1 /*
2 * linux/kernel/panic.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7 /*
8 * This function is used through-out the kernel (including mm and fs)
9 * to indicate a major problem.
10 */
11 #include <linux/debug_locks.h>
12 #include <linux/interrupt.h>
13 #include <linux/kmsg_dump.h>
14 #include <linux/kallsyms.h>
15 #include <linux/notifier.h>
16 #include <linux/module.h>
17 #include <linux/random.h>
18 #include <linux/ftrace.h>
19 #include <linux/reboot.h>
20 #include <linux/delay.h>
21 #include <linux/kexec.h>
22 #include <linux/sched.h>
23 #include <linux/sysrq.h>
24 #include <linux/init.h>
25 #include <linux/nmi.h>
26 #include <linux/console.h>
27
28 #define PANIC_TIMER_STEP 100
29 #define PANIC_BLINK_SPD 18
30
31 int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE;
32 static unsigned long tainted_mask;
33 static int pause_on_oops;
34 static int pause_on_oops_flag;
35 static DEFINE_SPINLOCK(pause_on_oops_lock);
36 bool crash_kexec_post_notifiers;
37 int panic_on_warn __read_mostly;
38
39 int panic_timeout = CONFIG_PANIC_TIMEOUT;
40 EXPORT_SYMBOL_GPL(panic_timeout);
41
42 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
43
44 EXPORT_SYMBOL(panic_notifier_list);
45
no_blink(int state)46 static long no_blink(int state)
47 {
48 return 0;
49 }
50
51 /* Returns how long it waited in ms */
52 long (*panic_blink)(int state);
53 EXPORT_SYMBOL(panic_blink);
54
55 /*
56 * Stop ourself in panic -- architecture code may override this
57 */
panic_smp_self_stop(void)58 void __weak panic_smp_self_stop(void)
59 {
60 while (1)
61 cpu_relax();
62 }
63
64 /**
65 * panic - halt the system
66 * @fmt: The text string to print
67 *
68 * Display a message, then perform cleanups.
69 *
70 * This function never returns.
71 */
panic(const char * fmt,...)72 void panic(const char *fmt, ...)
73 {
74 static DEFINE_SPINLOCK(panic_lock);
75 static char buf[1024];
76 va_list args;
77 long i, i_next = 0;
78 int state = 0;
79
80 /*
81 * Disable local interrupts. This will prevent panic_smp_self_stop
82 * from deadlocking the first cpu that invokes the panic, since
83 * there is nothing to prevent an interrupt handler (that runs
84 * after the panic_lock is acquired) from invoking panic again.
85 */
86 local_irq_disable();
87
88 /*
89 * It's possible to come here directly from a panic-assertion and
90 * not have preempt disabled. Some functions called from here want
91 * preempt to be disabled. No point enabling it later though...
92 *
93 * Only one CPU is allowed to execute the panic code from here. For
94 * multiple parallel invocations of panic, all other CPUs either
95 * stop themself or will wait until they are stopped by the 1st CPU
96 * with smp_send_stop().
97 */
98 if (!spin_trylock(&panic_lock))
99 panic_smp_self_stop();
100
101 console_verbose();
102 bust_spinlocks(1);
103 va_start(args, fmt);
104 vsnprintf(buf, sizeof(buf), fmt, args);
105 va_end(args);
106 pr_emerg("Kernel panic - not syncing: %s\n", buf);
107 #ifdef CONFIG_DEBUG_BUGVERBOSE
108 /*
109 * Avoid nested stack-dumping if a panic occurs during oops processing
110 */
111 if (!test_taint(TAINT_DIE) && oops_in_progress <= 1)
112 dump_stack();
113 #endif
114
115 /*
116 * If we have crashed and we have a crash kernel loaded let it handle
117 * everything else.
118 * If we want to run this after calling panic_notifiers, pass
119 * the "crash_kexec_post_notifiers" option to the kernel.
120 */
121 if (!crash_kexec_post_notifiers)
122 crash_kexec(NULL);
123
124 /*
125 * Note smp_send_stop is the usual smp shutdown function, which
126 * unfortunately means it may not be hardened to work in a panic
127 * situation.
128 */
129 smp_send_stop();
130
131 /*
132 * Run any panic handlers, including those that might need to
133 * add information to the kmsg dump output.
134 */
135 atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
136
137 kmsg_dump(KMSG_DUMP_PANIC);
138
139 /*
140 * If you doubt kdump always works fine in any situation,
141 * "crash_kexec_post_notifiers" offers you a chance to run
142 * panic_notifiers and dumping kmsg before kdump.
143 * Note: since some panic_notifiers can make crashed kernel
144 * more unstable, it can increase risks of the kdump failure too.
145 */
146 if (crash_kexec_post_notifiers)
147 crash_kexec(NULL);
148
149 bust_spinlocks(0);
150
151 /*
152 * We may have ended up stopping the CPU holding the lock (in
153 * smp_send_stop()) while still having some valuable data in the console
154 * buffer. Try to acquire the lock then release it regardless of the
155 * result. The release will also print the buffers out. Locks debug
156 * should be disabled to avoid reporting bad unlock balance when
157 * panic() is not being callled from OOPS.
158 */
159 debug_locks_off();
160 console_flush_on_panic();
161
162 if (!panic_blink)
163 panic_blink = no_blink;
164
165 if (panic_timeout > 0) {
166 /*
167 * Delay timeout seconds before rebooting the machine.
168 * We can't use the "normal" timers since we just panicked.
169 */
170 pr_emerg("Rebooting in %d seconds..", panic_timeout);
171
172 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
173 touch_nmi_watchdog();
174 if (i >= i_next) {
175 i += panic_blink(state ^= 1);
176 i_next = i + 3600 / PANIC_BLINK_SPD;
177 }
178 mdelay(PANIC_TIMER_STEP);
179 }
180 }
181 if (panic_timeout != 0) {
182 /*
183 * This will not be a clean reboot, with everything
184 * shutting down. But if there is a chance of
185 * rebooting the system it will be rebooted.
186 */
187 emergency_restart();
188 }
189 #ifdef __sparc__
190 {
191 extern int stop_a_enabled;
192 /* Make sure the user can actually press Stop-A (L1-A) */
193 stop_a_enabled = 1;
194 pr_emerg("Press Stop-A (L1-A) to return to the boot prom\n");
195 }
196 #endif
197 #if defined(CONFIG_S390)
198 {
199 unsigned long caller;
200
201 caller = (unsigned long)__builtin_return_address(0);
202 disabled_wait(caller);
203 }
204 #endif
205 pr_emerg("---[ end Kernel panic - not syncing: %s\n", buf);
206 local_irq_enable();
207 for (i = 0; ; i += PANIC_TIMER_STEP) {
208 touch_softlockup_watchdog();
209 if (i >= i_next) {
210 i += panic_blink(state ^= 1);
211 i_next = i + 3600 / PANIC_BLINK_SPD;
212 }
213 mdelay(PANIC_TIMER_STEP);
214 }
215 }
216
217 EXPORT_SYMBOL(panic);
218
219
220 struct tnt {
221 u8 bit;
222 char true;
223 char false;
224 };
225
226 static const struct tnt tnts[] = {
227 { TAINT_PROPRIETARY_MODULE, 'P', 'G' },
228 { TAINT_FORCED_MODULE, 'F', ' ' },
229 { TAINT_CPU_OUT_OF_SPEC, 'S', ' ' },
230 { TAINT_FORCED_RMMOD, 'R', ' ' },
231 { TAINT_MACHINE_CHECK, 'M', ' ' },
232 { TAINT_BAD_PAGE, 'B', ' ' },
233 { TAINT_USER, 'U', ' ' },
234 { TAINT_DIE, 'D', ' ' },
235 { TAINT_OVERRIDDEN_ACPI_TABLE, 'A', ' ' },
236 { TAINT_WARN, 'W', ' ' },
237 { TAINT_CRAP, 'C', ' ' },
238 { TAINT_FIRMWARE_WORKAROUND, 'I', ' ' },
239 { TAINT_OOT_MODULE, 'O', ' ' },
240 { TAINT_UNSIGNED_MODULE, 'E', ' ' },
241 { TAINT_SOFTLOCKUP, 'L', ' ' },
242 { TAINT_LIVEPATCH, 'K', ' ' },
243 };
244
245 /**
246 * print_tainted - return a string to represent the kernel taint state.
247 *
248 * 'P' - Proprietary module has been loaded.
249 * 'F' - Module has been forcibly loaded.
250 * 'S' - SMP with CPUs not designed for SMP.
251 * 'R' - User forced a module unload.
252 * 'M' - System experienced a machine check exception.
253 * 'B' - System has hit bad_page.
254 * 'U' - Userspace-defined naughtiness.
255 * 'D' - Kernel has oopsed before
256 * 'A' - ACPI table overridden.
257 * 'W' - Taint on warning.
258 * 'C' - modules from drivers/staging are loaded.
259 * 'I' - Working around severe firmware bug.
260 * 'O' - Out-of-tree module has been loaded.
261 * 'E' - Unsigned module has been loaded.
262 * 'L' - A soft lockup has previously occurred.
263 * 'K' - Kernel has been live patched.
264 *
265 * The string is overwritten by the next call to print_tainted().
266 */
print_tainted(void)267 const char *print_tainted(void)
268 {
269 static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ")];
270
271 if (tainted_mask) {
272 char *s;
273 int i;
274
275 s = buf + sprintf(buf, "Tainted: ");
276 for (i = 0; i < ARRAY_SIZE(tnts); i++) {
277 const struct tnt *t = &tnts[i];
278 *s++ = test_bit(t->bit, &tainted_mask) ?
279 t->true : t->false;
280 }
281 *s = 0;
282 } else
283 snprintf(buf, sizeof(buf), "Not tainted");
284
285 return buf;
286 }
287
test_taint(unsigned flag)288 int test_taint(unsigned flag)
289 {
290 return test_bit(flag, &tainted_mask);
291 }
292 EXPORT_SYMBOL(test_taint);
293
get_taint(void)294 unsigned long get_taint(void)
295 {
296 return tainted_mask;
297 }
298
299 /**
300 * add_taint: add a taint flag if not already set.
301 * @flag: one of the TAINT_* constants.
302 * @lockdep_ok: whether lock debugging is still OK.
303 *
304 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
305 * some notewortht-but-not-corrupting cases, it can be set to true.
306 */
add_taint(unsigned flag,enum lockdep_ok lockdep_ok)307 void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
308 {
309 if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
310 pr_warn("Disabling lock debugging due to kernel taint\n");
311
312 set_bit(flag, &tainted_mask);
313 }
314 EXPORT_SYMBOL(add_taint);
315
spin_msec(int msecs)316 static void spin_msec(int msecs)
317 {
318 int i;
319
320 for (i = 0; i < msecs; i++) {
321 touch_nmi_watchdog();
322 mdelay(1);
323 }
324 }
325
326 /*
327 * It just happens that oops_enter() and oops_exit() are identically
328 * implemented...
329 */
do_oops_enter_exit(void)330 static void do_oops_enter_exit(void)
331 {
332 unsigned long flags;
333 static int spin_counter;
334
335 if (!pause_on_oops)
336 return;
337
338 spin_lock_irqsave(&pause_on_oops_lock, flags);
339 if (pause_on_oops_flag == 0) {
340 /* This CPU may now print the oops message */
341 pause_on_oops_flag = 1;
342 } else {
343 /* We need to stall this CPU */
344 if (!spin_counter) {
345 /* This CPU gets to do the counting */
346 spin_counter = pause_on_oops;
347 do {
348 spin_unlock(&pause_on_oops_lock);
349 spin_msec(MSEC_PER_SEC);
350 spin_lock(&pause_on_oops_lock);
351 } while (--spin_counter);
352 pause_on_oops_flag = 0;
353 } else {
354 /* This CPU waits for a different one */
355 while (spin_counter) {
356 spin_unlock(&pause_on_oops_lock);
357 spin_msec(1);
358 spin_lock(&pause_on_oops_lock);
359 }
360 }
361 }
362 spin_unlock_irqrestore(&pause_on_oops_lock, flags);
363 }
364
365 /*
366 * Return true if the calling CPU is allowed to print oops-related info.
367 * This is a bit racy..
368 */
oops_may_print(void)369 int oops_may_print(void)
370 {
371 return pause_on_oops_flag == 0;
372 }
373
374 /*
375 * Called when the architecture enters its oops handler, before it prints
376 * anything. If this is the first CPU to oops, and it's oopsing the first
377 * time then let it proceed.
378 *
379 * This is all enabled by the pause_on_oops kernel boot option. We do all
380 * this to ensure that oopses don't scroll off the screen. It has the
381 * side-effect of preventing later-oopsing CPUs from mucking up the display,
382 * too.
383 *
384 * It turns out that the CPU which is allowed to print ends up pausing for
385 * the right duration, whereas all the other CPUs pause for twice as long:
386 * once in oops_enter(), once in oops_exit().
387 */
oops_enter(void)388 void oops_enter(void)
389 {
390 tracing_off();
391 /* can't trust the integrity of the kernel anymore: */
392 debug_locks_off();
393 do_oops_enter_exit();
394 }
395
396 /*
397 * 64-bit random ID for oopses:
398 */
399 static u64 oops_id;
400
init_oops_id(void)401 static int init_oops_id(void)
402 {
403 if (!oops_id)
404 get_random_bytes(&oops_id, sizeof(oops_id));
405 else
406 oops_id++;
407
408 return 0;
409 }
410 late_initcall(init_oops_id);
411
print_oops_end_marker(void)412 void print_oops_end_marker(void)
413 {
414 init_oops_id();
415 pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id);
416 }
417
418 /*
419 * Called when the architecture exits its oops handler, after printing
420 * everything.
421 */
oops_exit(void)422 void oops_exit(void)
423 {
424 do_oops_enter_exit();
425 print_oops_end_marker();
426 kmsg_dump(KMSG_DUMP_OOPS);
427 }
428
429 #ifdef WANT_WARN_ON_SLOWPATH
430 struct slowpath_args {
431 const char *fmt;
432 va_list args;
433 };
434
warn_slowpath_common(const char * file,int line,void * caller,unsigned taint,struct slowpath_args * args)435 static void warn_slowpath_common(const char *file, int line, void *caller,
436 unsigned taint, struct slowpath_args *args)
437 {
438 disable_trace_on_warning();
439
440 pr_warn("------------[ cut here ]------------\n");
441 pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS()\n",
442 raw_smp_processor_id(), current->pid, file, line, caller);
443
444 if (args)
445 vprintk(args->fmt, args->args);
446
447 if (panic_on_warn) {
448 /*
449 * This thread may hit another WARN() in the panic path.
450 * Resetting this prevents additional WARN() from panicking the
451 * system on this thread. Other threads are blocked by the
452 * panic_mutex in panic().
453 */
454 panic_on_warn = 0;
455 panic("panic_on_warn set ...\n");
456 }
457
458 print_modules();
459 dump_stack();
460 print_oops_end_marker();
461 /* Just a warning, don't kill lockdep. */
462 add_taint(taint, LOCKDEP_STILL_OK);
463 }
464
warn_slowpath_fmt(const char * file,int line,const char * fmt,...)465 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
466 {
467 struct slowpath_args args;
468
469 args.fmt = fmt;
470 va_start(args.args, fmt);
471 warn_slowpath_common(file, line, __builtin_return_address(0),
472 TAINT_WARN, &args);
473 va_end(args.args);
474 }
475 EXPORT_SYMBOL(warn_slowpath_fmt);
476
warn_slowpath_fmt_taint(const char * file,int line,unsigned taint,const char * fmt,...)477 void warn_slowpath_fmt_taint(const char *file, int line,
478 unsigned taint, const char *fmt, ...)
479 {
480 struct slowpath_args args;
481
482 args.fmt = fmt;
483 va_start(args.args, fmt);
484 warn_slowpath_common(file, line, __builtin_return_address(0),
485 taint, &args);
486 va_end(args.args);
487 }
488 EXPORT_SYMBOL(warn_slowpath_fmt_taint);
489
warn_slowpath_null(const char * file,int line)490 void warn_slowpath_null(const char *file, int line)
491 {
492 warn_slowpath_common(file, line, __builtin_return_address(0),
493 TAINT_WARN, NULL);
494 }
495 EXPORT_SYMBOL(warn_slowpath_null);
496 #endif
497
498 #ifdef CONFIG_CC_STACKPROTECTOR
499
500 /*
501 * Called when gcc's -fstack-protector feature is used, and
502 * gcc detects corruption of the on-stack canary value
503 */
__stack_chk_fail(void)504 __visible void __stack_chk_fail(void)
505 {
506 panic("stack-protector: Kernel stack is corrupted in: %p\n",
507 __builtin_return_address(0));
508 }
509 EXPORT_SYMBOL(__stack_chk_fail);
510
511 #endif
512
513 core_param(panic, panic_timeout, int, 0644);
514 core_param(pause_on_oops, pause_on_oops, int, 0644);
515 core_param(panic_on_warn, panic_on_warn, int, 0644);
516
setup_crash_kexec_post_notifiers(char * s)517 static int __init setup_crash_kexec_post_notifiers(char *s)
518 {
519 crash_kexec_post_notifiers = true;
520 return 0;
521 }
522 early_param("crash_kexec_post_notifiers", setup_crash_kexec_post_notifiers);
523
oops_setup(char * s)524 static int __init oops_setup(char *s)
525 {
526 if (!s)
527 return -EINVAL;
528 if (!strcmp(s, "panic"))
529 panic_on_oops = 1;
530 return 0;
531 }
532 early_param("oops", oops_setup);
533