * Only one CPU is allowed to execute the panic code from here. For
157 * multiple parallel invocations of panic, all other CPUs either
158 * stop themself or will wait until they are stopped by the 1st CPU
159 * with smp_send_stop().
161 * `old_cpu == PANIC_CPU_INVALID' means this is the 1st CPU which
162 * comes here, so go ahead.
163 * `old_cpu == this_cpu' means we came from nmi_panic() which sets
164 * panic_cpu to this CPU. In this case, this is also the 1st CPU.
166 this_cpu = raw_smp_processor_id();
167 old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, this_cpu);
169 if (old_cpu != PANIC_CPU_INVALID && old_cpu != this_cpu)
170 panic_smp_self_stop();
175 vsnprintf(buf, sizeof(buf), fmt, args);
177 pr_emerg("Kernel panic - not syncing: %s\n", buf);
178 #ifdef CONFIG_DEBUG_BUGVERBOSE
180 * Avoid nested stack-dumping if a panic occurs during oops processing
182 if (!test_taint(TAINT_DIE) && oops_in_progress <= 1)
187 * If we have crashed and we have a crash kernel loaded let it handle
189 * If we want to run this after calling panic_notifiers, pass
190 * the "crash_kexec_post_notifiers" option to the kernel.
192 * Bypass the panic_cpu check and call __crash_kexec directly.
194 if (!_crash_kexec_post_notifiers) {
195 printk_safe_flush_on_panic();
199 * Note smp_send_stop is the usual smp shutdown function, which
200 * unfortunately means it may not be hardened to work in a
206 * If we want to do crash dump after notifier calls and
207 * kmsg_dump, we will need architecture dependent extra
208 * works in addition to stopping other CPUs.
210 crash_smp_send_stop();
214 * Run any panic handlers, including those that might need to
215 * add information to the kmsg dump output.
217 atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
219 /* Call flush even twice. It tries harder with a single online CPU */
220 printk_safe_flush_on_panic();
221 kmsg_dump(KMSG_DUMP_PANIC);
224 * If you doubt kdump always works fine in any situation,
225 * "crash_kexec_post_notifiers" offers you a chance to run
226 * panic_notifiers and dumping kmsg before kdump.
227 * Note: since some panic_notifiers can make crashed kernel
228 * more unstable, it can increase risks of the kdump failure too.
230 * Bypass the panic_cpu check and call __crash_kexec directly.
232 if (_crash_kexec_post_notifiers)
241 * We may have ended up stopping the CPU holding the lock (in
242 * smp_send_stop()) while still having some valuable data in the console
243 * buffer. Try to acquire the lock then release it regardless of the
244 * result. The release will also print the buffers out. Locks debug
245 * should be disabled to avoid reporting bad unlock balance when
246 * panic() is not being callled from OOPS.
249 console_flush_on_panic();
252 panic_blink = no_blink;
254 if (panic_timeout > 0) {
256 * Delay timeout seconds before rebooting the machine.
257 * We can't use the "normal" timers since we just panicked.
259 pr_emerg("Rebooting in %d seconds..\n", panic_timeout);
261 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
262 touch_nmi_watchdog();
264 i += panic_blink(state ^= 1);
265 i_next = i + 3600 / PANIC_BLINK_SPD;
267 mdelay(PANIC_TIMER_STEP);
270 if (panic_timeout != 0) {
272 * This will not be a clean reboot, with everything
273 * shutting down. But if there is a chance of
274 * rebooting the system it will be rebooted.
280 extern int stop_a_enabled;
281 /* Make sure the user can actually press Stop-A (L1-A) */
283 pr_emerg("Press Stop-A (L1-A) from sun keyboard or send break\n"
284 "twice on console to return to the boot prom\n");
287 #if defined(CONFIG_S390)
289 unsigned long caller;
291 caller = (unsigned long)__builtin_return_address(0);
292 disabled_wait(caller);
295 pr_emerg("---[ end Kernel panic - not syncing: %s\n", buf);
297 for (i = 0; ; i += PANIC_TIMER_STEP) {
298 touch_softlockup_watchdog();
300 i += panic_blink(state ^= 1);
301 i_next = i + 3600 / PANIC_BLINK_SPD;
303 mdelay(PANIC_TIMER_STEP);
307 EXPORT_SYMBOL(panic);
310 * TAINT_FORCED_RMMOD could be a per-module flag but the module
311 * is being removed anyway.
313 const struct taint_flag taint_flags[TAINT_FLAGS_COUNT] = {
314 { 'P', 'G', true }, /* TAINT_PROPRIETARY_MODULE */
315 { 'F', ' ', true }, /* TAINT_FORCED_MODULE */
316 { 'S', ' ', false }, /* TAINT_CPU_OUT_OF_SPEC */
317 { 'R', ' ', false }, /* TAINT_FORCED_RMMOD */
318 { 'M', ' ', false }, /* TAINT_MACHINE_CHECK */
319 { 'B', ' ', false }, /* TAINT_BAD_PAGE */
320 { 'U', ' ', false }, /* TAINT_USER */
321 { 'D', ' ', false }, /* TAINT_DIE */
322 { 'A', ' ', false }, /* TAINT_OVERRIDDEN_ACPI_TABLE */
323 { 'W', ' ', false }, /* TAINT_WARN */
324 { 'C', ' ', true }, /* TAINT_CRAP */
325 { 'I', ' ', false }, /* TAINT_FIRMWARE_WORKAROUND */
326 { 'O', ' ', true }, /* TAINT_OOT_MODULE */
327 { 'E', ' ', true }, /* TAINT_UNSIGNED_MODULE */
328 { 'L', ' ', false }, /* TAINT_SOFTLOCKUP */
329 { 'K', ' ', true }, /* TAINT_LIVEPATCH */
333 * print_tainted - return a string to represent the kernel taint state.
335 * 'P' - Proprietary module has been loaded.
336 * 'F' - Module has been forcibly loaded.
337 * 'S' - SMP with CPUs not designed for SMP.
338 * 'R' - User forced a module unload.
339 * 'M' - System experienced a machine check exception.
340 * 'B' - System has hit bad_page.
341 * 'U' - Userspace-defined naughtiness.
342 * 'D' - Kernel has oopsed before
343 * 'A' - ACPI table overridden.
344 * 'W' - Taint on warning.
345 * 'C' - modules from drivers/staging are loaded.
346 * 'I' - Working around severe firmware bug.
347 * 'O' - Out-of-tree module has been loaded.
348 * 'E' - Unsigned module has been loaded.
349 * 'L' - A soft lockup has previously occurred.
350 * 'K' - Kernel has been live patched.
352 * The string is overwritten by the next call to print_tainted().
354 const char *print_tainted(void)
356 static char buf[TAINT_FLAGS_COUNT + sizeof("Tainted: ")];
362 s = buf + sprintf(buf, "Tainted: ");
363 for (i = 0; i < TAINT_FLAGS_COUNT; i++) {
364 const struct taint_flag *t = &taint_flags[i];
365 *s++ = test_bit(i, &tainted_mask) ?
366 t->c_true : t->c_false;
370 snprintf(buf, sizeof(buf), "Not tainted");
375 int test_taint(unsigned flag)
377 return test_bit(flag, &tainted_mask);
379 EXPORT_SYMBOL(test_taint);
381 unsigned long get_taint(void)
387 * add_taint: add a taint flag if not already set.
388 * @flag: one of the TAINT_* constants.
389 * @lockdep_ok: whether lock debugging is still OK.
391 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
392 * some notewortht-but-not-corrupting cases, it can be set to true.
394 void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
396 if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
397 pr_warn("Disabling lock debugging due to kernel taint\n");
399 set_bit(flag, &tainted_mask);
401 EXPORT_SYMBOL(add_taint);
403 static void spin_msec(int msecs)
407 for (i = 0; i < msecs; i++) {
408 touch_nmi_watchdog();
414 * It just happens that oops_enter() and oops_exit() are identically
417 static void do_oops_enter_exit(void)
420 static int spin_counter;
425 spin_lock_irqsave(&pause_on_oops_lock, flags);
426 if (pause_on_oops_flag == 0) {
427 /* This CPU may now print the oops message */
428 pause_on_oops_flag = 1;
430 /* We need to stall this CPU */
432 /* This CPU gets to do the counting */
433 spin_counter = pause_on_oops;
435 spin_unlock(&pause_on_oops_lock);
436 spin_msec(MSEC_PER_SEC);
437 spin_lock(&pause_on_oops_lock);
438 } while (--spin_counter);
439 pause_on_oops_flag = 0;
441 /* This CPU waits for a different one */
442 while (spin_counter) {
443 spin_unlock(&pause_on_oops_lock);
445 spin_lock(&pause_on_oops_lock);
449 spin_unlock_irqrestore(&pause_on_oops_lock, flags);
453 * Return true if the calling CPU is allowed to print oops-related info.
454 * This is a bit racy..
456 int oops_may_print(void)
458 return pause_on_oops_flag == 0;
462 * Called when the architecture enters its oops handler, before it prints
463 * anything. If this is the first CPU to oops, and it's oopsing the first
464 * time then let it proceed.
466 * This is all enabled by the pause_on_oops kernel boot option. We do all
467 * this to ensure that oopses don't scroll off the screen. It has the
468 * side-effect of preventing later-oopsing CPUs from mucking up the display,
471 * It turns out that the CPU which is allowed to print ends up pausing for
472 * the right duration, whereas all the other CPUs pause for twice as long:
473 * once in oops_enter(), once in oops_exit().
475 void oops_enter(void)
478 /* can't trust the integrity of the kernel anymore: */
480 do_oops_enter_exit();
484 * 64-bit random ID for oopses:
488 static int init_oops_id(void)
491 get_random_bytes(&oops_id, sizeof(oops_id));
497 late_initcall(init_oops_id);
499 void print_oops_end_marker(void)
502 pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id);
506 * Called when the architecture exits its oops handler, after printing
511 do_oops_enter_exit();
512 print_oops_end_marker();
513 kmsg_dump(KMSG_DUMP_OOPS);
521 void __warn(const char *file, int line, void *caller, unsigned taint,
522 struct pt_regs *regs, struct warn_args *args)
524 disable_trace_on_warning();
526 pr_warn("------------[ cut here ]------------\n");
529 pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS\n",
530 raw_smp_processor_id(), current->pid, file, line,
533 pr_warn("WARNING: CPU: %d PID: %d at %pS\n",
534 raw_smp_processor_id(), current->pid, caller);
537 vprintk(args->fmt, args->args);
541 * This thread may hit another WARN() in the panic path.
542 * Resetting this prevents additional WARN() from panicking the
543 * system on this thread. Other threads are blocked by the
544 * panic_mutex in panic().
547 panic("panic_on_warn set ...\n");
557 print_oops_end_marker();
559 /* Just a warning, don't kill lockdep. */
560 add_taint(taint, LOCKDEP_STILL_OK);
563 #ifdef WANT_WARN_ON_SLOWPATH
564 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
566 struct warn_args args;
569 va_start(args.args, fmt);
570 __warn(file, line, __builtin_return_address(0), TAINT_WARN, NULL,
574 EXPORT_SYMBOL(warn_slowpath_fmt);
576 void warn_slowpath_fmt_taint(const char *file, int line,
577 unsigned taint, const char *fmt, ...)
579 struct warn_args args;
582 va_start(args.args, fmt);
583 __warn(file, line, __builtin_return_address(0), taint, NULL, &args);
586 EXPORT_SYMBOL(warn_slowpath_fmt_taint);
588 void warn_slowpath_null(const char *file, int line)
590 __warn(file, line, __builtin_return_address(0), TAINT_WARN, NULL, NULL);
592 EXPORT_SYMBOL(warn_slowpath_null);
595 #ifdef CONFIG_CC_STACKPROTECTOR
598 * Called when gcc's -fstack-protector feature is used, and
599 * gcc detects corruption of the on-stack canary value
601 __visible void __stack_chk_fail(void)
603 panic("stack-protector: Kernel stack is corrupted in: %p\n",
604 __builtin_return_address(0));
606 EXPORT_SYMBOL(__stack_chk_fail);
610 #ifdef CONFIG_ARCH_HAS_REFCOUNT
611 void refcount_error_report(struct pt_regs *regs, const char *err)
613 WARN_RATELIMIT(1, "refcount_t %s at %pB in %s[%d], uid/euid: %u/%u\n",
614 err, (void *)instruction_pointer(regs),
615 current->comm, task_pid_nr(current),
616 from_kuid_munged(&init_user_ns, current_uid()),
617 from_kuid_munged(&init_user_ns, current_euid()));
621 core_param(panic, panic_timeout, int, 0644);
622 core_param(pause_on_oops, pause_on_oops, int, 0644);
623 core_param(panic_on_warn, panic_on_warn, int, 0644);
624 core_param(crash_kexec_post_notifiers, crash_kexec_post_notifiers, bool, 0644);
626 static int __init oops_setup(char *s)
630 if (!strcmp(s, "panic"))
634 early_param("oops", oops_setup);