~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/kernel/panic.c

Version: ~ [ linux-5.3-rc5 ] ~ [ linux-5.2.9 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.67 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.139 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.189 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.189 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.72 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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/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/module.h>
 18 #include <linux/random.h>
 19 #include <linux/ftrace.h>
 20 #include <linux/reboot.h>
 21 #include <linux/delay.h>
 22 #include <linux/kexec.h>
 23 #include <linux/sched.h>
 24 #include <linux/sysrq.h>
 25 #include <linux/init.h>
 26 #include <linux/nmi.h>
 27 #include <linux/console.h>
 28 #include <linux/bug.h>
 29 
 30 #define PANIC_TIMER_STEP 100
 31 #define PANIC_BLINK_SPD 18
 32 
 33 int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE;
 34 static unsigned long tainted_mask;
 35 static int pause_on_oops;
 36 static int pause_on_oops_flag;
 37 static DEFINE_SPINLOCK(pause_on_oops_lock);
 38 bool crash_kexec_post_notifiers;
 39 int panic_on_warn __read_mostly;
 40 
 41 int panic_timeout = CONFIG_PANIC_TIMEOUT;
 42 EXPORT_SYMBOL_GPL(panic_timeout);
 43 
 44 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
 45 
 46 EXPORT_SYMBOL(panic_notifier_list);
 47 
 48 static long no_blink(int state)
 49 {
 50         return 0;
 51 }
 52 
 53 /* Returns how long it waited in ms */
 54 long (*panic_blink)(int state);
 55 EXPORT_SYMBOL(panic_blink);
 56 
 57 /*
 58  * Stop ourself in panic -- architecture code may override this
 59  */
 60 void __weak panic_smp_self_stop(void)
 61 {
 62         while (1)
 63                 cpu_relax();
 64 }
 65 
 66 /*
 67  * Stop ourselves in NMI context if another CPU has already panicked. Arch code
 68  * may override this to prepare for crash dumping, e.g. save regs info.
 69  */
 70 void __weak nmi_panic_self_stop(struct pt_regs *regs)
 71 {
 72         panic_smp_self_stop();
 73 }
 74 
 75 /*
 76  * Stop other CPUs in panic.  Architecture dependent code may override this
 77  * with more suitable version.  For example, if the architecture supports
 78  * crash dump, it should save registers of each stopped CPU and disable
 79  * per-CPU features such as virtualization extensions.
 80  */
 81 void __weak crash_smp_send_stop(void)
 82 {
 83         static int cpus_stopped;
 84 
 85         /*
 86          * This function can be called twice in panic path, but obviously
 87          * we execute this only once.
 88          */
 89         if (cpus_stopped)
 90                 return;
 91 
 92         /*
 93          * Note smp_send_stop is the usual smp shutdown function, which
 94          * unfortunately means it may not be hardened to work in a panic
 95          * situation.
 96          */
 97         smp_send_stop();
 98         cpus_stopped = 1;
 99 }
100 
101 atomic_t panic_cpu = ATOMIC_INIT(PANIC_CPU_INVALID);
102 
103 /*
104  * A variant of panic() called from NMI context. We return if we've already
105  * panicked on this CPU. If another CPU already panicked, loop in
106  * nmi_panic_self_stop() which can provide architecture dependent code such
107  * as saving register state for crash dump.
108  */
109 void nmi_panic(struct pt_regs *regs, const char *msg)
110 {
111         int old_cpu, cpu;
112 
113         cpu = raw_smp_processor_id();
114         old_cpu = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, cpu);
115 
116         if (old_cpu == PANIC_CPU_INVALID)
117                 panic("%s", msg);
118         else if (old_cpu != cpu)
119                 nmi_panic_self_stop(regs);
120 }
121 EXPORT_SYMBOL(nmi_panic);
122 
123 /**
124  *      panic - halt the system
125  *      @fmt: The text string to print
126  *
127  *      Display a message, then perform cleanups.
128  *
129  *      This function never returns.
130  */
131 void panic(const char *fmt, ...)
132 {
133         static char buf[1024];
134         va_list args;
135         long i, i_next = 0;
136         int state = 0;
137         int old_cpu, this_cpu;
138         bool _crash_kexec_post_notifiers = crash_kexec_post_notifiers;
139 
140         /*
141          * Disable local interrupts. This will prevent panic_smp_self_stop
142          * from deadlocking the first cpu that invokes the panic, since
143          * there is nothing to prevent an interrupt handler (that runs
144          * after setting panic_cpu) from invoking panic() again.
145          */
146         local_irq_disable();
147 
148         /*
149          * It's possible to come here directly from a panic-assertion and
150          * not have preempt disabled. Some functions called from here want
151          * preempt to be disabled. No point enabling it later though...
152          *
153          * Only one CPU is allowed to execute the panic code from here. For
154          * multiple parallel invocations of panic, all other CPUs either
155          * stop themself or will wait until they are stopped by the 1st CPU
156          * with smp_send_stop().
157          *
158          * `old_cpu == PANIC_CPU_INVALID' means this is the 1st CPU which
159          * comes here, so go ahead.
160          * `old_cpu == this_cpu' means we came from nmi_panic() which sets
161          * panic_cpu to this CPU.  In this case, this is also the 1st CPU.
162          */
163         this_cpu = raw_smp_processor_id();
164         old_cpu  = atomic_cmpxchg(&panic_cpu, PANIC_CPU_INVALID, this_cpu);
165 
166         if (old_cpu != PANIC_CPU_INVALID && old_cpu != this_cpu)
167                 panic_smp_self_stop();
168 
169         console_verbose();
170         bust_spinlocks(1);
171         va_start(args, fmt);
172         vsnprintf(buf, sizeof(buf), fmt, args);
173         va_end(args);
174         pr_emerg("Kernel panic - not syncing: %s\n", buf);
175 #ifdef CONFIG_DEBUG_BUGVERBOSE
176         /*
177          * Avoid nested stack-dumping if a panic occurs during oops processing
178          */
179         if (!test_taint(TAINT_DIE) && oops_in_progress <= 1)
180                 dump_stack();
181 #endif
182 
183         /*
184          * If we have crashed and we have a crash kernel loaded let it handle
185          * everything else.
186          * If we want to run this after calling panic_notifiers, pass
187          * the "crash_kexec_post_notifiers" option to the kernel.
188          *
189          * Bypass the panic_cpu check and call __crash_kexec directly.
190          */
191         if (!_crash_kexec_post_notifiers) {
192                 printk_safe_flush_on_panic();
193                 __crash_kexec(NULL);
194 
195                 /*
196                  * Note smp_send_stop is the usual smp shutdown function, which
197                  * unfortunately means it may not be hardened to work in a
198                  * panic situation.
199                  */
200                 smp_send_stop();
201         } else {
202                 /*
203                  * If we want to do crash dump after notifier calls and
204                  * kmsg_dump, we will need architecture dependent extra
205                  * works in addition to stopping other CPUs.
206                  */
207                 crash_smp_send_stop();
208         }
209 
210         /*
211          * Run any panic handlers, including those that might need to
212          * add information to the kmsg dump output.
213          */
214         atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
215 
216         /* Call flush even twice. It tries harder with a single online CPU */
217         printk_safe_flush_on_panic();
218         kmsg_dump(KMSG_DUMP_PANIC);
219 
220         /*
221          * If you doubt kdump always works fine in any situation,
222          * "crash_kexec_post_notifiers" offers you a chance to run
223          * panic_notifiers and dumping kmsg before kdump.
224          * Note: since some panic_notifiers can make crashed kernel
225          * more unstable, it can increase risks of the kdump failure too.
226          *
227          * Bypass the panic_cpu check and call __crash_kexec directly.
228          */
229         if (_crash_kexec_post_notifiers)
230                 __crash_kexec(NULL);
231 
232         bust_spinlocks(0);
233 
234         /*
235          * We may have ended up stopping the CPU holding the lock (in
236          * smp_send_stop()) while still having some valuable data in the console
237          * buffer.  Try to acquire the lock then release it regardless of the
238          * result.  The release will also print the buffers out.  Locks debug
239          * should be disabled to avoid reporting bad unlock balance when
240          * panic() is not being callled from OOPS.
241          */
242         debug_locks_off();
243         console_flush_on_panic();
244 
245         if (!panic_blink)
246                 panic_blink = no_blink;
247 
248         if (panic_timeout > 0) {
249                 /*
250                  * Delay timeout seconds before rebooting the machine.
251                  * We can't use the "normal" timers since we just panicked.
252                  */
253                 pr_emerg("Rebooting in %d seconds..\n", panic_timeout);
254 
255                 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
256                         touch_nmi_watchdog();
257                         if (i >= i_next) {
258                                 i += panic_blink(state ^= 1);
259                                 i_next = i + 3600 / PANIC_BLINK_SPD;
260                         }
261                         mdelay(PANIC_TIMER_STEP);
262                 }
263         }
264         if (panic_timeout != 0) {
265                 /*
266                  * This will not be a clean reboot, with everything
267                  * shutting down.  But if there is a chance of
268                  * rebooting the system it will be rebooted.
269                  */
270                 emergency_restart();
271         }
272 #ifdef __sparc__
273         {
274                 extern int stop_a_enabled;
275                 /* Make sure the user can actually press Stop-A (L1-A) */
276                 stop_a_enabled = 1;
277                 pr_emerg("Press Stop-A (L1-A) from sun keyboard or send break\n"
278                          "twice on console to return to the boot prom\n");
279         }
280 #endif
281 #if defined(CONFIG_S390)
282         {
283                 unsigned long caller;
284 
285                 caller = (unsigned long)__builtin_return_address(0);
286                 disabled_wait(caller);
287         }
288 #endif
289         pr_emerg("---[ end Kernel panic - not syncing: %s\n", buf);
290         local_irq_enable();
291         for (i = 0; ; i += PANIC_TIMER_STEP) {
292                 touch_softlockup_watchdog();
293                 if (i >= i_next) {
294                         i += panic_blink(state ^= 1);
295                         i_next = i + 3600 / PANIC_BLINK_SPD;
296                 }
297                 mdelay(PANIC_TIMER_STEP);
298         }
299 }
300 
301 EXPORT_SYMBOL(panic);
302 
303 /*
304  * TAINT_FORCED_RMMOD could be a per-module flag but the module
305  * is being removed anyway.
306  */
307 const struct taint_flag taint_flags[TAINT_FLAGS_COUNT] = {
308         { 'P', 'G', true },     /* TAINT_PROPRIETARY_MODULE */
309         { 'F', ' ', true },     /* TAINT_FORCED_MODULE */
310         { 'S', ' ', false },    /* TAINT_CPU_OUT_OF_SPEC */
311         { 'R', ' ', false },    /* TAINT_FORCED_RMMOD */
312         { 'M', ' ', false },    /* TAINT_MACHINE_CHECK */
313         { 'B', ' ', false },    /* TAINT_BAD_PAGE */
314         { 'U', ' ', false },    /* TAINT_USER */
315         { 'D', ' ', false },    /* TAINT_DIE */
316         { 'A', ' ', false },    /* TAINT_OVERRIDDEN_ACPI_TABLE */
317         { 'W', ' ', false },    /* TAINT_WARN */
318         { 'C', ' ', true },     /* TAINT_CRAP */
319         { 'I', ' ', false },    /* TAINT_FIRMWARE_WORKAROUND */
320         { 'O', ' ', true },     /* TAINT_OOT_MODULE */
321         { 'E', ' ', true },     /* TAINT_UNSIGNED_MODULE */
322         { 'L', ' ', false },    /* TAINT_SOFTLOCKUP */
323         { 'K', ' ', true },     /* TAINT_LIVEPATCH */
324 };
325 
326 /**
327  *      print_tainted - return a string to represent the kernel taint state.
328  *
329  *  'P' - Proprietary module has been loaded.
330  *  'F' - Module has been forcibly loaded.
331  *  'S' - SMP with CPUs not designed for SMP.
332  *  'R' - User forced a module unload.
333  *  'M' - System experienced a machine check exception.
334  *  'B' - System has hit bad_page.
335  *  'U' - Userspace-defined naughtiness.
336  *  'D' - Kernel has oopsed before
337  *  'A' - ACPI table overridden.
338  *  'W' - Taint on warning.
339  *  'C' - modules from drivers/staging are loaded.
340  *  'I' - Working around severe firmware bug.
341  *  'O' - Out-of-tree module has been loaded.
342  *  'E' - Unsigned module has been loaded.
343  *  'L' - A soft lockup has previously occurred.
344  *  'K' - Kernel has been live patched.
345  *
346  *      The string is overwritten by the next call to print_tainted().
347  */
348 const char *print_tainted(void)
349 {
350         static char buf[TAINT_FLAGS_COUNT + sizeof("Tainted: ")];
351 
352         if (tainted_mask) {
353                 char *s;
354                 int i;
355 
356                 s = buf + sprintf(buf, "Tainted: ");
357                 for (i = 0; i < TAINT_FLAGS_COUNT; i++) {
358                         const struct taint_flag *t = &taint_flags[i];
359                         *s++ = test_bit(i, &tainted_mask) ?
360                                         t->c_true : t->c_false;
361                 }
362                 *s = 0;
363         } else
364                 snprintf(buf, sizeof(buf), "Not tainted");
365 
366         return buf;
367 }
368 
369 int test_taint(unsigned flag)
370 {
371         return test_bit(flag, &tainted_mask);
372 }
373 EXPORT_SYMBOL(test_taint);
374 
375 unsigned long get_taint(void)
376 {
377         return tainted_mask;
378 }
379 
380 /**
381  * add_taint: add a taint flag if not already set.
382  * @flag: one of the TAINT_* constants.
383  * @lockdep_ok: whether lock debugging is still OK.
384  *
385  * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
386  * some notewortht-but-not-corrupting cases, it can be set to true.
387  */
388 void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
389 {
390         if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
391                 pr_warn("Disabling lock debugging due to kernel taint\n");
392 
393         set_bit(flag, &tainted_mask);
394 }
395 EXPORT_SYMBOL(add_taint);
396 
397 static void spin_msec(int msecs)
398 {
399         int i;
400 
401         for (i = 0; i < msecs; i++) {
402                 touch_nmi_watchdog();
403                 mdelay(1);
404         }
405 }
406 
407 /*
408  * It just happens that oops_enter() and oops_exit() are identically
409  * implemented...
410  */
411 static void do_oops_enter_exit(void)
412 {
413         unsigned long flags;
414         static int spin_counter;
415 
416         if (!pause_on_oops)
417                 return;
418 
419         spin_lock_irqsave(&pause_on_oops_lock, flags);
420         if (pause_on_oops_flag == 0) {
421                 /* This CPU may now print the oops message */
422                 pause_on_oops_flag = 1;
423         } else {
424                 /* We need to stall this CPU */
425                 if (!spin_counter) {
426                         /* This CPU gets to do the counting */
427                         spin_counter = pause_on_oops;
428                         do {
429                                 spin_unlock(&pause_on_oops_lock);
430                                 spin_msec(MSEC_PER_SEC);
431                                 spin_lock(&pause_on_oops_lock);
432                         } while (--spin_counter);
433                         pause_on_oops_flag = 0;
434                 } else {
435                         /* This CPU waits for a different one */
436                         while (spin_counter) {
437                                 spin_unlock(&pause_on_oops_lock);
438                                 spin_msec(1);
439                                 spin_lock(&pause_on_oops_lock);
440                         }
441                 }
442         }
443         spin_unlock_irqrestore(&pause_on_oops_lock, flags);
444 }
445 
446 /*
447  * Return true if the calling CPU is allowed to print oops-related info.
448  * This is a bit racy..
449  */
450 int oops_may_print(void)
451 {
452         return pause_on_oops_flag == 0;
453 }
454 
455 /*
456  * Called when the architecture enters its oops handler, before it prints
457  * anything.  If this is the first CPU to oops, and it's oopsing the first
458  * time then let it proceed.
459  *
460  * This is all enabled by the pause_on_oops kernel boot option.  We do all
461  * this to ensure that oopses don't scroll off the screen.  It has the
462  * side-effect of preventing later-oopsing CPUs from mucking up the display,
463  * too.
464  *
465  * It turns out that the CPU which is allowed to print ends up pausing for
466  * the right duration, whereas all the other CPUs pause for twice as long:
467  * once in oops_enter(), once in oops_exit().
468  */
469 void oops_enter(void)
470 {
471         tracing_off();
472         /* can't trust the integrity of the kernel anymore: */
473         debug_locks_off();
474         do_oops_enter_exit();
475 }
476 
477 /*
478  * 64-bit random ID for oopses:
479  */
480 static u64 oops_id;
481 
482 static int init_oops_id(void)
483 {
484         if (!oops_id)
485                 get_random_bytes(&oops_id, sizeof(oops_id));
486         else
487                 oops_id++;
488 
489         return 0;
490 }
491 late_initcall(init_oops_id);
492 
493 void print_oops_end_marker(void)
494 {
495         init_oops_id();
496         pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id);
497 }
498 
499 /*
500  * Called when the architecture exits its oops handler, after printing
501  * everything.
502  */
503 void oops_exit(void)
504 {
505         do_oops_enter_exit();
506         print_oops_end_marker();
507         kmsg_dump(KMSG_DUMP_OOPS);
508 }
509 
510 struct warn_args {
511         const char *fmt;
512         va_list args;
513 };
514 
515 void __warn(const char *file, int line, void *caller, unsigned taint,
516             struct pt_regs *regs, struct warn_args *args)
517 {
518         disable_trace_on_warning();
519 
520         pr_warn("------------[ cut here ]------------\n");
521 
522         if (file)
523                 pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS\n",
524                         raw_smp_processor_id(), current->pid, file, line,
525                         caller);
526         else
527                 pr_warn("WARNING: CPU: %d PID: %d at %pS\n",
528                         raw_smp_processor_id(), current->pid, caller);
529 
530         if (args)
531                 vprintk(args->fmt, args->args);
532 
533         if (panic_on_warn) {
534                 /*
535                  * This thread may hit another WARN() in the panic path.
536                  * Resetting this prevents additional WARN() from panicking the
537                  * system on this thread.  Other threads are blocked by the
538                  * panic_mutex in panic().
539                  */
540                 panic_on_warn = 0;
541                 panic("panic_on_warn set ...\n");
542         }
543 
544         print_modules();
545 
546         if (regs)
547                 show_regs(regs);
548         else
549                 dump_stack();
550 
551         print_oops_end_marker();
552 
553         /* Just a warning, don't kill lockdep. */
554         add_taint(taint, LOCKDEP_STILL_OK);
555 }
556 
557 #ifdef WANT_WARN_ON_SLOWPATH
558 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
559 {
560         struct warn_args args;
561 
562         args.fmt = fmt;
563         va_start(args.args, fmt);
564         __warn(file, line, __builtin_return_address(0), TAINT_WARN, NULL,
565                &args);
566         va_end(args.args);
567 }
568 EXPORT_SYMBOL(warn_slowpath_fmt);
569 
570 void warn_slowpath_fmt_taint(const char *file, int line,
571                              unsigned taint, const char *fmt, ...)
572 {
573         struct warn_args args;
574 
575         args.fmt = fmt;
576         va_start(args.args, fmt);
577         __warn(file, line, __builtin_return_address(0), taint, NULL, &args);
578         va_end(args.args);
579 }
580 EXPORT_SYMBOL(warn_slowpath_fmt_taint);
581 
582 void warn_slowpath_null(const char *file, int line)
583 {
584         __warn(file, line, __builtin_return_address(0), TAINT_WARN, NULL, NULL);
585 }
586 EXPORT_SYMBOL(warn_slowpath_null);
587 #endif
588 
589 #ifdef CONFIG_CC_STACKPROTECTOR
590 
591 /*
592  * Called when gcc's -fstack-protector feature is used, and
593  * gcc detects corruption of the on-stack canary value
594  */
595 __visible void __stack_chk_fail(void)
596 {
597         panic("stack-protector: Kernel stack is corrupted in: %p\n",
598                 __builtin_return_address(0));
599 }
600 EXPORT_SYMBOL(__stack_chk_fail);
601 
602 #endif
603 
604 core_param(panic, panic_timeout, int, 0644);
605 core_param(pause_on_oops, pause_on_oops, int, 0644);
606 core_param(panic_on_warn, panic_on_warn, int, 0644);
607 core_param(crash_kexec_post_notifiers, crash_kexec_post_notifiers, bool, 0644);
608 
609 static int __init oops_setup(char *s)
610 {
611         if (!s)
612                 return -EINVAL;
613         if (!strcmp(s, "panic"))
614                 panic_on_oops = 1;
615         return 0;
616 }
617 early_param("oops", oops_setup);
618 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp