Merge tag 'dmaengine-fix-4.2-rc5' of git://git.infradead.org/users/vkoul/slave-dma
[linux-2.6-block.git] / kernel / panic.c
... / ...
CommitLineData
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
27#define PANIC_TIMER_STEP 100
28#define PANIC_BLINK_SPD 18
29
30int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE;
31static unsigned long tainted_mask;
32static int pause_on_oops;
33static int pause_on_oops_flag;
34static DEFINE_SPINLOCK(pause_on_oops_lock);
35bool crash_kexec_post_notifiers;
36int panic_on_warn __read_mostly;
37
38int panic_timeout = CONFIG_PANIC_TIMEOUT;
39EXPORT_SYMBOL_GPL(panic_timeout);
40
41ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
42
43EXPORT_SYMBOL(panic_notifier_list);
44
45static long no_blink(int state)
46{
47 return 0;
48}
49
50/* Returns how long it waited in ms */
51long (*panic_blink)(int state);
52EXPORT_SYMBOL(panic_blink);
53
54/*
55 * Stop ourself in panic -- architecture code may override this
56 */
57void __weak panic_smp_self_stop(void)
58{
59 while (1)
60 cpu_relax();
61}
62
63/**
64 * panic - halt the system
65 * @fmt: The text string to print
66 *
67 * Display a message, then perform cleanups.
68 *
69 * This function never returns.
70 */
71void panic(const char *fmt, ...)
72{
73 static DEFINE_SPINLOCK(panic_lock);
74 static char buf[1024];
75 va_list args;
76 long i, i_next = 0;
77 int state = 0;
78
79 /*
80 * Disable local interrupts. This will prevent panic_smp_self_stop
81 * from deadlocking the first cpu that invokes the panic, since
82 * there is nothing to prevent an interrupt handler (that runs
83 * after the panic_lock is acquired) from invoking panic again.
84 */
85 local_irq_disable();
86
87 /*
88 * It's possible to come here directly from a panic-assertion and
89 * not have preempt disabled. Some functions called from here want
90 * preempt to be disabled. No point enabling it later though...
91 *
92 * Only one CPU is allowed to execute the panic code from here. For
93 * multiple parallel invocations of panic, all other CPUs either
94 * stop themself or will wait until they are stopped by the 1st CPU
95 * with smp_send_stop().
96 */
97 if (!spin_trylock(&panic_lock))
98 panic_smp_self_stop();
99
100 console_verbose();
101 bust_spinlocks(1);
102 va_start(args, fmt);
103 vsnprintf(buf, sizeof(buf), fmt, args);
104 va_end(args);
105 pr_emerg("Kernel panic - not syncing: %s\n", buf);
106#ifdef CONFIG_DEBUG_BUGVERBOSE
107 /*
108 * Avoid nested stack-dumping if a panic occurs during oops processing
109 */
110 if (!test_taint(TAINT_DIE) && oops_in_progress <= 1)
111 dump_stack();
112#endif
113
114 /*
115 * If we have crashed and we have a crash kernel loaded let it handle
116 * everything else.
117 * If we want to run this after calling panic_notifiers, pass
118 * the "crash_kexec_post_notifiers" option to the kernel.
119 */
120 if (!crash_kexec_post_notifiers)
121 crash_kexec(NULL);
122
123 /*
124 * Note smp_send_stop is the usual smp shutdown function, which
125 * unfortunately means it may not be hardened to work in a panic
126 * situation.
127 */
128 smp_send_stop();
129
130 /*
131 * Run any panic handlers, including those that might need to
132 * add information to the kmsg dump output.
133 */
134 atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
135
136 kmsg_dump(KMSG_DUMP_PANIC);
137
138 /*
139 * If you doubt kdump always works fine in any situation,
140 * "crash_kexec_post_notifiers" offers you a chance to run
141 * panic_notifiers and dumping kmsg before kdump.
142 * Note: since some panic_notifiers can make crashed kernel
143 * more unstable, it can increase risks of the kdump failure too.
144 */
145 if (crash_kexec_post_notifiers)
146 crash_kexec(NULL);
147
148 bust_spinlocks(0);
149
150 if (!panic_blink)
151 panic_blink = no_blink;
152
153 if (panic_timeout > 0) {
154 /*
155 * Delay timeout seconds before rebooting the machine.
156 * We can't use the "normal" timers since we just panicked.
157 */
158 pr_emerg("Rebooting in %d seconds..", panic_timeout);
159
160 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
161 touch_nmi_watchdog();
162 if (i >= i_next) {
163 i += panic_blink(state ^= 1);
164 i_next = i + 3600 / PANIC_BLINK_SPD;
165 }
166 mdelay(PANIC_TIMER_STEP);
167 }
168 }
169 if (panic_timeout != 0) {
170 /*
171 * This will not be a clean reboot, with everything
172 * shutting down. But if there is a chance of
173 * rebooting the system it will be rebooted.
174 */
175 emergency_restart();
176 }
177#ifdef __sparc__
178 {
179 extern int stop_a_enabled;
180 /* Make sure the user can actually press Stop-A (L1-A) */
181 stop_a_enabled = 1;
182 pr_emerg("Press Stop-A (L1-A) to return to the boot prom\n");
183 }
184#endif
185#if defined(CONFIG_S390)
186 {
187 unsigned long caller;
188
189 caller = (unsigned long)__builtin_return_address(0);
190 disabled_wait(caller);
191 }
192#endif
193 pr_emerg("---[ end Kernel panic - not syncing: %s\n", buf);
194 local_irq_enable();
195 for (i = 0; ; i += PANIC_TIMER_STEP) {
196 touch_softlockup_watchdog();
197 if (i >= i_next) {
198 i += panic_blink(state ^= 1);
199 i_next = i + 3600 / PANIC_BLINK_SPD;
200 }
201 mdelay(PANIC_TIMER_STEP);
202 }
203}
204
205EXPORT_SYMBOL(panic);
206
207
208struct tnt {
209 u8 bit;
210 char true;
211 char false;
212};
213
214static const struct tnt tnts[] = {
215 { TAINT_PROPRIETARY_MODULE, 'P', 'G' },
216 { TAINT_FORCED_MODULE, 'F', ' ' },
217 { TAINT_CPU_OUT_OF_SPEC, 'S', ' ' },
218 { TAINT_FORCED_RMMOD, 'R', ' ' },
219 { TAINT_MACHINE_CHECK, 'M', ' ' },
220 { TAINT_BAD_PAGE, 'B', ' ' },
221 { TAINT_USER, 'U', ' ' },
222 { TAINT_DIE, 'D', ' ' },
223 { TAINT_OVERRIDDEN_ACPI_TABLE, 'A', ' ' },
224 { TAINT_WARN, 'W', ' ' },
225 { TAINT_CRAP, 'C', ' ' },
226 { TAINT_FIRMWARE_WORKAROUND, 'I', ' ' },
227 { TAINT_OOT_MODULE, 'O', ' ' },
228 { TAINT_UNSIGNED_MODULE, 'E', ' ' },
229 { TAINT_SOFTLOCKUP, 'L', ' ' },
230 { TAINT_LIVEPATCH, 'K', ' ' },
231};
232
233/**
234 * print_tainted - return a string to represent the kernel taint state.
235 *
236 * 'P' - Proprietary module has been loaded.
237 * 'F' - Module has been forcibly loaded.
238 * 'S' - SMP with CPUs not designed for SMP.
239 * 'R' - User forced a module unload.
240 * 'M' - System experienced a machine check exception.
241 * 'B' - System has hit bad_page.
242 * 'U' - Userspace-defined naughtiness.
243 * 'D' - Kernel has oopsed before
244 * 'A' - ACPI table overridden.
245 * 'W' - Taint on warning.
246 * 'C' - modules from drivers/staging are loaded.
247 * 'I' - Working around severe firmware bug.
248 * 'O' - Out-of-tree module has been loaded.
249 * 'E' - Unsigned module has been loaded.
250 * 'L' - A soft lockup has previously occurred.
251 * 'K' - Kernel has been live patched.
252 *
253 * The string is overwritten by the next call to print_tainted().
254 */
255const char *print_tainted(void)
256{
257 static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ")];
258
259 if (tainted_mask) {
260 char *s;
261 int i;
262
263 s = buf + sprintf(buf, "Tainted: ");
264 for (i = 0; i < ARRAY_SIZE(tnts); i++) {
265 const struct tnt *t = &tnts[i];
266 *s++ = test_bit(t->bit, &tainted_mask) ?
267 t->true : t->false;
268 }
269 *s = 0;
270 } else
271 snprintf(buf, sizeof(buf), "Not tainted");
272
273 return buf;
274}
275
276int test_taint(unsigned flag)
277{
278 return test_bit(flag, &tainted_mask);
279}
280EXPORT_SYMBOL(test_taint);
281
282unsigned long get_taint(void)
283{
284 return tainted_mask;
285}
286
287/**
288 * add_taint: add a taint flag if not already set.
289 * @flag: one of the TAINT_* constants.
290 * @lockdep_ok: whether lock debugging is still OK.
291 *
292 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
293 * some notewortht-but-not-corrupting cases, it can be set to true.
294 */
295void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
296{
297 if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
298 pr_warn("Disabling lock debugging due to kernel taint\n");
299
300 set_bit(flag, &tainted_mask);
301}
302EXPORT_SYMBOL(add_taint);
303
304static void spin_msec(int msecs)
305{
306 int i;
307
308 for (i = 0; i < msecs; i++) {
309 touch_nmi_watchdog();
310 mdelay(1);
311 }
312}
313
314/*
315 * It just happens that oops_enter() and oops_exit() are identically
316 * implemented...
317 */
318static void do_oops_enter_exit(void)
319{
320 unsigned long flags;
321 static int spin_counter;
322
323 if (!pause_on_oops)
324 return;
325
326 spin_lock_irqsave(&pause_on_oops_lock, flags);
327 if (pause_on_oops_flag == 0) {
328 /* This CPU may now print the oops message */
329 pause_on_oops_flag = 1;
330 } else {
331 /* We need to stall this CPU */
332 if (!spin_counter) {
333 /* This CPU gets to do the counting */
334 spin_counter = pause_on_oops;
335 do {
336 spin_unlock(&pause_on_oops_lock);
337 spin_msec(MSEC_PER_SEC);
338 spin_lock(&pause_on_oops_lock);
339 } while (--spin_counter);
340 pause_on_oops_flag = 0;
341 } else {
342 /* This CPU waits for a different one */
343 while (spin_counter) {
344 spin_unlock(&pause_on_oops_lock);
345 spin_msec(1);
346 spin_lock(&pause_on_oops_lock);
347 }
348 }
349 }
350 spin_unlock_irqrestore(&pause_on_oops_lock, flags);
351}
352
353/*
354 * Return true if the calling CPU is allowed to print oops-related info.
355 * This is a bit racy..
356 */
357int oops_may_print(void)
358{
359 return pause_on_oops_flag == 0;
360}
361
362/*
363 * Called when the architecture enters its oops handler, before it prints
364 * anything. If this is the first CPU to oops, and it's oopsing the first
365 * time then let it proceed.
366 *
367 * This is all enabled by the pause_on_oops kernel boot option. We do all
368 * this to ensure that oopses don't scroll off the screen. It has the
369 * side-effect of preventing later-oopsing CPUs from mucking up the display,
370 * too.
371 *
372 * It turns out that the CPU which is allowed to print ends up pausing for
373 * the right duration, whereas all the other CPUs pause for twice as long:
374 * once in oops_enter(), once in oops_exit().
375 */
376void oops_enter(void)
377{
378 tracing_off();
379 /* can't trust the integrity of the kernel anymore: */
380 debug_locks_off();
381 do_oops_enter_exit();
382}
383
384/*
385 * 64-bit random ID for oopses:
386 */
387static u64 oops_id;
388
389static int init_oops_id(void)
390{
391 if (!oops_id)
392 get_random_bytes(&oops_id, sizeof(oops_id));
393 else
394 oops_id++;
395
396 return 0;
397}
398late_initcall(init_oops_id);
399
400void print_oops_end_marker(void)
401{
402 init_oops_id();
403 pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id);
404}
405
406/*
407 * Called when the architecture exits its oops handler, after printing
408 * everything.
409 */
410void oops_exit(void)
411{
412 do_oops_enter_exit();
413 print_oops_end_marker();
414 kmsg_dump(KMSG_DUMP_OOPS);
415}
416
417#ifdef WANT_WARN_ON_SLOWPATH
418struct slowpath_args {
419 const char *fmt;
420 va_list args;
421};
422
423static void warn_slowpath_common(const char *file, int line, void *caller,
424 unsigned taint, struct slowpath_args *args)
425{
426 disable_trace_on_warning();
427
428 pr_warn("------------[ cut here ]------------\n");
429 pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS()\n",
430 raw_smp_processor_id(), current->pid, file, line, caller);
431
432 if (args)
433 vprintk(args->fmt, args->args);
434
435 if (panic_on_warn) {
436 /*
437 * This thread may hit another WARN() in the panic path.
438 * Resetting this prevents additional WARN() from panicking the
439 * system on this thread. Other threads are blocked by the
440 * panic_mutex in panic().
441 */
442 panic_on_warn = 0;
443 panic("panic_on_warn set ...\n");
444 }
445
446 print_modules();
447 dump_stack();
448 print_oops_end_marker();
449 /* Just a warning, don't kill lockdep. */
450 add_taint(taint, LOCKDEP_STILL_OK);
451}
452
453void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
454{
455 struct slowpath_args args;
456
457 args.fmt = fmt;
458 va_start(args.args, fmt);
459 warn_slowpath_common(file, line, __builtin_return_address(0),
460 TAINT_WARN, &args);
461 va_end(args.args);
462}
463EXPORT_SYMBOL(warn_slowpath_fmt);
464
465void warn_slowpath_fmt_taint(const char *file, int line,
466 unsigned taint, const char *fmt, ...)
467{
468 struct slowpath_args args;
469
470 args.fmt = fmt;
471 va_start(args.args, fmt);
472 warn_slowpath_common(file, line, __builtin_return_address(0),
473 taint, &args);
474 va_end(args.args);
475}
476EXPORT_SYMBOL(warn_slowpath_fmt_taint);
477
478void warn_slowpath_null(const char *file, int line)
479{
480 warn_slowpath_common(file, line, __builtin_return_address(0),
481 TAINT_WARN, NULL);
482}
483EXPORT_SYMBOL(warn_slowpath_null);
484#endif
485
486#ifdef CONFIG_CC_STACKPROTECTOR
487
488/*
489 * Called when gcc's -fstack-protector feature is used, and
490 * gcc detects corruption of the on-stack canary value
491 */
492__visible void __stack_chk_fail(void)
493{
494 panic("stack-protector: Kernel stack is corrupted in: %p\n",
495 __builtin_return_address(0));
496}
497EXPORT_SYMBOL(__stack_chk_fail);
498
499#endif
500
501core_param(panic, panic_timeout, int, 0644);
502core_param(pause_on_oops, pause_on_oops, int, 0644);
503core_param(panic_on_warn, panic_on_warn, int, 0644);
504
505static int __init setup_crash_kexec_post_notifiers(char *s)
506{
507 crash_kexec_post_notifiers = true;
508 return 0;
509}
510early_param("crash_kexec_post_notifiers", setup_crash_kexec_post_notifiers);
511
512static int __init oops_setup(char *s)
513{
514 if (!s)
515 return -EINVAL;
516 if (!strcmp(s, "panic"))
517 panic_on_oops = 1;
518 return 0;
519}
520early_param("oops", oops_setup);