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