Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
[linux-2.6.git] / net / rfkill / core.c
blob1cec5e4f3a5e62ff4503d937d78d435c1761cd52
1 /*
2 * Copyright (C) 2006 - 2007 Ivo van Doorn
3 * Copyright (C) 2007 Dmitry Torokhov
4 * Copyright 2009 Johannes Berg <johannes@sipsolutions.net>
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.
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.
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.
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>
39 #include "rfkill.h"
41 #define POLL_INTERVAL (5 * HZ)
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)
51 struct rfkill {
52 spinlock_t lock;
54 const char *name;
55 enum rfkill_type type;
57 unsigned long state;
59 u32 idx;
61 bool registered;
62 bool persistent;
64 const struct rfkill_ops *ops;
65 void *data;
67 #ifdef CONFIG_RFKILL_LEDS
68 struct led_trigger led_trigger;
69 const char *ledtrigname;
70 #endif
72 struct device dev;
73 struct list_head node;
75 struct delayed_work poll_work;
76 struct work_struct uevent_work;
77 struct work_struct sync_work;
79 #define to_rfkill(d) container_of(d, struct rfkill, dev)
81 struct rfkill_int_event {
82 struct list_head list;
83 struct rfkill_event ev;
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;
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");
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.
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.
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 */
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");
121 static struct {
122 bool cur, sav;
123 } rfkill_global_states[NUM_RFKILL_TYPES];
125 static bool rfkill_epo_lock_active;
128 #ifdef CONFIG_RFKILL_LEDS
129 static void rfkill_led_trigger_event(struct rfkill *rfkill)
131 struct led_trigger *trigger;
133 if (!rfkill->registered)
134 return;
136 trigger = &rfkill->led_trigger;
138 if (rfkill->state & RFKILL_BLOCK_ANY)
139 led_trigger_event(trigger, LED_OFF);
140 else
141 led_trigger_event(trigger, LED_FULL);
144 static void rfkill_led_trigger_activate(struct led_classdev *led)
146 struct rfkill *rfkill;
148 rfkill = container_of(led->trigger, struct rfkill, led_trigger);
150 rfkill_led_trigger_event(rfkill);
153 const char *rfkill_get_led_trigger_name(struct rfkill *rfkill)
155 return rfkill->led_trigger.name;
157 EXPORT_SYMBOL(rfkill_get_led_trigger_name);
159 void rfkill_set_led_trigger_name(struct rfkill *rfkill, const char *name)
161 BUG_ON(!rfkill);
163 rfkill->ledtrigname = name;
165 EXPORT_SYMBOL(rfkill_set_led_trigger_name);
167 static int rfkill_led_trigger_register(struct rfkill *rfkill)
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);
175 static void rfkill_led_trigger_unregister(struct rfkill *rfkill)
177 led_trigger_unregister(&rfkill->led_trigger);
179 #else
180 static void rfkill_led_trigger_event(struct rfkill *rfkill)
184 static inline int rfkill_led_trigger_register(struct rfkill *rfkill)
186 return 0;
189 static inline void rfkill_led_trigger_unregister(struct rfkill *rfkill)
192 #endif /* CONFIG_RFKILL_LEDS */
194 static void rfkill_fill_event(struct rfkill_event *ev, struct rfkill *rfkill,
195 enum rfkill_operation op)
197 unsigned long flags;
199 ev->idx = rfkill->idx;
200 ev->type = rfkill->type;
201 ev->op = op;
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);
210 static void rfkill_send_events(struct rfkill *rfkill, enum rfkill_operation op)
212 struct rfkill_data *data;
213 struct rfkill_int_event *ev;
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);
227 static void rfkill_event(struct rfkill *rfkill)
229 if (!rfkill->registered)
230 return;
232 kobject_uevent(&rfkill->dev.kobj, KOBJ_CHANGE);
234 /* also send event to /dev/rfkill */
235 rfkill_send_events(rfkill, RFKILL_OP_CHANGE);
238 static bool __rfkill_set_hw_state(struct rfkill *rfkill,
239 bool blocked, bool *change)
241 unsigned long flags;
242 bool prev, any;
244 BUG_ON(!rfkill);
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);
256 rfkill_led_trigger_event(rfkill);
258 return any;
262 * rfkill_set_block - wrapper for set_block method
264 * @rfkill: the rfkill struct to use
265 * @blocked: the new software state
267 * Calls the set_block method (when applicable) and handles notifications
268 * etc. as well.
270 static void rfkill_set_block(struct rfkill *rfkill, bool blocked)
272 unsigned long flags;
273 bool prev, curr;
274 int err;
276 if (unlikely(rfkill->dev.power.power_state.event & PM_EVENT_SLEEP))
277 return;
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.
284 if (rfkill->ops->query)
285 rfkill->ops->query(rfkill, rfkill->data);
287 spin_lock_irqsave(&rfkill->lock, flags);
288 prev = rfkill->state & RFKILL_BLOCK_SW;
290 if (rfkill->state & RFKILL_BLOCK_SW)
291 rfkill->state |= RFKILL_BLOCK_SW_PREV;
292 else
293 rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
295 if (blocked)
296 rfkill->state |= RFKILL_BLOCK_SW;
297 else
298 rfkill->state &= ~RFKILL_BLOCK_SW;
300 rfkill->state |= RFKILL_BLOCK_SW_SETCALL;
301 spin_unlock_irqrestore(&rfkill->lock, flags);
303 err = rfkill->ops->set_block(rfkill->data, blocked);
305 spin_lock_irqsave(&rfkill->lock, flags);
306 if (err) {
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.
312 if (rfkill->state & RFKILL_BLOCK_SW_PREV)
313 rfkill->state |= RFKILL_BLOCK_SW;
314 else
315 rfkill->state &= ~RFKILL_BLOCK_SW;
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);
322 rfkill_led_trigger_event(rfkill);
324 if (prev != curr)
325 rfkill_event(rfkill);
328 #ifdef CONFIG_RFKILL_INPUT
329 static atomic_t rfkill_input_disabled = ATOMIC_INIT(0);
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
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.
340 * Caller must have acquired rfkill_global_mutex.
342 static void __rfkill_switch_all(const enum rfkill_type type, bool blocked)
344 struct rfkill *rfkill;
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;
351 rfkill_set_block(rfkill, blocked);
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
360 * Acquires rfkill_global_mutex and calls __rfkill_switch_all(@type, @state).
361 * Please refer to __rfkill_switch_all() for details.
363 * Does nothing if the EPO lock is active.
365 void rfkill_switch_all(enum rfkill_type type, bool blocked)
367 if (atomic_read(&rfkill_input_disabled))
368 return;
370 mutex_lock(&rfkill_global_mutex);
372 if (!rfkill_epo_lock_active)
373 __rfkill_switch_all(type, blocked);
375 mutex_unlock(&rfkill_global_mutex);
379 * rfkill_epo - emergency power off all transmitters
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.
384 * The global state before the EPO is saved and can be restored later
385 * using rfkill_restore_states().
387 void rfkill_epo(void)
389 struct rfkill *rfkill;
390 int i;
392 if (atomic_read(&rfkill_input_disabled))
393 return;
395 mutex_lock(&rfkill_global_mutex);
397 rfkill_epo_lock_active = true;
398 list_for_each_entry(rfkill, &rfkill_list, node)
399 rfkill_set_block(rfkill, true);
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;
406 mutex_unlock(&rfkill_global_mutex);
410 * rfkill_restore_states - restore global states
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().
416 void rfkill_restore_states(void)
418 int i;
420 if (atomic_read(&rfkill_input_disabled))
421 return;
423 mutex_lock(&rfkill_global_mutex);
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);
432 * rfkill_remove_epo_lock - unlock state changes
434 * Used by rfkill-input manually unlock state changes, when
435 * the EPO switch is deactivated.
437 void rfkill_remove_epo_lock(void)
439 if (atomic_read(&rfkill_input_disabled))
440 return;
442 mutex_lock(&rfkill_global_mutex);
443 rfkill_epo_lock_active = false;
444 mutex_unlock(&rfkill_global_mutex);
448 * rfkill_is_epo_lock_active - returns true EPO is active
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.
454 * Can be called in atomic context.
456 bool rfkill_is_epo_lock_active(void)
458 return rfkill_epo_lock_active;
462 * rfkill_get_global_sw_state - returns global state for a type
463 * @type: the type to get the global state of
465 * Returns the current global state for a given wireless
466 * device type.
468 bool rfkill_get_global_sw_state(const enum rfkill_type type)
470 return rfkill_global_states[type].cur;
472 #endif
475 bool rfkill_set_hw_state(struct rfkill *rfkill, bool blocked)
477 bool ret, change;
479 ret = __rfkill_set_hw_state(rfkill, blocked, &change);
481 if (!rfkill->registered)
482 return ret;
484 if (change)
485 schedule_work(&rfkill->uevent_work);
487 return ret;
489 EXPORT_SYMBOL(rfkill_set_hw_state);
491 static void __rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
493 u32 bit = RFKILL_BLOCK_SW;
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;
499 if (blocked)
500 rfkill->state |= bit;
501 else
502 rfkill->state &= ~bit;
505 bool rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
507 unsigned long flags;
508 bool prev, hwblock;
510 BUG_ON(!rfkill);
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);
519 if (!rfkill->registered)
520 return blocked;
522 if (prev != blocked && !hwblock)
523 schedule_work(&rfkill->uevent_work);
525 rfkill_led_trigger_event(rfkill);
527 return blocked;
529 EXPORT_SYMBOL(rfkill_set_sw_state);
531 void rfkill_init_sw_state(struct rfkill *rfkill, bool blocked)
533 unsigned long flags;
535 BUG_ON(!rfkill);
536 BUG_ON(rfkill->registered);
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);
543 EXPORT_SYMBOL(rfkill_init_sw_state);
545 void rfkill_set_states(struct rfkill *rfkill, bool sw, bool hw)
547 unsigned long flags;
548 bool swprev, hwprev;
550 BUG_ON(!rfkill);
552 spin_lock_irqsave(&rfkill->lock, flags);
555 * No need to care about prev/setblock ... this is for uevent only
556 * and that will get triggered by rfkill_set_block anyway.
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;
566 spin_unlock_irqrestore(&rfkill->lock, flags);
568 if (!rfkill->registered) {
569 rfkill->persistent = true;
570 } else {
571 if (swprev != sw || hwprev != hw)
572 schedule_work(&rfkill->uevent_work);
574 rfkill_led_trigger_event(rfkill);
577 EXPORT_SYMBOL(rfkill_set_states);
579 static ssize_t rfkill_name_show(struct device *dev,
580 struct device_attribute *attr,
581 char *buf)
583 struct rfkill *rfkill = to_rfkill(dev);
585 return sprintf(buf, "%s\n", rfkill->name);
588 static const char *rfkill_get_type_str(enum rfkill_type type)
590 BUILD_BUG_ON(NUM_RFKILL_TYPES != RFKILL_TYPE_NFC + 1);
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();
614 static ssize_t rfkill_type_show(struct device *dev,
615 struct device_attribute *attr,
616 char *buf)
618 struct rfkill *rfkill = to_rfkill(dev);
620 return sprintf(buf, "%s\n", rfkill_get_type_str(rfkill->type));
623 static ssize_t rfkill_idx_show(struct device *dev,
624 struct device_attribute *attr,
625 char *buf)
627 struct rfkill *rfkill = to_rfkill(dev);
629 return sprintf(buf, "%d\n", rfkill->idx);
632 static ssize_t rfkill_persistent_show(struct device *dev,
633 struct device_attribute *attr,
634 char *buf)
636 struct rfkill *rfkill = to_rfkill(dev);
638 return sprintf(buf, "%d\n", rfkill->persistent);
641 static ssize_t rfkill_hard_show(struct device *dev,
642 struct device_attribute *attr,
643 char *buf)
645 struct rfkill *rfkill = to_rfkill(dev);
647 return sprintf(buf, "%d\n", (rfkill->state & RFKILL_BLOCK_HW) ? 1 : 0 );
650 static ssize_t rfkill_soft_show(struct device *dev,
651 struct device_attribute *attr,
652 char *buf)
654 struct rfkill *rfkill = to_rfkill(dev);
656 return sprintf(buf, "%d\n", (rfkill->state & RFKILL_BLOCK_SW) ? 1 : 0 );
659 static ssize_t rfkill_soft_store(struct device *dev,
660 struct device_attribute *attr,
661 const char *buf, size_t count)
663 struct rfkill *rfkill = to_rfkill(dev);
664 unsigned long state;
665 int err;
667 if (!capable(CAP_NET_ADMIN))
668 return -EPERM;
670 err = kstrtoul(buf, 0, &state);
671 if (err)
672 return err;
674 if (state > 1 )
675 return -EINVAL;
677 mutex_lock(&rfkill_global_mutex);
678 rfkill_set_block(rfkill, state);
679 mutex_unlock(&rfkill_global_mutex);
681 return count;
684 static u8 user_state_from_blocked(unsigned long state)
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;
691 return RFKILL_USER_STATE_UNBLOCKED;
694 static ssize_t rfkill_state_show(struct device *dev,
695 struct device_attribute *attr,
696 char *buf)
698 struct rfkill *rfkill = to_rfkill(dev);
700 return sprintf(buf, "%d\n", user_state_from_blocked(rfkill->state));
703 static ssize_t rfkill_state_store(struct device *dev,
704 struct device_attribute *attr,
705 const char *buf, size_t count)
707 struct rfkill *rfkill = to_rfkill(dev);
708 unsigned long state;
709 int err;
711 if (!capable(CAP_NET_ADMIN))
712 return -EPERM;
714 err = kstrtoul(buf, 0, &state);
715 if (err)
716 return err;
718 if (state != RFKILL_USER_STATE_SOFT_BLOCKED &&
719 state != RFKILL_USER_STATE_UNBLOCKED)
720 return -EINVAL;
722 mutex_lock(&rfkill_global_mutex);
723 rfkill_set_block(rfkill, state == RFKILL_USER_STATE_SOFT_BLOCKED);
724 mutex_unlock(&rfkill_global_mutex);
726 return count;
729 static ssize_t rfkill_claim_show(struct device *dev,
730 struct device_attribute *attr,
731 char *buf)
733 return sprintf(buf, "%d\n", 0);
736 static ssize_t rfkill_claim_store(struct device *dev,
737 struct device_attribute *attr,
738 const char *buf, size_t count)
740 return -EOPNOTSUPP;
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
755 static void rfkill_release(struct device *dev)
757 struct rfkill *rfkill = to_rfkill(dev);
759 kfree(rfkill);
762 static int rfkill_dev_uevent(struct device *dev, struct kobj_uevent_env *env)
764 struct rfkill *rfkill = to_rfkill(dev);
765 unsigned long flags;
766 u32 state;
767 int error;
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;
784 void rfkill_pause_polling(struct rfkill *rfkill)
786 BUG_ON(!rfkill);
788 if (!rfkill->ops->poll)
789 return;
791 cancel_delayed_work_sync(&rfkill->poll_work);
793 EXPORT_SYMBOL(rfkill_pause_polling);
795 void rfkill_resume_polling(struct rfkill *rfkill)
797 BUG_ON(!rfkill);
799 if (!rfkill->ops->poll)
800 return;
802 schedule_work(&rfkill->poll_work.work);
804 EXPORT_SYMBOL(rfkill_resume_polling);
806 static int rfkill_suspend(struct device *dev, pm_message_t state)
808 struct rfkill *rfkill = to_rfkill(dev);
810 rfkill_pause_polling(rfkill);
812 return 0;
815 static int rfkill_resume(struct device *dev)
817 struct rfkill *rfkill = to_rfkill(dev);
818 bool cur;
820 if (!rfkill->persistent) {
821 cur = !!(rfkill->state & RFKILL_BLOCK_SW);
822 rfkill_set_block(rfkill, cur);
825 rfkill_resume_polling(rfkill);
827 return 0;
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,
839 bool rfkill_blocked(struct rfkill *rfkill)
841 unsigned long flags;
842 u32 state;
844 spin_lock_irqsave(&rfkill->lock, flags);
845 state = rfkill->state;
846 spin_unlock_irqrestore(&rfkill->lock, flags);
848 return !!(state & RFKILL_BLOCK_ANY);
850 EXPORT_SYMBOL(rfkill_blocked);
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)
859 struct rfkill *rfkill;
860 struct device *dev;
862 if (WARN_ON(!ops))
863 return NULL;
865 if (WARN_ON(!ops->set_block))
866 return NULL;
868 if (WARN_ON(!name))
869 return NULL;
871 if (WARN_ON(type == RFKILL_TYPE_ALL || type >= NUM_RFKILL_TYPES))
872 return NULL;
874 rfkill = kzalloc(sizeof(*rfkill), GFP_KERNEL);
875 if (!rfkill)
876 return NULL;
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;
885 dev = &rfkill->dev;
886 dev->class = &rfkill_class;
887 dev->parent = parent;
888 device_initialize(dev);
890 return rfkill;
892 EXPORT_SYMBOL(rfkill_alloc);
894 static void rfkill_poll(struct work_struct *work)
896 struct rfkill *rfkill;
898 rfkill = container_of(work, struct rfkill, poll_work.work);
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.
905 rfkill->ops->poll(rfkill, rfkill->data);
907 schedule_delayed_work(&rfkill->poll_work,
908 round_jiffies_relative(POLL_INTERVAL));
911 static void rfkill_uevent_work(struct work_struct *work)
913 struct rfkill *rfkill;
915 rfkill = container_of(work, struct rfkill, uevent_work);
917 mutex_lock(&rfkill_global_mutex);
918 rfkill_event(rfkill);
919 mutex_unlock(&rfkill_global_mutex);
922 static void rfkill_sync_work(struct work_struct *work)
924 struct rfkill *rfkill;
925 bool cur;
927 rfkill = container_of(work, struct rfkill, sync_work);
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);
935 int __must_check rfkill_register(struct rfkill *rfkill)
937 static unsigned long rfkill_no;
938 struct device *dev = &rfkill->dev;
939 int error;
941 BUG_ON(!rfkill);
943 mutex_lock(&rfkill_global_mutex);
945 if (rfkill->registered) {
946 error = -EALREADY;
947 goto unlock;
950 rfkill->idx = rfkill_no;
951 dev_set_name(dev, "rfkill%lu", rfkill_no);
952 rfkill_no++;
954 list_add_tail(&rfkill->node, &rfkill_list);
956 error = device_add(dev);
957 if (error)
958 goto remove;
960 error = rfkill_led_trigger_register(rfkill);
961 if (error)
962 goto devdel;
964 rfkill->registered = true;
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);
970 if (rfkill->ops->poll)
971 schedule_delayed_work(&rfkill->poll_work,
972 round_jiffies_relative(POLL_INTERVAL));
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);
980 if (!atomic_read(&rfkill_input_disabled))
981 __rfkill_switch_all(rfkill->type, soft_blocked);
982 #endif
985 rfkill_send_events(rfkill, RFKILL_OP_ADD);
987 mutex_unlock(&rfkill_global_mutex);
988 return 0;
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;
998 EXPORT_SYMBOL(rfkill_register);
1000 void rfkill_unregister(struct rfkill *rfkill)
1002 BUG_ON(!rfkill);
1004 if (rfkill->ops->poll)
1005 cancel_delayed_work_sync(&rfkill->poll_work);
1007 cancel_work_sync(&rfkill->uevent_work);
1008 cancel_work_sync(&rfkill->sync_work);
1010 rfkill->registered = false;
1012 device_del(&rfkill->dev);
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);
1019 rfkill_led_trigger_unregister(rfkill);
1021 EXPORT_SYMBOL(rfkill_unregister);
1023 void rfkill_destroy(struct rfkill *rfkill)
1025 if (rfkill)
1026 put_device(&rfkill->dev);
1028 EXPORT_SYMBOL(rfkill_destroy);
1030 static int rfkill_fop_open(struct inode *inode, struct file *file)
1032 struct rfkill_data *data;
1033 struct rfkill *rfkill;
1034 struct rfkill_int_event *ev, *tmp;
1036 data = kzalloc(sizeof(*data), GFP_KERNEL);
1037 if (!data)
1038 return -ENOMEM;
1040 INIT_LIST_HEAD(&data->events);
1041 mutex_init(&data->mtx);
1042 init_waitqueue_head(&data->read_wait);
1044 mutex_lock(&rfkill_global_mutex);
1045 mutex_lock(&data->mtx);
1047 * start getting events from elsewhere but hold mtx to get
1048 * startup events added first
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);
1058 list_add(&data->list, &rfkill_fds);
1059 mutex_unlock(&data->mtx);
1060 mutex_unlock(&rfkill_global_mutex);
1062 file->private_data = data;
1064 return nonseekable_open(inode, file);
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;
1076 static unsigned int rfkill_fop_poll(struct file *file, poll_table *wait)
1078 struct rfkill_data *data = file->private_data;
1079 unsigned int res = POLLOUT | POLLWRNORM;
1081 poll_wait(file, &data->read_wait, wait);
1083 mutex_lock(&data->mtx);
1084 if (!list_empty(&data->events))
1085 res = POLLIN | POLLRDNORM;
1086 mutex_unlock(&data->mtx);
1088 return res;
1091 static bool rfkill_readable(struct rfkill_data *data)
1093 bool r;
1095 mutex_lock(&data->mtx);
1096 r = !list_empty(&data->events);
1097 mutex_unlock(&data->mtx);
1099 return r;
1102 static ssize_t rfkill_fop_read(struct file *file, char __user *buf,
1103 size_t count, loff_t *pos)
1105 struct rfkill_data *data = file->private_data;
1106 struct rfkill_int_event *ev;
1107 unsigned long sz;
1108 int ret;
1110 mutex_lock(&data->mtx);
1112 while (list_empty(&data->events)) {
1113 if (file->f_flags & O_NONBLOCK) {
1114 ret = -EAGAIN;
1115 goto out;
1117 mutex_unlock(&data->mtx);
1118 ret = wait_event_interruptible(data->read_wait,
1119 rfkill_readable(data));
1120 mutex_lock(&data->mtx);
1122 if (ret)
1123 goto out;
1126 ev = list_first_entry(&data->events, struct rfkill_int_event,
1127 list);
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;
1134 list_del(&ev->list);
1135 kfree(ev);
1136 out:
1137 mutex_unlock(&data->mtx);
1138 return ret;
1141 static ssize_t rfkill_fop_write(struct file *file, const char __user *buf,
1142 size_t count, loff_t *pos)
1144 struct rfkill *rfkill;
1145 struct rfkill_event ev;
1147 /* we don't need the 'hard' variable but accept it */
1148 if (count < RFKILL_EVENT_SIZE_V1 - 1)
1149 return -EINVAL;
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.
1156 count = min(count, sizeof(ev));
1157 if (copy_from_user(&ev, buf, count))
1158 return -EFAULT;
1160 if (ev.op != RFKILL_OP_CHANGE && ev.op != RFKILL_OP_CHANGE_ALL)
1161 return -EINVAL;
1163 if (ev.type >= NUM_RFKILL_TYPES)
1164 return -EINVAL;
1166 mutex_lock(&rfkill_global_mutex);
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;
1178 list_for_each_entry(rfkill, &rfkill_list, node) {
1179 if (rfkill->idx != ev.idx && ev.op != RFKILL_OP_CHANGE_ALL)
1180 continue;
1182 if (rfkill->type != ev.type && ev.type != RFKILL_TYPE_ALL)
1183 continue;
1185 rfkill_set_block(rfkill, ev.soft);
1187 mutex_unlock(&rfkill_global_mutex);
1189 return count;
1192 static int rfkill_fop_release(struct inode *inode, struct file *file)
1194 struct rfkill_data *data = file->private_data;
1195 struct rfkill_int_event *ev, *tmp;
1197 mutex_lock(&rfkill_global_mutex);
1198 list_del(&data->list);
1199 mutex_unlock(&rfkill_global_mutex);
1201 mutex_destroy(&data->mtx);
1202 list_for_each_entry_safe(ev, tmp, &data->events, list)
1203 kfree(ev);
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
1211 kfree(data);
1213 return 0;
1216 #ifdef CONFIG_RFKILL_INPUT
1217 static long rfkill_fop_ioctl(struct file *file, unsigned int cmd,
1218 unsigned long arg)
1220 struct rfkill_data *data = file->private_data;
1222 if (_IOC_TYPE(cmd) != RFKILL_IOC_MAGIC)
1223 return -ENOSYS;
1225 if (_IOC_NR(cmd) != RFKILL_IOC_NOINPUT)
1226 return -ENOSYS;
1228 mutex_lock(&data->mtx);
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;
1236 mutex_unlock(&data->mtx);
1238 return 0;
1240 #endif
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,
1256 static struct miscdevice rfkill_miscdev = {
1257 .name = "rfkill",
1258 .fops = &rfkill_fops,
1259 .minor = MISC_DYNAMIC_MINOR,
1262 static int __init rfkill_init(void)
1264 int error;
1265 int i;
1267 for (i = 0; i < NUM_RFKILL_TYPES; i++)
1268 rfkill_global_states[i].cur = !rfkill_default_state;
1270 error = class_register(&rfkill_class);
1271 if (error)
1272 goto out;
1274 error = misc_register(&rfkill_miscdev);
1275 if (error) {
1276 class_unregister(&rfkill_class);
1277 goto out;
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;
1287 #endif
1289 out:
1290 return error;
1292 subsys_initcall(rfkill_init);
1294 static void __exit rfkill_exit(void)
1296 #ifdef CONFIG_RFKILL_INPUT
1297 rfkill_handler_exit();
1298 #endif
1299 misc_deregister(&rfkill_miscdev);
1300 class_unregister(&rfkill_class);
1302 module_exit(rfkill_exit);