xref: /openbmc/linux/kernel/panic.c (revision 8ac727c1)
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 		 * This will not be a clean reboot, with everything
124 		 * shutting down.  But if there is a chance of
125 		 * rebooting the system it will be rebooted.
126 		 */
127 		emergency_restart();
128 	}
129 #ifdef __sparc__
130 	{
131 		extern int stop_a_enabled;
132 		/* Make sure the user can actually press Stop-A (L1-A) */
133 		stop_a_enabled = 1;
134 		printk(KERN_EMERG "Press Stop-A (L1-A) to return to the boot prom\n");
135 	}
136 #endif
137 #if defined(CONFIG_S390)
138 	{
139 		unsigned long caller;
140 
141 		caller = (unsigned long)__builtin_return_address(0);
142 		disabled_wait(caller);
143 	}
144 #endif
145 	local_irq_enable();
146 	for (i = 0; ; i += PANIC_TIMER_STEP) {
147 		touch_softlockup_watchdog();
148 		if (i >= i_next) {
149 			i += panic_blink(state ^= 1);
150 			i_next = i + 3600 / PANIC_BLINK_SPD;
151 		}
152 		mdelay(PANIC_TIMER_STEP);
153 	}
154 }
155 
156 EXPORT_SYMBOL(panic);
157 
158 
159 struct tnt {
160 	u8	bit;
161 	char	true;
162 	char	false;
163 };
164 
165 static const struct tnt tnts[] = {
166 	{ TAINT_PROPRIETARY_MODULE,	'P', 'G' },
167 	{ TAINT_FORCED_MODULE,		'F', ' ' },
168 	{ TAINT_UNSAFE_SMP,		'S', ' ' },
169 	{ TAINT_FORCED_RMMOD,		'R', ' ' },
170 	{ TAINT_MACHINE_CHECK,		'M', ' ' },
171 	{ TAINT_BAD_PAGE,		'B', ' ' },
172 	{ TAINT_USER,			'U', ' ' },
173 	{ TAINT_DIE,			'D', ' ' },
174 	{ TAINT_OVERRIDDEN_ACPI_TABLE,	'A', ' ' },
175 	{ TAINT_WARN,			'W', ' ' },
176 	{ TAINT_CRAP,			'C', ' ' },
177 	{ TAINT_FIRMWARE_WORKAROUND,	'I', ' ' },
178 };
179 
180 /**
181  *	print_tainted - return a string to represent the kernel taint state.
182  *
183  *  'P' - Proprietary module has been loaded.
184  *  'F' - Module has been forcibly loaded.
185  *  'S' - SMP with CPUs not designed for SMP.
186  *  'R' - User forced a module unload.
187  *  'M' - System experienced a machine check exception.
188  *  'B' - System has hit bad_page.
189  *  'U' - Userspace-defined naughtiness.
190  *  'D' - Kernel has oopsed before
191  *  'A' - ACPI table overridden.
192  *  'W' - Taint on warning.
193  *  'C' - modules from drivers/staging are loaded.
194  *  'I' - Working around severe firmware bug.
195  *
196  *	The string is overwritten by the next call to print_tainted().
197  */
198 const char *print_tainted(void)
199 {
200 	static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ") + 1];
201 
202 	if (tainted_mask) {
203 		char *s;
204 		int i;
205 
206 		s = buf + sprintf(buf, "Tainted: ");
207 		for (i = 0; i < ARRAY_SIZE(tnts); i++) {
208 			const struct tnt *t = &tnts[i];
209 			*s++ = test_bit(t->bit, &tainted_mask) ?
210 					t->true : t->false;
211 		}
212 		*s = 0;
213 	} else
214 		snprintf(buf, sizeof(buf), "Not tainted");
215 
216 	return buf;
217 }
218 
219 int test_taint(unsigned flag)
220 {
221 	return test_bit(flag, &tainted_mask);
222 }
223 EXPORT_SYMBOL(test_taint);
224 
225 unsigned long get_taint(void)
226 {
227 	return tainted_mask;
228 }
229 
230 void add_taint(unsigned flag)
231 {
232 	/*
233 	 * Can't trust the integrity of the kernel anymore.
234 	 * We don't call directly debug_locks_off() because the issue
235 	 * is not necessarily serious enough to set oops_in_progress to 1
236 	 * Also we want to keep up lockdep for staging development and
237 	 * post-warning case.
238 	 */
239 	if (flag != TAINT_CRAP && flag != TAINT_WARN && __debug_locks_off())
240 		printk(KERN_WARNING "Disabling lock debugging due to kernel taint\n");
241 
242 	set_bit(flag, &tainted_mask);
243 }
244 EXPORT_SYMBOL(add_taint);
245 
246 static void spin_msec(int msecs)
247 {
248 	int i;
249 
250 	for (i = 0; i < msecs; i++) {
251 		touch_nmi_watchdog();
252 		mdelay(1);
253 	}
254 }
255 
256 /*
257  * It just happens that oops_enter() and oops_exit() are identically
258  * implemented...
259  */
260 static void do_oops_enter_exit(void)
261 {
262 	unsigned long flags;
263 	static int spin_counter;
264 
265 	if (!pause_on_oops)
266 		return;
267 
268 	spin_lock_irqsave(&pause_on_oops_lock, flags);
269 	if (pause_on_oops_flag == 0) {
270 		/* This CPU may now print the oops message */
271 		pause_on_oops_flag = 1;
272 	} else {
273 		/* We need to stall this CPU */
274 		if (!spin_counter) {
275 			/* This CPU gets to do the counting */
276 			spin_counter = pause_on_oops;
277 			do {
278 				spin_unlock(&pause_on_oops_lock);
279 				spin_msec(MSEC_PER_SEC);
280 				spin_lock(&pause_on_oops_lock);
281 			} while (--spin_counter);
282 			pause_on_oops_flag = 0;
283 		} else {
284 			/* This CPU waits for a different one */
285 			while (spin_counter) {
286 				spin_unlock(&pause_on_oops_lock);
287 				spin_msec(1);
288 				spin_lock(&pause_on_oops_lock);
289 			}
290 		}
291 	}
292 	spin_unlock_irqrestore(&pause_on_oops_lock, flags);
293 }
294 
295 /*
296  * Return true if the calling CPU is allowed to print oops-related info.
297  * This is a bit racy..
298  */
299 int oops_may_print(void)
300 {
301 	return pause_on_oops_flag == 0;
302 }
303 
304 /*
305  * Called when the architecture enters its oops handler, before it prints
306  * anything.  If this is the first CPU to oops, and it's oopsing the first
307  * time then let it proceed.
308  *
309  * This is all enabled by the pause_on_oops kernel boot option.  We do all
310  * this to ensure that oopses don't scroll off the screen.  It has the
311  * side-effect of preventing later-oopsing CPUs from mucking up the display,
312  * too.
313  *
314  * It turns out that the CPU which is allowed to print ends up pausing for
315  * the right duration, whereas all the other CPUs pause for twice as long:
316  * once in oops_enter(), once in oops_exit().
317  */
318 void oops_enter(void)
319 {
320 	tracing_off();
321 	/* can't trust the integrity of the kernel anymore: */
322 	debug_locks_off();
323 	do_oops_enter_exit();
324 }
325 
326 /*
327  * 64-bit random ID for oopses:
328  */
329 static u64 oops_id;
330 
331 static int init_oops_id(void)
332 {
333 	if (!oops_id)
334 		get_random_bytes(&oops_id, sizeof(oops_id));
335 	else
336 		oops_id++;
337 
338 	return 0;
339 }
340 late_initcall(init_oops_id);
341 
342 void print_oops_end_marker(void)
343 {
344 	init_oops_id();
345 	printk(KERN_WARNING "---[ end trace %016llx ]---\n",
346 		(unsigned long long)oops_id);
347 }
348 
349 /*
350  * Called when the architecture exits its oops handler, after printing
351  * everything.
352  */
353 void oops_exit(void)
354 {
355 	do_oops_enter_exit();
356 	print_oops_end_marker();
357 	kmsg_dump(KMSG_DUMP_OOPS);
358 }
359 
360 #ifdef WANT_WARN_ON_SLOWPATH
361 struct slowpath_args {
362 	const char *fmt;
363 	va_list args;
364 };
365 
366 static void warn_slowpath_common(const char *file, int line, void *caller,
367 				 unsigned taint, struct slowpath_args *args)
368 {
369 	const char *board;
370 
371 	printk(KERN_WARNING "------------[ cut here ]------------\n");
372 	printk(KERN_WARNING "WARNING: at %s:%d %pS()\n", file, line, caller);
373 	board = dmi_get_system_info(DMI_PRODUCT_NAME);
374 	if (board)
375 		printk(KERN_WARNING "Hardware name: %s\n", board);
376 
377 	if (args)
378 		vprintk(args->fmt, args->args);
379 
380 	print_modules();
381 	dump_stack();
382 	print_oops_end_marker();
383 	add_taint(taint);
384 }
385 
386 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
387 {
388 	struct slowpath_args args;
389 
390 	args.fmt = fmt;
391 	va_start(args.args, fmt);
392 	warn_slowpath_common(file, line, __builtin_return_address(0),
393 			     TAINT_WARN, &args);
394 	va_end(args.args);
395 }
396 EXPORT_SYMBOL(warn_slowpath_fmt);
397 
398 void warn_slowpath_fmt_taint(const char *file, int line,
399 			     unsigned taint, const char *fmt, ...)
400 {
401 	struct slowpath_args args;
402 
403 	args.fmt = fmt;
404 	va_start(args.args, fmt);
405 	warn_slowpath_common(file, line, __builtin_return_address(0),
406 			     taint, &args);
407 	va_end(args.args);
408 }
409 EXPORT_SYMBOL(warn_slowpath_fmt_taint);
410 
411 void warn_slowpath_null(const char *file, int line)
412 {
413 	warn_slowpath_common(file, line, __builtin_return_address(0),
414 			     TAINT_WARN, NULL);
415 }
416 EXPORT_SYMBOL(warn_slowpath_null);
417 #endif
418 
419 #ifdef CONFIG_CC_STACKPROTECTOR
420 
421 /*
422  * Called when gcc's -fstack-protector feature is used, and
423  * gcc detects corruption of the on-stack canary value
424  */
425 void __stack_chk_fail(void)
426 {
427 	panic("stack-protector: Kernel stack is corrupted in: %p\n",
428 		__builtin_return_address(0));
429 }
430 EXPORT_SYMBOL(__stack_chk_fail);
431 
432 #endif
433 
434 core_param(panic, panic_timeout, int, 0644);
435 core_param(pause_on_oops, pause_on_oops, int, 0644);
436