xref: /openbmc/linux/net/rfkill/core.c (revision 09a4f6f5)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * Copyright (C) 2006 - 2007 Ivo van Doorn
4  * Copyright (C) 2007 Dmitry Torokhov
5  * Copyright 2009 Johannes Berg <johannes@sipsolutions.net>
6  */
7 
8 #include <linux/kernel.h>
9 #include <linux/module.h>
10 #include <linux/init.h>
11 #include <linux/workqueue.h>
12 #include <linux/capability.h>
13 #include <linux/list.h>
14 #include <linux/mutex.h>
15 #include <linux/rfkill.h>
16 #include <linux/sched.h>
17 #include <linux/spinlock.h>
18 #include <linux/device.h>
19 #include <linux/miscdevice.h>
20 #include <linux/wait.h>
21 #include <linux/poll.h>
22 #include <linux/fs.h>
23 #include <linux/slab.h>
24 
25 #include "rfkill.h"
26 
27 #define POLL_INTERVAL		(5 * HZ)
28 
29 #define RFKILL_BLOCK_HW		BIT(0)
30 #define RFKILL_BLOCK_SW		BIT(1)
31 #define RFKILL_BLOCK_SW_PREV	BIT(2)
32 #define RFKILL_BLOCK_ANY	(RFKILL_BLOCK_HW |\
33 				 RFKILL_BLOCK_SW |\
34 				 RFKILL_BLOCK_SW_PREV)
35 #define RFKILL_BLOCK_SW_SETCALL	BIT(31)
36 
37 struct rfkill {
38 	spinlock_t		lock;
39 
40 	enum rfkill_type	type;
41 
42 	unsigned long		state;
43 	unsigned long		hard_block_reasons;
44 
45 	u32			idx;
46 
47 	bool			registered;
48 	bool			persistent;
49 	bool			polling_paused;
50 	bool			suspended;
51 
52 	const struct rfkill_ops	*ops;
53 	void			*data;
54 
55 #ifdef CONFIG_RFKILL_LEDS
56 	struct led_trigger	led_trigger;
57 	const char		*ledtrigname;
58 #endif
59 
60 	struct device		dev;
61 	struct list_head	node;
62 
63 	struct delayed_work	poll_work;
64 	struct work_struct	uevent_work;
65 	struct work_struct	sync_work;
66 	char			name[];
67 };
68 #define to_rfkill(d)	container_of(d, struct rfkill, dev)
69 
70 struct rfkill_int_event {
71 	struct list_head	list;
72 	struct rfkill_event	ev;
73 };
74 
75 struct rfkill_data {
76 	struct list_head	list;
77 	struct list_head	events;
78 	struct mutex		mtx;
79 	wait_queue_head_t	read_wait;
80 	bool			input_handler;
81 };
82 
83 
84 MODULE_AUTHOR("Ivo van Doorn <IvDoorn@gmail.com>");
85 MODULE_AUTHOR("Johannes Berg <johannes@sipsolutions.net>");
86 MODULE_DESCRIPTION("RF switch support");
87 MODULE_LICENSE("GPL");
88 
89 
90 /*
91  * The locking here should be made much smarter, we currently have
92  * a bit of a stupid situation because drivers might want to register
93  * the rfkill struct under their own lock, and take this lock during
94  * rfkill method calls -- which will cause an AB-BA deadlock situation.
95  *
96  * To fix that, we need to rework this code here to be mostly lock-free
97  * and only use the mutex for list manipulations, not to protect the
98  * various other global variables. Then we can avoid holding the mutex
99  * around driver operations, and all is happy.
100  */
101 static LIST_HEAD(rfkill_list);	/* list of registered rf switches */
102 static DEFINE_MUTEX(rfkill_global_mutex);
103 static LIST_HEAD(rfkill_fds);	/* list of open fds of /dev/rfkill */
104 
105 static unsigned int rfkill_default_state = 1;
106 module_param_named(default_state, rfkill_default_state, uint, 0444);
107 MODULE_PARM_DESC(default_state,
108 		 "Default initial state for all radio types, 0 = radio off");
109 
110 static struct {
111 	bool cur, sav;
112 } rfkill_global_states[NUM_RFKILL_TYPES];
113 
114 static bool rfkill_epo_lock_active;
115 
116 
117 #ifdef CONFIG_RFKILL_LEDS
118 static void rfkill_led_trigger_event(struct rfkill *rfkill)
119 {
120 	struct led_trigger *trigger;
121 
122 	if (!rfkill->registered)
123 		return;
124 
125 	trigger = &rfkill->led_trigger;
126 
127 	if (rfkill->state & RFKILL_BLOCK_ANY)
128 		led_trigger_event(trigger, LED_OFF);
129 	else
130 		led_trigger_event(trigger, LED_FULL);
131 }
132 
133 static int rfkill_led_trigger_activate(struct led_classdev *led)
134 {
135 	struct rfkill *rfkill;
136 
137 	rfkill = container_of(led->trigger, struct rfkill, led_trigger);
138 
139 	rfkill_led_trigger_event(rfkill);
140 
141 	return 0;
142 }
143 
144 const char *rfkill_get_led_trigger_name(struct rfkill *rfkill)
145 {
146 	return rfkill->led_trigger.name;
147 }
148 EXPORT_SYMBOL(rfkill_get_led_trigger_name);
149 
150 void rfkill_set_led_trigger_name(struct rfkill *rfkill, const char *name)
151 {
152 	BUG_ON(!rfkill);
153 
154 	rfkill->ledtrigname = name;
155 }
156 EXPORT_SYMBOL(rfkill_set_led_trigger_name);
157 
158 static int rfkill_led_trigger_register(struct rfkill *rfkill)
159 {
160 	rfkill->led_trigger.name = rfkill->ledtrigname
161 					? : dev_name(&rfkill->dev);
162 	rfkill->led_trigger.activate = rfkill_led_trigger_activate;
163 	return led_trigger_register(&rfkill->led_trigger);
164 }
165 
166 static void rfkill_led_trigger_unregister(struct rfkill *rfkill)
167 {
168 	led_trigger_unregister(&rfkill->led_trigger);
169 }
170 
171 static struct led_trigger rfkill_any_led_trigger;
172 static struct led_trigger rfkill_none_led_trigger;
173 static struct work_struct rfkill_global_led_trigger_work;
174 
175 static void rfkill_global_led_trigger_worker(struct work_struct *work)
176 {
177 	enum led_brightness brightness = LED_OFF;
178 	struct rfkill *rfkill;
179 
180 	mutex_lock(&rfkill_global_mutex);
181 	list_for_each_entry(rfkill, &rfkill_list, node) {
182 		if (!(rfkill->state & RFKILL_BLOCK_ANY)) {
183 			brightness = LED_FULL;
184 			break;
185 		}
186 	}
187 	mutex_unlock(&rfkill_global_mutex);
188 
189 	led_trigger_event(&rfkill_any_led_trigger, brightness);
190 	led_trigger_event(&rfkill_none_led_trigger,
191 			  brightness == LED_OFF ? LED_FULL : LED_OFF);
192 }
193 
194 static void rfkill_global_led_trigger_event(void)
195 {
196 	schedule_work(&rfkill_global_led_trigger_work);
197 }
198 
199 static int rfkill_global_led_trigger_register(void)
200 {
201 	int ret;
202 
203 	INIT_WORK(&rfkill_global_led_trigger_work,
204 			rfkill_global_led_trigger_worker);
205 
206 	rfkill_any_led_trigger.name = "rfkill-any";
207 	ret = led_trigger_register(&rfkill_any_led_trigger);
208 	if (ret)
209 		return ret;
210 
211 	rfkill_none_led_trigger.name = "rfkill-none";
212 	ret = led_trigger_register(&rfkill_none_led_trigger);
213 	if (ret)
214 		led_trigger_unregister(&rfkill_any_led_trigger);
215 	else
216 		/* Delay activation until all global triggers are registered */
217 		rfkill_global_led_trigger_event();
218 
219 	return ret;
220 }
221 
222 static void rfkill_global_led_trigger_unregister(void)
223 {
224 	led_trigger_unregister(&rfkill_none_led_trigger);
225 	led_trigger_unregister(&rfkill_any_led_trigger);
226 	cancel_work_sync(&rfkill_global_led_trigger_work);
227 }
228 #else
229 static void rfkill_led_trigger_event(struct rfkill *rfkill)
230 {
231 }
232 
233 static inline int rfkill_led_trigger_register(struct rfkill *rfkill)
234 {
235 	return 0;
236 }
237 
238 static inline void rfkill_led_trigger_unregister(struct rfkill *rfkill)
239 {
240 }
241 
242 static void rfkill_global_led_trigger_event(void)
243 {
244 }
245 
246 static int rfkill_global_led_trigger_register(void)
247 {
248 	return 0;
249 }
250 
251 static void rfkill_global_led_trigger_unregister(void)
252 {
253 }
254 #endif /* CONFIG_RFKILL_LEDS */
255 
256 static void rfkill_fill_event(struct rfkill_event *ev, struct rfkill *rfkill,
257 			      enum rfkill_operation op)
258 {
259 	unsigned long flags;
260 
261 	ev->idx = rfkill->idx;
262 	ev->type = rfkill->type;
263 	ev->op = op;
264 
265 	spin_lock_irqsave(&rfkill->lock, flags);
266 	ev->hard = !!(rfkill->state & RFKILL_BLOCK_HW);
267 	ev->soft = !!(rfkill->state & (RFKILL_BLOCK_SW |
268 					RFKILL_BLOCK_SW_PREV));
269 	ev->hard_block_reasons = rfkill->hard_block_reasons;
270 	spin_unlock_irqrestore(&rfkill->lock, flags);
271 }
272 
273 static void rfkill_send_events(struct rfkill *rfkill, enum rfkill_operation op)
274 {
275 	struct rfkill_data *data;
276 	struct rfkill_int_event *ev;
277 
278 	list_for_each_entry(data, &rfkill_fds, list) {
279 		ev = kzalloc(sizeof(*ev), GFP_KERNEL);
280 		if (!ev)
281 			continue;
282 		rfkill_fill_event(&ev->ev, rfkill, op);
283 		mutex_lock(&data->mtx);
284 		list_add_tail(&ev->list, &data->events);
285 		mutex_unlock(&data->mtx);
286 		wake_up_interruptible(&data->read_wait);
287 	}
288 }
289 
290 static void rfkill_event(struct rfkill *rfkill)
291 {
292 	if (!rfkill->registered)
293 		return;
294 
295 	kobject_uevent(&rfkill->dev.kobj, KOBJ_CHANGE);
296 
297 	/* also send event to /dev/rfkill */
298 	rfkill_send_events(rfkill, RFKILL_OP_CHANGE);
299 }
300 
301 /**
302  * rfkill_set_block - wrapper for set_block method
303  *
304  * @rfkill: the rfkill struct to use
305  * @blocked: the new software state
306  *
307  * Calls the set_block method (when applicable) and handles notifications
308  * etc. as well.
309  */
310 static void rfkill_set_block(struct rfkill *rfkill, bool blocked)
311 {
312 	unsigned long flags;
313 	bool prev, curr;
314 	int err;
315 
316 	if (unlikely(rfkill->dev.power.power_state.event & PM_EVENT_SLEEP))
317 		return;
318 
319 	/*
320 	 * Some platforms (...!) generate input events which affect the
321 	 * _hard_ kill state -- whenever something tries to change the
322 	 * current software state query the hardware state too.
323 	 */
324 	if (rfkill->ops->query)
325 		rfkill->ops->query(rfkill, rfkill->data);
326 
327 	spin_lock_irqsave(&rfkill->lock, flags);
328 	prev = rfkill->state & RFKILL_BLOCK_SW;
329 
330 	if (prev)
331 		rfkill->state |= RFKILL_BLOCK_SW_PREV;
332 	else
333 		rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
334 
335 	if (blocked)
336 		rfkill->state |= RFKILL_BLOCK_SW;
337 	else
338 		rfkill->state &= ~RFKILL_BLOCK_SW;
339 
340 	rfkill->state |= RFKILL_BLOCK_SW_SETCALL;
341 	spin_unlock_irqrestore(&rfkill->lock, flags);
342 
343 	err = rfkill->ops->set_block(rfkill->data, blocked);
344 
345 	spin_lock_irqsave(&rfkill->lock, flags);
346 	if (err) {
347 		/*
348 		 * Failed -- reset status to _PREV, which may be different
349 		 * from what we have set _PREV to earlier in this function
350 		 * if rfkill_set_sw_state was invoked.
351 		 */
352 		if (rfkill->state & RFKILL_BLOCK_SW_PREV)
353 			rfkill->state |= RFKILL_BLOCK_SW;
354 		else
355 			rfkill->state &= ~RFKILL_BLOCK_SW;
356 	}
357 	rfkill->state &= ~RFKILL_BLOCK_SW_SETCALL;
358 	rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
359 	curr = rfkill->state & RFKILL_BLOCK_SW;
360 	spin_unlock_irqrestore(&rfkill->lock, flags);
361 
362 	rfkill_led_trigger_event(rfkill);
363 	rfkill_global_led_trigger_event();
364 
365 	if (prev != curr)
366 		rfkill_event(rfkill);
367 }
368 
369 static void rfkill_update_global_state(enum rfkill_type type, bool blocked)
370 {
371 	int i;
372 
373 	if (type != RFKILL_TYPE_ALL) {
374 		rfkill_global_states[type].cur = blocked;
375 		return;
376 	}
377 
378 	for (i = 0; i < NUM_RFKILL_TYPES; i++)
379 		rfkill_global_states[i].cur = blocked;
380 }
381 
382 #ifdef CONFIG_RFKILL_INPUT
383 static atomic_t rfkill_input_disabled = ATOMIC_INIT(0);
384 
385 /**
386  * __rfkill_switch_all - Toggle state of all switches of given type
387  * @type: type of interfaces to be affected
388  * @blocked: the new state
389  *
390  * This function sets the state of all switches of given type,
391  * unless a specific switch is suspended.
392  *
393  * Caller must have acquired rfkill_global_mutex.
394  */
395 static void __rfkill_switch_all(const enum rfkill_type type, bool blocked)
396 {
397 	struct rfkill *rfkill;
398 
399 	rfkill_update_global_state(type, blocked);
400 	list_for_each_entry(rfkill, &rfkill_list, node) {
401 		if (rfkill->type != type && type != RFKILL_TYPE_ALL)
402 			continue;
403 
404 		rfkill_set_block(rfkill, blocked);
405 	}
406 }
407 
408 /**
409  * rfkill_switch_all - Toggle state of all switches of given type
410  * @type: type of interfaces to be affected
411  * @blocked: the new state
412  *
413  * Acquires rfkill_global_mutex and calls __rfkill_switch_all(@type, @state).
414  * Please refer to __rfkill_switch_all() for details.
415  *
416  * Does nothing if the EPO lock is active.
417  */
418 void rfkill_switch_all(enum rfkill_type type, bool blocked)
419 {
420 	if (atomic_read(&rfkill_input_disabled))
421 		return;
422 
423 	mutex_lock(&rfkill_global_mutex);
424 
425 	if (!rfkill_epo_lock_active)
426 		__rfkill_switch_all(type, blocked);
427 
428 	mutex_unlock(&rfkill_global_mutex);
429 }
430 
431 /**
432  * rfkill_epo - emergency power off all transmitters
433  *
434  * This kicks all non-suspended rfkill devices to RFKILL_STATE_SOFT_BLOCKED,
435  * ignoring everything in its path but rfkill_global_mutex and rfkill->mutex.
436  *
437  * The global state before the EPO is saved and can be restored later
438  * using rfkill_restore_states().
439  */
440 void rfkill_epo(void)
441 {
442 	struct rfkill *rfkill;
443 	int i;
444 
445 	if (atomic_read(&rfkill_input_disabled))
446 		return;
447 
448 	mutex_lock(&rfkill_global_mutex);
449 
450 	rfkill_epo_lock_active = true;
451 	list_for_each_entry(rfkill, &rfkill_list, node)
452 		rfkill_set_block(rfkill, true);
453 
454 	for (i = 0; i < NUM_RFKILL_TYPES; i++) {
455 		rfkill_global_states[i].sav = rfkill_global_states[i].cur;
456 		rfkill_global_states[i].cur = true;
457 	}
458 
459 	mutex_unlock(&rfkill_global_mutex);
460 }
461 
462 /**
463  * rfkill_restore_states - restore global states
464  *
465  * Restore (and sync switches to) the global state from the
466  * states in rfkill_default_states.  This can undo the effects of
467  * a call to rfkill_epo().
468  */
469 void rfkill_restore_states(void)
470 {
471 	int i;
472 
473 	if (atomic_read(&rfkill_input_disabled))
474 		return;
475 
476 	mutex_lock(&rfkill_global_mutex);
477 
478 	rfkill_epo_lock_active = false;
479 	for (i = 0; i < NUM_RFKILL_TYPES; i++)
480 		__rfkill_switch_all(i, rfkill_global_states[i].sav);
481 	mutex_unlock(&rfkill_global_mutex);
482 }
483 
484 /**
485  * rfkill_remove_epo_lock - unlock state changes
486  *
487  * Used by rfkill-input manually unlock state changes, when
488  * the EPO switch is deactivated.
489  */
490 void rfkill_remove_epo_lock(void)
491 {
492 	if (atomic_read(&rfkill_input_disabled))
493 		return;
494 
495 	mutex_lock(&rfkill_global_mutex);
496 	rfkill_epo_lock_active = false;
497 	mutex_unlock(&rfkill_global_mutex);
498 }
499 
500 /**
501  * rfkill_is_epo_lock_active - returns true EPO is active
502  *
503  * Returns 0 (false) if there is NOT an active EPO condition,
504  * and 1 (true) if there is an active EPO condition, which
505  * locks all radios in one of the BLOCKED states.
506  *
507  * Can be called in atomic context.
508  */
509 bool rfkill_is_epo_lock_active(void)
510 {
511 	return rfkill_epo_lock_active;
512 }
513 
514 /**
515  * rfkill_get_global_sw_state - returns global state for a type
516  * @type: the type to get the global state of
517  *
518  * Returns the current global state for a given wireless
519  * device type.
520  */
521 bool rfkill_get_global_sw_state(const enum rfkill_type type)
522 {
523 	return rfkill_global_states[type].cur;
524 }
525 #endif
526 
527 bool rfkill_set_hw_state_reason(struct rfkill *rfkill,
528 				bool blocked, unsigned long reason)
529 {
530 	unsigned long flags;
531 	bool ret, prev;
532 
533 	BUG_ON(!rfkill);
534 
535 	if (WARN(reason &
536 	    ~(RFKILL_HARD_BLOCK_SIGNAL | RFKILL_HARD_BLOCK_NOT_OWNER),
537 	    "hw_state reason not supported: 0x%lx", reason))
538 		return blocked;
539 
540 	spin_lock_irqsave(&rfkill->lock, flags);
541 	prev = !!(rfkill->hard_block_reasons & reason);
542 	if (blocked) {
543 		rfkill->state |= RFKILL_BLOCK_HW;
544 		rfkill->hard_block_reasons |= reason;
545 	} else {
546 		rfkill->hard_block_reasons &= ~reason;
547 		if (!rfkill->hard_block_reasons)
548 			rfkill->state &= ~RFKILL_BLOCK_HW;
549 	}
550 	ret = !!(rfkill->state & RFKILL_BLOCK_ANY);
551 	spin_unlock_irqrestore(&rfkill->lock, flags);
552 
553 	rfkill_led_trigger_event(rfkill);
554 	rfkill_global_led_trigger_event();
555 
556 	if (rfkill->registered && prev != blocked)
557 		schedule_work(&rfkill->uevent_work);
558 
559 	return ret;
560 }
561 EXPORT_SYMBOL(rfkill_set_hw_state_reason);
562 
563 static void __rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
564 {
565 	u32 bit = RFKILL_BLOCK_SW;
566 
567 	/* if in a ops->set_block right now, use other bit */
568 	if (rfkill->state & RFKILL_BLOCK_SW_SETCALL)
569 		bit = RFKILL_BLOCK_SW_PREV;
570 
571 	if (blocked)
572 		rfkill->state |= bit;
573 	else
574 		rfkill->state &= ~bit;
575 }
576 
577 bool rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
578 {
579 	unsigned long flags;
580 	bool prev, hwblock;
581 
582 	BUG_ON(!rfkill);
583 
584 	spin_lock_irqsave(&rfkill->lock, flags);
585 	prev = !!(rfkill->state & RFKILL_BLOCK_SW);
586 	__rfkill_set_sw_state(rfkill, blocked);
587 	hwblock = !!(rfkill->state & RFKILL_BLOCK_HW);
588 	blocked = blocked || hwblock;
589 	spin_unlock_irqrestore(&rfkill->lock, flags);
590 
591 	if (!rfkill->registered)
592 		return blocked;
593 
594 	if (prev != blocked && !hwblock)
595 		schedule_work(&rfkill->uevent_work);
596 
597 	rfkill_led_trigger_event(rfkill);
598 	rfkill_global_led_trigger_event();
599 
600 	return blocked;
601 }
602 EXPORT_SYMBOL(rfkill_set_sw_state);
603 
604 void rfkill_init_sw_state(struct rfkill *rfkill, bool blocked)
605 {
606 	unsigned long flags;
607 
608 	BUG_ON(!rfkill);
609 	BUG_ON(rfkill->registered);
610 
611 	spin_lock_irqsave(&rfkill->lock, flags);
612 	__rfkill_set_sw_state(rfkill, blocked);
613 	rfkill->persistent = true;
614 	spin_unlock_irqrestore(&rfkill->lock, flags);
615 }
616 EXPORT_SYMBOL(rfkill_init_sw_state);
617 
618 void rfkill_set_states(struct rfkill *rfkill, bool sw, bool hw)
619 {
620 	unsigned long flags;
621 	bool swprev, hwprev;
622 
623 	BUG_ON(!rfkill);
624 
625 	spin_lock_irqsave(&rfkill->lock, flags);
626 
627 	/*
628 	 * No need to care about prev/setblock ... this is for uevent only
629 	 * and that will get triggered by rfkill_set_block anyway.
630 	 */
631 	swprev = !!(rfkill->state & RFKILL_BLOCK_SW);
632 	hwprev = !!(rfkill->state & RFKILL_BLOCK_HW);
633 	__rfkill_set_sw_state(rfkill, sw);
634 	if (hw)
635 		rfkill->state |= RFKILL_BLOCK_HW;
636 	else
637 		rfkill->state &= ~RFKILL_BLOCK_HW;
638 
639 	spin_unlock_irqrestore(&rfkill->lock, flags);
640 
641 	if (!rfkill->registered) {
642 		rfkill->persistent = true;
643 	} else {
644 		if (swprev != sw || hwprev != hw)
645 			schedule_work(&rfkill->uevent_work);
646 
647 		rfkill_led_trigger_event(rfkill);
648 		rfkill_global_led_trigger_event();
649 	}
650 }
651 EXPORT_SYMBOL(rfkill_set_states);
652 
653 static const char * const rfkill_types[] = {
654 	NULL, /* RFKILL_TYPE_ALL */
655 	"wlan",
656 	"bluetooth",
657 	"ultrawideband",
658 	"wimax",
659 	"wwan",
660 	"gps",
661 	"fm",
662 	"nfc",
663 };
664 
665 enum rfkill_type rfkill_find_type(const char *name)
666 {
667 	int i;
668 
669 	BUILD_BUG_ON(ARRAY_SIZE(rfkill_types) != NUM_RFKILL_TYPES);
670 
671 	if (!name)
672 		return RFKILL_TYPE_ALL;
673 
674 	for (i = 1; i < NUM_RFKILL_TYPES; i++)
675 		if (!strcmp(name, rfkill_types[i]))
676 			return i;
677 	return RFKILL_TYPE_ALL;
678 }
679 EXPORT_SYMBOL(rfkill_find_type);
680 
681 static ssize_t name_show(struct device *dev, struct device_attribute *attr,
682 			 char *buf)
683 {
684 	struct rfkill *rfkill = to_rfkill(dev);
685 
686 	return sprintf(buf, "%s\n", rfkill->name);
687 }
688 static DEVICE_ATTR_RO(name);
689 
690 static ssize_t type_show(struct device *dev, struct device_attribute *attr,
691 			 char *buf)
692 {
693 	struct rfkill *rfkill = to_rfkill(dev);
694 
695 	return sprintf(buf, "%s\n", rfkill_types[rfkill->type]);
696 }
697 static DEVICE_ATTR_RO(type);
698 
699 static ssize_t index_show(struct device *dev, struct device_attribute *attr,
700 			  char *buf)
701 {
702 	struct rfkill *rfkill = to_rfkill(dev);
703 
704 	return sprintf(buf, "%d\n", rfkill->idx);
705 }
706 static DEVICE_ATTR_RO(index);
707 
708 static ssize_t persistent_show(struct device *dev,
709 			       struct device_attribute *attr, char *buf)
710 {
711 	struct rfkill *rfkill = to_rfkill(dev);
712 
713 	return sprintf(buf, "%d\n", rfkill->persistent);
714 }
715 static DEVICE_ATTR_RO(persistent);
716 
717 static ssize_t hard_show(struct device *dev, struct device_attribute *attr,
718 			 char *buf)
719 {
720 	struct rfkill *rfkill = to_rfkill(dev);
721 
722 	return sprintf(buf, "%d\n", (rfkill->state & RFKILL_BLOCK_HW) ? 1 : 0 );
723 }
724 static DEVICE_ATTR_RO(hard);
725 
726 static ssize_t soft_show(struct device *dev, struct device_attribute *attr,
727 			 char *buf)
728 {
729 	struct rfkill *rfkill = to_rfkill(dev);
730 
731 	return sprintf(buf, "%d\n", (rfkill->state & RFKILL_BLOCK_SW) ? 1 : 0 );
732 }
733 
734 static ssize_t soft_store(struct device *dev, struct device_attribute *attr,
735 			  const char *buf, size_t count)
736 {
737 	struct rfkill *rfkill = to_rfkill(dev);
738 	unsigned long state;
739 	int err;
740 
741 	if (!capable(CAP_NET_ADMIN))
742 		return -EPERM;
743 
744 	err = kstrtoul(buf, 0, &state);
745 	if (err)
746 		return err;
747 
748 	if (state > 1 )
749 		return -EINVAL;
750 
751 	mutex_lock(&rfkill_global_mutex);
752 	rfkill_set_block(rfkill, state);
753 	mutex_unlock(&rfkill_global_mutex);
754 
755 	return count;
756 }
757 static DEVICE_ATTR_RW(soft);
758 
759 static ssize_t hard_block_reasons_show(struct device *dev,
760 				       struct device_attribute *attr,
761 				       char *buf)
762 {
763 	struct rfkill *rfkill = to_rfkill(dev);
764 
765 	return sprintf(buf, "0x%lx\n", rfkill->hard_block_reasons);
766 }
767 static DEVICE_ATTR_RO(hard_block_reasons);
768 
769 static u8 user_state_from_blocked(unsigned long state)
770 {
771 	if (state & RFKILL_BLOCK_HW)
772 		return RFKILL_USER_STATE_HARD_BLOCKED;
773 	if (state & RFKILL_BLOCK_SW)
774 		return RFKILL_USER_STATE_SOFT_BLOCKED;
775 
776 	return RFKILL_USER_STATE_UNBLOCKED;
777 }
778 
779 static ssize_t state_show(struct device *dev, struct device_attribute *attr,
780 			  char *buf)
781 {
782 	struct rfkill *rfkill = to_rfkill(dev);
783 
784 	return sprintf(buf, "%d\n", user_state_from_blocked(rfkill->state));
785 }
786 
787 static ssize_t state_store(struct device *dev, struct device_attribute *attr,
788 			   const char *buf, size_t count)
789 {
790 	struct rfkill *rfkill = to_rfkill(dev);
791 	unsigned long state;
792 	int err;
793 
794 	if (!capable(CAP_NET_ADMIN))
795 		return -EPERM;
796 
797 	err = kstrtoul(buf, 0, &state);
798 	if (err)
799 		return err;
800 
801 	if (state != RFKILL_USER_STATE_SOFT_BLOCKED &&
802 	    state != RFKILL_USER_STATE_UNBLOCKED)
803 		return -EINVAL;
804 
805 	mutex_lock(&rfkill_global_mutex);
806 	rfkill_set_block(rfkill, state == RFKILL_USER_STATE_SOFT_BLOCKED);
807 	mutex_unlock(&rfkill_global_mutex);
808 
809 	return count;
810 }
811 static DEVICE_ATTR_RW(state);
812 
813 static struct attribute *rfkill_dev_attrs[] = {
814 	&dev_attr_name.attr,
815 	&dev_attr_type.attr,
816 	&dev_attr_index.attr,
817 	&dev_attr_persistent.attr,
818 	&dev_attr_state.attr,
819 	&dev_attr_soft.attr,
820 	&dev_attr_hard.attr,
821 	&dev_attr_hard_block_reasons.attr,
822 	NULL,
823 };
824 ATTRIBUTE_GROUPS(rfkill_dev);
825 
826 static void rfkill_release(struct device *dev)
827 {
828 	struct rfkill *rfkill = to_rfkill(dev);
829 
830 	kfree(rfkill);
831 }
832 
833 static int rfkill_dev_uevent(struct device *dev, struct kobj_uevent_env *env)
834 {
835 	struct rfkill *rfkill = to_rfkill(dev);
836 	unsigned long flags;
837 	unsigned long reasons;
838 	u32 state;
839 	int error;
840 
841 	error = add_uevent_var(env, "RFKILL_NAME=%s", rfkill->name);
842 	if (error)
843 		return error;
844 	error = add_uevent_var(env, "RFKILL_TYPE=%s",
845 			       rfkill_types[rfkill->type]);
846 	if (error)
847 		return error;
848 	spin_lock_irqsave(&rfkill->lock, flags);
849 	state = rfkill->state;
850 	reasons = rfkill->hard_block_reasons;
851 	spin_unlock_irqrestore(&rfkill->lock, flags);
852 	error = add_uevent_var(env, "RFKILL_STATE=%d",
853 			       user_state_from_blocked(state));
854 	if (error)
855 		return error;
856 	return add_uevent_var(env, "RFKILL_HW_BLOCK_REASON=0x%lx", reasons);
857 }
858 
859 void rfkill_pause_polling(struct rfkill *rfkill)
860 {
861 	BUG_ON(!rfkill);
862 
863 	if (!rfkill->ops->poll)
864 		return;
865 
866 	rfkill->polling_paused = true;
867 	cancel_delayed_work_sync(&rfkill->poll_work);
868 }
869 EXPORT_SYMBOL(rfkill_pause_polling);
870 
871 void rfkill_resume_polling(struct rfkill *rfkill)
872 {
873 	BUG_ON(!rfkill);
874 
875 	if (!rfkill->ops->poll)
876 		return;
877 
878 	rfkill->polling_paused = false;
879 
880 	if (rfkill->suspended)
881 		return;
882 
883 	queue_delayed_work(system_power_efficient_wq,
884 			   &rfkill->poll_work, 0);
885 }
886 EXPORT_SYMBOL(rfkill_resume_polling);
887 
888 #ifdef CONFIG_PM_SLEEP
889 static int rfkill_suspend(struct device *dev)
890 {
891 	struct rfkill *rfkill = to_rfkill(dev);
892 
893 	rfkill->suspended = true;
894 	cancel_delayed_work_sync(&rfkill->poll_work);
895 
896 	return 0;
897 }
898 
899 static int rfkill_resume(struct device *dev)
900 {
901 	struct rfkill *rfkill = to_rfkill(dev);
902 	bool cur;
903 
904 	rfkill->suspended = false;
905 
906 	if (!rfkill->registered)
907 		return 0;
908 
909 	if (!rfkill->persistent) {
910 		cur = !!(rfkill->state & RFKILL_BLOCK_SW);
911 		rfkill_set_block(rfkill, cur);
912 	}
913 
914 	if (rfkill->ops->poll && !rfkill->polling_paused)
915 		queue_delayed_work(system_power_efficient_wq,
916 				   &rfkill->poll_work, 0);
917 
918 	return 0;
919 }
920 
921 static SIMPLE_DEV_PM_OPS(rfkill_pm_ops, rfkill_suspend, rfkill_resume);
922 #define RFKILL_PM_OPS (&rfkill_pm_ops)
923 #else
924 #define RFKILL_PM_OPS NULL
925 #endif
926 
927 static struct class rfkill_class = {
928 	.name		= "rfkill",
929 	.dev_release	= rfkill_release,
930 	.dev_groups	= rfkill_dev_groups,
931 	.dev_uevent	= rfkill_dev_uevent,
932 	.pm		= RFKILL_PM_OPS,
933 };
934 
935 bool rfkill_blocked(struct rfkill *rfkill)
936 {
937 	unsigned long flags;
938 	u32 state;
939 
940 	spin_lock_irqsave(&rfkill->lock, flags);
941 	state = rfkill->state;
942 	spin_unlock_irqrestore(&rfkill->lock, flags);
943 
944 	return !!(state & RFKILL_BLOCK_ANY);
945 }
946 EXPORT_SYMBOL(rfkill_blocked);
947 
948 
949 struct rfkill * __must_check rfkill_alloc(const char *name,
950 					  struct device *parent,
951 					  const enum rfkill_type type,
952 					  const struct rfkill_ops *ops,
953 					  void *ops_data)
954 {
955 	struct rfkill *rfkill;
956 	struct device *dev;
957 
958 	if (WARN_ON(!ops))
959 		return NULL;
960 
961 	if (WARN_ON(!ops->set_block))
962 		return NULL;
963 
964 	if (WARN_ON(!name))
965 		return NULL;
966 
967 	if (WARN_ON(type == RFKILL_TYPE_ALL || type >= NUM_RFKILL_TYPES))
968 		return NULL;
969 
970 	rfkill = kzalloc(sizeof(*rfkill) + strlen(name) + 1, GFP_KERNEL);
971 	if (!rfkill)
972 		return NULL;
973 
974 	spin_lock_init(&rfkill->lock);
975 	INIT_LIST_HEAD(&rfkill->node);
976 	rfkill->type = type;
977 	strcpy(rfkill->name, name);
978 	rfkill->ops = ops;
979 	rfkill->data = ops_data;
980 
981 	dev = &rfkill->dev;
982 	dev->class = &rfkill_class;
983 	dev->parent = parent;
984 	device_initialize(dev);
985 
986 	return rfkill;
987 }
988 EXPORT_SYMBOL(rfkill_alloc);
989 
990 static void rfkill_poll(struct work_struct *work)
991 {
992 	struct rfkill *rfkill;
993 
994 	rfkill = container_of(work, struct rfkill, poll_work.work);
995 
996 	/*
997 	 * Poll hardware state -- driver will use one of the
998 	 * rfkill_set{,_hw,_sw}_state functions and use its
999 	 * return value to update the current status.
1000 	 */
1001 	rfkill->ops->poll(rfkill, rfkill->data);
1002 
1003 	queue_delayed_work(system_power_efficient_wq,
1004 		&rfkill->poll_work,
1005 		round_jiffies_relative(POLL_INTERVAL));
1006 }
1007 
1008 static void rfkill_uevent_work(struct work_struct *work)
1009 {
1010 	struct rfkill *rfkill;
1011 
1012 	rfkill = container_of(work, struct rfkill, uevent_work);
1013 
1014 	mutex_lock(&rfkill_global_mutex);
1015 	rfkill_event(rfkill);
1016 	mutex_unlock(&rfkill_global_mutex);
1017 }
1018 
1019 static void rfkill_sync_work(struct work_struct *work)
1020 {
1021 	struct rfkill *rfkill;
1022 	bool cur;
1023 
1024 	rfkill = container_of(work, struct rfkill, sync_work);
1025 
1026 	mutex_lock(&rfkill_global_mutex);
1027 	cur = rfkill_global_states[rfkill->type].cur;
1028 	rfkill_set_block(rfkill, cur);
1029 	mutex_unlock(&rfkill_global_mutex);
1030 }
1031 
1032 int __must_check rfkill_register(struct rfkill *rfkill)
1033 {
1034 	static unsigned long rfkill_no;
1035 	struct device *dev;
1036 	int error;
1037 
1038 	if (!rfkill)
1039 		return -EINVAL;
1040 
1041 	dev = &rfkill->dev;
1042 
1043 	mutex_lock(&rfkill_global_mutex);
1044 
1045 	if (rfkill->registered) {
1046 		error = -EALREADY;
1047 		goto unlock;
1048 	}
1049 
1050 	rfkill->idx = rfkill_no;
1051 	dev_set_name(dev, "rfkill%lu", rfkill_no);
1052 	rfkill_no++;
1053 
1054 	list_add_tail(&rfkill->node, &rfkill_list);
1055 
1056 	error = device_add(dev);
1057 	if (error)
1058 		goto remove;
1059 
1060 	error = rfkill_led_trigger_register(rfkill);
1061 	if (error)
1062 		goto devdel;
1063 
1064 	rfkill->registered = true;
1065 
1066 	INIT_DELAYED_WORK(&rfkill->poll_work, rfkill_poll);
1067 	INIT_WORK(&rfkill->uevent_work, rfkill_uevent_work);
1068 	INIT_WORK(&rfkill->sync_work, rfkill_sync_work);
1069 
1070 	if (rfkill->ops->poll)
1071 		queue_delayed_work(system_power_efficient_wq,
1072 			&rfkill->poll_work,
1073 			round_jiffies_relative(POLL_INTERVAL));
1074 
1075 	if (!rfkill->persistent || rfkill_epo_lock_active) {
1076 		schedule_work(&rfkill->sync_work);
1077 	} else {
1078 #ifdef CONFIG_RFKILL_INPUT
1079 		bool soft_blocked = !!(rfkill->state & RFKILL_BLOCK_SW);
1080 
1081 		if (!atomic_read(&rfkill_input_disabled))
1082 			__rfkill_switch_all(rfkill->type, soft_blocked);
1083 #endif
1084 	}
1085 
1086 	rfkill_global_led_trigger_event();
1087 	rfkill_send_events(rfkill, RFKILL_OP_ADD);
1088 
1089 	mutex_unlock(&rfkill_global_mutex);
1090 	return 0;
1091 
1092  devdel:
1093 	device_del(&rfkill->dev);
1094  remove:
1095 	list_del_init(&rfkill->node);
1096  unlock:
1097 	mutex_unlock(&rfkill_global_mutex);
1098 	return error;
1099 }
1100 EXPORT_SYMBOL(rfkill_register);
1101 
1102 void rfkill_unregister(struct rfkill *rfkill)
1103 {
1104 	BUG_ON(!rfkill);
1105 
1106 	if (rfkill->ops->poll)
1107 		cancel_delayed_work_sync(&rfkill->poll_work);
1108 
1109 	cancel_work_sync(&rfkill->uevent_work);
1110 	cancel_work_sync(&rfkill->sync_work);
1111 
1112 	rfkill->registered = false;
1113 
1114 	device_del(&rfkill->dev);
1115 
1116 	mutex_lock(&rfkill_global_mutex);
1117 	rfkill_send_events(rfkill, RFKILL_OP_DEL);
1118 	list_del_init(&rfkill->node);
1119 	rfkill_global_led_trigger_event();
1120 	mutex_unlock(&rfkill_global_mutex);
1121 
1122 	rfkill_led_trigger_unregister(rfkill);
1123 }
1124 EXPORT_SYMBOL(rfkill_unregister);
1125 
1126 void rfkill_destroy(struct rfkill *rfkill)
1127 {
1128 	if (rfkill)
1129 		put_device(&rfkill->dev);
1130 }
1131 EXPORT_SYMBOL(rfkill_destroy);
1132 
1133 static int rfkill_fop_open(struct inode *inode, struct file *file)
1134 {
1135 	struct rfkill_data *data;
1136 	struct rfkill *rfkill;
1137 	struct rfkill_int_event *ev, *tmp;
1138 
1139 	data = kzalloc(sizeof(*data), GFP_KERNEL);
1140 	if (!data)
1141 		return -ENOMEM;
1142 
1143 	INIT_LIST_HEAD(&data->events);
1144 	mutex_init(&data->mtx);
1145 	init_waitqueue_head(&data->read_wait);
1146 
1147 	mutex_lock(&rfkill_global_mutex);
1148 	mutex_lock(&data->mtx);
1149 	/*
1150 	 * start getting events from elsewhere but hold mtx to get
1151 	 * startup events added first
1152 	 */
1153 
1154 	list_for_each_entry(rfkill, &rfkill_list, node) {
1155 		ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1156 		if (!ev)
1157 			goto free;
1158 		rfkill_fill_event(&ev->ev, rfkill, RFKILL_OP_ADD);
1159 		list_add_tail(&ev->list, &data->events);
1160 	}
1161 	list_add(&data->list, &rfkill_fds);
1162 	mutex_unlock(&data->mtx);
1163 	mutex_unlock(&rfkill_global_mutex);
1164 
1165 	file->private_data = data;
1166 
1167 	return stream_open(inode, file);
1168 
1169  free:
1170 	mutex_unlock(&data->mtx);
1171 	mutex_unlock(&rfkill_global_mutex);
1172 	mutex_destroy(&data->mtx);
1173 	list_for_each_entry_safe(ev, tmp, &data->events, list)
1174 		kfree(ev);
1175 	kfree(data);
1176 	return -ENOMEM;
1177 }
1178 
1179 static __poll_t rfkill_fop_poll(struct file *file, poll_table *wait)
1180 {
1181 	struct rfkill_data *data = file->private_data;
1182 	__poll_t res = EPOLLOUT | EPOLLWRNORM;
1183 
1184 	poll_wait(file, &data->read_wait, wait);
1185 
1186 	mutex_lock(&data->mtx);
1187 	if (!list_empty(&data->events))
1188 		res = EPOLLIN | EPOLLRDNORM;
1189 	mutex_unlock(&data->mtx);
1190 
1191 	return res;
1192 }
1193 
1194 static ssize_t rfkill_fop_read(struct file *file, char __user *buf,
1195 			       size_t count, loff_t *pos)
1196 {
1197 	struct rfkill_data *data = file->private_data;
1198 	struct rfkill_int_event *ev;
1199 	unsigned long sz;
1200 	int ret;
1201 
1202 	mutex_lock(&data->mtx);
1203 
1204 	while (list_empty(&data->events)) {
1205 		if (file->f_flags & O_NONBLOCK) {
1206 			ret = -EAGAIN;
1207 			goto out;
1208 		}
1209 		mutex_unlock(&data->mtx);
1210 		/* since we re-check and it just compares pointers,
1211 		 * using !list_empty() without locking isn't a problem
1212 		 */
1213 		ret = wait_event_interruptible(data->read_wait,
1214 					       !list_empty(&data->events));
1215 		mutex_lock(&data->mtx);
1216 
1217 		if (ret)
1218 			goto out;
1219 	}
1220 
1221 	ev = list_first_entry(&data->events, struct rfkill_int_event,
1222 				list);
1223 
1224 	sz = min_t(unsigned long, sizeof(ev->ev), count);
1225 	ret = sz;
1226 	if (copy_to_user(buf, &ev->ev, sz))
1227 		ret = -EFAULT;
1228 
1229 	list_del(&ev->list);
1230 	kfree(ev);
1231  out:
1232 	mutex_unlock(&data->mtx);
1233 	return ret;
1234 }
1235 
1236 static ssize_t rfkill_fop_write(struct file *file, const char __user *buf,
1237 				size_t count, loff_t *pos)
1238 {
1239 	struct rfkill *rfkill;
1240 	struct rfkill_event ev;
1241 	int ret;
1242 
1243 	/* we don't need the 'hard' variable but accept it */
1244 	if (count < RFKILL_EVENT_SIZE_V1 - 1)
1245 		return -EINVAL;
1246 
1247 	/*
1248 	 * Copy as much data as we can accept into our 'ev' buffer,
1249 	 * but tell userspace how much we've copied so it can determine
1250 	 * our API version even in a write() call, if it cares.
1251 	 */
1252 	count = min(count, sizeof(ev));
1253 	if (copy_from_user(&ev, buf, count))
1254 		return -EFAULT;
1255 
1256 	if (ev.type >= NUM_RFKILL_TYPES)
1257 		return -EINVAL;
1258 
1259 	mutex_lock(&rfkill_global_mutex);
1260 
1261 	switch (ev.op) {
1262 	case RFKILL_OP_CHANGE_ALL:
1263 		rfkill_update_global_state(ev.type, ev.soft);
1264 		list_for_each_entry(rfkill, &rfkill_list, node)
1265 			if (rfkill->type == ev.type ||
1266 			    ev.type == RFKILL_TYPE_ALL)
1267 				rfkill_set_block(rfkill, ev.soft);
1268 		ret = 0;
1269 		break;
1270 	case RFKILL_OP_CHANGE:
1271 		list_for_each_entry(rfkill, &rfkill_list, node)
1272 			if (rfkill->idx == ev.idx &&
1273 			    (rfkill->type == ev.type ||
1274 			     ev.type == RFKILL_TYPE_ALL))
1275 				rfkill_set_block(rfkill, ev.soft);
1276 		ret = 0;
1277 		break;
1278 	default:
1279 		ret = -EINVAL;
1280 		break;
1281 	}
1282 
1283 	mutex_unlock(&rfkill_global_mutex);
1284 
1285 	return ret ?: count;
1286 }
1287 
1288 static int rfkill_fop_release(struct inode *inode, struct file *file)
1289 {
1290 	struct rfkill_data *data = file->private_data;
1291 	struct rfkill_int_event *ev, *tmp;
1292 
1293 	mutex_lock(&rfkill_global_mutex);
1294 	list_del(&data->list);
1295 	mutex_unlock(&rfkill_global_mutex);
1296 
1297 	mutex_destroy(&data->mtx);
1298 	list_for_each_entry_safe(ev, tmp, &data->events, list)
1299 		kfree(ev);
1300 
1301 #ifdef CONFIG_RFKILL_INPUT
1302 	if (data->input_handler)
1303 		if (atomic_dec_return(&rfkill_input_disabled) == 0)
1304 			printk(KERN_DEBUG "rfkill: input handler enabled\n");
1305 #endif
1306 
1307 	kfree(data);
1308 
1309 	return 0;
1310 }
1311 
1312 #ifdef CONFIG_RFKILL_INPUT
1313 static long rfkill_fop_ioctl(struct file *file, unsigned int cmd,
1314 			     unsigned long arg)
1315 {
1316 	struct rfkill_data *data = file->private_data;
1317 
1318 	if (_IOC_TYPE(cmd) != RFKILL_IOC_MAGIC)
1319 		return -ENOSYS;
1320 
1321 	if (_IOC_NR(cmd) != RFKILL_IOC_NOINPUT)
1322 		return -ENOSYS;
1323 
1324 	mutex_lock(&data->mtx);
1325 
1326 	if (!data->input_handler) {
1327 		if (atomic_inc_return(&rfkill_input_disabled) == 1)
1328 			printk(KERN_DEBUG "rfkill: input handler disabled\n");
1329 		data->input_handler = true;
1330 	}
1331 
1332 	mutex_unlock(&data->mtx);
1333 
1334 	return 0;
1335 }
1336 #endif
1337 
1338 static const struct file_operations rfkill_fops = {
1339 	.owner		= THIS_MODULE,
1340 	.open		= rfkill_fop_open,
1341 	.read		= rfkill_fop_read,
1342 	.write		= rfkill_fop_write,
1343 	.poll		= rfkill_fop_poll,
1344 	.release	= rfkill_fop_release,
1345 #ifdef CONFIG_RFKILL_INPUT
1346 	.unlocked_ioctl	= rfkill_fop_ioctl,
1347 	.compat_ioctl	= compat_ptr_ioctl,
1348 #endif
1349 	.llseek		= no_llseek,
1350 };
1351 
1352 #define RFKILL_NAME "rfkill"
1353 
1354 static struct miscdevice rfkill_miscdev = {
1355 	.fops	= &rfkill_fops,
1356 	.name	= RFKILL_NAME,
1357 	.minor	= RFKILL_MINOR,
1358 };
1359 
1360 static int __init rfkill_init(void)
1361 {
1362 	int error;
1363 
1364 	rfkill_update_global_state(RFKILL_TYPE_ALL, !rfkill_default_state);
1365 
1366 	error = class_register(&rfkill_class);
1367 	if (error)
1368 		goto error_class;
1369 
1370 	error = misc_register(&rfkill_miscdev);
1371 	if (error)
1372 		goto error_misc;
1373 
1374 	error = rfkill_global_led_trigger_register();
1375 	if (error)
1376 		goto error_led_trigger;
1377 
1378 #ifdef CONFIG_RFKILL_INPUT
1379 	error = rfkill_handler_init();
1380 	if (error)
1381 		goto error_input;
1382 #endif
1383 
1384 	return 0;
1385 
1386 #ifdef CONFIG_RFKILL_INPUT
1387 error_input:
1388 	rfkill_global_led_trigger_unregister();
1389 #endif
1390 error_led_trigger:
1391 	misc_deregister(&rfkill_miscdev);
1392 error_misc:
1393 	class_unregister(&rfkill_class);
1394 error_class:
1395 	return error;
1396 }
1397 subsys_initcall(rfkill_init);
1398 
1399 static void __exit rfkill_exit(void)
1400 {
1401 #ifdef CONFIG_RFKILL_INPUT
1402 	rfkill_handler_exit();
1403 #endif
1404 	rfkill_global_led_trigger_unregister();
1405 	misc_deregister(&rfkill_miscdev);
1406 	class_unregister(&rfkill_class);
1407 }
1408 module_exit(rfkill_exit);
1409 
1410 MODULE_ALIAS_MISCDEV(RFKILL_MINOR);
1411 MODULE_ALIAS("devname:" RFKILL_NAME);
1412