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 unsigned long panic_print;
71 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
73 EXPORT_SYMBOL(panic_notifier_list);
75 static long no_blink(int state)
80 /* Returns how long it waited in ms */
81 long (*panic_blink)(int state);
82 EXPORT_SYMBOL(panic_blink);
85 * Stop ourself in panic -- architecture code may override this
87 void __weak panic_smp_self_stop(void)
94 * Stop ourselves in NMI context if another CPU has already panicked. Arch code
95 * may override this to prepare for crash dumping, e.g. save regs info.
97 void __weak nmi_panic_self_stop(struct pt_regs *regs)
99 panic_smp_self_stop();
103 * Stop other CPUs in panic. Architecture dependent code may override this
104 * with more suitable version. For example, if the architecture supports
105 * crash dump, it should save registers of each stopped CPU and disable
106 * per-CPU features such as virtualization extensions.
108 void __weak crash_smp_send_stop(void)
110 static int cpus_stopped;
113 * This function can be called twice in panic path, but obviously
114 * we execute this only once.
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
128 atomic_t panic_cpu = ATOMIC_INIT(PANIC_CPU_INVALID);
131 * A variant of panic() called from NMI context. We return if we've already
132 * panicked on this CPU. If another CPU already panicked, loop in
133 * nmi_panic_self_stop() which can provide architecture dependent code such
134 * as saving register state for crash dump.
136 void nmi_panic(struct pt_regs *regs, const char *msg)
140 cpu = raw_smp_processor_id();
141 old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, cpu);
143 if (old_cpu == PANIC_CPU_INVALID)
145 else if (old_cpu != cpu)
146 nmi_panic_self_stop(regs);
148 EXPORT_SYMBOL(nmi_panic);
150 static void panic_print_sys_info(void)
152 if (panic_print & PANIC_PRINT_ALL_PRINTK_MSG)
153 console_flush_on_panic(CONSOLE_REPLAY_ALL);
155 if (panic_print & PANIC_PRINT_TASK_INFO)
158 if (panic_print & PANIC_PRINT_MEM_INFO)
161 if (panic_print & PANIC_PRINT_TIMER_INFO)
162 sysrq_timer_list_show();
164 if (panic_print & PANIC_PRINT_LOCK_INFO)
165 debug_show_all_locks();
167 if (panic_print & PANIC_PRINT_FTRACE_INFO)
168 ftrace_dump(DUMP_ALL);
172 * panic - halt the system
173 * @fmt: The text string to print
175 * Display a message, then perform cleanups.
177 * This function never returns.
179 void panic(const char *fmt, ...)
181 static char buf[1024];
183 long i, i_next = 0, len;
185 int old_cpu, this_cpu;
186 bool _crash_kexec_post_notifiers = crash_kexec_post_notifiers;
189 * Disable local interrupts. This will prevent panic_smp_self_stop
190 * from deadlocking the first cpu that invokes the panic, since
191 * there is nothing to prevent an interrupt handler (that runs
192 * after setting panic_cpu) from invoking panic() again.
195 preempt_disable_notrace();
198 * It's possible to come here directly from a panic-assertion and
199 * not have preempt disabled. Some functions called from here want
200 * preempt to be disabled. No point enabling it later though...
202 * Only one CPU is allowed to execute the panic code from here. For
203 * multiple parallel invocations of panic, all other CPUs either
204 * stop themself or will wait until they are stopped by the 1st CPU
205 * with smp_send_stop().
207 * `old_cpu == PANIC_CPU_INVALID' means this is the 1st CPU which
208 * comes here, so go ahead.
209 * `old_cpu == this_cpu' means we came from nmi_panic() which sets
210 * panic_cpu to this CPU. In this case, this is also the 1st CPU.
212 this_cpu = raw_smp_processor_id();
213 old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, this_cpu);
215 if (old_cpu != PANIC_CPU_INVALID && old_cpu != this_cpu)
216 panic_smp_self_stop();
221 len = vscnprintf(buf, sizeof(buf), fmt, args);
224 if (len && buf[len - 1] == '\n')
227 pr_emerg("Kernel panic - not syncing: %s\n", buf);
228 #ifdef CONFIG_DEBUG_BUGVERBOSE
230 * Avoid nested stack-dumping if a panic occurs during oops processing
232 if (!test_taint(TAINT_DIE) && oops_in_progress <= 1)
237 * If kgdb is enabled, give it a chance to run before we stop all
238 * the other CPUs or else we won't be able to debug processes left
244 * If we have crashed and we have a crash kernel loaded let it handle
246 * If we want to run this after calling panic_notifiers, pass
247 * the "crash_kexec_post_notifiers" option to the kernel.
249 * Bypass the panic_cpu check and call __crash_kexec directly.
251 if (!_crash_kexec_post_notifiers) {
255 * Note smp_send_stop is the usual smp shutdown function, which
256 * unfortunately means it may not be hardened to work in a
262 * If we want to do crash dump after notifier calls and
263 * kmsg_dump, we will need architecture dependent extra
264 * works in addition to stopping other CPUs.
266 crash_smp_send_stop();
270 * Run any panic handlers, including those that might need to
271 * add information to the kmsg dump output.
273 atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
275 kmsg_dump(KMSG_DUMP_PANIC);
278 * If you doubt kdump always works fine in any situation,
279 * "crash_kexec_post_notifiers" offers you a chance to run
280 * panic_notifiers and dumping kmsg before kdump.
281 * Note: since some panic_notifiers can make crashed kernel
282 * more unstable, it can increase risks of the kdump failure too.
284 * Bypass the panic_cpu check and call __crash_kexec directly.
286 if (_crash_kexec_post_notifiers)
295 * We may have ended up stopping the CPU holding the lock (in
296 * smp_send_stop()) while still having some valuable data in the console
297 * buffer. Try to acquire the lock then release it regardless of the
298 * result. The release will also print the buffers out. Locks debug
299 * should be disabled to avoid reporting bad unlock balance when
300 * panic() is not being callled from OOPS.
303 console_flush_on_panic(CONSOLE_FLUSH_PENDING);
305 panic_print_sys_info();
308 panic_blink = no_blink;
310 if (panic_timeout > 0) {
312 * Delay timeout seconds before rebooting the machine.
313 * We can't use the "normal" timers since we just panicked.
315 pr_emerg("Rebooting in %d seconds..\n", panic_timeout);
317 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
318 touch_nmi_watchdog();
320 i += panic_blink(state ^= 1);
321 i_next = i + 3600 / PANIC_BLINK_SPD;
323 mdelay(PANIC_TIMER_STEP);
326 if (panic_timeout != 0) {
328 * This will not be a clean reboot, with everything
329 * shutting down. But if there is a chance of
330 * rebooting the system it will be rebooted.
332 if (panic_reboot_mode != REBOOT_UNDEFINED)
333 reboot_mode = panic_reboot_mode;
338 extern int stop_a_enabled;
339 /* Make sure the user can actually press Stop-A (L1-A) */
341 pr_emerg("Press Stop-A (L1-A) from sun keyboard or send break\n"
342 "twice on console to return to the boot prom\n");
345 #if defined(CONFIG_S390)
348 pr_emerg("---[ end Kernel panic - not syncing: %s ]---\n", buf);
350 /* Do not scroll important messages printed above */
353 for (i = 0; ; i += PANIC_TIMER_STEP) {
354 touch_softlockup_watchdog();
356 i += panic_blink(state ^= 1);
357 i_next = i + 3600 / PANIC_BLINK_SPD;
359 mdelay(PANIC_TIMER_STEP);
363 EXPORT_SYMBOL(panic);
366 * TAINT_FORCED_RMMOD could be a per-module flag but the module
367 * is being removed anyway.
369 const struct taint_flag taint_flags[TAINT_FLAGS_COUNT] = {
370 [ TAINT_PROPRIETARY_MODULE ] = { 'P', 'G', true },
371 [ TAINT_FORCED_MODULE ] = { 'F', ' ', true },
372 [ TAINT_CPU_OUT_OF_SPEC ] = { 'S', ' ', false },
373 [ TAINT_FORCED_RMMOD ] = { 'R', ' ', false },
374 [ TAINT_MACHINE_CHECK ] = { 'M', ' ', false },
375 [ TAINT_BAD_PAGE ] = { 'B', ' ', false },
376 [ TAINT_USER ] = { 'U', ' ', false },
377 [ TAINT_DIE ] = { 'D', ' ', false },
378 [ TAINT_OVERRIDDEN_ACPI_TABLE ] = { 'A', ' ', false },
379 [ TAINT_WARN ] = { 'W', ' ', false },
380 [ TAINT_CRAP ] = { 'C', ' ', true },
381 [ TAINT_FIRMWARE_WORKAROUND ] = { 'I', ' ', false },
382 [ TAINT_OOT_MODULE ] = { 'O', ' ', true },
383 [ TAINT_UNSIGNED_MODULE ] = { 'E', ' ', true },
384 [ TAINT_SOFTLOCKUP ] = { 'L', ' ', false },
385 [ TAINT_LIVEPATCH ] = { 'K', ' ', true },
386 [ TAINT_AUX ] = { 'X', ' ', true },
387 [ TAINT_RANDSTRUCT ] = { 'T', ' ', true },
391 * print_tainted - return a string to represent the kernel taint state.
393 * For individual taint flag meanings, see Documentation/admin-guide/sysctl/kernel.rst
395 * The string is overwritten by the next call to print_tainted(),
396 * but is always NULL terminated.
398 const char *print_tainted(void)
400 static char buf[TAINT_FLAGS_COUNT + sizeof("Tainted: ")];
402 BUILD_BUG_ON(ARRAY_SIZE(taint_flags) != TAINT_FLAGS_COUNT);
408 s = buf + sprintf(buf, "Tainted: ");
409 for (i = 0; i < TAINT_FLAGS_COUNT; i++) {
410 const struct taint_flag *t = &taint_flags[i];
411 *s++ = test_bit(i, &tainted_mask) ?
412 t->c_true : t->c_false;
416 snprintf(buf, sizeof(buf), "Not tainted");
421 int test_taint(unsigned flag)
423 return test_bit(flag, &tainted_mask);
425 EXPORT_SYMBOL(test_taint);
427 unsigned long get_taint(void)
433 * add_taint: add a taint flag if not already set.
434 * @flag: one of the TAINT_* constants.
435 * @lockdep_ok: whether lock debugging is still OK.
437 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
438 * some notewortht-but-not-corrupting cases, it can be set to true.
440 void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
442 if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
443 pr_warn("Disabling lock debugging due to kernel taint\n");
445 set_bit(flag, &tainted_mask);
447 if (tainted_mask & panic_on_taint) {
449 panic("panic_on_taint set ...");
452 EXPORT_SYMBOL(add_taint);
454 static void spin_msec(int msecs)
458 for (i = 0; i < msecs; i++) {
459 touch_nmi_watchdog();
465 * It just happens that oops_enter() and oops_exit() are identically
468 static void do_oops_enter_exit(void)
471 static int spin_counter;
476 spin_lock_irqsave(&pause_on_oops_lock, flags);
477 if (pause_on_oops_flag == 0) {
478 /* This CPU may now print the oops message */
479 pause_on_oops_flag = 1;
481 /* We need to stall this CPU */
483 /* This CPU gets to do the counting */
484 spin_counter = pause_on_oops;
486 spin_unlock(&pause_on_oops_lock);
487 spin_msec(MSEC_PER_SEC);
488 spin_lock(&pause_on_oops_lock);
489 } while (--spin_counter);
490 pause_on_oops_flag = 0;
492 /* This CPU waits for a different one */
493 while (spin_counter) {
494 spin_unlock(&pause_on_oops_lock);
496 spin_lock(&pause_on_oops_lock);
500 spin_unlock_irqrestore(&pause_on_oops_lock, flags);
504 * Return true if the calling CPU is allowed to print oops-related info.
505 * This is a bit racy..
507 bool oops_may_print(void)
509 return pause_on_oops_flag == 0;
513 * Called when the architecture enters its oops handler, before it prints
514 * anything. If this is the first CPU to oops, and it's oopsing the first
515 * time then let it proceed.
517 * This is all enabled by the pause_on_oops kernel boot option. We do all
518 * this to ensure that oopses don't scroll off the screen. It has the
519 * side-effect of preventing later-oopsing CPUs from mucking up the display,
522 * It turns out that the CPU which is allowed to print ends up pausing for
523 * the right duration, whereas all the other CPUs pause for twice as long:
524 * once in oops_enter(), once in oops_exit().
526 void oops_enter(void)
529 /* can't trust the integrity of the kernel anymore: */
531 do_oops_enter_exit();
533 if (sysctl_oops_all_cpu_backtrace)
534 trigger_all_cpu_backtrace();
538 * 64-bit random ID for oopses:
542 static int init_oops_id(void)
545 get_random_bytes(&oops_id, sizeof(oops_id));
551 late_initcall(init_oops_id);
553 static void print_oops_end_marker(void)
556 pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id);
560 * Called when the architecture exits its oops handler, after printing
565 do_oops_enter_exit();
566 print_oops_end_marker();
567 kmsg_dump(KMSG_DUMP_OOPS);
575 void __warn(const char *file, int line, void *caller, unsigned taint,
576 struct pt_regs *regs, struct warn_args *args)
578 disable_trace_on_warning();
581 pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS\n",
582 raw_smp_processor_id(), current->pid, file, line,
585 pr_warn("WARNING: CPU: %d PID: %d at %pS\n",
586 raw_smp_processor_id(), current->pid, caller);
589 vprintk(args->fmt, args->args);
598 * This thread may hit another WARN() in the panic path.
599 * Resetting this prevents additional WARN() from panicking the
600 * system on this thread. Other threads are blocked by the
601 * panic_mutex in panic().
604 panic("panic_on_warn set ...\n");
610 print_irqtrace_events(current);
612 print_oops_end_marker();
613 trace_error_report_end(ERROR_DETECTOR_WARN, (unsigned long)caller);
615 /* Just a warning, don't kill lockdep. */
616 add_taint(taint, LOCKDEP_STILL_OK);
620 void warn_slowpath_fmt(const char *file, int line, unsigned taint,
621 const char *fmt, ...)
623 struct warn_args args;
628 __warn(file, line, __builtin_return_address(0), taint,
634 va_start(args.args, fmt);
635 __warn(file, line, __builtin_return_address(0), taint, NULL, &args);
638 EXPORT_SYMBOL(warn_slowpath_fmt);
640 void __warn_printk(const char *fmt, ...)
650 EXPORT_SYMBOL(__warn_printk);
655 /* Support resetting WARN*_ONCE state */
657 static int clear_warn_once_set(void *data, u64 val)
659 generic_bug_clear_once();
660 memset(__start_once, 0, __end_once - __start_once);
664 DEFINE_DEBUGFS_ATTRIBUTE(clear_warn_once_fops, NULL, clear_warn_once_set,
667 static __init int register_warn_debugfs(void)
669 /* Don't care about failure */
670 debugfs_create_file_unsafe("clear_warn_once", 0200, NULL, NULL,
671 &clear_warn_once_fops);
675 device_initcall(register_warn_debugfs);
678 #ifdef CONFIG_STACKPROTECTOR
681 * Called when gcc's -fstack-protector feature is used, and
682 * gcc detects corruption of the on-stack canary value
684 __visible noinstr void __stack_chk_fail(void)
686 instrumentation_begin();
687 panic("stack-protector: Kernel stack is corrupted in: %pB",
688 __builtin_return_address(0));
689 instrumentation_end();
691 EXPORT_SYMBOL(__stack_chk_fail);
695 core_param(panic, panic_timeout, int, 0644);
696 core_param(panic_print, panic_print, ulong, 0644);
697 core_param(pause_on_oops, pause_on_oops, int, 0644);
698 core_param(panic_on_warn, panic_on_warn, int, 0644);
699 core_param(crash_kexec_post_notifiers, crash_kexec_post_notifiers, bool, 0644);
701 static int __init oops_setup(char *s)
705 if (!strcmp(s, "panic"))
709 early_param("oops", oops_setup);
711 static int __init panic_on_taint_setup(char *s)
718 taint_str = strsep(&s, ",");
719 if (kstrtoul(taint_str, 16, &panic_on_taint))
722 /* make sure panic_on_taint doesn't hold out-of-range TAINT flags */
723 panic_on_taint &= TAINT_FLAGS_MAX;
728 if (s && !strcmp(s, "nousertaint"))
729 panic_on_taint_nousertaint = true;
731 pr_info("panic_on_taint: bitmask=0x%lx nousertaint_mode=%sabled\n",
732 panic_on_taint, panic_on_taint_nousertaint ? "en" : "dis");
736 early_param("panic_on_taint", panic_on_taint_setup);