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/debug-snapshot.h>
33 #define PANIC_TIMER_STEP 100
34 #define PANIC_BLINK_SPD 18
36 int panic_on_oops
= CONFIG_PANIC_ON_OOPS_VALUE
;
37 static unsigned long tainted_mask
;
38 static int pause_on_oops
;
39 static int pause_on_oops_flag
;
40 static DEFINE_SPINLOCK(pause_on_oops_lock
);
41 bool crash_kexec_post_notifiers
;
42 int panic_on_warn __read_mostly
;
44 int panic_timeout
= CONFIG_PANIC_TIMEOUT
;
45 EXPORT_SYMBOL_GPL(panic_timeout
);
47 ATOMIC_NOTIFIER_HEAD(panic_notifier_list
);
49 EXPORT_SYMBOL(panic_notifier_list
);
51 static long no_blink(int state
)
56 /* Returns how long it waited in ms */
57 long (*panic_blink
)(int state
);
58 EXPORT_SYMBOL(panic_blink
);
61 * Stop ourself in panic -- architecture code may override this
63 void __weak
panic_smp_self_stop(void)
70 * Stop ourselves in NMI context if another CPU has already panicked. Arch code
71 * may override this to prepare for crash dumping, e.g. save regs info.
73 void __weak
nmi_panic_self_stop(struct pt_regs
*regs
)
75 panic_smp_self_stop();
79 * Stop other CPUs in panic. Architecture dependent code may override this
80 * with more suitable version. For example, if the architecture supports
81 * crash dump, it should save registers of each stopped CPU and disable
82 * per-CPU features such as virtualization extensions.
84 void __weak
crash_smp_send_stop(void)
86 static int cpus_stopped
;
89 * This function can be called twice in panic path, but obviously
90 * we execute this only once.
96 * Note smp_send_stop is the usual smp shutdown function, which
97 * unfortunately means it may not be hardened to work in a panic
104 atomic_t panic_cpu
= ATOMIC_INIT(PANIC_CPU_INVALID
);
107 * A variant of panic() called from NMI context. We return if we've already
108 * panicked on this CPU. If another CPU already panicked, loop in
109 * nmi_panic_self_stop() which can provide architecture dependent code such
110 * as saving register state for crash dump.
112 void nmi_panic(struct pt_regs
*regs
, const char *msg
)
116 cpu
= raw_smp_processor_id();
117 old_cpu
= atomic_cmpxchg(&panic_cpu
, PANIC_CPU_INVALID
, cpu
);
119 if (old_cpu
== PANIC_CPU_INVALID
)
121 else if (old_cpu
!= cpu
)
122 nmi_panic_self_stop(regs
);
124 EXPORT_SYMBOL(nmi_panic
);
127 * panic - halt the system
128 * @fmt: The text string to print
130 * Display a message, then perform cleanups.
132 * This function never returns.
134 void panic(const char *fmt
, ...)
136 static char buf
[1024];
140 int old_cpu
, this_cpu
;
141 bool _crash_kexec_post_notifiers
= crash_kexec_post_notifiers
;
144 * dbg_snapshot_early_panic is for supporting wapper functions
145 * to users need to run SoC specific function in NOT interrupt
148 dbg_snapshot_early_panic();
151 * Disable local interrupts. This will prevent panic_smp_self_stop
152 * from deadlocking the first cpu that invokes the panic, since
153 * there is nothing to prevent an interrupt handler (that runs
154 * after setting panic_cpu) from invoking panic() again.
159 * It's possible to come here directly from a panic-assertion and
160 * not have preempt disabled. Some functions called from here want
161 * preempt to be disabled. No point enabling it later though...
163 * Only one CPU is allowed to execute the panic code from here. For
164 * multiple parallel invocations of panic, all other CPUs either
165 * stop themself or will wait until they are stopped by the 1st CPU
166 * with smp_send_stop().
168 * `old_cpu == PANIC_CPU_INVALID' means this is the 1st CPU which
169 * comes here, so go ahead.
170 * `old_cpu == this_cpu' means we came from nmi_panic() which sets
171 * panic_cpu to this CPU. In this case, this is also the 1st CPU.
173 this_cpu
= raw_smp_processor_id();
174 old_cpu
= atomic_cmpxchg(&panic_cpu
, PANIC_CPU_INVALID
, this_cpu
);
176 if (old_cpu
!= PANIC_CPU_INVALID
) {
177 dbg_snapshot_hook_hardlockup_exit();
178 panic_smp_self_stop();
184 vsnprintf(buf
, sizeof(buf
), fmt
, args
);
186 pr_emerg("Kernel panic - not syncing: %s\n", buf
);
188 dbg_snapshot_prepare_panic();
189 dbg_snapshot_dump_panic(buf
, (size_t)strnlen(buf
, sizeof(buf
)));
190 #ifdef CONFIG_DEBUG_BUGVERBOSE
192 * Avoid nested stack-dumping if a panic occurs during oops processing
194 if (!test_taint(TAINT_DIE
) && oops_in_progress
<= 1)
197 //sysrq_sched_debug_show();
200 * If we have crashed and we have a crash kernel loaded let it handle
202 * If we want to run this after calling panic_notifiers, pass
203 * the "crash_kexec_post_notifiers" option to the kernel.
205 * Bypass the panic_cpu check and call __crash_kexec directly.
207 if (!_crash_kexec_post_notifiers
) {
208 printk_safe_flush_on_panic();
212 * Note smp_send_stop is the usual smp shutdown function, which
213 * unfortunately means it may not be hardened to work in a
219 * If we want to do crash dump after notifier calls and
220 * kmsg_dump, we will need architecture dependent extra
221 * works in addition to stopping other CPUs.
223 crash_smp_send_stop();
227 * Run any panic handlers, including those that might need to
228 * add information to the kmsg dump output.
230 atomic_notifier_call_chain(&panic_notifier_list
, 0, buf
);
232 /* Call flush even twice. It tries harder with a single online CPU */
233 printk_safe_flush_on_panic();
234 kmsg_dump(KMSG_DUMP_PANIC
);
236 dbg_snapshot_post_panic();
239 * If you doubt kdump always works fine in any situation,
240 * "crash_kexec_post_notifiers" offers you a chance to run
241 * panic_notifiers and dumping kmsg before kdump.
242 * Note: since some panic_notifiers can make crashed kernel
243 * more unstable, it can increase risks of the kdump failure too.
245 * Bypass the panic_cpu check and call __crash_kexec directly.
247 if (_crash_kexec_post_notifiers
)
256 * We may have ended up stopping the CPU holding the lock (in
257 * smp_send_stop()) while still having some valuable data in the console
258 * buffer. Try to acquire the lock then release it regardless of the
259 * result. The release will also print the buffers out. Locks debug
260 * should be disabled to avoid reporting bad unlock balance when
261 * panic() is not being callled from OOPS.
264 console_flush_on_panic();
267 panic_blink
= no_blink
;
269 if (panic_timeout
> 0) {
271 * Delay timeout seconds before rebooting the machine.
272 * We can't use the "normal" timers since we just panicked.
274 pr_emerg("Rebooting in %d seconds..\n", panic_timeout
);
276 for (i
= 0; i
< panic_timeout
* 1000; i
+= PANIC_TIMER_STEP
) {
277 touch_nmi_watchdog();
279 i
+= panic_blink(state
^= 1);
280 i_next
= i
+ 3600 / PANIC_BLINK_SPD
;
282 mdelay(PANIC_TIMER_STEP
);
285 if (panic_timeout
!= 0) {
287 * This will not be a clean reboot, with everything
288 * shutting down. But if there is a chance of
289 * rebooting the system it will be rebooted.
295 extern int stop_a_enabled
;
296 /* Make sure the user can actually press Stop-A (L1-A) */
298 pr_emerg("Press Stop-A (L1-A) from sun keyboard or send break\n"
299 "twice on console to return to the boot prom\n");
302 #if defined(CONFIG_S390)
304 unsigned long caller
;
306 caller
= (unsigned long)__builtin_return_address(0);
307 disabled_wait(caller
);
310 pr_emerg("---[ end Kernel panic - not syncing: %s\n", buf
);
312 for (i
= 0; ; i
+= PANIC_TIMER_STEP
) {
313 touch_softlockup_watchdog();
315 i
+= panic_blink(state
^= 1);
316 i_next
= i
+ 3600 / PANIC_BLINK_SPD
;
318 mdelay(PANIC_TIMER_STEP
);
322 EXPORT_SYMBOL(panic
);
325 * TAINT_FORCED_RMMOD could be a per-module flag but the module
326 * is being removed anyway.
328 const struct taint_flag taint_flags
[TAINT_FLAGS_COUNT
] = {
329 { 'P', 'G', true }, /* TAINT_PROPRIETARY_MODULE */
330 { 'F', ' ', true }, /* TAINT_FORCED_MODULE */
331 { 'S', ' ', false }, /* TAINT_CPU_OUT_OF_SPEC */
332 { 'R', ' ', false }, /* TAINT_FORCED_RMMOD */
333 { 'M', ' ', false }, /* TAINT_MACHINE_CHECK */
334 { 'B', ' ', false }, /* TAINT_BAD_PAGE */
335 { 'U', ' ', false }, /* TAINT_USER */
336 { 'D', ' ', false }, /* TAINT_DIE */
337 { 'A', ' ', false }, /* TAINT_OVERRIDDEN_ACPI_TABLE */
338 { 'W', ' ', false }, /* TAINT_WARN */
339 { 'C', ' ', true }, /* TAINT_CRAP */
340 { 'I', ' ', false }, /* TAINT_FIRMWARE_WORKAROUND */
341 { 'O', ' ', true }, /* TAINT_OOT_MODULE */
342 { 'E', ' ', true }, /* TAINT_UNSIGNED_MODULE */
343 { 'L', ' ', false }, /* TAINT_SOFTLOCKUP */
344 { 'K', ' ', true }, /* TAINT_LIVEPATCH */
348 * print_tainted - return a string to represent the kernel taint state.
350 * 'P' - Proprietary module has been loaded.
351 * 'F' - Module has been forcibly loaded.
352 * 'S' - SMP with CPUs not designed for SMP.
353 * 'R' - User forced a module unload.
354 * 'M' - System experienced a machine check exception.
355 * 'B' - System has hit bad_page.
356 * 'U' - Userspace-defined naughtiness.
357 * 'D' - Kernel has oopsed before
358 * 'A' - ACPI table overridden.
359 * 'W' - Taint on warning.
360 * 'C' - modules from drivers/staging are loaded.
361 * 'I' - Working around severe firmware bug.
362 * 'O' - Out-of-tree module has been loaded.
363 * 'E' - Unsigned module has been loaded.
364 * 'L' - A soft lockup has previously occurred.
365 * 'K' - Kernel has been live patched.
367 * The string is overwritten by the next call to print_tainted().
369 const char *print_tainted(void)
371 static char buf
[TAINT_FLAGS_COUNT
+ sizeof("Tainted: ")];
377 s
= buf
+ sprintf(buf
, "Tainted: ");
378 for (i
= 0; i
< TAINT_FLAGS_COUNT
; i
++) {
379 const struct taint_flag
*t
= &taint_flags
[i
];
380 *s
++ = test_bit(i
, &tainted_mask
) ?
381 t
->c_true
: t
->c_false
;
385 snprintf(buf
, sizeof(buf
), "Not tainted");
390 int test_taint(unsigned flag
)
392 return test_bit(flag
, &tainted_mask
);
394 EXPORT_SYMBOL(test_taint
);
396 unsigned long get_taint(void)
402 * add_taint: add a taint flag if not already set.
403 * @flag: one of the TAINT_* constants.
404 * @lockdep_ok: whether lock debugging is still OK.
406 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
407 * some notewortht-but-not-corrupting cases, it can be set to true.
409 void add_taint(unsigned flag
, enum lockdep_ok lockdep_ok
)
411 if (lockdep_ok
== LOCKDEP_NOW_UNRELIABLE
&& __debug_locks_off())
412 pr_warn("Disabling lock debugging due to kernel taint\n");
414 set_bit(flag
, &tainted_mask
);
416 EXPORT_SYMBOL(add_taint
);
418 static void spin_msec(int msecs
)
422 for (i
= 0; i
< msecs
; i
++) {
423 touch_nmi_watchdog();
429 * It just happens that oops_enter() and oops_exit() are identically
432 static void do_oops_enter_exit(void)
435 static int spin_counter
;
440 spin_lock_irqsave(&pause_on_oops_lock
, flags
);
441 if (pause_on_oops_flag
== 0) {
442 /* This CPU may now print the oops message */
443 pause_on_oops_flag
= 1;
445 /* We need to stall this CPU */
447 /* This CPU gets to do the counting */
448 spin_counter
= pause_on_oops
;
450 spin_unlock(&pause_on_oops_lock
);
451 spin_msec(MSEC_PER_SEC
);
452 spin_lock(&pause_on_oops_lock
);
453 } while (--spin_counter
);
454 pause_on_oops_flag
= 0;
456 /* This CPU waits for a different one */
457 while (spin_counter
) {
458 spin_unlock(&pause_on_oops_lock
);
460 spin_lock(&pause_on_oops_lock
);
464 spin_unlock_irqrestore(&pause_on_oops_lock
, flags
);
468 * Return true if the calling CPU is allowed to print oops-related info.
469 * This is a bit racy..
471 int oops_may_print(void)
473 return pause_on_oops_flag
== 0;
477 * Called when the architecture enters its oops handler, before it prints
478 * anything. If this is the first CPU to oops, and it's oopsing the first
479 * time then let it proceed.
481 * This is all enabled by the pause_on_oops kernel boot option. We do all
482 * this to ensure that oopses don't scroll off the screen. It has the
483 * side-effect of preventing later-oopsing CPUs from mucking up the display,
486 * It turns out that the CPU which is allowed to print ends up pausing for
487 * the right duration, whereas all the other CPUs pause for twice as long:
488 * once in oops_enter(), once in oops_exit().
490 void oops_enter(void)
493 /* can't trust the integrity of the kernel anymore: */
495 do_oops_enter_exit();
499 * 64-bit random ID for oopses:
503 static int init_oops_id(void)
506 get_random_bytes(&oops_id
, sizeof(oops_id
));
512 late_initcall(init_oops_id
);
514 void print_oops_end_marker(void)
517 pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id
);
521 * Called when the architecture exits its oops handler, after printing
526 do_oops_enter_exit();
527 print_oops_end_marker();
528 kmsg_dump(KMSG_DUMP_OOPS
);
536 void __warn(const char *file
, int line
, void *caller
, unsigned taint
,
537 struct pt_regs
*regs
, struct warn_args
*args
)
539 disable_trace_on_warning();
541 pr_warn("------------[ cut here ]------------\n");
544 pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS\n",
545 raw_smp_processor_id(), current
->pid
, file
, line
,
548 pr_warn("WARNING: CPU: %d PID: %d at %pS\n",
549 raw_smp_processor_id(), current
->pid
, caller
);
552 vprintk(args
->fmt
, args
->args
);
556 * This thread may hit another WARN() in the panic path.
557 * Resetting this prevents additional WARN() from panicking the
558 * system on this thread. Other threads are blocked by the
559 * panic_mutex in panic().
562 panic("panic_on_warn set ...\n");
569 print_oops_end_marker();
571 /* Just a warning, don't kill lockdep. */
572 add_taint(taint
, LOCKDEP_STILL_OK
);
575 #ifdef WANT_WARN_ON_SLOWPATH
576 void warn_slowpath_fmt(const char *file
, int line
, const char *fmt
, ...)
578 struct warn_args args
;
581 va_start(args
.args
, fmt
);
582 __warn(file
, line
, __builtin_return_address(0), TAINT_WARN
, NULL
,
586 EXPORT_SYMBOL(warn_slowpath_fmt
);
588 void warn_slowpath_fmt_taint(const char *file
, int line
,
589 unsigned taint
, const char *fmt
, ...)
591 struct warn_args args
;
594 va_start(args
.args
, fmt
);
595 __warn(file
, line
, __builtin_return_address(0), taint
, NULL
, &args
);
598 EXPORT_SYMBOL(warn_slowpath_fmt_taint
);
600 void warn_slowpath_null(const char *file
, int line
)
602 __warn(file
, line
, __builtin_return_address(0), TAINT_WARN
, NULL
, NULL
);
604 EXPORT_SYMBOL(warn_slowpath_null
);
607 #ifdef CONFIG_CC_STACKPROTECTOR
610 * Called when gcc's -fstack-protector feature is used, and
611 * gcc detects corruption of the on-stack canary value
613 __visible
void __stack_chk_fail(void)
615 panic("stack-protector: Kernel stack is corrupted in: %p\n",
616 __builtin_return_address(0));
618 EXPORT_SYMBOL(__stack_chk_fail
);
622 #ifdef CONFIG_ARCH_HAS_REFCOUNT
623 void refcount_error_report(struct pt_regs
*regs
, const char *err
)
625 WARN_RATELIMIT(1, "refcount_t %s at %pB in %s[%d], uid/euid: %u/%u\n",
626 err
, (void *)instruction_pointer(regs
),
627 current
->comm
, task_pid_nr(current
),
628 from_kuid_munged(&init_user_ns
, current_uid()),
629 from_kuid_munged(&init_user_ns
, current_euid()));
633 core_param(panic
, panic_timeout
, int, 0644);
634 core_param(pause_on_oops
, pause_on_oops
, int, 0644);
635 core_param(panic_on_warn
, panic_on_warn
, int, 0644);
636 core_param(crash_kexec_post_notifiers
, crash_kexec_post_notifiers
, bool, 0644);
638 static int __init
oops_setup(char *s
)
642 if (!strcmp(s
, "panic"))
646 early_param("oops", oops_setup
);