xref: /openbmc/linux/kernel/panic.c (revision 75f25bd3)
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/reboot.h>
19 #include <linux/delay.h>
20 #include <linux/kexec.h>
21 #include <linux/sched.h>
22 #include <linux/sysrq.h>
23 #include <linux/init.h>
24 #include <linux/nmi.h>
25 #include <linux/dmi.h>
26 
27 #define PANIC_TIMER_STEP 100
28 #define PANIC_BLINK_SPD 18
29 
30 int panic_on_oops;
31 static unsigned long tainted_mask;
32 static int pause_on_oops;
33 static int pause_on_oops_flag;
34 static DEFINE_SPINLOCK(pause_on_oops_lock);
35 
36 int panic_timeout;
37 EXPORT_SYMBOL_GPL(panic_timeout);
38 
39 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
40 
41 EXPORT_SYMBOL(panic_notifier_list);
42 
43 static long no_blink(int state)
44 {
45 	return 0;
46 }
47 
48 /* Returns how long it waited in ms */
49 long (*panic_blink)(int state);
50 EXPORT_SYMBOL(panic_blink);
51 
52 /**
53  *	panic - halt the system
54  *	@fmt: The text string to print
55  *
56  *	Display a message, then perform cleanups.
57  *
58  *	This function never returns.
59  */
60 NORET_TYPE void panic(const char * fmt, ...)
61 {
62 	static char buf[1024];
63 	va_list args;
64 	long i, i_next = 0;
65 	int state = 0;
66 
67 	/*
68 	 * It's possible to come here directly from a panic-assertion and
69 	 * not have preempt disabled. Some functions called from here want
70 	 * preempt to be disabled. No point enabling it later though...
71 	 */
72 	preempt_disable();
73 
74 	console_verbose();
75 	bust_spinlocks(1);
76 	va_start(args, fmt);
77 	vsnprintf(buf, sizeof(buf), fmt, args);
78 	va_end(args);
79 	printk(KERN_EMERG "Kernel panic - not syncing: %s\n",buf);
80 #ifdef CONFIG_DEBUG_BUGVERBOSE
81 	dump_stack();
82 #endif
83 
84 	/*
85 	 * If we have crashed and we have a crash kernel loaded let it handle
86 	 * everything else.
87 	 * Do we want to call this before we try to display a message?
88 	 */
89 	crash_kexec(NULL);
90 
91 	kmsg_dump(KMSG_DUMP_PANIC);
92 
93 	/*
94 	 * Note smp_send_stop is the usual smp shutdown function, which
95 	 * unfortunately means it may not be hardened to work in a panic
96 	 * situation.
97 	 */
98 	smp_send_stop();
99 
100 	atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
101 
102 	bust_spinlocks(0);
103 
104 	if (!panic_blink)
105 		panic_blink = no_blink;
106 
107 	if (panic_timeout > 0) {
108 		/*
109 		 * Delay timeout seconds before rebooting the machine.
110 		 * We can't use the "normal" timers since we just panicked.
111 		 */
112 		printk(KERN_EMERG "Rebooting in %d seconds..", panic_timeout);
113 
114 		for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
115 			touch_nmi_watchdog();
116 			if (i >= i_next) {
117 				i += panic_blink(state ^= 1);
118 				i_next = i + 3600 / PANIC_BLINK_SPD;
119 			}
120 			mdelay(PANIC_TIMER_STEP);
121 		}
122 	}
123 	if (panic_timeout != 0) {
124 		/*
125 		 * This will not be a clean reboot, with everything
126 		 * shutting down.  But if there is a chance of
127 		 * rebooting the system it will be rebooted.
128 		 */
129 		emergency_restart();
130 	}
131 #ifdef __sparc__
132 	{
133 		extern int stop_a_enabled;
134 		/* Make sure the user can actually press Stop-A (L1-A) */
135 		stop_a_enabled = 1;
136 		printk(KERN_EMERG "Press Stop-A (L1-A) to return to the boot prom\n");
137 	}
138 #endif
139 #if defined(CONFIG_S390)
140 	{
141 		unsigned long caller;
142 
143 		caller = (unsigned long)__builtin_return_address(0);
144 		disabled_wait(caller);
145 	}
146 #endif
147 	local_irq_enable();
148 	for (i = 0; ; i += PANIC_TIMER_STEP) {
149 		touch_softlockup_watchdog();
150 		if (i >= i_next) {
151 			i += panic_blink(state ^= 1);
152 			i_next = i + 3600 / PANIC_BLINK_SPD;
153 		}
154 		mdelay(PANIC_TIMER_STEP);
155 	}
156 }
157 
158 EXPORT_SYMBOL(panic);
159 
160 
161 struct tnt {
162 	u8	bit;
163 	char	true;
164 	char	false;
165 };
166 
167 static const struct tnt tnts[] = {
168 	{ TAINT_PROPRIETARY_MODULE,	'P', 'G' },
169 	{ TAINT_FORCED_MODULE,		'F', ' ' },
170 	{ TAINT_UNSAFE_SMP,		'S', ' ' },
171 	{ TAINT_FORCED_RMMOD,		'R', ' ' },
172 	{ TAINT_MACHINE_CHECK,		'M', ' ' },
173 	{ TAINT_BAD_PAGE,		'B', ' ' },
174 	{ TAINT_USER,			'U', ' ' },
175 	{ TAINT_DIE,			'D', ' ' },
176 	{ TAINT_OVERRIDDEN_ACPI_TABLE,	'A', ' ' },
177 	{ TAINT_WARN,			'W', ' ' },
178 	{ TAINT_CRAP,			'C', ' ' },
179 	{ TAINT_FIRMWARE_WORKAROUND,	'I', ' ' },
180 };
181 
182 /**
183  *	print_tainted - return a string to represent the kernel taint state.
184  *
185  *  'P' - Proprietary module has been loaded.
186  *  'F' - Module has been forcibly loaded.
187  *  'S' - SMP with CPUs not designed for SMP.
188  *  'R' - User forced a module unload.
189  *  'M' - System experienced a machine check exception.
190  *  'B' - System has hit bad_page.
191  *  'U' - Userspace-defined naughtiness.
192  *  'D' - Kernel has oopsed before
193  *  'A' - ACPI table overridden.
194  *  'W' - Taint on warning.
195  *  'C' - modules from drivers/staging are loaded.
196  *  'I' - Working around severe firmware bug.
197  *
198  *	The string is overwritten by the next call to print_tainted().
199  */
200 const char *print_tainted(void)
201 {
202 	static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
203 
204 	if (tainted_mask) {
205 		char *s;
206 		int i;
207 
208 		s = buf + sprintf(buf, "Tainted: ");
209 		for (i = 0; i < ARRAY_SIZE(tnts); i++) {
210 			const struct tnt *t = &tnts[i];
211 			*s++ = test_bit(t->bit, &tainted_mask) ?
212 					t->true : t->false;
213 		}
214 		*s = 0;
215 	} else
216 		snprintf(buf, sizeof(buf), "Not tainted");
217 
218 	return buf;
219 }
220 
221 int test_taint(unsigned flag)
222 {
223 	return test_bit(flag, &tainted_mask);
224 }
225 EXPORT_SYMBOL(test_taint);
226 
227 unsigned long get_taint(void)
228 {
229 	return tainted_mask;
230 }
231 
232 void add_taint(unsigned flag)
233 {
234 	/*
235 	 * Can't trust the integrity of the kernel anymore.
236 	 * We don't call directly debug_locks_off() because the issue
237 	 * is not necessarily serious enough to set oops_in_progress to 1
238 	 * Also we want to keep up lockdep for staging development and
239 	 * post-warning case.
240 	 */
241 	if (flag != TAINT_CRAP && flag != TAINT_WARN && __debug_locks_off())
242 		printk(KERN_WARNING "Disabling lock debugging due to kernel taint\n");
243 
244 	set_bit(flag, &tainted_mask);
245 }
246 EXPORT_SYMBOL(add_taint);
247 
248 static void spin_msec(int msecs)
249 {
250 	int i;
251 
252 	for (i = 0; i < msecs; i++) {
253 		touch_nmi_watchdog();
254 		mdelay(1);
255 	}
256 }
257 
258 /*
259  * It just happens that oops_enter() and oops_exit() are identically
260  * implemented...
261  */
262 static void do_oops_enter_exit(void)
263 {
264 	unsigned long flags;
265 	static int spin_counter;
266 
267 	if (!pause_on_oops)
268 		return;
269 
270 	spin_lock_irqsave(&pause_on_oops_lock, flags);
271 	if (pause_on_oops_flag == 0) {
272 		/* This CPU may now print the oops message */
273 		pause_on_oops_flag = 1;
274 	} else {
275 		/* We need to stall this CPU */
276 		if (!spin_counter) {
277 			/* This CPU gets to do the counting */
278 			spin_counter = pause_on_oops;
279 			do {
280 				spin_unlock(&pause_on_oops_lock);
281 				spin_msec(MSEC_PER_SEC);
282 				spin_lock(&pause_on_oops_lock);
283 			} while (--spin_counter);
284 			pause_on_oops_flag = 0;
285 		} else {
286 			/* This CPU waits for a different one */
287 			while (spin_counter) {
288 				spin_unlock(&pause_on_oops_lock);
289 				spin_msec(1);
290 				spin_lock(&pause_on_oops_lock);
291 			}
292 		}
293 	}
294 	spin_unlock_irqrestore(&pause_on_oops_lock, flags);
295 }
296 
297 /*
298  * Return true if the calling CPU is allowed to print oops-related info.
299  * This is a bit racy..
300  */
301 int oops_may_print(void)
302 {
303 	return pause_on_oops_flag == 0;
304 }
305 
306 /*
307  * Called when the architecture enters its oops handler, before it prints
308  * anything.  If this is the first CPU to oops, and it's oopsing the first
309  * time then let it proceed.
310  *
311  * This is all enabled by the pause_on_oops kernel boot option.  We do all
312  * this to ensure that oopses don't scroll off the screen.  It has the
313  * side-effect of preventing later-oopsing CPUs from mucking up the display,
314  * too.
315  *
316  * It turns out that the CPU which is allowed to print ends up pausing for
317  * the right duration, whereas all the other CPUs pause for twice as long:
318  * once in oops_enter(), once in oops_exit().
319  */
320 void oops_enter(void)
321 {
322 	tracing_off();
323 	/* can't trust the integrity of the kernel anymore: */
324 	debug_locks_off();
325 	do_oops_enter_exit();
326 }
327 
328 /*
329  * 64-bit random ID for oopses:
330  */
331 static u64 oops_id;
332 
333 static int init_oops_id(void)
334 {
335 	if (!oops_id)
336 		get_random_bytes(&oops_id, sizeof(oops_id));
337 	else
338 		oops_id++;
339 
340 	return 0;
341 }
342 late_initcall(init_oops_id);
343 
344 void print_oops_end_marker(void)
345 {
346 	init_oops_id();
347 	printk(KERN_WARNING "---[ end trace %016llx ]---\n",
348 		(unsigned long long)oops_id);
349 }
350 
351 /*
352  * Called when the architecture exits its oops handler, after printing
353  * everything.
354  */
355 void oops_exit(void)
356 {
357 	do_oops_enter_exit();
358 	print_oops_end_marker();
359 	kmsg_dump(KMSG_DUMP_OOPS);
360 }
361 
362 #ifdef WANT_WARN_ON_SLOWPATH
363 struct slowpath_args {
364 	const char *fmt;
365 	va_list args;
366 };
367 
368 static void warn_slowpath_common(const char *file, int line, void *caller,
369 				 unsigned taint, struct slowpath_args *args)
370 {
371 	const char *board;
372 
373 	printk(KERN_WARNING "------------[ cut here ]------------\n");
374 	printk(KERN_WARNING "WARNING: at %s:%d %pS()\n", file, line, caller);
375 	board = dmi_get_system_info(DMI_PRODUCT_NAME);
376 	if (board)
377 		printk(KERN_WARNING "Hardware name: %s\n", board);
378 
379 	if (args)
380 		vprintk(args->fmt, args->args);
381 
382 	print_modules();
383 	dump_stack();
384 	print_oops_end_marker();
385 	add_taint(taint);
386 }
387 
388 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
389 {
390 	struct slowpath_args args;
391 
392 	args.fmt = fmt;
393 	va_start(args.args, fmt);
394 	warn_slowpath_common(file, line, __builtin_return_address(0),
395 			     TAINT_WARN, &args);
396 	va_end(args.args);
397 }
398 EXPORT_SYMBOL(warn_slowpath_fmt);
399 
400 void warn_slowpath_fmt_taint(const char *file, int line,
401 			     unsigned taint, const char *fmt, ...)
402 {
403 	struct slowpath_args args;
404 
405 	args.fmt = fmt;
406 	va_start(args.args, fmt);
407 	warn_slowpath_common(file, line, __builtin_return_address(0),
408 			     taint, &args);
409 	va_end(args.args);
410 }
411 EXPORT_SYMBOL(warn_slowpath_fmt_taint);
412 
413 void warn_slowpath_null(const char *file, int line)
414 {
415 	warn_slowpath_common(file, line, __builtin_return_address(0),
416 			     TAINT_WARN, NULL);
417 }
418 EXPORT_SYMBOL(warn_slowpath_null);
419 #endif
420 
421 #ifdef CONFIG_CC_STACKPROTECTOR
422 
423 /*
424  * Called when gcc's -fstack-protector feature is used, and
425  * gcc detects corruption of the on-stack canary value
426  */
427 void __stack_chk_fail(void)
428 {
429 	panic("stack-protector: Kernel stack is corrupted in: %p\n",
430 		__builtin_return_address(0));
431 }
432 EXPORT_SYMBOL(__stack_chk_fail);
433 
434 #endif
435 
436 core_param(panic, panic_timeout, int, 0644);
437 core_param(pause_on_oops, pause_on_oops, int, 0644);
438 
439 static int __init oops_setup(char *s)
440 {
441 	if (!s)
442 		return -EINVAL;
443 	if (!strcmp(s, "panic"))
444 		panic_on_oops = 1;
445 	return 0;
446 }
447 early_param("oops", oops_setup);
448