xref: /openbmc/linux/net/rfkill/core.c (revision 39b6f3aa)
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/device.h>
33 #include <linux/miscdevice.h>
34 #include <linux/wait.h>
35 #include <linux/poll.h>
36 #include <linux/fs.h>
37 #include <linux/slab.h>
38 
39 #include "rfkill.h"
40 
41 #define POLL_INTERVAL		(5 * HZ)
42 
43 #define RFKILL_BLOCK_HW		BIT(0)
44 #define RFKILL_BLOCK_SW		BIT(1)
45 #define RFKILL_BLOCK_SW_PREV	BIT(2)
46 #define RFKILL_BLOCK_ANY	(RFKILL_BLOCK_HW |\
47 				 RFKILL_BLOCK_SW |\
48 				 RFKILL_BLOCK_SW_PREV)
49 #define RFKILL_BLOCK_SW_SETCALL	BIT(31)
50 
51 struct rfkill {
52 	spinlock_t		lock;
53 
54 	const char		*name;
55 	enum rfkill_type	type;
56 
57 	unsigned long		state;
58 
59 	u32			idx;
60 
61 	bool			registered;
62 	bool			persistent;
63 
64 	const struct rfkill_ops	*ops;
65 	void			*data;
66 
67 #ifdef CONFIG_RFKILL_LEDS
68 	struct led_trigger	led_trigger;
69 	const char		*ledtrigname;
70 #endif
71 
72 	struct device		dev;
73 	struct list_head	node;
74 
75 	struct delayed_work	poll_work;
76 	struct work_struct	uevent_work;
77 	struct work_struct	sync_work;
78 };
79 #define to_rfkill(d)	container_of(d, struct rfkill, dev)
80 
81 struct rfkill_int_event {
82 	struct list_head	list;
83 	struct rfkill_event	ev;
84 };
85 
86 struct rfkill_data {
87 	struct list_head	list;
88 	struct list_head	events;
89 	struct mutex		mtx;
90 	wait_queue_head_t	read_wait;
91 	bool			input_handler;
92 };
93 
94 
95 MODULE_AUTHOR("Ivo van Doorn <IvDoorn@gmail.com>");
96 MODULE_AUTHOR("Johannes Berg <johannes@sipsolutions.net>");
97 MODULE_DESCRIPTION("RF switch support");
98 MODULE_LICENSE("GPL");
99 
100 
101 /*
102  * The locking here should be made much smarter, we currently have
103  * a bit of a stupid situation because drivers might want to register
104  * the rfkill struct under their own lock, and take this lock during
105  * rfkill method calls -- which will cause an AB-BA deadlock situation.
106  *
107  * To fix that, we need to rework this code here to be mostly lock-free
108  * and only use the mutex for list manipulations, not to protect the
109  * various other global variables. Then we can avoid holding the mutex
110  * around driver operations, and all is happy.
111  */
112 static LIST_HEAD(rfkill_list);	/* list of registered rf switches */
113 static DEFINE_MUTEX(rfkill_global_mutex);
114 static LIST_HEAD(rfkill_fds);	/* list of open fds of /dev/rfkill */
115 
116 static unsigned int rfkill_default_state = 1;
117 module_param_named(default_state, rfkill_default_state, uint, 0444);
118 MODULE_PARM_DESC(default_state,
119 		 "Default initial state for all radio types, 0 = radio off");
120 
121 static struct {
122 	bool cur, sav;
123 } rfkill_global_states[NUM_RFKILL_TYPES];
124 
125 static bool rfkill_epo_lock_active;
126 
127 
128 #ifdef CONFIG_RFKILL_LEDS
129 static void rfkill_led_trigger_event(struct rfkill *rfkill)
130 {
131 	struct led_trigger *trigger;
132 
133 	if (!rfkill->registered)
134 		return;
135 
136 	trigger = &rfkill->led_trigger;
137 
138 	if (rfkill->state & RFKILL_BLOCK_ANY)
139 		led_trigger_event(trigger, LED_OFF);
140 	else
141 		led_trigger_event(trigger, LED_FULL);
142 }
143 
144 static void rfkill_led_trigger_activate(struct led_classdev *led)
145 {
146 	struct rfkill *rfkill;
147 
148 	rfkill = container_of(led->trigger, struct rfkill, led_trigger);
149 
150 	rfkill_led_trigger_event(rfkill);
151 }
152 
153 const char *rfkill_get_led_trigger_name(struct rfkill *rfkill)
154 {
155 	return rfkill->led_trigger.name;
156 }
157 EXPORT_SYMBOL(rfkill_get_led_trigger_name);
158 
159 void rfkill_set_led_trigger_name(struct rfkill *rfkill, const char *name)
160 {
161 	BUG_ON(!rfkill);
162 
163 	rfkill->ledtrigname = name;
164 }
165 EXPORT_SYMBOL(rfkill_set_led_trigger_name);
166 
167 static int rfkill_led_trigger_register(struct rfkill *rfkill)
168 {
169 	rfkill->led_trigger.name = rfkill->ledtrigname
170 					? : dev_name(&rfkill->dev);
171 	rfkill->led_trigger.activate = rfkill_led_trigger_activate;
172 	return led_trigger_register(&rfkill->led_trigger);
173 }
174 
175 static void rfkill_led_trigger_unregister(struct rfkill *rfkill)
176 {
177 	led_trigger_unregister(&rfkill->led_trigger);
178 }
179 #else
180 static void rfkill_led_trigger_event(struct rfkill *rfkill)
181 {
182 }
183 
184 static inline int rfkill_led_trigger_register(struct rfkill *rfkill)
185 {
186 	return 0;
187 }
188 
189 static inline void rfkill_led_trigger_unregister(struct rfkill *rfkill)
190 {
191 }
192 #endif /* CONFIG_RFKILL_LEDS */
193 
194 static void rfkill_fill_event(struct rfkill_event *ev, struct rfkill *rfkill,
195 			      enum rfkill_operation op)
196 {
197 	unsigned long flags;
198 
199 	ev->idx = rfkill->idx;
200 	ev->type = rfkill->type;
201 	ev->op = op;
202 
203 	spin_lock_irqsave(&rfkill->lock, flags);
204 	ev->hard = !!(rfkill->state & RFKILL_BLOCK_HW);
205 	ev->soft = !!(rfkill->state & (RFKILL_BLOCK_SW |
206 					RFKILL_BLOCK_SW_PREV));
207 	spin_unlock_irqrestore(&rfkill->lock, flags);
208 }
209 
210 static void rfkill_send_events(struct rfkill *rfkill, enum rfkill_operation op)
211 {
212 	struct rfkill_data *data;
213 	struct rfkill_int_event *ev;
214 
215 	list_for_each_entry(data, &rfkill_fds, list) {
216 		ev = kzalloc(sizeof(*ev), GFP_KERNEL);
217 		if (!ev)
218 			continue;
219 		rfkill_fill_event(&ev->ev, rfkill, op);
220 		mutex_lock(&data->mtx);
221 		list_add_tail(&ev->list, &data->events);
222 		mutex_unlock(&data->mtx);
223 		wake_up_interruptible(&data->read_wait);
224 	}
225 }
226 
227 static void rfkill_event(struct rfkill *rfkill)
228 {
229 	if (!rfkill->registered)
230 		return;
231 
232 	kobject_uevent(&rfkill->dev.kobj, KOBJ_CHANGE);
233 
234 	/* also send event to /dev/rfkill */
235 	rfkill_send_events(rfkill, RFKILL_OP_CHANGE);
236 }
237 
238 static bool __rfkill_set_hw_state(struct rfkill *rfkill,
239 				  bool blocked, bool *change)
240 {
241 	unsigned long flags;
242 	bool prev, any;
243 
244 	BUG_ON(!rfkill);
245 
246 	spin_lock_irqsave(&rfkill->lock, flags);
247 	prev = !!(rfkill->state & RFKILL_BLOCK_HW);
248 	if (blocked)
249 		rfkill->state |= RFKILL_BLOCK_HW;
250 	else
251 		rfkill->state &= ~RFKILL_BLOCK_HW;
252 	*change = prev != blocked;
253 	any = !!(rfkill->state & RFKILL_BLOCK_ANY);
254 	spin_unlock_irqrestore(&rfkill->lock, flags);
255 
256 	rfkill_led_trigger_event(rfkill);
257 
258 	return any;
259 }
260 
261 /**
262  * rfkill_set_block - wrapper for set_block method
263  *
264  * @rfkill: the rfkill struct to use
265  * @blocked: the new software state
266  *
267  * Calls the set_block method (when applicable) and handles notifications
268  * etc. as well.
269  */
270 static void rfkill_set_block(struct rfkill *rfkill, bool blocked)
271 {
272 	unsigned long flags;
273 	bool prev, curr;
274 	int err;
275 
276 	if (unlikely(rfkill->dev.power.power_state.event & PM_EVENT_SLEEP))
277 		return;
278 
279 	/*
280 	 * Some platforms (...!) generate input events which affect the
281 	 * _hard_ kill state -- whenever something tries to change the
282 	 * current software state query the hardware state too.
283 	 */
284 	if (rfkill->ops->query)
285 		rfkill->ops->query(rfkill, rfkill->data);
286 
287 	spin_lock_irqsave(&rfkill->lock, flags);
288 	prev = rfkill->state & RFKILL_BLOCK_SW;
289 
290 	if (rfkill->state & RFKILL_BLOCK_SW)
291 		rfkill->state |= RFKILL_BLOCK_SW_PREV;
292 	else
293 		rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
294 
295 	if (blocked)
296 		rfkill->state |= RFKILL_BLOCK_SW;
297 	else
298 		rfkill->state &= ~RFKILL_BLOCK_SW;
299 
300 	rfkill->state |= RFKILL_BLOCK_SW_SETCALL;
301 	spin_unlock_irqrestore(&rfkill->lock, flags);
302 
303 	err = rfkill->ops->set_block(rfkill->data, blocked);
304 
305 	spin_lock_irqsave(&rfkill->lock, flags);
306 	if (err) {
307 		/*
308 		 * Failed -- reset status to _prev, this may be different
309 		 * from what set set _PREV to earlier in this function
310 		 * if rfkill_set_sw_state was invoked.
311 		 */
312 		if (rfkill->state & RFKILL_BLOCK_SW_PREV)
313 			rfkill->state |= RFKILL_BLOCK_SW;
314 		else
315 			rfkill->state &= ~RFKILL_BLOCK_SW;
316 	}
317 	rfkill->state &= ~RFKILL_BLOCK_SW_SETCALL;
318 	rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
319 	curr = rfkill->state & RFKILL_BLOCK_SW;
320 	spin_unlock_irqrestore(&rfkill->lock, flags);
321 
322 	rfkill_led_trigger_event(rfkill);
323 
324 	if (prev != curr)
325 		rfkill_event(rfkill);
326 }
327 
328 #ifdef CONFIG_RFKILL_INPUT
329 static atomic_t rfkill_input_disabled = ATOMIC_INIT(0);
330 
331 /**
332  * __rfkill_switch_all - Toggle state of all switches of given type
333  * @type: type of interfaces to be affected
334  * @state: the new state
335  *
336  * This function sets the state of all switches of given type,
337  * unless a specific switch is claimed by userspace (in which case,
338  * that switch is left alone) or suspended.
339  *
340  * Caller must have acquired rfkill_global_mutex.
341  */
342 static void __rfkill_switch_all(const enum rfkill_type type, bool blocked)
343 {
344 	struct rfkill *rfkill;
345 
346 	rfkill_global_states[type].cur = blocked;
347 	list_for_each_entry(rfkill, &rfkill_list, node) {
348 		if (rfkill->type != type && type != RFKILL_TYPE_ALL)
349 			continue;
350 
351 		rfkill_set_block(rfkill, blocked);
352 	}
353 }
354 
355 /**
356  * rfkill_switch_all - Toggle state of all switches of given type
357  * @type: type of interfaces to be affected
358  * @state: the new state
359  *
360  * Acquires rfkill_global_mutex and calls __rfkill_switch_all(@type, @state).
361  * Please refer to __rfkill_switch_all() for details.
362  *
363  * Does nothing if the EPO lock is active.
364  */
365 void rfkill_switch_all(enum rfkill_type type, bool blocked)
366 {
367 	if (atomic_read(&rfkill_input_disabled))
368 		return;
369 
370 	mutex_lock(&rfkill_global_mutex);
371 
372 	if (!rfkill_epo_lock_active)
373 		__rfkill_switch_all(type, blocked);
374 
375 	mutex_unlock(&rfkill_global_mutex);
376 }
377 
378 /**
379  * rfkill_epo - emergency power off all transmitters
380  *
381  * This kicks all non-suspended rfkill devices to RFKILL_STATE_SOFT_BLOCKED,
382  * ignoring everything in its path but rfkill_global_mutex and rfkill->mutex.
383  *
384  * The global state before the EPO is saved and can be restored later
385  * using rfkill_restore_states().
386  */
387 void rfkill_epo(void)
388 {
389 	struct rfkill *rfkill;
390 	int i;
391 
392 	if (atomic_read(&rfkill_input_disabled))
393 		return;
394 
395 	mutex_lock(&rfkill_global_mutex);
396 
397 	rfkill_epo_lock_active = true;
398 	list_for_each_entry(rfkill, &rfkill_list, node)
399 		rfkill_set_block(rfkill, true);
400 
401 	for (i = 0; i < NUM_RFKILL_TYPES; i++) {
402 		rfkill_global_states[i].sav = rfkill_global_states[i].cur;
403 		rfkill_global_states[i].cur = true;
404 	}
405 
406 	mutex_unlock(&rfkill_global_mutex);
407 }
408 
409 /**
410  * rfkill_restore_states - restore global states
411  *
412  * Restore (and sync switches to) the global state from the
413  * states in rfkill_default_states.  This can undo the effects of
414  * a call to rfkill_epo().
415  */
416 void rfkill_restore_states(void)
417 {
418 	int i;
419 
420 	if (atomic_read(&rfkill_input_disabled))
421 		return;
422 
423 	mutex_lock(&rfkill_global_mutex);
424 
425 	rfkill_epo_lock_active = false;
426 	for (i = 0; i < NUM_RFKILL_TYPES; i++)
427 		__rfkill_switch_all(i, rfkill_global_states[i].sav);
428 	mutex_unlock(&rfkill_global_mutex);
429 }
430 
431 /**
432  * rfkill_remove_epo_lock - unlock state changes
433  *
434  * Used by rfkill-input manually unlock state changes, when
435  * the EPO switch is deactivated.
436  */
437 void rfkill_remove_epo_lock(void)
438 {
439 	if (atomic_read(&rfkill_input_disabled))
440 		return;
441 
442 	mutex_lock(&rfkill_global_mutex);
443 	rfkill_epo_lock_active = false;
444 	mutex_unlock(&rfkill_global_mutex);
445 }
446 
447 /**
448  * rfkill_is_epo_lock_active - returns true EPO is active
449  *
450  * Returns 0 (false) if there is NOT an active EPO contidion,
451  * and 1 (true) if there is an active EPO contition, which
452  * locks all radios in one of the BLOCKED states.
453  *
454  * Can be called in atomic context.
455  */
456 bool rfkill_is_epo_lock_active(void)
457 {
458 	return rfkill_epo_lock_active;
459 }
460 
461 /**
462  * rfkill_get_global_sw_state - returns global state for a type
463  * @type: the type to get the global state of
464  *
465  * Returns the current global state for a given wireless
466  * device type.
467  */
468 bool rfkill_get_global_sw_state(const enum rfkill_type type)
469 {
470 	return rfkill_global_states[type].cur;
471 }
472 #endif
473 
474 
475 bool rfkill_set_hw_state(struct rfkill *rfkill, bool blocked)
476 {
477 	bool ret, change;
478 
479 	ret = __rfkill_set_hw_state(rfkill, blocked, &change);
480 
481 	if (!rfkill->registered)
482 		return ret;
483 
484 	if (change)
485 		schedule_work(&rfkill->uevent_work);
486 
487 	return ret;
488 }
489 EXPORT_SYMBOL(rfkill_set_hw_state);
490 
491 static void __rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
492 {
493 	u32 bit = RFKILL_BLOCK_SW;
494 
495 	/* if in a ops->set_block right now, use other bit */
496 	if (rfkill->state & RFKILL_BLOCK_SW_SETCALL)
497 		bit = RFKILL_BLOCK_SW_PREV;
498 
499 	if (blocked)
500 		rfkill->state |= bit;
501 	else
502 		rfkill->state &= ~bit;
503 }
504 
505 bool rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
506 {
507 	unsigned long flags;
508 	bool prev, hwblock;
509 
510 	BUG_ON(!rfkill);
511 
512 	spin_lock_irqsave(&rfkill->lock, flags);
513 	prev = !!(rfkill->state & RFKILL_BLOCK_SW);
514 	__rfkill_set_sw_state(rfkill, blocked);
515 	hwblock = !!(rfkill->state & RFKILL_BLOCK_HW);
516 	blocked = blocked || hwblock;
517 	spin_unlock_irqrestore(&rfkill->lock, flags);
518 
519 	if (!rfkill->registered)
520 		return blocked;
521 
522 	if (prev != blocked && !hwblock)
523 		schedule_work(&rfkill->uevent_work);
524 
525 	rfkill_led_trigger_event(rfkill);
526 
527 	return blocked;
528 }
529 EXPORT_SYMBOL(rfkill_set_sw_state);
530 
531 void rfkill_init_sw_state(struct rfkill *rfkill, bool blocked)
532 {
533 	unsigned long flags;
534 
535 	BUG_ON(!rfkill);
536 	BUG_ON(rfkill->registered);
537 
538 	spin_lock_irqsave(&rfkill->lock, flags);
539 	__rfkill_set_sw_state(rfkill, blocked);
540 	rfkill->persistent = true;
541 	spin_unlock_irqrestore(&rfkill->lock, flags);
542 }
543 EXPORT_SYMBOL(rfkill_init_sw_state);
544 
545 void rfkill_set_states(struct rfkill *rfkill, bool sw, bool hw)
546 {
547 	unsigned long flags;
548 	bool swprev, hwprev;
549 
550 	BUG_ON(!rfkill);
551 
552 	spin_lock_irqsave(&rfkill->lock, flags);
553 
554 	/*
555 	 * No need to care about prev/setblock ... this is for uevent only
556 	 * and that will get triggered by rfkill_set_block anyway.
557 	 */
558 	swprev = !!(rfkill->state & RFKILL_BLOCK_SW);
559 	hwprev = !!(rfkill->state & RFKILL_BLOCK_HW);
560 	__rfkill_set_sw_state(rfkill, sw);
561 	if (hw)
562 		rfkill->state |= RFKILL_BLOCK_HW;
563 	else
564 		rfkill->state &= ~RFKILL_BLOCK_HW;
565 
566 	spin_unlock_irqrestore(&rfkill->lock, flags);
567 
568 	if (!rfkill->registered) {
569 		rfkill->persistent = true;
570 	} else {
571 		if (swprev != sw || hwprev != hw)
572 			schedule_work(&rfkill->uevent_work);
573 
574 		rfkill_led_trigger_event(rfkill);
575 	}
576 }
577 EXPORT_SYMBOL(rfkill_set_states);
578 
579 static ssize_t rfkill_name_show(struct device *dev,
580 				struct device_attribute *attr,
581 				char *buf)
582 {
583 	struct rfkill *rfkill = to_rfkill(dev);
584 
585 	return sprintf(buf, "%s\n", rfkill->name);
586 }
587 
588 static const char *rfkill_get_type_str(enum rfkill_type type)
589 {
590 	BUILD_BUG_ON(NUM_RFKILL_TYPES != RFKILL_TYPE_NFC + 1);
591 
592 	switch (type) {
593 	case RFKILL_TYPE_WLAN:
594 		return "wlan";
595 	case RFKILL_TYPE_BLUETOOTH:
596 		return "bluetooth";
597 	case RFKILL_TYPE_UWB:
598 		return "ultrawideband";
599 	case RFKILL_TYPE_WIMAX:
600 		return "wimax";
601 	case RFKILL_TYPE_WWAN:
602 		return "wwan";
603 	case RFKILL_TYPE_GPS:
604 		return "gps";
605 	case RFKILL_TYPE_FM:
606 		return "fm";
607 	case RFKILL_TYPE_NFC:
608 		return "nfc";
609 	default:
610 		BUG();
611 	}
612 }
613 
614 static ssize_t rfkill_type_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, "%s\n", rfkill_get_type_str(rfkill->type));
621 }
622 
623 static ssize_t rfkill_idx_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->idx);
630 }
631 
632 static ssize_t rfkill_persistent_show(struct device *dev,
633 			       struct device_attribute *attr,
634 			       char *buf)
635 {
636 	struct rfkill *rfkill = to_rfkill(dev);
637 
638 	return sprintf(buf, "%d\n", rfkill->persistent);
639 }
640 
641 static ssize_t rfkill_hard_show(struct device *dev,
642 				 struct device_attribute *attr,
643 				 char *buf)
644 {
645 	struct rfkill *rfkill = to_rfkill(dev);
646 
647 	return sprintf(buf, "%d\n", (rfkill->state & RFKILL_BLOCK_HW) ? 1 : 0 );
648 }
649 
650 static ssize_t rfkill_soft_show(struct device *dev,
651 				 struct device_attribute *attr,
652 				 char *buf)
653 {
654 	struct rfkill *rfkill = to_rfkill(dev);
655 
656 	return sprintf(buf, "%d\n", (rfkill->state & RFKILL_BLOCK_SW) ? 1 : 0 );
657 }
658 
659 static ssize_t rfkill_soft_store(struct device *dev,
660 				  struct device_attribute *attr,
661 				  const char *buf, size_t count)
662 {
663 	struct rfkill *rfkill = to_rfkill(dev);
664 	unsigned long state;
665 	int err;
666 
667 	if (!capable(CAP_NET_ADMIN))
668 		return -EPERM;
669 
670 	err = kstrtoul(buf, 0, &state);
671 	if (err)
672 		return err;
673 
674 	if (state > 1 )
675 		return -EINVAL;
676 
677 	mutex_lock(&rfkill_global_mutex);
678 	rfkill_set_block(rfkill, state);
679 	mutex_unlock(&rfkill_global_mutex);
680 
681 	return count;
682 }
683 
684 static u8 user_state_from_blocked(unsigned long state)
685 {
686 	if (state & RFKILL_BLOCK_HW)
687 		return RFKILL_USER_STATE_HARD_BLOCKED;
688 	if (state & RFKILL_BLOCK_SW)
689 		return RFKILL_USER_STATE_SOFT_BLOCKED;
690 
691 	return RFKILL_USER_STATE_UNBLOCKED;
692 }
693 
694 static ssize_t rfkill_state_show(struct device *dev,
695 				 struct device_attribute *attr,
696 				 char *buf)
697 {
698 	struct rfkill *rfkill = to_rfkill(dev);
699 
700 	return sprintf(buf, "%d\n", user_state_from_blocked(rfkill->state));
701 }
702 
703 static ssize_t rfkill_state_store(struct device *dev,
704 				  struct device_attribute *attr,
705 				  const char *buf, size_t count)
706 {
707 	struct rfkill *rfkill = to_rfkill(dev);
708 	unsigned long state;
709 	int err;
710 
711 	if (!capable(CAP_NET_ADMIN))
712 		return -EPERM;
713 
714 	err = kstrtoul(buf, 0, &state);
715 	if (err)
716 		return err;
717 
718 	if (state != RFKILL_USER_STATE_SOFT_BLOCKED &&
719 	    state != RFKILL_USER_STATE_UNBLOCKED)
720 		return -EINVAL;
721 
722 	mutex_lock(&rfkill_global_mutex);
723 	rfkill_set_block(rfkill, state == RFKILL_USER_STATE_SOFT_BLOCKED);
724 	mutex_unlock(&rfkill_global_mutex);
725 
726 	return count;
727 }
728 
729 static ssize_t rfkill_claim_show(struct device *dev,
730 				 struct device_attribute *attr,
731 				 char *buf)
732 {
733 	return sprintf(buf, "%d\n", 0);
734 }
735 
736 static ssize_t rfkill_claim_store(struct device *dev,
737 				  struct device_attribute *attr,
738 				  const char *buf, size_t count)
739 {
740 	return -EOPNOTSUPP;
741 }
742 
743 static struct device_attribute rfkill_dev_attrs[] = {
744 	__ATTR(name, S_IRUGO, rfkill_name_show, NULL),
745 	__ATTR(type, S_IRUGO, rfkill_type_show, NULL),
746 	__ATTR(index, S_IRUGO, rfkill_idx_show, NULL),
747 	__ATTR(persistent, S_IRUGO, rfkill_persistent_show, NULL),
748 	__ATTR(state, S_IRUGO|S_IWUSR, rfkill_state_show, rfkill_state_store),
749 	__ATTR(claim, S_IRUGO|S_IWUSR, rfkill_claim_show, rfkill_claim_store),
750 	__ATTR(soft, S_IRUGO|S_IWUSR, rfkill_soft_show, rfkill_soft_store),
751 	__ATTR(hard, S_IRUGO, rfkill_hard_show, NULL),
752 	__ATTR_NULL
753 };
754 
755 static void rfkill_release(struct device *dev)
756 {
757 	struct rfkill *rfkill = to_rfkill(dev);
758 
759 	kfree(rfkill);
760 }
761 
762 static int rfkill_dev_uevent(struct device *dev, struct kobj_uevent_env *env)
763 {
764 	struct rfkill *rfkill = to_rfkill(dev);
765 	unsigned long flags;
766 	u32 state;
767 	int error;
768 
769 	error = add_uevent_var(env, "RFKILL_NAME=%s", rfkill->name);
770 	if (error)
771 		return error;
772 	error = add_uevent_var(env, "RFKILL_TYPE=%s",
773 			       rfkill_get_type_str(rfkill->type));
774 	if (error)
775 		return error;
776 	spin_lock_irqsave(&rfkill->lock, flags);
777 	state = rfkill->state;
778 	spin_unlock_irqrestore(&rfkill->lock, flags);
779 	error = add_uevent_var(env, "RFKILL_STATE=%d",
780 			       user_state_from_blocked(state));
781 	return error;
782 }
783 
784 void rfkill_pause_polling(struct rfkill *rfkill)
785 {
786 	BUG_ON(!rfkill);
787 
788 	if (!rfkill->ops->poll)
789 		return;
790 
791 	cancel_delayed_work_sync(&rfkill->poll_work);
792 }
793 EXPORT_SYMBOL(rfkill_pause_polling);
794 
795 void rfkill_resume_polling(struct rfkill *rfkill)
796 {
797 	BUG_ON(!rfkill);
798 
799 	if (!rfkill->ops->poll)
800 		return;
801 
802 	schedule_work(&rfkill->poll_work.work);
803 }
804 EXPORT_SYMBOL(rfkill_resume_polling);
805 
806 static int rfkill_suspend(struct device *dev, pm_message_t state)
807 {
808 	struct rfkill *rfkill = to_rfkill(dev);
809 
810 	rfkill_pause_polling(rfkill);
811 
812 	return 0;
813 }
814 
815 static int rfkill_resume(struct device *dev)
816 {
817 	struct rfkill *rfkill = to_rfkill(dev);
818 	bool cur;
819 
820 	if (!rfkill->persistent) {
821 		cur = !!(rfkill->state & RFKILL_BLOCK_SW);
822 		rfkill_set_block(rfkill, cur);
823 	}
824 
825 	rfkill_resume_polling(rfkill);
826 
827 	return 0;
828 }
829 
830 static struct class rfkill_class = {
831 	.name		= "rfkill",
832 	.dev_release	= rfkill_release,
833 	.dev_attrs	= rfkill_dev_attrs,
834 	.dev_uevent	= rfkill_dev_uevent,
835 	.suspend	= rfkill_suspend,
836 	.resume		= rfkill_resume,
837 };
838 
839 bool rfkill_blocked(struct rfkill *rfkill)
840 {
841 	unsigned long flags;
842 	u32 state;
843 
844 	spin_lock_irqsave(&rfkill->lock, flags);
845 	state = rfkill->state;
846 	spin_unlock_irqrestore(&rfkill->lock, flags);
847 
848 	return !!(state & RFKILL_BLOCK_ANY);
849 }
850 EXPORT_SYMBOL(rfkill_blocked);
851 
852 
853 struct rfkill * __must_check rfkill_alloc(const char *name,
854 					  struct device *parent,
855 					  const enum rfkill_type type,
856 					  const struct rfkill_ops *ops,
857 					  void *ops_data)
858 {
859 	struct rfkill *rfkill;
860 	struct device *dev;
861 
862 	if (WARN_ON(!ops))
863 		return NULL;
864 
865 	if (WARN_ON(!ops->set_block))
866 		return NULL;
867 
868 	if (WARN_ON(!name))
869 		return NULL;
870 
871 	if (WARN_ON(type == RFKILL_TYPE_ALL || type >= NUM_RFKILL_TYPES))
872 		return NULL;
873 
874 	rfkill = kzalloc(sizeof(*rfkill), GFP_KERNEL);
875 	if (!rfkill)
876 		return NULL;
877 
878 	spin_lock_init(&rfkill->lock);
879 	INIT_LIST_HEAD(&rfkill->node);
880 	rfkill->type = type;
881 	rfkill->name = name;
882 	rfkill->ops = ops;
883 	rfkill->data = ops_data;
884 
885 	dev = &rfkill->dev;
886 	dev->class = &rfkill_class;
887 	dev->parent = parent;
888 	device_initialize(dev);
889 
890 	return rfkill;
891 }
892 EXPORT_SYMBOL(rfkill_alloc);
893 
894 static void rfkill_poll(struct work_struct *work)
895 {
896 	struct rfkill *rfkill;
897 
898 	rfkill = container_of(work, struct rfkill, poll_work.work);
899 
900 	/*
901 	 * Poll hardware state -- driver will use one of the
902 	 * rfkill_set{,_hw,_sw}_state functions and use its
903 	 * return value to update the current status.
904 	 */
905 	rfkill->ops->poll(rfkill, rfkill->data);
906 
907 	schedule_delayed_work(&rfkill->poll_work,
908 		round_jiffies_relative(POLL_INTERVAL));
909 }
910 
911 static void rfkill_uevent_work(struct work_struct *work)
912 {
913 	struct rfkill *rfkill;
914 
915 	rfkill = container_of(work, struct rfkill, uevent_work);
916 
917 	mutex_lock(&rfkill_global_mutex);
918 	rfkill_event(rfkill);
919 	mutex_unlock(&rfkill_global_mutex);
920 }
921 
922 static void rfkill_sync_work(struct work_struct *work)
923 {
924 	struct rfkill *rfkill;
925 	bool cur;
926 
927 	rfkill = container_of(work, struct rfkill, sync_work);
928 
929 	mutex_lock(&rfkill_global_mutex);
930 	cur = rfkill_global_states[rfkill->type].cur;
931 	rfkill_set_block(rfkill, cur);
932 	mutex_unlock(&rfkill_global_mutex);
933 }
934 
935 int __must_check rfkill_register(struct rfkill *rfkill)
936 {
937 	static unsigned long rfkill_no;
938 	struct device *dev = &rfkill->dev;
939 	int error;
940 
941 	BUG_ON(!rfkill);
942 
943 	mutex_lock(&rfkill_global_mutex);
944 
945 	if (rfkill->registered) {
946 		error = -EALREADY;
947 		goto unlock;
948 	}
949 
950 	rfkill->idx = rfkill_no;
951 	dev_set_name(dev, "rfkill%lu", rfkill_no);
952 	rfkill_no++;
953 
954 	list_add_tail(&rfkill->node, &rfkill_list);
955 
956 	error = device_add(dev);
957 	if (error)
958 		goto remove;
959 
960 	error = rfkill_led_trigger_register(rfkill);
961 	if (error)
962 		goto devdel;
963 
964 	rfkill->registered = true;
965 
966 	INIT_DELAYED_WORK(&rfkill->poll_work, rfkill_poll);
967 	INIT_WORK(&rfkill->uevent_work, rfkill_uevent_work);
968 	INIT_WORK(&rfkill->sync_work, rfkill_sync_work);
969 
970 	if (rfkill->ops->poll)
971 		schedule_delayed_work(&rfkill->poll_work,
972 			round_jiffies_relative(POLL_INTERVAL));
973 
974 	if (!rfkill->persistent || rfkill_epo_lock_active) {
975 		schedule_work(&rfkill->sync_work);
976 	} else {
977 #ifdef CONFIG_RFKILL_INPUT
978 		bool soft_blocked = !!(rfkill->state & RFKILL_BLOCK_SW);
979 
980 		if (!atomic_read(&rfkill_input_disabled))
981 			__rfkill_switch_all(rfkill->type, soft_blocked);
982 #endif
983 	}
984 
985 	rfkill_send_events(rfkill, RFKILL_OP_ADD);
986 
987 	mutex_unlock(&rfkill_global_mutex);
988 	return 0;
989 
990  devdel:
991 	device_del(&rfkill->dev);
992  remove:
993 	list_del_init(&rfkill->node);
994  unlock:
995 	mutex_unlock(&rfkill_global_mutex);
996 	return error;
997 }
998 EXPORT_SYMBOL(rfkill_register);
999 
1000 void rfkill_unregister(struct rfkill *rfkill)
1001 {
1002 	BUG_ON(!rfkill);
1003 
1004 	if (rfkill->ops->poll)
1005 		cancel_delayed_work_sync(&rfkill->poll_work);
1006 
1007 	cancel_work_sync(&rfkill->uevent_work);
1008 	cancel_work_sync(&rfkill->sync_work);
1009 
1010 	rfkill->registered = false;
1011 
1012 	device_del(&rfkill->dev);
1013 
1014 	mutex_lock(&rfkill_global_mutex);
1015 	rfkill_send_events(rfkill, RFKILL_OP_DEL);
1016 	list_del_init(&rfkill->node);
1017 	mutex_unlock(&rfkill_global_mutex);
1018 
1019 	rfkill_led_trigger_unregister(rfkill);
1020 }
1021 EXPORT_SYMBOL(rfkill_unregister);
1022 
1023 void rfkill_destroy(struct rfkill *rfkill)
1024 {
1025 	if (rfkill)
1026 		put_device(&rfkill->dev);
1027 }
1028 EXPORT_SYMBOL(rfkill_destroy);
1029 
1030 static int rfkill_fop_open(struct inode *inode, struct file *file)
1031 {
1032 	struct rfkill_data *data;
1033 	struct rfkill *rfkill;
1034 	struct rfkill_int_event *ev, *tmp;
1035 
1036 	data = kzalloc(sizeof(*data), GFP_KERNEL);
1037 	if (!data)
1038 		return -ENOMEM;
1039 
1040 	INIT_LIST_HEAD(&data->events);
1041 	mutex_init(&data->mtx);
1042 	init_waitqueue_head(&data->read_wait);
1043 
1044 	mutex_lock(&rfkill_global_mutex);
1045 	mutex_lock(&data->mtx);
1046 	/*
1047 	 * start getting events from elsewhere but hold mtx to get
1048 	 * startup events added first
1049 	 */
1050 
1051 	list_for_each_entry(rfkill, &rfkill_list, node) {
1052 		ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1053 		if (!ev)
1054 			goto free;
1055 		rfkill_fill_event(&ev->ev, rfkill, RFKILL_OP_ADD);
1056 		list_add_tail(&ev->list, &data->events);
1057 	}
1058 	list_add(&data->list, &rfkill_fds);
1059 	mutex_unlock(&data->mtx);
1060 	mutex_unlock(&rfkill_global_mutex);
1061 
1062 	file->private_data = data;
1063 
1064 	return nonseekable_open(inode, file);
1065 
1066  free:
1067 	mutex_unlock(&data->mtx);
1068 	mutex_unlock(&rfkill_global_mutex);
1069 	mutex_destroy(&data->mtx);
1070 	list_for_each_entry_safe(ev, tmp, &data->events, list)
1071 		kfree(ev);
1072 	kfree(data);
1073 	return -ENOMEM;
1074 }
1075 
1076 static unsigned int rfkill_fop_poll(struct file *file, poll_table *wait)
1077 {
1078 	struct rfkill_data *data = file->private_data;
1079 	unsigned int res = POLLOUT | POLLWRNORM;
1080 
1081 	poll_wait(file, &data->read_wait, wait);
1082 
1083 	mutex_lock(&data->mtx);
1084 	if (!list_empty(&data->events))
1085 		res = POLLIN | POLLRDNORM;
1086 	mutex_unlock(&data->mtx);
1087 
1088 	return res;
1089 }
1090 
1091 static bool rfkill_readable(struct rfkill_data *data)
1092 {
1093 	bool r;
1094 
1095 	mutex_lock(&data->mtx);
1096 	r = !list_empty(&data->events);
1097 	mutex_unlock(&data->mtx);
1098 
1099 	return r;
1100 }
1101 
1102 static ssize_t rfkill_fop_read(struct file *file, char __user *buf,
1103 			       size_t count, loff_t *pos)
1104 {
1105 	struct rfkill_data *data = file->private_data;
1106 	struct rfkill_int_event *ev;
1107 	unsigned long sz;
1108 	int ret;
1109 
1110 	mutex_lock(&data->mtx);
1111 
1112 	while (list_empty(&data->events)) {
1113 		if (file->f_flags & O_NONBLOCK) {
1114 			ret = -EAGAIN;
1115 			goto out;
1116 		}
1117 		mutex_unlock(&data->mtx);
1118 		ret = wait_event_interruptible(data->read_wait,
1119 					       rfkill_readable(data));
1120 		mutex_lock(&data->mtx);
1121 
1122 		if (ret)
1123 			goto out;
1124 	}
1125 
1126 	ev = list_first_entry(&data->events, struct rfkill_int_event,
1127 				list);
1128 
1129 	sz = min_t(unsigned long, sizeof(ev->ev), count);
1130 	ret = sz;
1131 	if (copy_to_user(buf, &ev->ev, sz))
1132 		ret = -EFAULT;
1133 
1134 	list_del(&ev->list);
1135 	kfree(ev);
1136  out:
1137 	mutex_unlock(&data->mtx);
1138 	return ret;
1139 }
1140 
1141 static ssize_t rfkill_fop_write(struct file *file, const char __user *buf,
1142 				size_t count, loff_t *pos)
1143 {
1144 	struct rfkill *rfkill;
1145 	struct rfkill_event ev;
1146 
1147 	/* we don't need the 'hard' variable but accept it */
1148 	if (count < RFKILL_EVENT_SIZE_V1 - 1)
1149 		return -EINVAL;
1150 
1151 	/*
1152 	 * Copy as much data as we can accept into our 'ev' buffer,
1153 	 * but tell userspace how much we've copied so it can determine
1154 	 * our API version even in a write() call, if it cares.
1155 	 */
1156 	count = min(count, sizeof(ev));
1157 	if (copy_from_user(&ev, buf, count))
1158 		return -EFAULT;
1159 
1160 	if (ev.op != RFKILL_OP_CHANGE && ev.op != RFKILL_OP_CHANGE_ALL)
1161 		return -EINVAL;
1162 
1163 	if (ev.type >= NUM_RFKILL_TYPES)
1164 		return -EINVAL;
1165 
1166 	mutex_lock(&rfkill_global_mutex);
1167 
1168 	if (ev.op == RFKILL_OP_CHANGE_ALL) {
1169 		if (ev.type == RFKILL_TYPE_ALL) {
1170 			enum rfkill_type i;
1171 			for (i = 0; i < NUM_RFKILL_TYPES; i++)
1172 				rfkill_global_states[i].cur = ev.soft;
1173 		} else {
1174 			rfkill_global_states[ev.type].cur = ev.soft;
1175 		}
1176 	}
1177 
1178 	list_for_each_entry(rfkill, &rfkill_list, node) {
1179 		if (rfkill->idx != ev.idx && ev.op != RFKILL_OP_CHANGE_ALL)
1180 			continue;
1181 
1182 		if (rfkill->type != ev.type && ev.type != RFKILL_TYPE_ALL)
1183 			continue;
1184 
1185 		rfkill_set_block(rfkill, ev.soft);
1186 	}
1187 	mutex_unlock(&rfkill_global_mutex);
1188 
1189 	return count;
1190 }
1191 
1192 static int rfkill_fop_release(struct inode *inode, struct file *file)
1193 {
1194 	struct rfkill_data *data = file->private_data;
1195 	struct rfkill_int_event *ev, *tmp;
1196 
1197 	mutex_lock(&rfkill_global_mutex);
1198 	list_del(&data->list);
1199 	mutex_unlock(&rfkill_global_mutex);
1200 
1201 	mutex_destroy(&data->mtx);
1202 	list_for_each_entry_safe(ev, tmp, &data->events, list)
1203 		kfree(ev);
1204 
1205 #ifdef CONFIG_RFKILL_INPUT
1206 	if (data->input_handler)
1207 		if (atomic_dec_return(&rfkill_input_disabled) == 0)
1208 			printk(KERN_DEBUG "rfkill: input handler enabled\n");
1209 #endif
1210 
1211 	kfree(data);
1212 
1213 	return 0;
1214 }
1215 
1216 #ifdef CONFIG_RFKILL_INPUT
1217 static long rfkill_fop_ioctl(struct file *file, unsigned int cmd,
1218 			     unsigned long arg)
1219 {
1220 	struct rfkill_data *data = file->private_data;
1221 
1222 	if (_IOC_TYPE(cmd) != RFKILL_IOC_MAGIC)
1223 		return -ENOSYS;
1224 
1225 	if (_IOC_NR(cmd) != RFKILL_IOC_NOINPUT)
1226 		return -ENOSYS;
1227 
1228 	mutex_lock(&data->mtx);
1229 
1230 	if (!data->input_handler) {
1231 		if (atomic_inc_return(&rfkill_input_disabled) == 1)
1232 			printk(KERN_DEBUG "rfkill: input handler disabled\n");
1233 		data->input_handler = true;
1234 	}
1235 
1236 	mutex_unlock(&data->mtx);
1237 
1238 	return 0;
1239 }
1240 #endif
1241 
1242 static const struct file_operations rfkill_fops = {
1243 	.owner		= THIS_MODULE,
1244 	.open		= rfkill_fop_open,
1245 	.read		= rfkill_fop_read,
1246 	.write		= rfkill_fop_write,
1247 	.poll		= rfkill_fop_poll,
1248 	.release	= rfkill_fop_release,
1249 #ifdef CONFIG_RFKILL_INPUT
1250 	.unlocked_ioctl	= rfkill_fop_ioctl,
1251 	.compat_ioctl	= rfkill_fop_ioctl,
1252 #endif
1253 	.llseek		= no_llseek,
1254 };
1255 
1256 static struct miscdevice rfkill_miscdev = {
1257 	.name	= "rfkill",
1258 	.fops	= &rfkill_fops,
1259 	.minor	= MISC_DYNAMIC_MINOR,
1260 };
1261 
1262 static int __init rfkill_init(void)
1263 {
1264 	int error;
1265 	int i;
1266 
1267 	for (i = 0; i < NUM_RFKILL_TYPES; i++)
1268 		rfkill_global_states[i].cur = !rfkill_default_state;
1269 
1270 	error = class_register(&rfkill_class);
1271 	if (error)
1272 		goto out;
1273 
1274 	error = misc_register(&rfkill_miscdev);
1275 	if (error) {
1276 		class_unregister(&rfkill_class);
1277 		goto out;
1278 	}
1279 
1280 #ifdef CONFIG_RFKILL_INPUT
1281 	error = rfkill_handler_init();
1282 	if (error) {
1283 		misc_deregister(&rfkill_miscdev);
1284 		class_unregister(&rfkill_class);
1285 		goto out;
1286 	}
1287 #endif
1288 
1289  out:
1290 	return error;
1291 }
1292 subsys_initcall(rfkill_init);
1293 
1294 static void __exit rfkill_exit(void)
1295 {
1296 #ifdef CONFIG_RFKILL_INPUT
1297 	rfkill_handler_exit();
1298 #endif
1299 	misc_deregister(&rfkill_miscdev);
1300 	class_unregister(&rfkill_class);
1301 }
1302 module_exit(rfkill_exit);
1303