1 // SPDX-License-Identifier: GPL-2.0-only
5 * Copyright (C) 1991, 1992 Linus Torvalds
9 * This function is used through-out the kernel (including mm and fs)
10 * to indicate a major problem.
12 #include <linux/debug_locks.h>
13 #include <linux/sched/debug.h>
14 #include <linux/interrupt.h>
15 #include <linux/kgdb.h>
16 #include <linux/kmsg_dump.h>
17 #include <linux/kallsyms.h>
18 #include <linux/notifier.h>
19 #include <linux/vt_kern.h>
20 #include <linux/module.h>
21 #include <linux/random.h>
22 #include <linux/ftrace.h>
23 #include <linux/reboot.h>
24 #include <linux/delay.h>
25 #include <linux/kexec.h>
26 #include <linux/panic_notifier.h>
27 #include <linux/sched.h>
28 #include <linux/sysrq.h>
29 #include <linux/init.h>
30 #include <linux/nmi.h>
31 #include <linux/console.h>
32 #include <linux/bug.h>
33 #include <linux/ratelimit.h>
34 #include <linux/debugfs.h>
35 #include <trace/events/error_report.h>
36 #include <asm/sections.h>
38 #define PANIC_TIMER_STEP 100
39 #define PANIC_BLINK_SPD 18
43 * Should we dump all CPUs backtraces in an oops event?
44 * Defaults to 0, can be changed via sysctl.
46 unsigned int __read_mostly sysctl_oops_all_cpu_backtrace;
47 #endif /* CONFIG_SMP */
49 int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE;
50 static unsigned long tainted_mask =
51 IS_ENABLED(CONFIG_GCC_PLUGIN_RANDSTRUCT) ? (1 << TAINT_RANDSTRUCT) : 0;
52 static int pause_on_oops;
53 static int pause_on_oops_flag;
54 static DEFINE_SPINLOCK(pause_on_oops_lock);
55 bool crash_kexec_post_notifiers;
56 int panic_on_warn __read_mostly;
57 unsigned long panic_on_taint;
58 bool panic_on_taint_nousertaint = false;
60 int panic_timeout = CONFIG_PANIC_TIMEOUT;
61 EXPORT_SYMBOL_GPL(panic_timeout);
63 #define PANIC_PRINT_TASK_INFO 0x00000001
64 #define PANIC_PRINT_MEM_INFO 0x00000002
65 #define PANIC_PRINT_TIMER_INFO 0x00000004
66 #define PANIC_PRINT_LOCK_INFO 0x00000008
67 #define PANIC_PRINT_FTRACE_INFO 0x00000010
68 #define PANIC_PRINT_ALL_PRINTK_MSG 0x00000020
69 #define PANIC_PRINT_ALL_CPU_BT 0x00000040
70 unsigned long panic_print;
72 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
74 EXPORT_SYMBOL(panic_notifier_list);
76 static long no_blink(int state)
81 /* Returns how long it waited in ms */
82 long (*panic_blink)(int state);
83 EXPORT_SYMBOL(panic_blink);
86 * Stop ourself in panic -- architecture code may override this
88 void __weak panic_smp_self_stop(void)
95 * Stop ourselves in NMI context if another CPU has already panicked. Arch code
96 * may override this to prepare for crash dumping, e.g. save regs info.
98 void __weak nmi_panic_self_stop(struct pt_regs *regs)
100 panic_smp_self_stop();
104 * Stop other CPUs in panic. Architecture dependent code may override this
105 * with more suitable version. For example, if the architecture supports
106 * crash dump, it should save registers of each stopped CPU and disable
107 * per-CPU features such as virtualization extensions.
109 void __weak crash_smp_send_stop(void)
111 static int cpus_stopped;
114 * This function can be called twice in panic path, but obviously
115 * we execute this only once.
121 * Note smp_send_stop is the usual smp shutdown function, which
122 * unfortunately means it may not be hardened to work in a panic
129 atomic_t panic_cpu = ATOMIC_INIT(PANIC_CPU_INVALID);
132 * A variant of panic() called from NMI context. We return if we've already
133 * panicked on this CPU. If another CPU already panicked, loop in
134 * nmi_panic_self_stop() which can provide architecture dependent code such
135 * as saving register state for crash dump.
137 void nmi_panic(struct pt_regs *regs, const char *msg)
141 cpu = raw_smp_processor_id();
142 old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, cpu);
144 if (old_cpu == PANIC_CPU_INVALID)
146 else if (old_cpu != cpu)
147 nmi_panic_self_stop(regs);
149 EXPORT_SYMBOL(nmi_panic);
151 static void panic_print_sys_info(bool console_flush)
154 if (panic_print & PANIC_PRINT_ALL_PRINTK_MSG)
155 console_flush_on_panic(CONSOLE_REPLAY_ALL);
159 if (panic_print & PANIC_PRINT_ALL_CPU_BT)
160 trigger_all_cpu_backtrace();
162 if (panic_print & PANIC_PRINT_TASK_INFO)
165 if (panic_print & PANIC_PRINT_MEM_INFO)
168 if (panic_print & PANIC_PRINT_TIMER_INFO)
169 sysrq_timer_list_show();
171 if (panic_print & PANIC_PRINT_LOCK_INFO)
172 debug_show_all_locks();
174 if (panic_print & PANIC_PRINT_FTRACE_INFO)
175 ftrace_dump(DUMP_ALL);
179 * panic - halt the system
180 * @fmt: The text string to print
182 * Display a message, then perform cleanups.
184 * This function never returns.
186 void panic(const char *fmt, ...)
188 static char buf[1024];
190 long i, i_next = 0, len;
192 int old_cpu, this_cpu;
193 bool _crash_kexec_post_notifiers = crash_kexec_post_notifiers;
197 * This thread may hit another WARN() in the panic path.
198 * Resetting this prevents additional WARN() from panicking the
199 * system on this thread. Other threads are blocked by the
200 * panic_mutex in panic().
206 * Disable local interrupts. This will prevent panic_smp_self_stop
207 * from deadlocking the first cpu that invokes the panic, since
208 * there is nothing to prevent an interrupt handler (that runs
209 * after setting panic_cpu) from invoking panic() again.
212 preempt_disable_notrace();
215 * It's possible to come here directly from a panic-assertion and
216 * not have preempt disabled. Some functions called from here want
217 * preempt to be disabled. No point enabling it later though...
219 * Only one CPU is allowed to execute the panic code from here. For
220 * multiple parallel invocations of panic, all other CPUs either
221 * stop themself or will wait until they are stopped by the 1st CPU
222 * with smp_send_stop().
224 * `old_cpu == PANIC_CPU_INVALID' means this is the 1st CPU which
225 * comes here, so go ahead.
226 * `old_cpu == this_cpu' means we came from nmi_panic() which sets
227 * panic_cpu to this CPU. In this case, this is also the 1st CPU.
229 this_cpu = raw_smp_processor_id();
230 old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, this_cpu);
232 if (old_cpu != PANIC_CPU_INVALID && old_cpu != this_cpu)
233 panic_smp_self_stop();
238 len = vscnprintf(buf, sizeof(buf), fmt, args);
241 if (len && buf[len - 1] == '\n')
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 kgdb is enabled, give it a chance to run before we stop all
255 * the other CPUs or else we won't be able to debug processes left
261 * If we have crashed and we have a crash kernel loaded let it handle
263 * If we want to run this after calling panic_notifiers, pass
264 * the "crash_kexec_post_notifiers" option to the kernel.
266 * Bypass the panic_cpu check and call __crash_kexec directly.
268 if (!_crash_kexec_post_notifiers) {
272 * Note smp_send_stop is the usual smp shutdown function, which
273 * unfortunately means it may not be hardened to work in a
279 * If we want to do crash dump after notifier calls and
280 * kmsg_dump, we will need architecture dependent extra
281 * works in addition to stopping other CPUs.
283 crash_smp_send_stop();
287 * Run any panic handlers, including those that might need to
288 * add information to the kmsg dump output.
290 atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
292 panic_print_sys_info(false);
294 kmsg_dump(KMSG_DUMP_PANIC);
297 * If you doubt kdump always works fine in any situation,
298 * "crash_kexec_post_notifiers" offers you a chance to run
299 * panic_notifiers and dumping kmsg before kdump.
300 * Note: since some panic_notifiers can make crashed kernel
301 * more unstable, it can increase risks of the kdump failure too.
303 * Bypass the panic_cpu check and call __crash_kexec directly.
305 if (_crash_kexec_post_notifiers)
314 * We may have ended up stopping the CPU holding the lock (in
315 * smp_send_stop()) while still having some valuable data in the console
316 * buffer. Try to acquire the lock then release it regardless of the
317 * result. The release will also print the buffers out. Locks debug
318 * should be disabled to avoid reporting bad unlock balance when
319 * panic() is not being callled from OOPS.
322 console_flush_on_panic(CONSOLE_FLUSH_PENDING);
324 panic_print_sys_info(true);
327 panic_blink = no_blink;
329 if (panic_timeout > 0) {
331 * Delay timeout seconds before rebooting the machine.
332 * We can't use the "normal" timers since we just panicked.
334 pr_emerg("Rebooting in %d seconds..\n", panic_timeout);
336 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
337 touch_nmi_watchdog();
339 i += panic_blink(state ^= 1);
340 i_next = i + 3600 / PANIC_BLINK_SPD;
342 mdelay(PANIC_TIMER_STEP);
345 if (panic_timeout != 0) {
347 * This will not be a clean reboot, with everything
348 * shutting down. But if there is a chance of
349 * rebooting the system it will be rebooted.
351 if (panic_reboot_mode != REBOOT_UNDEFINED)
352 reboot_mode = panic_reboot_mode;
357 extern int stop_a_enabled;
358 /* Make sure the user can actually press Stop-A (L1-A) */
360 pr_emerg("Press Stop-A (L1-A) from sun keyboard or send break\n"
361 "twice on console to return to the boot prom\n");
364 #if defined(CONFIG_S390)
367 pr_emerg("---[ end Kernel panic - not syncing: %s ]---\n", buf);
369 /* Do not scroll important messages printed above */
372 for (i = 0; ; i += PANIC_TIMER_STEP) {
373 touch_softlockup_watchdog();
375 i += panic_blink(state ^= 1);
376 i_next = i + 3600 / PANIC_BLINK_SPD;
378 mdelay(PANIC_TIMER_STEP);
382 EXPORT_SYMBOL(panic);
385 * TAINT_FORCED_RMMOD could be a per-module flag but the module
386 * is being removed anyway.
388 const struct taint_flag taint_flags[TAINT_FLAGS_COUNT] = {
389 [ TAINT_PROPRIETARY_MODULE ] = { 'P', 'G', true },
390 [ TAINT_FORCED_MODULE ] = { 'F', ' ', true },
391 [ TAINT_CPU_OUT_OF_SPEC ] = { 'S', ' ', false },
392 [ TAINT_FORCED_RMMOD ] = { 'R', ' ', false },
393 [ TAINT_MACHINE_CHECK ] = { 'M', ' ', false },
394 [ TAINT_BAD_PAGE ] = { 'B', ' ', false },
395 [ TAINT_USER ] = { 'U', ' ', false },
396 [ TAINT_DIE ] = { 'D', ' ', false },
397 [ TAINT_OVERRIDDEN_ACPI_TABLE ] = { 'A', ' ', false },
398 [ TAINT_WARN ] = { 'W', ' ', false },
399 [ TAINT_CRAP ] = { 'C', ' ', true },
400 [ TAINT_FIRMWARE_WORKAROUND ] = { 'I', ' ', false },
401 [ TAINT_OOT_MODULE ] = { 'O', ' ', true },
402 [ TAINT_UNSIGNED_MODULE ] = { 'E', ' ', true },
403 [ TAINT_SOFTLOCKUP ] = { 'L', ' ', false },
404 [ TAINT_LIVEPATCH ] = { 'K', ' ', true },
405 [ TAINT_AUX ] = { 'X', ' ', true },
406 [ TAINT_RANDSTRUCT ] = { 'T', ' ', true },
410 * print_tainted - return a string to represent the kernel taint state.
412 * For individual taint flag meanings, see Documentation/admin-guide/sysctl/kernel.rst
414 * The string is overwritten by the next call to print_tainted(),
415 * but is always NULL terminated.
417 const char *print_tainted(void)
419 static char buf[TAINT_FLAGS_COUNT + sizeof("Tainted: ")];
421 BUILD_BUG_ON(ARRAY_SIZE(taint_flags) != TAINT_FLAGS_COUNT);
427 s = buf + sprintf(buf, "Tainted: ");
428 for (i = 0; i < TAINT_FLAGS_COUNT; i++) {
429 const struct taint_flag *t = &taint_flags[i];
430 *s++ = test_bit(i, &tainted_mask) ?
431 t->c_true : t->c_false;
435 snprintf(buf, sizeof(buf), "Not tainted");
440 int test_taint(unsigned flag)
442 return test_bit(flag, &tainted_mask);
444 EXPORT_SYMBOL(test_taint);
446 unsigned long get_taint(void)
452 * add_taint: add a taint flag if not already set.
453 * @flag: one of the TAINT_* constants.
454 * @lockdep_ok: whether lock debugging is still OK.
456 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
457 * some notewortht-but-not-corrupting cases, it can be set to true.
459 void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
461 if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
462 pr_warn("Disabling lock debugging due to kernel taint\n");
464 set_bit(flag, &tainted_mask);
466 if (tainted_mask & panic_on_taint) {
468 panic("panic_on_taint set ...");
471 EXPORT_SYMBOL(add_taint);
473 static void spin_msec(int msecs)
477 for (i = 0; i < msecs; i++) {
478 touch_nmi_watchdog();
484 * It just happens that oops_enter() and oops_exit() are identically
487 static void do_oops_enter_exit(void)
490 static int spin_counter;
495 spin_lock_irqsave(&pause_on_oops_lock, flags);
496 if (pause_on_oops_flag == 0) {
497 /* This CPU may now print the oops message */
498 pause_on_oops_flag = 1;
500 /* We need to stall this CPU */
502 /* This CPU gets to do the counting */
503 spin_counter = pause_on_oops;
505 spin_unlock(&pause_on_oops_lock);
506 spin_msec(MSEC_PER_SEC);
507 spin_lock(&pause_on_oops_lock);
508 } while (--spin_counter);
509 pause_on_oops_flag = 0;
511 /* This CPU waits for a different one */
512 while (spin_counter) {
513 spin_unlock(&pause_on_oops_lock);
515 spin_lock(&pause_on_oops_lock);
519 spin_unlock_irqrestore(&pause_on_oops_lock, flags);
523 * Return true if the calling CPU is allowed to print oops-related info.
524 * This is a bit racy..
526 bool oops_may_print(void)
528 return pause_on_oops_flag == 0;
532 * Called when the architecture enters its oops handler, before it prints
533 * anything. If this is the first CPU to oops, and it's oopsing the first
534 * time then let it proceed.
536 * This is all enabled by the pause_on_oops kernel boot option. We do all
537 * this to ensure that oopses don't scroll off the screen. It has the
538 * side-effect of preventing later-oopsing CPUs from mucking up the display,
541 * It turns out that the CPU which is allowed to print ends up pausing for
542 * the right duration, whereas all the other CPUs pause for twice as long:
543 * once in oops_enter(), once in oops_exit().
545 void oops_enter(void)
548 /* can't trust the integrity of the kernel anymore: */
550 do_oops_enter_exit();
552 if (sysctl_oops_all_cpu_backtrace)
553 trigger_all_cpu_backtrace();
556 static void print_oops_end_marker(void)
558 pr_warn("---[ end trace %016llx ]---\n", 0ULL);
562 * Called when the architecture exits its oops handler, after printing
567 do_oops_enter_exit();
568 print_oops_end_marker();
569 kmsg_dump(KMSG_DUMP_OOPS);
577 void __warn(const char *file, int line, void *caller, unsigned taint,
578 struct pt_regs *regs, struct warn_args *args)
580 disable_trace_on_warning();
583 pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS\n",
584 raw_smp_processor_id(), current->pid, file, line,
587 pr_warn("WARNING: CPU: %d PID: %d at %pS\n",
588 raw_smp_processor_id(), current->pid, caller);
591 vprintk(args->fmt, args->args);
599 panic("panic_on_warn set ...\n");
604 print_irqtrace_events(current);
606 print_oops_end_marker();
607 trace_error_report_end(ERROR_DETECTOR_WARN, (unsigned long)caller);
609 /* Just a warning, don't kill lockdep. */
610 add_taint(taint, LOCKDEP_STILL_OK);
614 void warn_slowpath_fmt(const char *file, int line, unsigned taint,
615 const char *fmt, ...)
617 struct warn_args args;
622 __warn(file, line, __builtin_return_address(0), taint,
628 va_start(args.args, fmt);
629 __warn(file, line, __builtin_return_address(0), taint, NULL, &args);
632 EXPORT_SYMBOL(warn_slowpath_fmt);
634 void __warn_printk(const char *fmt, ...)
644 EXPORT_SYMBOL(__warn_printk);
649 /* Support resetting WARN*_ONCE state */
651 static int clear_warn_once_set(void *data, u64 val)
653 generic_bug_clear_once();
654 memset(__start_once, 0, __end_once - __start_once);
658 DEFINE_DEBUGFS_ATTRIBUTE(clear_warn_once_fops, NULL, clear_warn_once_set,
661 static __init int register_warn_debugfs(void)
663 /* Don't care about failure */
664 debugfs_create_file_unsafe("clear_warn_once", 0200, NULL, NULL,
665 &clear_warn_once_fops);
669 device_initcall(register_warn_debugfs);
672 #ifdef CONFIG_STACKPROTECTOR
675 * Called when gcc's -fstack-protector feature is used, and
676 * gcc detects corruption of the on-stack canary value
678 __visible noinstr void __stack_chk_fail(void)
680 instrumentation_begin();
681 panic("stack-protector: Kernel stack is corrupted in: %pB",
682 __builtin_return_address(0));
683 instrumentation_end();
685 EXPORT_SYMBOL(__stack_chk_fail);
689 core_param(panic, panic_timeout, int, 0644);
690 core_param(panic_print, panic_print, ulong, 0644);
691 core_param(pause_on_oops, pause_on_oops, int, 0644);
692 core_param(panic_on_warn, panic_on_warn, int, 0644);
693 core_param(crash_kexec_post_notifiers, crash_kexec_post_notifiers, bool, 0644);
695 static int __init oops_setup(char *s)
699 if (!strcmp(s, "panic"))
703 early_param("oops", oops_setup);
705 static int __init panic_on_taint_setup(char *s)
712 taint_str = strsep(&s, ",");
713 if (kstrtoul(taint_str, 16, &panic_on_taint))
716 /* make sure panic_on_taint doesn't hold out-of-range TAINT flags */
717 panic_on_taint &= TAINT_FLAGS_MAX;
722 if (s && !strcmp(s, "nousertaint"))
723 panic_on_taint_nousertaint = true;
725 pr_info("panic_on_taint: bitmask=0x%lx nousertaint_mode=%sabled\n",
726 panic_on_taint, panic_on_taint_nousertaint ? "en" : "dis");
730 early_param("panic_on_taint", panic_on_taint_setup);