xref: /openbmc/linux/net/rfkill/core.c (revision 26b32974)
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_ext	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 	u8			max_size;
82 };
83 
84 
85 MODULE_AUTHOR("Ivo van Doorn <IvDoorn@gmail.com>");
86 MODULE_AUTHOR("Johannes Berg <johannes@sipsolutions.net>");
87 MODULE_DESCRIPTION("RF switch support");
88 MODULE_LICENSE("GPL");
89 
90 
91 /*
92  * The locking here should be made much smarter, we currently have
93  * a bit of a stupid situation because drivers might want to register
94  * the rfkill struct under their own lock, and take this lock during
95  * rfkill method calls -- which will cause an AB-BA deadlock situation.
96  *
97  * To fix that, we need to rework this code here to be mostly lock-free
98  * and only use the mutex for list manipulations, not to protect the
99  * various other global variables. Then we can avoid holding the mutex
100  * around driver operations, and all is happy.
101  */
102 static LIST_HEAD(rfkill_list);	/* list of registered rf switches */
103 static DEFINE_MUTEX(rfkill_global_mutex);
104 static LIST_HEAD(rfkill_fds);	/* list of open fds of /dev/rfkill */
105 
106 static unsigned int rfkill_default_state = 1;
107 module_param_named(default_state, rfkill_default_state, uint, 0444);
108 MODULE_PARM_DESC(default_state,
109 		 "Default initial state for all radio types, 0 = radio off");
110 
111 static struct {
112 	bool cur, sav;
113 } rfkill_global_states[NUM_RFKILL_TYPES];
114 
115 static bool rfkill_epo_lock_active;
116 
117 
118 #ifdef CONFIG_RFKILL_LEDS
119 static void rfkill_led_trigger_event(struct rfkill *rfkill)
120 {
121 	struct led_trigger *trigger;
122 
123 	if (!rfkill->registered)
124 		return;
125 
126 	trigger = &rfkill->led_trigger;
127 
128 	if (rfkill->state & RFKILL_BLOCK_ANY)
129 		led_trigger_event(trigger, LED_OFF);
130 	else
131 		led_trigger_event(trigger, LED_FULL);
132 }
133 
134 static int rfkill_led_trigger_activate(struct led_classdev *led)
135 {
136 	struct rfkill *rfkill;
137 
138 	rfkill = container_of(led->trigger, struct rfkill, led_trigger);
139 
140 	rfkill_led_trigger_event(rfkill);
141 
142 	return 0;
143 }
144 
145 const char *rfkill_get_led_trigger_name(struct rfkill *rfkill)
146 {
147 	return rfkill->led_trigger.name;
148 }
149 EXPORT_SYMBOL(rfkill_get_led_trigger_name);
150 
151 void rfkill_set_led_trigger_name(struct rfkill *rfkill, const char *name)
152 {
153 	BUG_ON(!rfkill);
154 
155 	rfkill->ledtrigname = name;
156 }
157 EXPORT_SYMBOL(rfkill_set_led_trigger_name);
158 
159 static int rfkill_led_trigger_register(struct rfkill *rfkill)
160 {
161 	rfkill->led_trigger.name = rfkill->ledtrigname
162 					? : dev_name(&rfkill->dev);
163 	rfkill->led_trigger.activate = rfkill_led_trigger_activate;
164 	return led_trigger_register(&rfkill->led_trigger);
165 }
166 
167 static void rfkill_led_trigger_unregister(struct rfkill *rfkill)
168 {
169 	led_trigger_unregister(&rfkill->led_trigger);
170 }
171 
172 static struct led_trigger rfkill_any_led_trigger;
173 static struct led_trigger rfkill_none_led_trigger;
174 static struct work_struct rfkill_global_led_trigger_work;
175 
176 static void rfkill_global_led_trigger_worker(struct work_struct *work)
177 {
178 	enum led_brightness brightness = LED_OFF;
179 	struct rfkill *rfkill;
180 
181 	mutex_lock(&rfkill_global_mutex);
182 	list_for_each_entry(rfkill, &rfkill_list, node) {
183 		if (!(rfkill->state & RFKILL_BLOCK_ANY)) {
184 			brightness = LED_FULL;
185 			break;
186 		}
187 	}
188 	mutex_unlock(&rfkill_global_mutex);
189 
190 	led_trigger_event(&rfkill_any_led_trigger, brightness);
191 	led_trigger_event(&rfkill_none_led_trigger,
192 			  brightness == LED_OFF ? LED_FULL : LED_OFF);
193 }
194 
195 static void rfkill_global_led_trigger_event(void)
196 {
197 	schedule_work(&rfkill_global_led_trigger_work);
198 }
199 
200 static int rfkill_global_led_trigger_register(void)
201 {
202 	int ret;
203 
204 	INIT_WORK(&rfkill_global_led_trigger_work,
205 			rfkill_global_led_trigger_worker);
206 
207 	rfkill_any_led_trigger.name = "rfkill-any";
208 	ret = led_trigger_register(&rfkill_any_led_trigger);
209 	if (ret)
210 		return ret;
211 
212 	rfkill_none_led_trigger.name = "rfkill-none";
213 	ret = led_trigger_register(&rfkill_none_led_trigger);
214 	if (ret)
215 		led_trigger_unregister(&rfkill_any_led_trigger);
216 	else
217 		/* Delay activation until all global triggers are registered */
218 		rfkill_global_led_trigger_event();
219 
220 	return ret;
221 }
222 
223 static void rfkill_global_led_trigger_unregister(void)
224 {
225 	led_trigger_unregister(&rfkill_none_led_trigger);
226 	led_trigger_unregister(&rfkill_any_led_trigger);
227 	cancel_work_sync(&rfkill_global_led_trigger_work);
228 }
229 #else
230 static void rfkill_led_trigger_event(struct rfkill *rfkill)
231 {
232 }
233 
234 static inline int rfkill_led_trigger_register(struct rfkill *rfkill)
235 {
236 	return 0;
237 }
238 
239 static inline void rfkill_led_trigger_unregister(struct rfkill *rfkill)
240 {
241 }
242 
243 static void rfkill_global_led_trigger_event(void)
244 {
245 }
246 
247 static int rfkill_global_led_trigger_register(void)
248 {
249 	return 0;
250 }
251 
252 static void rfkill_global_led_trigger_unregister(void)
253 {
254 }
255 #endif /* CONFIG_RFKILL_LEDS */
256 
257 static void rfkill_fill_event(struct rfkill_event_ext *ev,
258 			      struct rfkill *rfkill,
259 			      enum rfkill_operation op)
260 {
261 	unsigned long flags;
262 
263 	ev->idx = rfkill->idx;
264 	ev->type = rfkill->type;
265 	ev->op = op;
266 
267 	spin_lock_irqsave(&rfkill->lock, flags);
268 	ev->hard = !!(rfkill->state & RFKILL_BLOCK_HW);
269 	ev->soft = !!(rfkill->state & (RFKILL_BLOCK_SW |
270 					RFKILL_BLOCK_SW_PREV));
271 	ev->hard_block_reasons = rfkill->hard_block_reasons;
272 	spin_unlock_irqrestore(&rfkill->lock, flags);
273 }
274 
275 static void rfkill_send_events(struct rfkill *rfkill, enum rfkill_operation op)
276 {
277 	struct rfkill_data *data;
278 	struct rfkill_int_event *ev;
279 
280 	list_for_each_entry(data, &rfkill_fds, list) {
281 		ev = kzalloc(sizeof(*ev), GFP_KERNEL);
282 		if (!ev)
283 			continue;
284 		rfkill_fill_event(&ev->ev, rfkill, op);
285 		mutex_lock(&data->mtx);
286 		list_add_tail(&ev->list, &data->events);
287 		mutex_unlock(&data->mtx);
288 		wake_up_interruptible(&data->read_wait);
289 	}
290 }
291 
292 static void rfkill_event(struct rfkill *rfkill)
293 {
294 	if (!rfkill->registered)
295 		return;
296 
297 	kobject_uevent(&rfkill->dev.kobj, KOBJ_CHANGE);
298 
299 	/* also send event to /dev/rfkill */
300 	rfkill_send_events(rfkill, RFKILL_OP_CHANGE);
301 }
302 
303 /**
304  * rfkill_set_block - wrapper for set_block method
305  *
306  * @rfkill: the rfkill struct to use
307  * @blocked: the new software state
308  *
309  * Calls the set_block method (when applicable) and handles notifications
310  * etc. as well.
311  */
312 static void rfkill_set_block(struct rfkill *rfkill, bool blocked)
313 {
314 	unsigned long flags;
315 	bool prev, curr;
316 	int err;
317 
318 	if (unlikely(rfkill->dev.power.power_state.event & PM_EVENT_SLEEP))
319 		return;
320 
321 	/*
322 	 * Some platforms (...!) generate input events which affect the
323 	 * _hard_ kill state -- whenever something tries to change the
324 	 * current software state query the hardware state too.
325 	 */
326 	if (rfkill->ops->query)
327 		rfkill->ops->query(rfkill, rfkill->data);
328 
329 	spin_lock_irqsave(&rfkill->lock, flags);
330 	prev = rfkill->state & RFKILL_BLOCK_SW;
331 
332 	if (prev)
333 		rfkill->state |= RFKILL_BLOCK_SW_PREV;
334 	else
335 		rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
336 
337 	if (blocked)
338 		rfkill->state |= RFKILL_BLOCK_SW;
339 	else
340 		rfkill->state &= ~RFKILL_BLOCK_SW;
341 
342 	rfkill->state |= RFKILL_BLOCK_SW_SETCALL;
343 	spin_unlock_irqrestore(&rfkill->lock, flags);
344 
345 	err = rfkill->ops->set_block(rfkill->data, blocked);
346 
347 	spin_lock_irqsave(&rfkill->lock, flags);
348 	if (err) {
349 		/*
350 		 * Failed -- reset status to _PREV, which may be different
351 		 * from what we have set _PREV to earlier in this function
352 		 * if rfkill_set_sw_state was invoked.
353 		 */
354 		if (rfkill->state & RFKILL_BLOCK_SW_PREV)
355 			rfkill->state |= RFKILL_BLOCK_SW;
356 		else
357 			rfkill->state &= ~RFKILL_BLOCK_SW;
358 	}
359 	rfkill->state &= ~RFKILL_BLOCK_SW_SETCALL;
360 	rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
361 	curr = rfkill->state & RFKILL_BLOCK_SW;
362 	spin_unlock_irqrestore(&rfkill->lock, flags);
363 
364 	rfkill_led_trigger_event(rfkill);
365 	rfkill_global_led_trigger_event();
366 
367 	if (prev != curr)
368 		rfkill_event(rfkill);
369 }
370 
371 static void rfkill_update_global_state(enum rfkill_type type, bool blocked)
372 {
373 	int i;
374 
375 	if (type != RFKILL_TYPE_ALL) {
376 		rfkill_global_states[type].cur = blocked;
377 		return;
378 	}
379 
380 	for (i = 0; i < NUM_RFKILL_TYPES; i++)
381 		rfkill_global_states[i].cur = blocked;
382 }
383 
384 #ifdef CONFIG_RFKILL_INPUT
385 static atomic_t rfkill_input_disabled = ATOMIC_INIT(0);
386 
387 /**
388  * __rfkill_switch_all - Toggle state of all switches of given type
389  * @type: type of interfaces to be affected
390  * @blocked: the new state
391  *
392  * This function sets the state of all switches of given type,
393  * unless a specific switch is suspended.
394  *
395  * Caller must have acquired rfkill_global_mutex.
396  */
397 static void __rfkill_switch_all(const enum rfkill_type type, bool blocked)
398 {
399 	struct rfkill *rfkill;
400 
401 	rfkill_update_global_state(type, blocked);
402 	list_for_each_entry(rfkill, &rfkill_list, node) {
403 		if (rfkill->type != type && type != RFKILL_TYPE_ALL)
404 			continue;
405 
406 		rfkill_set_block(rfkill, blocked);
407 	}
408 }
409 
410 /**
411  * rfkill_switch_all - Toggle state of all switches of given type
412  * @type: type of interfaces to be affected
413  * @blocked: the new state
414  *
415  * Acquires rfkill_global_mutex and calls __rfkill_switch_all(@type, @state).
416  * Please refer to __rfkill_switch_all() for details.
417  *
418  * Does nothing if the EPO lock is active.
419  */
420 void rfkill_switch_all(enum rfkill_type type, bool blocked)
421 {
422 	if (atomic_read(&rfkill_input_disabled))
423 		return;
424 
425 	mutex_lock(&rfkill_global_mutex);
426 
427 	if (!rfkill_epo_lock_active)
428 		__rfkill_switch_all(type, blocked);
429 
430 	mutex_unlock(&rfkill_global_mutex);
431 }
432 
433 /**
434  * rfkill_epo - emergency power off all transmitters
435  *
436  * This kicks all non-suspended rfkill devices to RFKILL_STATE_SOFT_BLOCKED,
437  * ignoring everything in its path but rfkill_global_mutex and rfkill->mutex.
438  *
439  * The global state before the EPO is saved and can be restored later
440  * using rfkill_restore_states().
441  */
442 void rfkill_epo(void)
443 {
444 	struct rfkill *rfkill;
445 	int i;
446 
447 	if (atomic_read(&rfkill_input_disabled))
448 		return;
449 
450 	mutex_lock(&rfkill_global_mutex);
451 
452 	rfkill_epo_lock_active = true;
453 	list_for_each_entry(rfkill, &rfkill_list, node)
454 		rfkill_set_block(rfkill, true);
455 
456 	for (i = 0; i < NUM_RFKILL_TYPES; i++) {
457 		rfkill_global_states[i].sav = rfkill_global_states[i].cur;
458 		rfkill_global_states[i].cur = true;
459 	}
460 
461 	mutex_unlock(&rfkill_global_mutex);
462 }
463 
464 /**
465  * rfkill_restore_states - restore global states
466  *
467  * Restore (and sync switches to) the global state from the
468  * states in rfkill_default_states.  This can undo the effects of
469  * a call to rfkill_epo().
470  */
471 void rfkill_restore_states(void)
472 {
473 	int i;
474 
475 	if (atomic_read(&rfkill_input_disabled))
476 		return;
477 
478 	mutex_lock(&rfkill_global_mutex);
479 
480 	rfkill_epo_lock_active = false;
481 	for (i = 0; i < NUM_RFKILL_TYPES; i++)
482 		__rfkill_switch_all(i, rfkill_global_states[i].sav);
483 	mutex_unlock(&rfkill_global_mutex);
484 }
485 
486 /**
487  * rfkill_remove_epo_lock - unlock state changes
488  *
489  * Used by rfkill-input manually unlock state changes, when
490  * the EPO switch is deactivated.
491  */
492 void rfkill_remove_epo_lock(void)
493 {
494 	if (atomic_read(&rfkill_input_disabled))
495 		return;
496 
497 	mutex_lock(&rfkill_global_mutex);
498 	rfkill_epo_lock_active = false;
499 	mutex_unlock(&rfkill_global_mutex);
500 }
501 
502 /**
503  * rfkill_is_epo_lock_active - returns true EPO is active
504  *
505  * Returns 0 (false) if there is NOT an active EPO condition,
506  * and 1 (true) if there is an active EPO condition, which
507  * locks all radios in one of the BLOCKED states.
508  *
509  * Can be called in atomic context.
510  */
511 bool rfkill_is_epo_lock_active(void)
512 {
513 	return rfkill_epo_lock_active;
514 }
515 
516 /**
517  * rfkill_get_global_sw_state - returns global state for a type
518  * @type: the type to get the global state of
519  *
520  * Returns the current global state for a given wireless
521  * device type.
522  */
523 bool rfkill_get_global_sw_state(const enum rfkill_type type)
524 {
525 	return rfkill_global_states[type].cur;
526 }
527 #endif
528 
529 bool rfkill_set_hw_state_reason(struct rfkill *rfkill,
530 				bool blocked, unsigned long reason)
531 {
532 	unsigned long flags;
533 	bool ret, prev;
534 
535 	BUG_ON(!rfkill);
536 
537 	if (WARN(reason &
538 	    ~(RFKILL_HARD_BLOCK_SIGNAL | RFKILL_HARD_BLOCK_NOT_OWNER),
539 	    "hw_state reason not supported: 0x%lx", reason))
540 		return blocked;
541 
542 	spin_lock_irqsave(&rfkill->lock, flags);
543 	prev = !!(rfkill->hard_block_reasons & reason);
544 	if (blocked) {
545 		rfkill->state |= RFKILL_BLOCK_HW;
546 		rfkill->hard_block_reasons |= reason;
547 	} else {
548 		rfkill->hard_block_reasons &= ~reason;
549 		if (!rfkill->hard_block_reasons)
550 			rfkill->state &= ~RFKILL_BLOCK_HW;
551 	}
552 	ret = !!(rfkill->state & RFKILL_BLOCK_ANY);
553 	spin_unlock_irqrestore(&rfkill->lock, flags);
554 
555 	rfkill_led_trigger_event(rfkill);
556 	rfkill_global_led_trigger_event();
557 
558 	if (rfkill->registered && prev != blocked)
559 		schedule_work(&rfkill->uevent_work);
560 
561 	return ret;
562 }
563 EXPORT_SYMBOL(rfkill_set_hw_state_reason);
564 
565 static void __rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
566 {
567 	u32 bit = RFKILL_BLOCK_SW;
568 
569 	/* if in a ops->set_block right now, use other bit */
570 	if (rfkill->state & RFKILL_BLOCK_SW_SETCALL)
571 		bit = RFKILL_BLOCK_SW_PREV;
572 
573 	if (blocked)
574 		rfkill->state |= bit;
575 	else
576 		rfkill->state &= ~bit;
577 }
578 
579 bool rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
580 {
581 	unsigned long flags;
582 	bool prev, hwblock;
583 
584 	BUG_ON(!rfkill);
585 
586 	spin_lock_irqsave(&rfkill->lock, flags);
587 	prev = !!(rfkill->state & RFKILL_BLOCK_SW);
588 	__rfkill_set_sw_state(rfkill, blocked);
589 	hwblock = !!(rfkill->state & RFKILL_BLOCK_HW);
590 	blocked = blocked || hwblock;
591 	spin_unlock_irqrestore(&rfkill->lock, flags);
592 
593 	if (!rfkill->registered)
594 		return blocked;
595 
596 	if (prev != blocked && !hwblock)
597 		schedule_work(&rfkill->uevent_work);
598 
599 	rfkill_led_trigger_event(rfkill);
600 	rfkill_global_led_trigger_event();
601 
602 	return blocked;
603 }
604 EXPORT_SYMBOL(rfkill_set_sw_state);
605 
606 void rfkill_init_sw_state(struct rfkill *rfkill, bool blocked)
607 {
608 	unsigned long flags;
609 
610 	BUG_ON(!rfkill);
611 	BUG_ON(rfkill->registered);
612 
613 	spin_lock_irqsave(&rfkill->lock, flags);
614 	__rfkill_set_sw_state(rfkill, blocked);
615 	rfkill->persistent = true;
616 	spin_unlock_irqrestore(&rfkill->lock, flags);
617 }
618 EXPORT_SYMBOL(rfkill_init_sw_state);
619 
620 void rfkill_set_states(struct rfkill *rfkill, bool sw, bool hw)
621 {
622 	unsigned long flags;
623 	bool swprev, hwprev;
624 
625 	BUG_ON(!rfkill);
626 
627 	spin_lock_irqsave(&rfkill->lock, flags);
628 
629 	/*
630 	 * No need to care about prev/setblock ... this is for uevent only
631 	 * and that will get triggered by rfkill_set_block anyway.
632 	 */
633 	swprev = !!(rfkill->state & RFKILL_BLOCK_SW);
634 	hwprev = !!(rfkill->state & RFKILL_BLOCK_HW);
635 	__rfkill_set_sw_state(rfkill, sw);
636 	if (hw)
637 		rfkill->state |= RFKILL_BLOCK_HW;
638 	else
639 		rfkill->state &= ~RFKILL_BLOCK_HW;
640 
641 	spin_unlock_irqrestore(&rfkill->lock, flags);
642 
643 	if (!rfkill->registered) {
644 		rfkill->persistent = true;
645 	} else {
646 		if (swprev != sw || hwprev != hw)
647 			schedule_work(&rfkill->uevent_work);
648 
649 		rfkill_led_trigger_event(rfkill);
650 		rfkill_global_led_trigger_event();
651 	}
652 }
653 EXPORT_SYMBOL(rfkill_set_states);
654 
655 static const char * const rfkill_types[] = {
656 	NULL, /* RFKILL_TYPE_ALL */
657 	"wlan",
658 	"bluetooth",
659 	"ultrawideband",
660 	"wimax",
661 	"wwan",
662 	"gps",
663 	"fm",
664 	"nfc",
665 };
666 
667 enum rfkill_type rfkill_find_type(const char *name)
668 {
669 	int i;
670 
671 	BUILD_BUG_ON(ARRAY_SIZE(rfkill_types) != NUM_RFKILL_TYPES);
672 
673 	if (!name)
674 		return RFKILL_TYPE_ALL;
675 
676 	for (i = 1; i < NUM_RFKILL_TYPES; i++)
677 		if (!strcmp(name, rfkill_types[i]))
678 			return i;
679 	return RFKILL_TYPE_ALL;
680 }
681 EXPORT_SYMBOL(rfkill_find_type);
682 
683 static ssize_t name_show(struct device *dev, struct device_attribute *attr,
684 			 char *buf)
685 {
686 	struct rfkill *rfkill = to_rfkill(dev);
687 
688 	return sysfs_emit(buf, "%s\n", rfkill->name);
689 }
690 static DEVICE_ATTR_RO(name);
691 
692 static ssize_t type_show(struct device *dev, struct device_attribute *attr,
693 			 char *buf)
694 {
695 	struct rfkill *rfkill = to_rfkill(dev);
696 
697 	return sysfs_emit(buf, "%s\n", rfkill_types[rfkill->type]);
698 }
699 static DEVICE_ATTR_RO(type);
700 
701 static ssize_t index_show(struct device *dev, struct device_attribute *attr,
702 			  char *buf)
703 {
704 	struct rfkill *rfkill = to_rfkill(dev);
705 
706 	return sysfs_emit(buf, "%d\n", rfkill->idx);
707 }
708 static DEVICE_ATTR_RO(index);
709 
710 static ssize_t persistent_show(struct device *dev,
711 			       struct device_attribute *attr, char *buf)
712 {
713 	struct rfkill *rfkill = to_rfkill(dev);
714 
715 	return sysfs_emit(buf, "%d\n", rfkill->persistent);
716 }
717 static DEVICE_ATTR_RO(persistent);
718 
719 static ssize_t hard_show(struct device *dev, struct device_attribute *attr,
720 			 char *buf)
721 {
722 	struct rfkill *rfkill = to_rfkill(dev);
723 
724 	return sysfs_emit(buf, "%d\n", (rfkill->state & RFKILL_BLOCK_HW) ? 1 : 0);
725 }
726 static DEVICE_ATTR_RO(hard);
727 
728 static ssize_t soft_show(struct device *dev, struct device_attribute *attr,
729 			 char *buf)
730 {
731 	struct rfkill *rfkill = to_rfkill(dev);
732 
733 	return sysfs_emit(buf, "%d\n", (rfkill->state & RFKILL_BLOCK_SW) ? 1 : 0);
734 }
735 
736 static ssize_t soft_store(struct device *dev, struct device_attribute *attr,
737 			  const char *buf, size_t count)
738 {
739 	struct rfkill *rfkill = to_rfkill(dev);
740 	unsigned long state;
741 	int err;
742 
743 	if (!capable(CAP_NET_ADMIN))
744 		return -EPERM;
745 
746 	err = kstrtoul(buf, 0, &state);
747 	if (err)
748 		return err;
749 
750 	if (state > 1 )
751 		return -EINVAL;
752 
753 	mutex_lock(&rfkill_global_mutex);
754 	rfkill_set_block(rfkill, state);
755 	mutex_unlock(&rfkill_global_mutex);
756 
757 	return count;
758 }
759 static DEVICE_ATTR_RW(soft);
760 
761 static ssize_t hard_block_reasons_show(struct device *dev,
762 				       struct device_attribute *attr,
763 				       char *buf)
764 {
765 	struct rfkill *rfkill = to_rfkill(dev);
766 
767 	return sysfs_emit(buf, "0x%lx\n", rfkill->hard_block_reasons);
768 }
769 static DEVICE_ATTR_RO(hard_block_reasons);
770 
771 static u8 user_state_from_blocked(unsigned long state)
772 {
773 	if (state & RFKILL_BLOCK_HW)
774 		return RFKILL_USER_STATE_HARD_BLOCKED;
775 	if (state & RFKILL_BLOCK_SW)
776 		return RFKILL_USER_STATE_SOFT_BLOCKED;
777 
778 	return RFKILL_USER_STATE_UNBLOCKED;
779 }
780 
781 static ssize_t state_show(struct device *dev, struct device_attribute *attr,
782 			  char *buf)
783 {
784 	struct rfkill *rfkill = to_rfkill(dev);
785 
786 	return sysfs_emit(buf, "%d\n", user_state_from_blocked(rfkill->state));
787 }
788 
789 static ssize_t state_store(struct device *dev, struct device_attribute *attr,
790 			   const char *buf, size_t count)
791 {
792 	struct rfkill *rfkill = to_rfkill(dev);
793 	unsigned long state;
794 	int err;
795 
796 	if (!capable(CAP_NET_ADMIN))
797 		return -EPERM;
798 
799 	err = kstrtoul(buf, 0, &state);
800 	if (err)
801 		return err;
802 
803 	if (state != RFKILL_USER_STATE_SOFT_BLOCKED &&
804 	    state != RFKILL_USER_STATE_UNBLOCKED)
805 		return -EINVAL;
806 
807 	mutex_lock(&rfkill_global_mutex);
808 	rfkill_set_block(rfkill, state == RFKILL_USER_STATE_SOFT_BLOCKED);
809 	mutex_unlock(&rfkill_global_mutex);
810 
811 	return count;
812 }
813 static DEVICE_ATTR_RW(state);
814 
815 static struct attribute *rfkill_dev_attrs[] = {
816 	&dev_attr_name.attr,
817 	&dev_attr_type.attr,
818 	&dev_attr_index.attr,
819 	&dev_attr_persistent.attr,
820 	&dev_attr_state.attr,
821 	&dev_attr_soft.attr,
822 	&dev_attr_hard.attr,
823 	&dev_attr_hard_block_reasons.attr,
824 	NULL,
825 };
826 ATTRIBUTE_GROUPS(rfkill_dev);
827 
828 static void rfkill_release(struct device *dev)
829 {
830 	struct rfkill *rfkill = to_rfkill(dev);
831 
832 	kfree(rfkill);
833 }
834 
835 static int rfkill_dev_uevent(const struct device *dev, struct kobj_uevent_env *env)
836 {
837 	struct rfkill *rfkill = to_rfkill(dev);
838 	unsigned long flags;
839 	unsigned long reasons;
840 	u32 state;
841 	int error;
842 
843 	error = add_uevent_var(env, "RFKILL_NAME=%s", rfkill->name);
844 	if (error)
845 		return error;
846 	error = add_uevent_var(env, "RFKILL_TYPE=%s",
847 			       rfkill_types[rfkill->type]);
848 	if (error)
849 		return error;
850 	spin_lock_irqsave(&rfkill->lock, flags);
851 	state = rfkill->state;
852 	reasons = rfkill->hard_block_reasons;
853 	spin_unlock_irqrestore(&rfkill->lock, flags);
854 	error = add_uevent_var(env, "RFKILL_STATE=%d",
855 			       user_state_from_blocked(state));
856 	if (error)
857 		return error;
858 	return add_uevent_var(env, "RFKILL_HW_BLOCK_REASON=0x%lx", reasons);
859 }
860 
861 void rfkill_pause_polling(struct rfkill *rfkill)
862 {
863 	BUG_ON(!rfkill);
864 
865 	if (!rfkill->ops->poll)
866 		return;
867 
868 	rfkill->polling_paused = true;
869 	cancel_delayed_work_sync(&rfkill->poll_work);
870 }
871 EXPORT_SYMBOL(rfkill_pause_polling);
872 
873 void rfkill_resume_polling(struct rfkill *rfkill)
874 {
875 	BUG_ON(!rfkill);
876 
877 	if (!rfkill->ops->poll)
878 		return;
879 
880 	rfkill->polling_paused = false;
881 
882 	if (rfkill->suspended)
883 		return;
884 
885 	queue_delayed_work(system_power_efficient_wq,
886 			   &rfkill->poll_work, 0);
887 }
888 EXPORT_SYMBOL(rfkill_resume_polling);
889 
890 #ifdef CONFIG_PM_SLEEP
891 static int rfkill_suspend(struct device *dev)
892 {
893 	struct rfkill *rfkill = to_rfkill(dev);
894 
895 	rfkill->suspended = true;
896 	cancel_delayed_work_sync(&rfkill->poll_work);
897 
898 	return 0;
899 }
900 
901 static int rfkill_resume(struct device *dev)
902 {
903 	struct rfkill *rfkill = to_rfkill(dev);
904 	bool cur;
905 
906 	rfkill->suspended = false;
907 
908 	if (!rfkill->registered)
909 		return 0;
910 
911 	if (!rfkill->persistent) {
912 		cur = !!(rfkill->state & RFKILL_BLOCK_SW);
913 		rfkill_set_block(rfkill, cur);
914 	}
915 
916 	if (rfkill->ops->poll && !rfkill->polling_paused)
917 		queue_delayed_work(system_power_efficient_wq,
918 				   &rfkill->poll_work, 0);
919 
920 	return 0;
921 }
922 
923 static SIMPLE_DEV_PM_OPS(rfkill_pm_ops, rfkill_suspend, rfkill_resume);
924 #define RFKILL_PM_OPS (&rfkill_pm_ops)
925 #else
926 #define RFKILL_PM_OPS NULL
927 #endif
928 
929 static struct class rfkill_class = {
930 	.name		= "rfkill",
931 	.dev_release	= rfkill_release,
932 	.dev_groups	= rfkill_dev_groups,
933 	.dev_uevent	= rfkill_dev_uevent,
934 	.pm		= RFKILL_PM_OPS,
935 };
936 
937 bool rfkill_blocked(struct rfkill *rfkill)
938 {
939 	unsigned long flags;
940 	u32 state;
941 
942 	spin_lock_irqsave(&rfkill->lock, flags);
943 	state = rfkill->state;
944 	spin_unlock_irqrestore(&rfkill->lock, flags);
945 
946 	return !!(state & RFKILL_BLOCK_ANY);
947 }
948 EXPORT_SYMBOL(rfkill_blocked);
949 
950 bool rfkill_soft_blocked(struct rfkill *rfkill)
951 {
952 	unsigned long flags;
953 	u32 state;
954 
955 	spin_lock_irqsave(&rfkill->lock, flags);
956 	state = rfkill->state;
957 	spin_unlock_irqrestore(&rfkill->lock, flags);
958 
959 	return !!(state & RFKILL_BLOCK_SW);
960 }
961 EXPORT_SYMBOL(rfkill_soft_blocked);
962 
963 struct rfkill * __must_check rfkill_alloc(const char *name,
964 					  struct device *parent,
965 					  const enum rfkill_type type,
966 					  const struct rfkill_ops *ops,
967 					  void *ops_data)
968 {
969 	struct rfkill *rfkill;
970 	struct device *dev;
971 
972 	if (WARN_ON(!ops))
973 		return NULL;
974 
975 	if (WARN_ON(!ops->set_block))
976 		return NULL;
977 
978 	if (WARN_ON(!name))
979 		return NULL;
980 
981 	if (WARN_ON(type == RFKILL_TYPE_ALL || type >= NUM_RFKILL_TYPES))
982 		return NULL;
983 
984 	rfkill = kzalloc(sizeof(*rfkill) + strlen(name) + 1, GFP_KERNEL);
985 	if (!rfkill)
986 		return NULL;
987 
988 	spin_lock_init(&rfkill->lock);
989 	INIT_LIST_HEAD(&rfkill->node);
990 	rfkill->type = type;
991 	strcpy(rfkill->name, name);
992 	rfkill->ops = ops;
993 	rfkill->data = ops_data;
994 
995 	dev = &rfkill->dev;
996 	dev->class = &rfkill_class;
997 	dev->parent = parent;
998 	device_initialize(dev);
999 
1000 	return rfkill;
1001 }
1002 EXPORT_SYMBOL(rfkill_alloc);
1003 
1004 static void rfkill_poll(struct work_struct *work)
1005 {
1006 	struct rfkill *rfkill;
1007 
1008 	rfkill = container_of(work, struct rfkill, poll_work.work);
1009 
1010 	/*
1011 	 * Poll hardware state -- driver will use one of the
1012 	 * rfkill_set{,_hw,_sw}_state functions and use its
1013 	 * return value to update the current status.
1014 	 */
1015 	rfkill->ops->poll(rfkill, rfkill->data);
1016 
1017 	queue_delayed_work(system_power_efficient_wq,
1018 		&rfkill->poll_work,
1019 		round_jiffies_relative(POLL_INTERVAL));
1020 }
1021 
1022 static void rfkill_uevent_work(struct work_struct *work)
1023 {
1024 	struct rfkill *rfkill;
1025 
1026 	rfkill = container_of(work, struct rfkill, uevent_work);
1027 
1028 	mutex_lock(&rfkill_global_mutex);
1029 	rfkill_event(rfkill);
1030 	mutex_unlock(&rfkill_global_mutex);
1031 }
1032 
1033 static void rfkill_sync_work(struct work_struct *work)
1034 {
1035 	struct rfkill *rfkill;
1036 	bool cur;
1037 
1038 	rfkill = container_of(work, struct rfkill, sync_work);
1039 
1040 	mutex_lock(&rfkill_global_mutex);
1041 	cur = rfkill_global_states[rfkill->type].cur;
1042 	rfkill_set_block(rfkill, cur);
1043 	mutex_unlock(&rfkill_global_mutex);
1044 }
1045 
1046 int __must_check rfkill_register(struct rfkill *rfkill)
1047 {
1048 	static unsigned long rfkill_no;
1049 	struct device *dev;
1050 	int error;
1051 
1052 	if (!rfkill)
1053 		return -EINVAL;
1054 
1055 	dev = &rfkill->dev;
1056 
1057 	mutex_lock(&rfkill_global_mutex);
1058 
1059 	if (rfkill->registered) {
1060 		error = -EALREADY;
1061 		goto unlock;
1062 	}
1063 
1064 	rfkill->idx = rfkill_no;
1065 	dev_set_name(dev, "rfkill%lu", rfkill_no);
1066 	rfkill_no++;
1067 
1068 	list_add_tail(&rfkill->node, &rfkill_list);
1069 
1070 	error = device_add(dev);
1071 	if (error)
1072 		goto remove;
1073 
1074 	error = rfkill_led_trigger_register(rfkill);
1075 	if (error)
1076 		goto devdel;
1077 
1078 	rfkill->registered = true;
1079 
1080 	INIT_DELAYED_WORK(&rfkill->poll_work, rfkill_poll);
1081 	INIT_WORK(&rfkill->uevent_work, rfkill_uevent_work);
1082 	INIT_WORK(&rfkill->sync_work, rfkill_sync_work);
1083 
1084 	if (rfkill->ops->poll)
1085 		queue_delayed_work(system_power_efficient_wq,
1086 			&rfkill->poll_work,
1087 			round_jiffies_relative(POLL_INTERVAL));
1088 
1089 	if (!rfkill->persistent || rfkill_epo_lock_active) {
1090 		schedule_work(&rfkill->sync_work);
1091 	} else {
1092 #ifdef CONFIG_RFKILL_INPUT
1093 		bool soft_blocked = !!(rfkill->state & RFKILL_BLOCK_SW);
1094 
1095 		if (!atomic_read(&rfkill_input_disabled))
1096 			__rfkill_switch_all(rfkill->type, soft_blocked);
1097 #endif
1098 	}
1099 
1100 	rfkill_global_led_trigger_event();
1101 	rfkill_send_events(rfkill, RFKILL_OP_ADD);
1102 
1103 	mutex_unlock(&rfkill_global_mutex);
1104 	return 0;
1105 
1106  devdel:
1107 	device_del(&rfkill->dev);
1108  remove:
1109 	list_del_init(&rfkill->node);
1110  unlock:
1111 	mutex_unlock(&rfkill_global_mutex);
1112 	return error;
1113 }
1114 EXPORT_SYMBOL(rfkill_register);
1115 
1116 void rfkill_unregister(struct rfkill *rfkill)
1117 {
1118 	BUG_ON(!rfkill);
1119 
1120 	if (rfkill->ops->poll)
1121 		cancel_delayed_work_sync(&rfkill->poll_work);
1122 
1123 	cancel_work_sync(&rfkill->uevent_work);
1124 	cancel_work_sync(&rfkill->sync_work);
1125 
1126 	rfkill->registered = false;
1127 
1128 	device_del(&rfkill->dev);
1129 
1130 	mutex_lock(&rfkill_global_mutex);
1131 	rfkill_send_events(rfkill, RFKILL_OP_DEL);
1132 	list_del_init(&rfkill->node);
1133 	rfkill_global_led_trigger_event();
1134 	mutex_unlock(&rfkill_global_mutex);
1135 
1136 	rfkill_led_trigger_unregister(rfkill);
1137 }
1138 EXPORT_SYMBOL(rfkill_unregister);
1139 
1140 void rfkill_destroy(struct rfkill *rfkill)
1141 {
1142 	if (rfkill)
1143 		put_device(&rfkill->dev);
1144 }
1145 EXPORT_SYMBOL(rfkill_destroy);
1146 
1147 static int rfkill_fop_open(struct inode *inode, struct file *file)
1148 {
1149 	struct rfkill_data *data;
1150 	struct rfkill *rfkill;
1151 	struct rfkill_int_event *ev, *tmp;
1152 
1153 	data = kzalloc(sizeof(*data), GFP_KERNEL);
1154 	if (!data)
1155 		return -ENOMEM;
1156 
1157 	data->max_size = RFKILL_EVENT_SIZE_V1;
1158 
1159 	INIT_LIST_HEAD(&data->events);
1160 	mutex_init(&data->mtx);
1161 	init_waitqueue_head(&data->read_wait);
1162 
1163 	mutex_lock(&rfkill_global_mutex);
1164 	mutex_lock(&data->mtx);
1165 	/*
1166 	 * start getting events from elsewhere but hold mtx to get
1167 	 * startup events added first
1168 	 */
1169 
1170 	list_for_each_entry(rfkill, &rfkill_list, node) {
1171 		ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1172 		if (!ev)
1173 			goto free;
1174 		rfkill_fill_event(&ev->ev, rfkill, RFKILL_OP_ADD);
1175 		list_add_tail(&ev->list, &data->events);
1176 	}
1177 	list_add(&data->list, &rfkill_fds);
1178 	mutex_unlock(&data->mtx);
1179 	mutex_unlock(&rfkill_global_mutex);
1180 
1181 	file->private_data = data;
1182 
1183 	return stream_open(inode, file);
1184 
1185  free:
1186 	mutex_unlock(&data->mtx);
1187 	mutex_unlock(&rfkill_global_mutex);
1188 	mutex_destroy(&data->mtx);
1189 	list_for_each_entry_safe(ev, tmp, &data->events, list)
1190 		kfree(ev);
1191 	kfree(data);
1192 	return -ENOMEM;
1193 }
1194 
1195 static __poll_t rfkill_fop_poll(struct file *file, poll_table *wait)
1196 {
1197 	struct rfkill_data *data = file->private_data;
1198 	__poll_t res = EPOLLOUT | EPOLLWRNORM;
1199 
1200 	poll_wait(file, &data->read_wait, wait);
1201 
1202 	mutex_lock(&data->mtx);
1203 	if (!list_empty(&data->events))
1204 		res = EPOLLIN | EPOLLRDNORM;
1205 	mutex_unlock(&data->mtx);
1206 
1207 	return res;
1208 }
1209 
1210 static ssize_t rfkill_fop_read(struct file *file, char __user *buf,
1211 			       size_t count, loff_t *pos)
1212 {
1213 	struct rfkill_data *data = file->private_data;
1214 	struct rfkill_int_event *ev;
1215 	unsigned long sz;
1216 	int ret;
1217 
1218 	mutex_lock(&data->mtx);
1219 
1220 	while (list_empty(&data->events)) {
1221 		if (file->f_flags & O_NONBLOCK) {
1222 			ret = -EAGAIN;
1223 			goto out;
1224 		}
1225 		mutex_unlock(&data->mtx);
1226 		/* since we re-check and it just compares pointers,
1227 		 * using !list_empty() without locking isn't a problem
1228 		 */
1229 		ret = wait_event_interruptible(data->read_wait,
1230 					       !list_empty(&data->events));
1231 		mutex_lock(&data->mtx);
1232 
1233 		if (ret)
1234 			goto out;
1235 	}
1236 
1237 	ev = list_first_entry(&data->events, struct rfkill_int_event,
1238 				list);
1239 
1240 	sz = min_t(unsigned long, sizeof(ev->ev), count);
1241 	sz = min_t(unsigned long, sz, data->max_size);
1242 	ret = sz;
1243 	if (copy_to_user(buf, &ev->ev, sz))
1244 		ret = -EFAULT;
1245 
1246 	list_del(&ev->list);
1247 	kfree(ev);
1248  out:
1249 	mutex_unlock(&data->mtx);
1250 	return ret;
1251 }
1252 
1253 static ssize_t rfkill_fop_write(struct file *file, const char __user *buf,
1254 				size_t count, loff_t *pos)
1255 {
1256 	struct rfkill_data *data = file->private_data;
1257 	struct rfkill *rfkill;
1258 	struct rfkill_event_ext ev;
1259 	int ret;
1260 
1261 	/* we don't need the 'hard' variable but accept it */
1262 	if (count < RFKILL_EVENT_SIZE_V1 - 1)
1263 		return -EINVAL;
1264 
1265 	/*
1266 	 * Copy as much data as we can accept into our 'ev' buffer,
1267 	 * but tell userspace how much we've copied so it can determine
1268 	 * our API version even in a write() call, if it cares.
1269 	 */
1270 	count = min(count, sizeof(ev));
1271 	count = min_t(size_t, count, data->max_size);
1272 	if (copy_from_user(&ev, buf, count))
1273 		return -EFAULT;
1274 
1275 	if (ev.type >= NUM_RFKILL_TYPES)
1276 		return -EINVAL;
1277 
1278 	mutex_lock(&rfkill_global_mutex);
1279 
1280 	switch (ev.op) {
1281 	case RFKILL_OP_CHANGE_ALL:
1282 		rfkill_update_global_state(ev.type, ev.soft);
1283 		list_for_each_entry(rfkill, &rfkill_list, node)
1284 			if (rfkill->type == ev.type ||
1285 			    ev.type == RFKILL_TYPE_ALL)
1286 				rfkill_set_block(rfkill, ev.soft);
1287 		ret = 0;
1288 		break;
1289 	case RFKILL_OP_CHANGE:
1290 		list_for_each_entry(rfkill, &rfkill_list, node)
1291 			if (rfkill->idx == ev.idx &&
1292 			    (rfkill->type == ev.type ||
1293 			     ev.type == RFKILL_TYPE_ALL))
1294 				rfkill_set_block(rfkill, ev.soft);
1295 		ret = 0;
1296 		break;
1297 	default:
1298 		ret = -EINVAL;
1299 		break;
1300 	}
1301 
1302 	mutex_unlock(&rfkill_global_mutex);
1303 
1304 	return ret ?: count;
1305 }
1306 
1307 static int rfkill_fop_release(struct inode *inode, struct file *file)
1308 {
1309 	struct rfkill_data *data = file->private_data;
1310 	struct rfkill_int_event *ev, *tmp;
1311 
1312 	mutex_lock(&rfkill_global_mutex);
1313 	list_del(&data->list);
1314 	mutex_unlock(&rfkill_global_mutex);
1315 
1316 	mutex_destroy(&data->mtx);
1317 	list_for_each_entry_safe(ev, tmp, &data->events, list)
1318 		kfree(ev);
1319 
1320 #ifdef CONFIG_RFKILL_INPUT
1321 	if (data->input_handler)
1322 		if (atomic_dec_return(&rfkill_input_disabled) == 0)
1323 			printk(KERN_DEBUG "rfkill: input handler enabled\n");
1324 #endif
1325 
1326 	kfree(data);
1327 
1328 	return 0;
1329 }
1330 
1331 static long rfkill_fop_ioctl(struct file *file, unsigned int cmd,
1332 			     unsigned long arg)
1333 {
1334 	struct rfkill_data *data = file->private_data;
1335 	int ret = -ENOSYS;
1336 	u32 size;
1337 
1338 	if (_IOC_TYPE(cmd) != RFKILL_IOC_MAGIC)
1339 		return -ENOSYS;
1340 
1341 	mutex_lock(&data->mtx);
1342 	switch (_IOC_NR(cmd)) {
1343 #ifdef CONFIG_RFKILL_INPUT
1344 	case RFKILL_IOC_NOINPUT:
1345 		if (!data->input_handler) {
1346 			if (atomic_inc_return(&rfkill_input_disabled) == 1)
1347 				printk(KERN_DEBUG "rfkill: input handler disabled\n");
1348 			data->input_handler = true;
1349 		}
1350 		ret = 0;
1351 		break;
1352 #endif
1353 	case RFKILL_IOC_MAX_SIZE:
1354 		if (get_user(size, (__u32 __user *)arg)) {
1355 			ret = -EFAULT;
1356 			break;
1357 		}
1358 		if (size < RFKILL_EVENT_SIZE_V1 || size > U8_MAX) {
1359 			ret = -EINVAL;
1360 			break;
1361 		}
1362 		data->max_size = size;
1363 		ret = 0;
1364 		break;
1365 	default:
1366 		break;
1367 	}
1368 	mutex_unlock(&data->mtx);
1369 
1370 	return ret;
1371 }
1372 
1373 static const struct file_operations rfkill_fops = {
1374 	.owner		= THIS_MODULE,
1375 	.open		= rfkill_fop_open,
1376 	.read		= rfkill_fop_read,
1377 	.write		= rfkill_fop_write,
1378 	.poll		= rfkill_fop_poll,
1379 	.release	= rfkill_fop_release,
1380 	.unlocked_ioctl	= rfkill_fop_ioctl,
1381 	.compat_ioctl	= compat_ptr_ioctl,
1382 	.llseek		= no_llseek,
1383 };
1384 
1385 #define RFKILL_NAME "rfkill"
1386 
1387 static struct miscdevice rfkill_miscdev = {
1388 	.fops	= &rfkill_fops,
1389 	.name	= RFKILL_NAME,
1390 	.minor	= RFKILL_MINOR,
1391 };
1392 
1393 static int __init rfkill_init(void)
1394 {
1395 	int error;
1396 
1397 	rfkill_update_global_state(RFKILL_TYPE_ALL, !rfkill_default_state);
1398 
1399 	error = class_register(&rfkill_class);
1400 	if (error)
1401 		goto error_class;
1402 
1403 	error = misc_register(&rfkill_miscdev);
1404 	if (error)
1405 		goto error_misc;
1406 
1407 	error = rfkill_global_led_trigger_register();
1408 	if (error)
1409 		goto error_led_trigger;
1410 
1411 #ifdef CONFIG_RFKILL_INPUT
1412 	error = rfkill_handler_init();
1413 	if (error)
1414 		goto error_input;
1415 #endif
1416 
1417 	return 0;
1418 
1419 #ifdef CONFIG_RFKILL_INPUT
1420 error_input:
1421 	rfkill_global_led_trigger_unregister();
1422 #endif
1423 error_led_trigger:
1424 	misc_deregister(&rfkill_miscdev);
1425 error_misc:
1426 	class_unregister(&rfkill_class);
1427 error_class:
1428 	return error;
1429 }
1430 subsys_initcall(rfkill_init);
1431 
1432 static void __exit rfkill_exit(void)
1433 {
1434 #ifdef CONFIG_RFKILL_INPUT
1435 	rfkill_handler_exit();
1436 #endif
1437 	rfkill_global_led_trigger_unregister();
1438 	misc_deregister(&rfkill_miscdev);
1439 	class_unregister(&rfkill_class);
1440 }
1441 module_exit(rfkill_exit);
1442 
1443 MODULE_ALIAS_MISCDEV(RFKILL_MINOR);
1444 MODULE_ALIAS("devname:" RFKILL_NAME);
1445