xref: /linux/net/rfkill/core.c (revision a33f32244d8550da8b4a26e277ce07d5c6d158b5)
1 /*
2  * Copyright (C) 2006 - 2007 Ivo van Doorn
3  * Copyright (C) 2007 Dmitry Torokhov
4  * Copyright 2009 Johannes Berg <johannes@sipsolutions.net>
5  *
6  * This program is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License as published by
8  * the Free Software Foundation; either version 2 of the License, or
9  * (at your option) any later version.
10  *
11  * This program is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14  * GNU General Public License for more details.
15  *
16  * You should have received a copy of the GNU General Public License
17  * along with this program; if not, write to the
18  * Free Software Foundation, Inc.,
19  * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
20  */
21 
22 #include <linux/kernel.h>
23 #include <linux/module.h>
24 #include <linux/init.h>
25 #include <linux/workqueue.h>
26 #include <linux/capability.h>
27 #include <linux/list.h>
28 #include <linux/mutex.h>
29 #include <linux/rfkill.h>
30 #include <linux/sched.h>
31 #include <linux/spinlock.h>
32 #include <linux/miscdevice.h>
33 #include <linux/wait.h>
34 #include <linux/poll.h>
35 #include <linux/fs.h>
36 #include <linux/slab.h>
37 
38 #include "rfkill.h"
39 
40 #define POLL_INTERVAL		(5 * HZ)
41 
42 #define RFKILL_BLOCK_HW		BIT(0)
43 #define RFKILL_BLOCK_SW		BIT(1)
44 #define RFKILL_BLOCK_SW_PREV	BIT(2)
45 #define RFKILL_BLOCK_ANY	(RFKILL_BLOCK_HW |\
46 				 RFKILL_BLOCK_SW |\
47 				 RFKILL_BLOCK_SW_PREV)
48 #define RFKILL_BLOCK_SW_SETCALL	BIT(31)
49 
50 struct rfkill {
51 	spinlock_t		lock;
52 
53 	const char		*name;
54 	enum rfkill_type	type;
55 
56 	unsigned long		state;
57 
58 	u32			idx;
59 
60 	bool			registered;
61 	bool			persistent;
62 
63 	const struct rfkill_ops	*ops;
64 	void			*data;
65 
66 #ifdef CONFIG_RFKILL_LEDS
67 	struct led_trigger	led_trigger;
68 	const char		*ledtrigname;
69 #endif
70 
71 	struct device		dev;
72 	struct list_head	node;
73 
74 	struct delayed_work	poll_work;
75 	struct work_struct	uevent_work;
76 	struct work_struct	sync_work;
77 };
78 #define to_rfkill(d)	container_of(d, struct rfkill, dev)
79 
80 struct rfkill_int_event {
81 	struct list_head	list;
82 	struct rfkill_event	ev;
83 };
84 
85 struct rfkill_data {
86 	struct list_head	list;
87 	struct list_head	events;
88 	struct mutex		mtx;
89 	wait_queue_head_t	read_wait;
90 	bool			input_handler;
91 };
92 
93 
94 MODULE_AUTHOR("Ivo van Doorn <IvDoorn@gmail.com>");
95 MODULE_AUTHOR("Johannes Berg <johannes@sipsolutions.net>");
96 MODULE_DESCRIPTION("RF switch support");
97 MODULE_LICENSE("GPL");
98 
99 
100 /*
101  * The locking here should be made much smarter, we currently have
102  * a bit of a stupid situation because drivers might want to register
103  * the rfkill struct under their own lock, and take this lock during
104  * rfkill method calls -- which will cause an AB-BA deadlock situation.
105  *
106  * To fix that, we need to rework this code here to be mostly lock-free
107  * and only use the mutex for list manipulations, not to protect the
108  * various other global variables. Then we can avoid holding the mutex
109  * around driver operations, and all is happy.
110  */
111 static LIST_HEAD(rfkill_list);	/* list of registered rf switches */
112 static DEFINE_MUTEX(rfkill_global_mutex);
113 static LIST_HEAD(rfkill_fds);	/* list of open fds of /dev/rfkill */
114 
115 static unsigned int rfkill_default_state = 1;
116 module_param_named(default_state, rfkill_default_state, uint, 0444);
117 MODULE_PARM_DESC(default_state,
118 		 "Default initial state for all radio types, 0 = radio off");
119 
120 static struct {
121 	bool cur, sav;
122 } rfkill_global_states[NUM_RFKILL_TYPES];
123 
124 static bool rfkill_epo_lock_active;
125 
126 
127 #ifdef CONFIG_RFKILL_LEDS
128 static void rfkill_led_trigger_event(struct rfkill *rfkill)
129 {
130 	struct led_trigger *trigger;
131 
132 	if (!rfkill->registered)
133 		return;
134 
135 	trigger = &rfkill->led_trigger;
136 
137 	if (rfkill->state & RFKILL_BLOCK_ANY)
138 		led_trigger_event(trigger, LED_OFF);
139 	else
140 		led_trigger_event(trigger, LED_FULL);
141 }
142 
143 static void rfkill_led_trigger_activate(struct led_classdev *led)
144 {
145 	struct rfkill *rfkill;
146 
147 	rfkill = container_of(led->trigger, struct rfkill, led_trigger);
148 
149 	rfkill_led_trigger_event(rfkill);
150 }
151 
152 const char *rfkill_get_led_trigger_name(struct rfkill *rfkill)
153 {
154 	return rfkill->led_trigger.name;
155 }
156 EXPORT_SYMBOL(rfkill_get_led_trigger_name);
157 
158 void rfkill_set_led_trigger_name(struct rfkill *rfkill, const char *name)
159 {
160 	BUG_ON(!rfkill);
161 
162 	rfkill->ledtrigname = name;
163 }
164 EXPORT_SYMBOL(rfkill_set_led_trigger_name);
165 
166 static int rfkill_led_trigger_register(struct rfkill *rfkill)
167 {
168 	rfkill->led_trigger.name = rfkill->ledtrigname
169 					? : dev_name(&rfkill->dev);
170 	rfkill->led_trigger.activate = rfkill_led_trigger_activate;
171 	return led_trigger_register(&rfkill->led_trigger);
172 }
173 
174 static void rfkill_led_trigger_unregister(struct rfkill *rfkill)
175 {
176 	led_trigger_unregister(&rfkill->led_trigger);
177 }
178 #else
179 static void rfkill_led_trigger_event(struct rfkill *rfkill)
180 {
181 }
182 
183 static inline int rfkill_led_trigger_register(struct rfkill *rfkill)
184 {
185 	return 0;
186 }
187 
188 static inline void rfkill_led_trigger_unregister(struct rfkill *rfkill)
189 {
190 }
191 #endif /* CONFIG_RFKILL_LEDS */
192 
193 static void rfkill_fill_event(struct rfkill_event *ev, struct rfkill *rfkill,
194 			      enum rfkill_operation op)
195 {
196 	unsigned long flags;
197 
198 	ev->idx = rfkill->idx;
199 	ev->type = rfkill->type;
200 	ev->op = op;
201 
202 	spin_lock_irqsave(&rfkill->lock, flags);
203 	ev->hard = !!(rfkill->state & RFKILL_BLOCK_HW);
204 	ev->soft = !!(rfkill->state & (RFKILL_BLOCK_SW |
205 					RFKILL_BLOCK_SW_PREV));
206 	spin_unlock_irqrestore(&rfkill->lock, flags);
207 }
208 
209 static void rfkill_send_events(struct rfkill *rfkill, enum rfkill_operation op)
210 {
211 	struct rfkill_data *data;
212 	struct rfkill_int_event *ev;
213 
214 	list_for_each_entry(data, &rfkill_fds, list) {
215 		ev = kzalloc(sizeof(*ev), GFP_KERNEL);
216 		if (!ev)
217 			continue;
218 		rfkill_fill_event(&ev->ev, rfkill, op);
219 		mutex_lock(&data->mtx);
220 		list_add_tail(&ev->list, &data->events);
221 		mutex_unlock(&data->mtx);
222 		wake_up_interruptible(&data->read_wait);
223 	}
224 }
225 
226 static void rfkill_event(struct rfkill *rfkill)
227 {
228 	if (!rfkill->registered)
229 		return;
230 
231 	kobject_uevent(&rfkill->dev.kobj, KOBJ_CHANGE);
232 
233 	/* also send event to /dev/rfkill */
234 	rfkill_send_events(rfkill, RFKILL_OP_CHANGE);
235 }
236 
237 static bool __rfkill_set_hw_state(struct rfkill *rfkill,
238 				  bool blocked, bool *change)
239 {
240 	unsigned long flags;
241 	bool prev, any;
242 
243 	BUG_ON(!rfkill);
244 
245 	spin_lock_irqsave(&rfkill->lock, flags);
246 	prev = !!(rfkill->state & RFKILL_BLOCK_HW);
247 	if (blocked)
248 		rfkill->state |= RFKILL_BLOCK_HW;
249 	else
250 		rfkill->state &= ~RFKILL_BLOCK_HW;
251 	*change = prev != blocked;
252 	any = rfkill->state & RFKILL_BLOCK_ANY;
253 	spin_unlock_irqrestore(&rfkill->lock, flags);
254 
255 	rfkill_led_trigger_event(rfkill);
256 
257 	return any;
258 }
259 
260 /**
261  * rfkill_set_block - wrapper for set_block method
262  *
263  * @rfkill: the rfkill struct to use
264  * @blocked: the new software state
265  *
266  * Calls the set_block method (when applicable) and handles notifications
267  * etc. as well.
268  */
269 static void rfkill_set_block(struct rfkill *rfkill, bool blocked)
270 {
271 	unsigned long flags;
272 	int err;
273 
274 	if (unlikely(rfkill->dev.power.power_state.event & PM_EVENT_SLEEP))
275 		return;
276 
277 	/*
278 	 * Some platforms (...!) generate input events which affect the
279 	 * _hard_ kill state -- whenever something tries to change the
280 	 * current software state query the hardware state too.
281 	 */
282 	if (rfkill->ops->query)
283 		rfkill->ops->query(rfkill, rfkill->data);
284 
285 	spin_lock_irqsave(&rfkill->lock, flags);
286 	if (rfkill->state & RFKILL_BLOCK_SW)
287 		rfkill->state |= RFKILL_BLOCK_SW_PREV;
288 	else
289 		rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
290 
291 	if (blocked)
292 		rfkill->state |= RFKILL_BLOCK_SW;
293 	else
294 		rfkill->state &= ~RFKILL_BLOCK_SW;
295 
296 	rfkill->state |= RFKILL_BLOCK_SW_SETCALL;
297 	spin_unlock_irqrestore(&rfkill->lock, flags);
298 
299 	err = rfkill->ops->set_block(rfkill->data, blocked);
300 
301 	spin_lock_irqsave(&rfkill->lock, flags);
302 	if (err) {
303 		/*
304 		 * Failed -- reset status to _prev, this may be different
305 		 * from what set set _PREV to earlier in this function
306 		 * if rfkill_set_sw_state was invoked.
307 		 */
308 		if (rfkill->state & RFKILL_BLOCK_SW_PREV)
309 			rfkill->state |= RFKILL_BLOCK_SW;
310 		else
311 			rfkill->state &= ~RFKILL_BLOCK_SW;
312 	}
313 	rfkill->state &= ~RFKILL_BLOCK_SW_SETCALL;
314 	rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
315 	spin_unlock_irqrestore(&rfkill->lock, flags);
316 
317 	rfkill_led_trigger_event(rfkill);
318 	rfkill_event(rfkill);
319 }
320 
321 #ifdef CONFIG_RFKILL_INPUT
322 static atomic_t rfkill_input_disabled = ATOMIC_INIT(0);
323 
324 /**
325  * __rfkill_switch_all - Toggle state of all switches of given type
326  * @type: type of interfaces to be affected
327  * @state: the new state
328  *
329  * This function sets the state of all switches of given type,
330  * unless a specific switch is claimed by userspace (in which case,
331  * that switch is left alone) or suspended.
332  *
333  * Caller must have acquired rfkill_global_mutex.
334  */
335 static void __rfkill_switch_all(const enum rfkill_type type, bool blocked)
336 {
337 	struct rfkill *rfkill;
338 
339 	rfkill_global_states[type].cur = blocked;
340 	list_for_each_entry(rfkill, &rfkill_list, node) {
341 		if (rfkill->type != type)
342 			continue;
343 
344 		rfkill_set_block(rfkill, blocked);
345 	}
346 }
347 
348 /**
349  * rfkill_switch_all - Toggle state of all switches of given type
350  * @type: type of interfaces to be affected
351  * @state: the new state
352  *
353  * Acquires rfkill_global_mutex and calls __rfkill_switch_all(@type, @state).
354  * Please refer to __rfkill_switch_all() for details.
355  *
356  * Does nothing if the EPO lock is active.
357  */
358 void rfkill_switch_all(enum rfkill_type type, bool blocked)
359 {
360 	if (atomic_read(&rfkill_input_disabled))
361 		return;
362 
363 	mutex_lock(&rfkill_global_mutex);
364 
365 	if (!rfkill_epo_lock_active)
366 		__rfkill_switch_all(type, blocked);
367 
368 	mutex_unlock(&rfkill_global_mutex);
369 }
370 
371 /**
372  * rfkill_epo - emergency power off all transmitters
373  *
374  * This kicks all non-suspended rfkill devices to RFKILL_STATE_SOFT_BLOCKED,
375  * ignoring everything in its path but rfkill_global_mutex and rfkill->mutex.
376  *
377  * The global state before the EPO is saved and can be restored later
378  * using rfkill_restore_states().
379  */
380 void rfkill_epo(void)
381 {
382 	struct rfkill *rfkill;
383 	int i;
384 
385 	if (atomic_read(&rfkill_input_disabled))
386 		return;
387 
388 	mutex_lock(&rfkill_global_mutex);
389 
390 	rfkill_epo_lock_active = true;
391 	list_for_each_entry(rfkill, &rfkill_list, node)
392 		rfkill_set_block(rfkill, true);
393 
394 	for (i = 0; i < NUM_RFKILL_TYPES; i++) {
395 		rfkill_global_states[i].sav = rfkill_global_states[i].cur;
396 		rfkill_global_states[i].cur = true;
397 	}
398 
399 	mutex_unlock(&rfkill_global_mutex);
400 }
401 
402 /**
403  * rfkill_restore_states - restore global states
404  *
405  * Restore (and sync switches to) the global state from the
406  * states in rfkill_default_states.  This can undo the effects of
407  * a call to rfkill_epo().
408  */
409 void rfkill_restore_states(void)
410 {
411 	int i;
412 
413 	if (atomic_read(&rfkill_input_disabled))
414 		return;
415 
416 	mutex_lock(&rfkill_global_mutex);
417 
418 	rfkill_epo_lock_active = false;
419 	for (i = 0; i < NUM_RFKILL_TYPES; i++)
420 		__rfkill_switch_all(i, rfkill_global_states[i].sav);
421 	mutex_unlock(&rfkill_global_mutex);
422 }
423 
424 /**
425  * rfkill_remove_epo_lock - unlock state changes
426  *
427  * Used by rfkill-input manually unlock state changes, when
428  * the EPO switch is deactivated.
429  */
430 void rfkill_remove_epo_lock(void)
431 {
432 	if (atomic_read(&rfkill_input_disabled))
433 		return;
434 
435 	mutex_lock(&rfkill_global_mutex);
436 	rfkill_epo_lock_active = false;
437 	mutex_unlock(&rfkill_global_mutex);
438 }
439 
440 /**
441  * rfkill_is_epo_lock_active - returns true EPO is active
442  *
443  * Returns 0 (false) if there is NOT an active EPO contidion,
444  * and 1 (true) if there is an active EPO contition, which
445  * locks all radios in one of the BLOCKED states.
446  *
447  * Can be called in atomic context.
448  */
449 bool rfkill_is_epo_lock_active(void)
450 {
451 	return rfkill_epo_lock_active;
452 }
453 
454 /**
455  * rfkill_get_global_sw_state - returns global state for a type
456  * @type: the type to get the global state of
457  *
458  * Returns the current global state for a given wireless
459  * device type.
460  */
461 bool rfkill_get_global_sw_state(const enum rfkill_type type)
462 {
463 	return rfkill_global_states[type].cur;
464 }
465 #endif
466 
467 
468 bool rfkill_set_hw_state(struct rfkill *rfkill, bool blocked)
469 {
470 	bool ret, change;
471 
472 	ret = __rfkill_set_hw_state(rfkill, blocked, &change);
473 
474 	if (!rfkill->registered)
475 		return ret;
476 
477 	if (change)
478 		schedule_work(&rfkill->uevent_work);
479 
480 	return ret;
481 }
482 EXPORT_SYMBOL(rfkill_set_hw_state);
483 
484 static void __rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
485 {
486 	u32 bit = RFKILL_BLOCK_SW;
487 
488 	/* if in a ops->set_block right now, use other bit */
489 	if (rfkill->state & RFKILL_BLOCK_SW_SETCALL)
490 		bit = RFKILL_BLOCK_SW_PREV;
491 
492 	if (blocked)
493 		rfkill->state |= bit;
494 	else
495 		rfkill->state &= ~bit;
496 }
497 
498 bool rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
499 {
500 	unsigned long flags;
501 	bool prev, hwblock;
502 
503 	BUG_ON(!rfkill);
504 
505 	spin_lock_irqsave(&rfkill->lock, flags);
506 	prev = !!(rfkill->state & RFKILL_BLOCK_SW);
507 	__rfkill_set_sw_state(rfkill, blocked);
508 	hwblock = !!(rfkill->state & RFKILL_BLOCK_HW);
509 	blocked = blocked || hwblock;
510 	spin_unlock_irqrestore(&rfkill->lock, flags);
511 
512 	if (!rfkill->registered)
513 		return blocked;
514 
515 	if (prev != blocked && !hwblock)
516 		schedule_work(&rfkill->uevent_work);
517 
518 	rfkill_led_trigger_event(rfkill);
519 
520 	return blocked;
521 }
522 EXPORT_SYMBOL(rfkill_set_sw_state);
523 
524 void rfkill_init_sw_state(struct rfkill *rfkill, bool blocked)
525 {
526 	unsigned long flags;
527 
528 	BUG_ON(!rfkill);
529 	BUG_ON(rfkill->registered);
530 
531 	spin_lock_irqsave(&rfkill->lock, flags);
532 	__rfkill_set_sw_state(rfkill, blocked);
533 	rfkill->persistent = true;
534 	spin_unlock_irqrestore(&rfkill->lock, flags);
535 }
536 EXPORT_SYMBOL(rfkill_init_sw_state);
537 
538 void rfkill_set_states(struct rfkill *rfkill, bool sw, bool hw)
539 {
540 	unsigned long flags;
541 	bool swprev, hwprev;
542 
543 	BUG_ON(!rfkill);
544 
545 	spin_lock_irqsave(&rfkill->lock, flags);
546 
547 	/*
548 	 * No need to care about prev/setblock ... this is for uevent only
549 	 * and that will get triggered by rfkill_set_block anyway.
550 	 */
551 	swprev = !!(rfkill->state & RFKILL_BLOCK_SW);
552 	hwprev = !!(rfkill->state & RFKILL_BLOCK_HW);
553 	__rfkill_set_sw_state(rfkill, sw);
554 	if (hw)
555 		rfkill->state |= RFKILL_BLOCK_HW;
556 	else
557 		rfkill->state &= ~RFKILL_BLOCK_HW;
558 
559 	spin_unlock_irqrestore(&rfkill->lock, flags);
560 
561 	if (!rfkill->registered) {
562 		rfkill->persistent = true;
563 	} else {
564 		if (swprev != sw || hwprev != hw)
565 			schedule_work(&rfkill->uevent_work);
566 
567 		rfkill_led_trigger_event(rfkill);
568 	}
569 }
570 EXPORT_SYMBOL(rfkill_set_states);
571 
572 static ssize_t rfkill_name_show(struct device *dev,
573 				struct device_attribute *attr,
574 				char *buf)
575 {
576 	struct rfkill *rfkill = to_rfkill(dev);
577 
578 	return sprintf(buf, "%s\n", rfkill->name);
579 }
580 
581 static const char *rfkill_get_type_str(enum rfkill_type type)
582 {
583 	BUILD_BUG_ON(NUM_RFKILL_TYPES != RFKILL_TYPE_FM + 1);
584 
585 	switch (type) {
586 	case RFKILL_TYPE_WLAN:
587 		return "wlan";
588 	case RFKILL_TYPE_BLUETOOTH:
589 		return "bluetooth";
590 	case RFKILL_TYPE_UWB:
591 		return "ultrawideband";
592 	case RFKILL_TYPE_WIMAX:
593 		return "wimax";
594 	case RFKILL_TYPE_WWAN:
595 		return "wwan";
596 	case RFKILL_TYPE_GPS:
597 		return "gps";
598 	case RFKILL_TYPE_FM:
599 		return "fm";
600 	default:
601 		BUG();
602 	}
603 }
604 
605 static ssize_t rfkill_type_show(struct device *dev,
606 				struct device_attribute *attr,
607 				char *buf)
608 {
609 	struct rfkill *rfkill = to_rfkill(dev);
610 
611 	return sprintf(buf, "%s\n", rfkill_get_type_str(rfkill->type));
612 }
613 
614 static ssize_t rfkill_idx_show(struct device *dev,
615 			       struct device_attribute *attr,
616 			       char *buf)
617 {
618 	struct rfkill *rfkill = to_rfkill(dev);
619 
620 	return sprintf(buf, "%d\n", rfkill->idx);
621 }
622 
623 static ssize_t rfkill_persistent_show(struct device *dev,
624 			       struct device_attribute *attr,
625 			       char *buf)
626 {
627 	struct rfkill *rfkill = to_rfkill(dev);
628 
629 	return sprintf(buf, "%d\n", rfkill->persistent);
630 }
631 
632 static u8 user_state_from_blocked(unsigned long state)
633 {
634 	if (state & RFKILL_BLOCK_HW)
635 		return RFKILL_USER_STATE_HARD_BLOCKED;
636 	if (state & RFKILL_BLOCK_SW)
637 		return RFKILL_USER_STATE_SOFT_BLOCKED;
638 
639 	return RFKILL_USER_STATE_UNBLOCKED;
640 }
641 
642 static ssize_t rfkill_state_show(struct device *dev,
643 				 struct device_attribute *attr,
644 				 char *buf)
645 {
646 	struct rfkill *rfkill = to_rfkill(dev);
647 	unsigned long flags;
648 	u32 state;
649 
650 	spin_lock_irqsave(&rfkill->lock, flags);
651 	state = rfkill->state;
652 	spin_unlock_irqrestore(&rfkill->lock, flags);
653 
654 	return sprintf(buf, "%d\n", user_state_from_blocked(state));
655 }
656 
657 static ssize_t rfkill_state_store(struct device *dev,
658 				  struct device_attribute *attr,
659 				  const char *buf, size_t count)
660 {
661 	struct rfkill *rfkill = to_rfkill(dev);
662 	unsigned long state;
663 	int err;
664 
665 	if (!capable(CAP_NET_ADMIN))
666 		return -EPERM;
667 
668 	err = strict_strtoul(buf, 0, &state);
669 	if (err)
670 		return err;
671 
672 	if (state != RFKILL_USER_STATE_SOFT_BLOCKED &&
673 	    state != RFKILL_USER_STATE_UNBLOCKED)
674 		return -EINVAL;
675 
676 	mutex_lock(&rfkill_global_mutex);
677 	rfkill_set_block(rfkill, state == RFKILL_USER_STATE_SOFT_BLOCKED);
678 	mutex_unlock(&rfkill_global_mutex);
679 
680 	return err ?: count;
681 }
682 
683 static ssize_t rfkill_claim_show(struct device *dev,
684 				 struct device_attribute *attr,
685 				 char *buf)
686 {
687 	return sprintf(buf, "%d\n", 0);
688 }
689 
690 static ssize_t rfkill_claim_store(struct device *dev,
691 				  struct device_attribute *attr,
692 				  const char *buf, size_t count)
693 {
694 	return -EOPNOTSUPP;
695 }
696 
697 static struct device_attribute rfkill_dev_attrs[] = {
698 	__ATTR(name, S_IRUGO, rfkill_name_show, NULL),
699 	__ATTR(type, S_IRUGO, rfkill_type_show, NULL),
700 	__ATTR(index, S_IRUGO, rfkill_idx_show, NULL),
701 	__ATTR(persistent, S_IRUGO, rfkill_persistent_show, NULL),
702 	__ATTR(state, S_IRUGO|S_IWUSR, rfkill_state_show, rfkill_state_store),
703 	__ATTR(claim, S_IRUGO|S_IWUSR, rfkill_claim_show, rfkill_claim_store),
704 	__ATTR_NULL
705 };
706 
707 static void rfkill_release(struct device *dev)
708 {
709 	struct rfkill *rfkill = to_rfkill(dev);
710 
711 	kfree(rfkill);
712 }
713 
714 static int rfkill_dev_uevent(struct device *dev, struct kobj_uevent_env *env)
715 {
716 	struct rfkill *rfkill = to_rfkill(dev);
717 	unsigned long flags;
718 	u32 state;
719 	int error;
720 
721 	error = add_uevent_var(env, "RFKILL_NAME=%s", rfkill->name);
722 	if (error)
723 		return error;
724 	error = add_uevent_var(env, "RFKILL_TYPE=%s",
725 			       rfkill_get_type_str(rfkill->type));
726 	if (error)
727 		return error;
728 	spin_lock_irqsave(&rfkill->lock, flags);
729 	state = rfkill->state;
730 	spin_unlock_irqrestore(&rfkill->lock, flags);
731 	error = add_uevent_var(env, "RFKILL_STATE=%d",
732 			       user_state_from_blocked(state));
733 	return error;
734 }
735 
736 void rfkill_pause_polling(struct rfkill *rfkill)
737 {
738 	BUG_ON(!rfkill);
739 
740 	if (!rfkill->ops->poll)
741 		return;
742 
743 	cancel_delayed_work_sync(&rfkill->poll_work);
744 }
745 EXPORT_SYMBOL(rfkill_pause_polling);
746 
747 void rfkill_resume_polling(struct rfkill *rfkill)
748 {
749 	BUG_ON(!rfkill);
750 
751 	if (!rfkill->ops->poll)
752 		return;
753 
754 	schedule_work(&rfkill->poll_work.work);
755 }
756 EXPORT_SYMBOL(rfkill_resume_polling);
757 
758 static int rfkill_suspend(struct device *dev, pm_message_t state)
759 {
760 	struct rfkill *rfkill = to_rfkill(dev);
761 
762 	rfkill_pause_polling(rfkill);
763 
764 	return 0;
765 }
766 
767 static int rfkill_resume(struct device *dev)
768 {
769 	struct rfkill *rfkill = to_rfkill(dev);
770 	bool cur;
771 
772 	if (!rfkill->persistent) {
773 		cur = !!(rfkill->state & RFKILL_BLOCK_SW);
774 		rfkill_set_block(rfkill, cur);
775 	}
776 
777 	rfkill_resume_polling(rfkill);
778 
779 	return 0;
780 }
781 
782 static struct class rfkill_class = {
783 	.name		= "rfkill",
784 	.dev_release	= rfkill_release,
785 	.dev_attrs	= rfkill_dev_attrs,
786 	.dev_uevent	= rfkill_dev_uevent,
787 	.suspend	= rfkill_suspend,
788 	.resume		= rfkill_resume,
789 };
790 
791 bool rfkill_blocked(struct rfkill *rfkill)
792 {
793 	unsigned long flags;
794 	u32 state;
795 
796 	spin_lock_irqsave(&rfkill->lock, flags);
797 	state = rfkill->state;
798 	spin_unlock_irqrestore(&rfkill->lock, flags);
799 
800 	return !!(state & RFKILL_BLOCK_ANY);
801 }
802 EXPORT_SYMBOL(rfkill_blocked);
803 
804 
805 struct rfkill * __must_check rfkill_alloc(const char *name,
806 					  struct device *parent,
807 					  const enum rfkill_type type,
808 					  const struct rfkill_ops *ops,
809 					  void *ops_data)
810 {
811 	struct rfkill *rfkill;
812 	struct device *dev;
813 
814 	if (WARN_ON(!ops))
815 		return NULL;
816 
817 	if (WARN_ON(!ops->set_block))
818 		return NULL;
819 
820 	if (WARN_ON(!name))
821 		return NULL;
822 
823 	if (WARN_ON(type == RFKILL_TYPE_ALL || type >= NUM_RFKILL_TYPES))
824 		return NULL;
825 
826 	rfkill = kzalloc(sizeof(*rfkill), GFP_KERNEL);
827 	if (!rfkill)
828 		return NULL;
829 
830 	spin_lock_init(&rfkill->lock);
831 	INIT_LIST_HEAD(&rfkill->node);
832 	rfkill->type = type;
833 	rfkill->name = name;
834 	rfkill->ops = ops;
835 	rfkill->data = ops_data;
836 
837 	dev = &rfkill->dev;
838 	dev->class = &rfkill_class;
839 	dev->parent = parent;
840 	device_initialize(dev);
841 
842 	return rfkill;
843 }
844 EXPORT_SYMBOL(rfkill_alloc);
845 
846 static void rfkill_poll(struct work_struct *work)
847 {
848 	struct rfkill *rfkill;
849 
850 	rfkill = container_of(work, struct rfkill, poll_work.work);
851 
852 	/*
853 	 * Poll hardware state -- driver will use one of the
854 	 * rfkill_set{,_hw,_sw}_state functions and use its
855 	 * return value to update the current status.
856 	 */
857 	rfkill->ops->poll(rfkill, rfkill->data);
858 
859 	schedule_delayed_work(&rfkill->poll_work,
860 		round_jiffies_relative(POLL_INTERVAL));
861 }
862 
863 static void rfkill_uevent_work(struct work_struct *work)
864 {
865 	struct rfkill *rfkill;
866 
867 	rfkill = container_of(work, struct rfkill, uevent_work);
868 
869 	mutex_lock(&rfkill_global_mutex);
870 	rfkill_event(rfkill);
871 	mutex_unlock(&rfkill_global_mutex);
872 }
873 
874 static void rfkill_sync_work(struct work_struct *work)
875 {
876 	struct rfkill *rfkill;
877 	bool cur;
878 
879 	rfkill = container_of(work, struct rfkill, sync_work);
880 
881 	mutex_lock(&rfkill_global_mutex);
882 	cur = rfkill_global_states[rfkill->type].cur;
883 	rfkill_set_block(rfkill, cur);
884 	mutex_unlock(&rfkill_global_mutex);
885 }
886 
887 int __must_check rfkill_register(struct rfkill *rfkill)
888 {
889 	static unsigned long rfkill_no;
890 	struct device *dev = &rfkill->dev;
891 	int error;
892 
893 	BUG_ON(!rfkill);
894 
895 	mutex_lock(&rfkill_global_mutex);
896 
897 	if (rfkill->registered) {
898 		error = -EALREADY;
899 		goto unlock;
900 	}
901 
902 	rfkill->idx = rfkill_no;
903 	dev_set_name(dev, "rfkill%lu", rfkill_no);
904 	rfkill_no++;
905 
906 	list_add_tail(&rfkill->node, &rfkill_list);
907 
908 	error = device_add(dev);
909 	if (error)
910 		goto remove;
911 
912 	error = rfkill_led_trigger_register(rfkill);
913 	if (error)
914 		goto devdel;
915 
916 	rfkill->registered = true;
917 
918 	INIT_DELAYED_WORK(&rfkill->poll_work, rfkill_poll);
919 	INIT_WORK(&rfkill->uevent_work, rfkill_uevent_work);
920 	INIT_WORK(&rfkill->sync_work, rfkill_sync_work);
921 
922 	if (rfkill->ops->poll)
923 		schedule_delayed_work(&rfkill->poll_work,
924 			round_jiffies_relative(POLL_INTERVAL));
925 
926 	if (!rfkill->persistent || rfkill_epo_lock_active) {
927 		schedule_work(&rfkill->sync_work);
928 	} else {
929 #ifdef CONFIG_RFKILL_INPUT
930 		bool soft_blocked = !!(rfkill->state & RFKILL_BLOCK_SW);
931 
932 		if (!atomic_read(&rfkill_input_disabled))
933 			__rfkill_switch_all(rfkill->type, soft_blocked);
934 #endif
935 	}
936 
937 	rfkill_send_events(rfkill, RFKILL_OP_ADD);
938 
939 	mutex_unlock(&rfkill_global_mutex);
940 	return 0;
941 
942  devdel:
943 	device_del(&rfkill->dev);
944  remove:
945 	list_del_init(&rfkill->node);
946  unlock:
947 	mutex_unlock(&rfkill_global_mutex);
948 	return error;
949 }
950 EXPORT_SYMBOL(rfkill_register);
951 
952 void rfkill_unregister(struct rfkill *rfkill)
953 {
954 	BUG_ON(!rfkill);
955 
956 	if (rfkill->ops->poll)
957 		cancel_delayed_work_sync(&rfkill->poll_work);
958 
959 	cancel_work_sync(&rfkill->uevent_work);
960 	cancel_work_sync(&rfkill->sync_work);
961 
962 	rfkill->registered = false;
963 
964 	device_del(&rfkill->dev);
965 
966 	mutex_lock(&rfkill_global_mutex);
967 	rfkill_send_events(rfkill, RFKILL_OP_DEL);
968 	list_del_init(&rfkill->node);
969 	mutex_unlock(&rfkill_global_mutex);
970 
971 	rfkill_led_trigger_unregister(rfkill);
972 }
973 EXPORT_SYMBOL(rfkill_unregister);
974 
975 void rfkill_destroy(struct rfkill *rfkill)
976 {
977 	if (rfkill)
978 		put_device(&rfkill->dev);
979 }
980 EXPORT_SYMBOL(rfkill_destroy);
981 
982 static int rfkill_fop_open(struct inode *inode, struct file *file)
983 {
984 	struct rfkill_data *data;
985 	struct rfkill *rfkill;
986 	struct rfkill_int_event *ev, *tmp;
987 
988 	data = kzalloc(sizeof(*data), GFP_KERNEL);
989 	if (!data)
990 		return -ENOMEM;
991 
992 	INIT_LIST_HEAD(&data->events);
993 	mutex_init(&data->mtx);
994 	init_waitqueue_head(&data->read_wait);
995 
996 	mutex_lock(&rfkill_global_mutex);
997 	mutex_lock(&data->mtx);
998 	/*
999 	 * start getting events from elsewhere but hold mtx to get
1000 	 * startup events added first
1001 	 */
1002 	list_add(&data->list, &rfkill_fds);
1003 
1004 	list_for_each_entry(rfkill, &rfkill_list, node) {
1005 		ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1006 		if (!ev)
1007 			goto free;
1008 		rfkill_fill_event(&ev->ev, rfkill, RFKILL_OP_ADD);
1009 		list_add_tail(&ev->list, &data->events);
1010 	}
1011 	mutex_unlock(&data->mtx);
1012 	mutex_unlock(&rfkill_global_mutex);
1013 
1014 	file->private_data = data;
1015 
1016 	return nonseekable_open(inode, file);
1017 
1018  free:
1019 	mutex_unlock(&data->mtx);
1020 	mutex_unlock(&rfkill_global_mutex);
1021 	mutex_destroy(&data->mtx);
1022 	list_for_each_entry_safe(ev, tmp, &data->events, list)
1023 		kfree(ev);
1024 	kfree(data);
1025 	return -ENOMEM;
1026 }
1027 
1028 static unsigned int rfkill_fop_poll(struct file *file, poll_table *wait)
1029 {
1030 	struct rfkill_data *data = file->private_data;
1031 	unsigned int res = POLLOUT | POLLWRNORM;
1032 
1033 	poll_wait(file, &data->read_wait, wait);
1034 
1035 	mutex_lock(&data->mtx);
1036 	if (!list_empty(&data->events))
1037 		res = POLLIN | POLLRDNORM;
1038 	mutex_unlock(&data->mtx);
1039 
1040 	return res;
1041 }
1042 
1043 static bool rfkill_readable(struct rfkill_data *data)
1044 {
1045 	bool r;
1046 
1047 	mutex_lock(&data->mtx);
1048 	r = !list_empty(&data->events);
1049 	mutex_unlock(&data->mtx);
1050 
1051 	return r;
1052 }
1053 
1054 static ssize_t rfkill_fop_read(struct file *file, char __user *buf,
1055 			       size_t count, loff_t *pos)
1056 {
1057 	struct rfkill_data *data = file->private_data;
1058 	struct rfkill_int_event *ev;
1059 	unsigned long sz;
1060 	int ret;
1061 
1062 	mutex_lock(&data->mtx);
1063 
1064 	while (list_empty(&data->events)) {
1065 		if (file->f_flags & O_NONBLOCK) {
1066 			ret = -EAGAIN;
1067 			goto out;
1068 		}
1069 		mutex_unlock(&data->mtx);
1070 		ret = wait_event_interruptible(data->read_wait,
1071 					       rfkill_readable(data));
1072 		mutex_lock(&data->mtx);
1073 
1074 		if (ret)
1075 			goto out;
1076 	}
1077 
1078 	ev = list_first_entry(&data->events, struct rfkill_int_event,
1079 				list);
1080 
1081 	sz = min_t(unsigned long, sizeof(ev->ev), count);
1082 	ret = sz;
1083 	if (copy_to_user(buf, &ev->ev, sz))
1084 		ret = -EFAULT;
1085 
1086 	list_del(&ev->list);
1087 	kfree(ev);
1088  out:
1089 	mutex_unlock(&data->mtx);
1090 	return ret;
1091 }
1092 
1093 static ssize_t rfkill_fop_write(struct file *file, const char __user *buf,
1094 				size_t count, loff_t *pos)
1095 {
1096 	struct rfkill *rfkill;
1097 	struct rfkill_event ev;
1098 
1099 	/* we don't need the 'hard' variable but accept it */
1100 	if (count < RFKILL_EVENT_SIZE_V1 - 1)
1101 		return -EINVAL;
1102 
1103 	/*
1104 	 * Copy as much data as we can accept into our 'ev' buffer,
1105 	 * but tell userspace how much we've copied so it can determine
1106 	 * our API version even in a write() call, if it cares.
1107 	 */
1108 	count = min(count, sizeof(ev));
1109 	if (copy_from_user(&ev, buf, count))
1110 		return -EFAULT;
1111 
1112 	if (ev.op != RFKILL_OP_CHANGE && ev.op != RFKILL_OP_CHANGE_ALL)
1113 		return -EINVAL;
1114 
1115 	if (ev.type >= NUM_RFKILL_TYPES)
1116 		return -EINVAL;
1117 
1118 	mutex_lock(&rfkill_global_mutex);
1119 
1120 	if (ev.op == RFKILL_OP_CHANGE_ALL) {
1121 		if (ev.type == RFKILL_TYPE_ALL) {
1122 			enum rfkill_type i;
1123 			for (i = 0; i < NUM_RFKILL_TYPES; i++)
1124 				rfkill_global_states[i].cur = ev.soft;
1125 		} else {
1126 			rfkill_global_states[ev.type].cur = ev.soft;
1127 		}
1128 	}
1129 
1130 	list_for_each_entry(rfkill, &rfkill_list, node) {
1131 		if (rfkill->idx != ev.idx && ev.op != RFKILL_OP_CHANGE_ALL)
1132 			continue;
1133 
1134 		if (rfkill->type != ev.type && ev.type != RFKILL_TYPE_ALL)
1135 			continue;
1136 
1137 		rfkill_set_block(rfkill, ev.soft);
1138 	}
1139 	mutex_unlock(&rfkill_global_mutex);
1140 
1141 	return count;
1142 }
1143 
1144 static int rfkill_fop_release(struct inode *inode, struct file *file)
1145 {
1146 	struct rfkill_data *data = file->private_data;
1147 	struct rfkill_int_event *ev, *tmp;
1148 
1149 	mutex_lock(&rfkill_global_mutex);
1150 	list_del(&data->list);
1151 	mutex_unlock(&rfkill_global_mutex);
1152 
1153 	mutex_destroy(&data->mtx);
1154 	list_for_each_entry_safe(ev, tmp, &data->events, list)
1155 		kfree(ev);
1156 
1157 #ifdef CONFIG_RFKILL_INPUT
1158 	if (data->input_handler)
1159 		if (atomic_dec_return(&rfkill_input_disabled) == 0)
1160 			printk(KERN_DEBUG "rfkill: input handler enabled\n");
1161 #endif
1162 
1163 	kfree(data);
1164 
1165 	return 0;
1166 }
1167 
1168 #ifdef CONFIG_RFKILL_INPUT
1169 static long rfkill_fop_ioctl(struct file *file, unsigned int cmd,
1170 			     unsigned long arg)
1171 {
1172 	struct rfkill_data *data = file->private_data;
1173 
1174 	if (_IOC_TYPE(cmd) != RFKILL_IOC_MAGIC)
1175 		return -ENOSYS;
1176 
1177 	if (_IOC_NR(cmd) != RFKILL_IOC_NOINPUT)
1178 		return -ENOSYS;
1179 
1180 	mutex_lock(&data->mtx);
1181 
1182 	if (!data->input_handler) {
1183 		if (atomic_inc_return(&rfkill_input_disabled) == 1)
1184 			printk(KERN_DEBUG "rfkill: input handler disabled\n");
1185 		data->input_handler = true;
1186 	}
1187 
1188 	mutex_unlock(&data->mtx);
1189 
1190 	return 0;
1191 }
1192 #endif
1193 
1194 static const struct file_operations rfkill_fops = {
1195 	.owner		= THIS_MODULE,
1196 	.open		= rfkill_fop_open,
1197 	.read		= rfkill_fop_read,
1198 	.write		= rfkill_fop_write,
1199 	.poll		= rfkill_fop_poll,
1200 	.release	= rfkill_fop_release,
1201 #ifdef CONFIG_RFKILL_INPUT
1202 	.unlocked_ioctl	= rfkill_fop_ioctl,
1203 	.compat_ioctl	= rfkill_fop_ioctl,
1204 #endif
1205 };
1206 
1207 static struct miscdevice rfkill_miscdev = {
1208 	.name	= "rfkill",
1209 	.fops	= &rfkill_fops,
1210 	.minor	= MISC_DYNAMIC_MINOR,
1211 };
1212 
1213 static int __init rfkill_init(void)
1214 {
1215 	int error;
1216 	int i;
1217 
1218 	for (i = 0; i < NUM_RFKILL_TYPES; i++)
1219 		rfkill_global_states[i].cur = !rfkill_default_state;
1220 
1221 	error = class_register(&rfkill_class);
1222 	if (error)
1223 		goto out;
1224 
1225 	error = misc_register(&rfkill_miscdev);
1226 	if (error) {
1227 		class_unregister(&rfkill_class);
1228 		goto out;
1229 	}
1230 
1231 #ifdef CONFIG_RFKILL_INPUT
1232 	error = rfkill_handler_init();
1233 	if (error) {
1234 		misc_deregister(&rfkill_miscdev);
1235 		class_unregister(&rfkill_class);
1236 		goto out;
1237 	}
1238 #endif
1239 
1240  out:
1241 	return error;
1242 }
1243 subsys_initcall(rfkill_init);
1244 
1245 static void __exit rfkill_exit(void)
1246 {
1247 #ifdef CONFIG_RFKILL_INPUT
1248 	rfkill_handler_exit();
1249 #endif
1250 	misc_deregister(&rfkill_miscdev);
1251 	class_unregister(&rfkill_class);
1252 }
1253 module_exit(rfkill_exit);
1254