xref: /openbmc/linux/kernel/panic.c (revision 92ed1a76)
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 
38 ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
39 
40 EXPORT_SYMBOL(panic_notifier_list);
41 
42 static long no_blink(int state)
43 {
44 	return 0;
45 }
46 
47 /* Returns how long it waited in ms */
48 long (*panic_blink)(int state);
49 EXPORT_SYMBOL(panic_blink);
50 
51 /**
52  *	panic - halt the system
53  *	@fmt: The text string to print
54  *
55  *	Display a message, then perform cleanups.
56  *
57  *	This function never returns.
58  */
59 NORET_TYPE void panic(const char * fmt, ...)
60 {
61 	static char buf[1024];
62 	va_list args;
63 	long i, i_next = 0;
64 	int state = 0;
65 
66 	/*
67 	 * It's possible to come here directly from a panic-assertion and
68 	 * not have preempt disabled. Some functions called from here want
69 	 * preempt to be disabled. No point enabling it later though...
70 	 */
71 	preempt_disable();
72 
73 	console_verbose();
74 	bust_spinlocks(1);
75 	va_start(args, fmt);
76 	vsnprintf(buf, sizeof(buf), fmt, args);
77 	va_end(args);
78 	printk(KERN_EMERG "Kernel panic - not syncing: %s\n",buf);
79 #ifdef CONFIG_DEBUG_BUGVERBOSE
80 	dump_stack();
81 #endif
82 
83 	/*
84 	 * If we have crashed and we have a crash kernel loaded let it handle
85 	 * everything else.
86 	 * Do we want to call this before we try to display a message?
87 	 */
88 	crash_kexec(NULL);
89 
90 	kmsg_dump(KMSG_DUMP_PANIC);
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 
99 	atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
100 
101 	bust_spinlocks(0);
102 
103 	if (!panic_blink)
104 		panic_blink = no_blink;
105 
106 	if (panic_timeout > 0) {
107 		/*
108 		 * Delay timeout seconds before rebooting the machine.
109 		 * We can't use the "normal" timers since we just panicked.
110 		 */
111 		printk(KERN_EMERG "Rebooting in %d seconds..", panic_timeout);
112 
113 		for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
114 			touch_nmi_watchdog();
115 			if (i >= i_next) {
116 				i += panic_blink(state ^= 1);
117 				i_next = i + 3600 / PANIC_BLINK_SPD;
118 			}
119 			mdelay(PANIC_TIMER_STEP);
120 		}
121 		/*
122 		 * This will not be a clean reboot, with everything
123 		 * shutting down.  But if there is a chance of
124 		 * rebooting the system it will be rebooted.
125 		 */
126 		emergency_restart();
127 	}
128 #ifdef __sparc__
129 	{
130 		extern int stop_a_enabled;
131 		/* Make sure the user can actually press Stop-A (L1-A) */
132 		stop_a_enabled = 1;
133 		printk(KERN_EMERG "Press Stop-A (L1-A) to return to the boot prom\n");
134 	}
135 #endif
136 #if defined(CONFIG_S390)
137 	{
138 		unsigned long caller;
139 
140 		caller = (unsigned long)__builtin_return_address(0);
141 		disabled_wait(caller);
142 	}
143 #endif
144 	local_irq_enable();
145 	for (i = 0; ; i += PANIC_TIMER_STEP) {
146 		touch_softlockup_watchdog();
147 		if (i >= i_next) {
148 			i += panic_blink(state ^= 1);
149 			i_next = i + 3600 / PANIC_BLINK_SPD;
150 		}
151 		mdelay(PANIC_TIMER_STEP);
152 	}
153 }
154 
155 EXPORT_SYMBOL(panic);
156 
157 
158 struct tnt {
159 	u8	bit;
160 	char	true;
161 	char	false;
162 };
163 
164 static const struct tnt tnts[] = {
165 	{ TAINT_PROPRIETARY_MODULE,	'P', 'G' },
166 	{ TAINT_FORCED_MODULE,		'F', ' ' },
167 	{ TAINT_UNSAFE_SMP,		'S', ' ' },
168 	{ TAINT_FORCED_RMMOD,		'R', ' ' },
169 	{ TAINT_MACHINE_CHECK,		'M', ' ' },
170 	{ TAINT_BAD_PAGE,		'B', ' ' },
171 	{ TAINT_USER,			'U', ' ' },
172 	{ TAINT_DIE,			'D', ' ' },
173 	{ TAINT_OVERRIDDEN_ACPI_TABLE,	'A', ' ' },
174 	{ TAINT_WARN,			'W', ' ' },
175 	{ TAINT_CRAP,			'C', ' ' },
176 	{ TAINT_FIRMWARE_WORKAROUND,	'I', ' ' },
177 };
178 
179 /**
180  *	print_tainted - return a string to represent the kernel taint state.
181  *
182  *  'P' - Proprietary module has been loaded.
183  *  'F' - Module has been forcibly loaded.
184  *  'S' - SMP with CPUs not designed for SMP.
185  *  'R' - User forced a module unload.
186  *  'M' - System experienced a machine check exception.
187  *  'B' - System has hit bad_page.
188  *  'U' - Userspace-defined naughtiness.
189  *  'D' - Kernel has oopsed before
190  *  'A' - ACPI table overridden.
191  *  'W' - Taint on warning.
192  *  'C' - modules from drivers/staging are loaded.
193  *  'I' - Working around severe firmware bug.
194  *
195  *	The string is overwritten by the next call to print_tainted().
196  */
197 const char *print_tainted(void)
198 {
199 	static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
200 
201 	if (tainted_mask) {
202 		char *s;
203 		int i;
204 
205 		s = buf + sprintf(buf, "Tainted: ");
206 		for (i = 0; i < ARRAY_SIZE(tnts); i++) {
207 			const struct tnt *t = &tnts[i];
208 			*s++ = test_bit(t->bit, &tainted_mask) ?
209 					t->true : t->false;
210 		}
211 		*s = 0;
212 	} else
213 		snprintf(buf, sizeof(buf), "Not tainted");
214 
215 	return buf;
216 }
217 
218 int test_taint(unsigned flag)
219 {
220 	return test_bit(flag, &tainted_mask);
221 }
222 EXPORT_SYMBOL(test_taint);
223 
224 unsigned long get_taint(void)
225 {
226 	return tainted_mask;
227 }
228 
229 void add_taint(unsigned flag)
230 {
231 	/*
232 	 * Can't trust the integrity of the kernel anymore.
233 	 * We don't call directly debug_locks_off() because the issue
234 	 * is not necessarily serious enough to set oops_in_progress to 1
235 	 * Also we want to keep up lockdep for staging development and
236 	 * post-warning case.
237 	 */
238 	if (flag != TAINT_CRAP && flag != TAINT_WARN && __debug_locks_off())
239 		printk(KERN_WARNING "Disabling lock debugging due to kernel taint\n");
240 
241 	set_bit(flag, &tainted_mask);
242 }
243 EXPORT_SYMBOL(add_taint);
244 
245 static void spin_msec(int msecs)
246 {
247 	int i;
248 
249 	for (i = 0; i < msecs; i++) {
250 		touch_nmi_watchdog();
251 		mdelay(1);
252 	}
253 }
254 
255 /*
256  * It just happens that oops_enter() and oops_exit() are identically
257  * implemented...
258  */
259 static void do_oops_enter_exit(void)
260 {
261 	unsigned long flags;
262 	static int spin_counter;
263 
264 	if (!pause_on_oops)
265 		return;
266 
267 	spin_lock_irqsave(&pause_on_oops_lock, flags);
268 	if (pause_on_oops_flag == 0) {
269 		/* This CPU may now print the oops message */
270 		pause_on_oops_flag = 1;
271 	} else {
272 		/* We need to stall this CPU */
273 		if (!spin_counter) {
274 			/* This CPU gets to do the counting */
275 			spin_counter = pause_on_oops;
276 			do {
277 				spin_unlock(&pause_on_oops_lock);
278 				spin_msec(MSEC_PER_SEC);
279 				spin_lock(&pause_on_oops_lock);
280 			} while (--spin_counter);
281 			pause_on_oops_flag = 0;
282 		} else {
283 			/* This CPU waits for a different one */
284 			while (spin_counter) {
285 				spin_unlock(&pause_on_oops_lock);
286 				spin_msec(1);
287 				spin_lock(&pause_on_oops_lock);
288 			}
289 		}
290 	}
291 	spin_unlock_irqrestore(&pause_on_oops_lock, flags);
292 }
293 
294 /*
295  * Return true if the calling CPU is allowed to print oops-related info.
296  * This is a bit racy..
297  */
298 int oops_may_print(void)
299 {
300 	return pause_on_oops_flag == 0;
301 }
302 
303 /*
304  * Called when the architecture enters its oops handler, before it prints
305  * anything.  If this is the first CPU to oops, and it's oopsing the first
306  * time then let it proceed.
307  *
308  * This is all enabled by the pause_on_oops kernel boot option.  We do all
309  * this to ensure that oopses don't scroll off the screen.  It has the
310  * side-effect of preventing later-oopsing CPUs from mucking up the display,
311  * too.
312  *
313  * It turns out that the CPU which is allowed to print ends up pausing for
314  * the right duration, whereas all the other CPUs pause for twice as long:
315  * once in oops_enter(), once in oops_exit().
316  */
317 void oops_enter(void)
318 {
319 	tracing_off();
320 	/* can't trust the integrity of the kernel anymore: */
321 	debug_locks_off();
322 	do_oops_enter_exit();
323 }
324 
325 /*
326  * 64-bit random ID for oopses:
327  */
328 static u64 oops_id;
329 
330 static int init_oops_id(void)
331 {
332 	if (!oops_id)
333 		get_random_bytes(&oops_id, sizeof(oops_id));
334 	else
335 		oops_id++;
336 
337 	return 0;
338 }
339 late_initcall(init_oops_id);
340 
341 void print_oops_end_marker(void)
342 {
343 	init_oops_id();
344 	printk(KERN_WARNING "---[ end trace %016llx ]---\n",
345 		(unsigned long long)oops_id);
346 }
347 
348 /*
349  * Called when the architecture exits its oops handler, after printing
350  * everything.
351  */
352 void oops_exit(void)
353 {
354 	do_oops_enter_exit();
355 	print_oops_end_marker();
356 	kmsg_dump(KMSG_DUMP_OOPS);
357 }
358 
359 #ifdef WANT_WARN_ON_SLOWPATH
360 struct slowpath_args {
361 	const char *fmt;
362 	va_list args;
363 };
364 
365 static void warn_slowpath_common(const char *file, int line, void *caller,
366 				 unsigned taint, struct slowpath_args *args)
367 {
368 	const char *board;
369 
370 	printk(KERN_WARNING "------------[ cut here ]------------\n");
371 	printk(KERN_WARNING "WARNING: at %s:%d %pS()\n", file, line, caller);
372 	board = dmi_get_system_info(DMI_PRODUCT_NAME);
373 	if (board)
374 		printk(KERN_WARNING "Hardware name: %s\n", board);
375 
376 	if (args)
377 		vprintk(args->fmt, args->args);
378 
379 	print_modules();
380 	dump_stack();
381 	print_oops_end_marker();
382 	add_taint(taint);
383 }
384 
385 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
386 {
387 	struct slowpath_args args;
388 
389 	args.fmt = fmt;
390 	va_start(args.args, fmt);
391 	warn_slowpath_common(file, line, __builtin_return_address(0),
392 			     TAINT_WARN, &args);
393 	va_end(args.args);
394 }
395 EXPORT_SYMBOL(warn_slowpath_fmt);
396 
397 void warn_slowpath_fmt_taint(const char *file, int line,
398 			     unsigned taint, const char *fmt, ...)
399 {
400 	struct slowpath_args args;
401 
402 	args.fmt = fmt;
403 	va_start(args.args, fmt);
404 	warn_slowpath_common(file, line, __builtin_return_address(0),
405 			     taint, &args);
406 	va_end(args.args);
407 }
408 EXPORT_SYMBOL(warn_slowpath_fmt_taint);
409 
410 void warn_slowpath_null(const char *file, int line)
411 {
412 	warn_slowpath_common(file, line, __builtin_return_address(0),
413 			     TAINT_WARN, NULL);
414 }
415 EXPORT_SYMBOL(warn_slowpath_null);
416 #endif
417 
418 #ifdef CONFIG_CC_STACKPROTECTOR
419 
420 /*
421  * Called when gcc's -fstack-protector feature is used, and
422  * gcc detects corruption of the on-stack canary value
423  */
424 void __stack_chk_fail(void)
425 {
426 	panic("stack-protector: Kernel stack is corrupted in: %p\n",
427 		__builtin_return_address(0));
428 }
429 EXPORT_SYMBOL(__stack_chk_fail);
430 
431 #endif
432 
433 core_param(panic, panic_timeout, int, 0644);
434 core_param(pause_on_oops, pause_on_oops, int, 0644);
435