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(void)
153 if (panic_print & PANIC_PRINT_ALL_PRINTK_MSG)
154 console_flush_on_panic(CONSOLE_REPLAY_ALL);
156 if (panic_print & PANIC_PRINT_ALL_CPU_BT)
157 trigger_all_cpu_backtrace();
159 if (panic_print & PANIC_PRINT_TASK_INFO)
162 if (panic_print & PANIC_PRINT_MEM_INFO)
165 if (panic_print & PANIC_PRINT_TIMER_INFO)
166 sysrq_timer_list_show();
168 if (panic_print & PANIC_PRINT_LOCK_INFO)
169 debug_show_all_locks();
171 if (panic_print & PANIC_PRINT_FTRACE_INFO)
172 ftrace_dump(DUMP_ALL);
176 * panic - halt the system
177 * @fmt: The text string to print
179 * Display a message, then perform cleanups.
181 * This function never returns.
183 void panic(const char *fmt, ...)
185 static char buf[1024];
187 long i, i_next = 0, len;
189 int old_cpu, this_cpu;
190 bool _crash_kexec_post_notifiers = crash_kexec_post_notifiers;
194 * This thread may hit another WARN() in the panic path.
195 * Resetting this prevents additional WARN() from panicking the
196 * system on this thread. Other threads are blocked by the
197 * panic_mutex in panic().
203 * Disable local interrupts. This will prevent panic_smp_self_stop
204 * from deadlocking the first cpu that invokes the panic, since
205 * there is nothing to prevent an interrupt handler (that runs
206 * after setting panic_cpu) from invoking panic() again.
209 preempt_disable_notrace();
212 * It's possible to come here directly from a panic-assertion and
213 * not have preempt disabled. Some functions called from here want
214 * preempt to be disabled. No point enabling it later though...
216 * Only one CPU is allowed to execute the panic code from here. For
217 * multiple parallel invocations of panic, all other CPUs either
218 * stop themself or will wait until they are stopped by the 1st CPU
219 * with smp_send_stop().
221 * `old_cpu == PANIC_CPU_INVALID' means this is the 1st CPU which
222 * comes here, so go ahead.
223 * `old_cpu == this_cpu' means we came from nmi_panic() which sets
224 * panic_cpu to this CPU. In this case, this is also the 1st CPU.
226 this_cpu = raw_smp_processor_id();
227 old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, this_cpu);
229 if (old_cpu != PANIC_CPU_INVALID && old_cpu != this_cpu)
230 panic_smp_self_stop();
235 len = vscnprintf(buf, sizeof(buf), fmt, args);
238 if (len && buf[len - 1] == '\n')
241 pr_emerg("Kernel panic - not syncing: %s\n", buf);
242 #ifdef CONFIG_DEBUG_BUGVERBOSE
244 * Avoid nested stack-dumping if a panic occurs during oops processing
246 if (!test_taint(TAINT_DIE) && oops_in_progress <= 1)
251 * If kgdb is enabled, give it a chance to run before we stop all
252 * the other CPUs or else we won't be able to debug processes left
258 * If we have crashed and we have a crash kernel loaded let it handle
260 * If we want to run this after calling panic_notifiers, pass
261 * the "crash_kexec_post_notifiers" option to the kernel.
263 * Bypass the panic_cpu check and call __crash_kexec directly.
265 if (!_crash_kexec_post_notifiers) {
269 * Note smp_send_stop is the usual smp shutdown function, which
270 * unfortunately means it may not be hardened to work in a
276 * If we want to do crash dump after notifier calls and
277 * kmsg_dump, we will need architecture dependent extra
278 * works in addition to stopping other CPUs.
280 crash_smp_send_stop();
284 * Run any panic handlers, including those that might need to
285 * add information to the kmsg dump output.
287 atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
289 kmsg_dump(KMSG_DUMP_PANIC);
292 * If you doubt kdump always works fine in any situation,
293 * "crash_kexec_post_notifiers" offers you a chance to run
294 * panic_notifiers and dumping kmsg before kdump.
295 * Note: since some panic_notifiers can make crashed kernel
296 * more unstable, it can increase risks of the kdump failure too.
298 * Bypass the panic_cpu check and call __crash_kexec directly.
300 if (_crash_kexec_post_notifiers)
309 * We may have ended up stopping the CPU holding the lock (in
310 * smp_send_stop()) while still having some valuable data in the console
311 * buffer. Try to acquire the lock then release it regardless of the
312 * result. The release will also print the buffers out. Locks debug
313 * should be disabled to avoid reporting bad unlock balance when
314 * panic() is not being callled from OOPS.
317 console_flush_on_panic(CONSOLE_FLUSH_PENDING);
319 panic_print_sys_info();
322 panic_blink = no_blink;
324 if (panic_timeout > 0) {
326 * Delay timeout seconds before rebooting the machine.
327 * We can't use the "normal" timers since we just panicked.
329 pr_emerg("Rebooting in %d seconds..\n", panic_timeout);
331 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
332 touch_nmi_watchdog();
334 i += panic_blink(state ^= 1);
335 i_next = i + 3600 / PANIC_BLINK_SPD;
337 mdelay(PANIC_TIMER_STEP);
340 if (panic_timeout != 0) {
342 * This will not be a clean reboot, with everything
343 * shutting down. But if there is a chance of
344 * rebooting the system it will be rebooted.
346 if (panic_reboot_mode != REBOOT_UNDEFINED)
347 reboot_mode = panic_reboot_mode;
352 extern int stop_a_enabled;
353 /* Make sure the user can actually press Stop-A (L1-A) */
355 pr_emerg("Press Stop-A (L1-A) from sun keyboard or send break\n"
356 "twice on console to return to the boot prom\n");
359 #if defined(CONFIG_S390)
362 pr_emerg("---[ end Kernel panic - not syncing: %s ]---\n", buf);
364 /* Do not scroll important messages printed above */
367 for (i = 0; ; i += PANIC_TIMER_STEP) {
368 touch_softlockup_watchdog();
370 i += panic_blink(state ^= 1);
371 i_next = i + 3600 / PANIC_BLINK_SPD;
373 mdelay(PANIC_TIMER_STEP);
377 EXPORT_SYMBOL(panic);
380 * TAINT_FORCED_RMMOD could be a per-module flag but the module
381 * is being removed anyway.
383 const struct taint_flag taint_flags[TAINT_FLAGS_COUNT] = {
384 [ TAINT_PROPRIETARY_MODULE ] = { 'P', 'G', true },
385 [ TAINT_FORCED_MODULE ] = { 'F', ' ', true },
386 [ TAINT_CPU_OUT_OF_SPEC ] = { 'S', ' ', false },
387 [ TAINT_FORCED_RMMOD ] = { 'R', ' ', false },
388 [ TAINT_MACHINE_CHECK ] = { 'M', ' ', false },
389 [ TAINT_BAD_PAGE ] = { 'B', ' ', false },
390 [ TAINT_USER ] = { 'U', ' ', false },
391 [ TAINT_DIE ] = { 'D', ' ', false },
392 [ TAINT_OVERRIDDEN_ACPI_TABLE ] = { 'A', ' ', false },
393 [ TAINT_WARN ] = { 'W', ' ', false },
394 [ TAINT_CRAP ] = { 'C', ' ', true },
395 [ TAINT_FIRMWARE_WORKAROUND ] = { 'I', ' ', false },
396 [ TAINT_OOT_MODULE ] = { 'O', ' ', true },
397 [ TAINT_UNSIGNED_MODULE ] = { 'E', ' ', true },
398 [ TAINT_SOFTLOCKUP ] = { 'L', ' ', false },
399 [ TAINT_LIVEPATCH ] = { 'K', ' ', true },
400 [ TAINT_AUX ] = { 'X', ' ', true },
401 [ TAINT_RANDSTRUCT ] = { 'T', ' ', true },
405 * print_tainted - return a string to represent the kernel taint state.
407 * For individual taint flag meanings, see Documentation/admin-guide/sysctl/kernel.rst
409 * The string is overwritten by the next call to print_tainted(),
410 * but is always NULL terminated.
412 const char *print_tainted(void)
414 static char buf[TAINT_FLAGS_COUNT + sizeof("Tainted: ")];
416 BUILD_BUG_ON(ARRAY_SIZE(taint_flags) != TAINT_FLAGS_COUNT);
422 s = buf + sprintf(buf, "Tainted: ");
423 for (i = 0; i < TAINT_FLAGS_COUNT; i++) {
424 const struct taint_flag *t = &taint_flags[i];
425 *s++ = test_bit(i, &tainted_mask) ?
426 t->c_true : t->c_false;
430 snprintf(buf, sizeof(buf), "Not tainted");
435 int test_taint(unsigned flag)
437 return test_bit(flag, &tainted_mask);
439 EXPORT_SYMBOL(test_taint);
441 unsigned long get_taint(void)
447 * add_taint: add a taint flag if not already set.
448 * @flag: one of the TAINT_* constants.
449 * @lockdep_ok: whether lock debugging is still OK.
451 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
452 * some notewortht-but-not-corrupting cases, it can be set to true.
454 void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
456 if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
457 pr_warn("Disabling lock debugging due to kernel taint\n");
459 set_bit(flag, &tainted_mask);
461 if (tainted_mask & panic_on_taint) {
463 panic("panic_on_taint set ...");
466 EXPORT_SYMBOL(add_taint);
468 static void spin_msec(int msecs)
472 for (i = 0; i < msecs; i++) {
473 touch_nmi_watchdog();
479 * It just happens that oops_enter() and oops_exit() are identically
482 static void do_oops_enter_exit(void)
485 static int spin_counter;
490 spin_lock_irqsave(&pause_on_oops_lock, flags);
491 if (pause_on_oops_flag == 0) {
492 /* This CPU may now print the oops message */
493 pause_on_oops_flag = 1;
495 /* We need to stall this CPU */
497 /* This CPU gets to do the counting */
498 spin_counter = pause_on_oops;
500 spin_unlock(&pause_on_oops_lock);
501 spin_msec(MSEC_PER_SEC);
502 spin_lock(&pause_on_oops_lock);
503 } while (--spin_counter);
504 pause_on_oops_flag = 0;
506 /* This CPU waits for a different one */
507 while (spin_counter) {
508 spin_unlock(&pause_on_oops_lock);
510 spin_lock(&pause_on_oops_lock);
514 spin_unlock_irqrestore(&pause_on_oops_lock, flags);
518 * Return true if the calling CPU is allowed to print oops-related info.
519 * This is a bit racy..
521 bool oops_may_print(void)
523 return pause_on_oops_flag == 0;
527 * Called when the architecture enters its oops handler, before it prints
528 * anything. If this is the first CPU to oops, and it's oopsing the first
529 * time then let it proceed.
531 * This is all enabled by the pause_on_oops kernel boot option. We do all
532 * this to ensure that oopses don't scroll off the screen. It has the
533 * side-effect of preventing later-oopsing CPUs from mucking up the display,
536 * It turns out that the CPU which is allowed to print ends up pausing for
537 * the right duration, whereas all the other CPUs pause for twice as long:
538 * once in oops_enter(), once in oops_exit().
540 void oops_enter(void)
543 /* can't trust the integrity of the kernel anymore: */
545 do_oops_enter_exit();
547 if (sysctl_oops_all_cpu_backtrace)
548 trigger_all_cpu_backtrace();
551 static void print_oops_end_marker(void)
553 pr_warn("---[ end trace %016llx ]---\n", 0ULL);
557 * Called when the architecture exits its oops handler, after printing
562 do_oops_enter_exit();
563 print_oops_end_marker();
564 kmsg_dump(KMSG_DUMP_OOPS);
572 void __warn(const char *file, int line, void *caller, unsigned taint,
573 struct pt_regs *regs, struct warn_args *args)
575 disable_trace_on_warning();
578 pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS\n",
579 raw_smp_processor_id(), current->pid, file, line,
582 pr_warn("WARNING: CPU: %d PID: %d at %pS\n",
583 raw_smp_processor_id(), current->pid, caller);
586 vprintk(args->fmt, args->args);
594 panic("panic_on_warn set ...\n");
599 print_irqtrace_events(current);
601 print_oops_end_marker();
602 trace_error_report_end(ERROR_DETECTOR_WARN, (unsigned long)caller);
604 /* Just a warning, don't kill lockdep. */
605 add_taint(taint, LOCKDEP_STILL_OK);
609 void warn_slowpath_fmt(const char *file, int line, unsigned taint,
610 const char *fmt, ...)
612 struct warn_args args;
617 __warn(file, line, __builtin_return_address(0), taint,
623 va_start(args.args, fmt);
624 __warn(file, line, __builtin_return_address(0), taint, NULL, &args);
627 EXPORT_SYMBOL(warn_slowpath_fmt);
629 void __warn_printk(const char *fmt, ...)
639 EXPORT_SYMBOL(__warn_printk);
644 /* Support resetting WARN*_ONCE state */
646 static int clear_warn_once_set(void *data, u64 val)
648 generic_bug_clear_once();
649 memset(__start_once, 0, __end_once - __start_once);
653 DEFINE_DEBUGFS_ATTRIBUTE(clear_warn_once_fops, NULL, clear_warn_once_set,
656 static __init int register_warn_debugfs(void)
658 /* Don't care about failure */
659 debugfs_create_file_unsafe("clear_warn_once", 0200, NULL, NULL,
660 &clear_warn_once_fops);
664 device_initcall(register_warn_debugfs);
667 #ifdef CONFIG_STACKPROTECTOR
670 * Called when gcc's -fstack-protector feature is used, and
671 * gcc detects corruption of the on-stack canary value
673 __visible noinstr void __stack_chk_fail(void)
675 instrumentation_begin();
676 panic("stack-protector: Kernel stack is corrupted in: %pB",
677 __builtin_return_address(0));
678 instrumentation_end();
680 EXPORT_SYMBOL(__stack_chk_fail);
684 core_param(panic, panic_timeout, int, 0644);
685 core_param(panic_print, panic_print, ulong, 0644);
686 core_param(pause_on_oops, pause_on_oops, int, 0644);
687 core_param(panic_on_warn, panic_on_warn, int, 0644);
688 core_param(crash_kexec_post_notifiers, crash_kexec_post_notifiers, bool, 0644);
690 static int __init oops_setup(char *s)
694 if (!strcmp(s, "panic"))
698 early_param("oops", oops_setup);
700 static int __init panic_on_taint_setup(char *s)
707 taint_str = strsep(&s, ",");
708 if (kstrtoul(taint_str, 16, &panic_on_taint))
711 /* make sure panic_on_taint doesn't hold out-of-range TAINT flags */
712 panic_on_taint &= TAINT_FLAGS_MAX;
717 if (s && !strcmp(s, "nousertaint"))
718 panic_on_taint_nousertaint = true;
720 pr_info("panic_on_taint: bitmask=0x%lx nousertaint_mode=%sabled\n",
721 panic_on_taint, panic_on_taint_nousertaint ? "en" : "dis");
725 early_param("panic_on_taint", panic_on_taint_setup);