~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/rfkill/core.c

Version: ~ [ linux-6.6-rc1 ] ~ [ linux-6.5.2 ] ~ [ linux-6.4.15 ] ~ [ linux-6.3.13 ] ~ [ linux-6.2.16 ] ~ [ linux-6.1.52 ] ~ [ linux-6.0.19 ] ~ [ linux-5.19.17 ] ~ [ linux-5.18.19 ] ~ [ linux-5.17.15 ] ~ [ linux-5.16.20 ] ~ [ linux-5.15.131 ] ~ [ linux-5.14.21 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.194 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.256 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.294 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.325 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.337 ] ~ [ linux-4.4.302 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.9 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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, see <http://www.gnu.org/licenses/>.
 18  */
 19 
 20 #include <linux/kernel.h>
 21 #include <linux/module.h>
 22 #include <linux/init.h>
 23 #include <linux/workqueue.h>
 24 #include <linux/capability.h>
 25 #include <linux/list.h>
 26 #include <linux/mutex.h>
 27 #include <linux/rfkill.h>
 28 #include <linux/sched.h>
 29 #include <linux/spinlock.h>
 30 #include <linux/device.h>
 31 #include <linux/miscdevice.h>
 32 #include <linux/wait.h>
 33 #include <linux/poll.h>
 34 #include <linux/fs.h>
 35 #include <linux/slab.h>
 36 
 37 #include "rfkill.h"
 38 
 39 #define POLL_INTERVAL           (5 * HZ)
 40 
 41 #define RFKILL_BLOCK_HW         BIT(0)
 42 #define RFKILL_BLOCK_SW         BIT(1)
 43 #define RFKILL_BLOCK_SW_PREV    BIT(2)
 44 #define RFKILL_BLOCK_ANY        (RFKILL_BLOCK_HW |\
 45                                  RFKILL_BLOCK_SW |\
 46                                  RFKILL_BLOCK_SW_PREV)
 47 #define RFKILL_BLOCK_SW_SETCALL BIT(31)
 48 
 49 struct rfkill {
 50         spinlock_t              lock;
 51 
 52         enum rfkill_type        type;
 53 
 54         unsigned long           state;
 55 
 56         u32                     idx;
 57 
 58         bool                    registered;
 59         bool                    persistent;
 60         bool                    polling_paused;
 61         bool                    suspended;
 62 
 63         const struct rfkill_ops *ops;
 64         void                    *data;
 65 
 66 #ifdef CONFIG_RFKILL_LEDS
 67         struct led_trigger      led_trigger;
 68         const char              *ledtrigname;
 69 #endif
 70 
 71         struct device           dev;
 72         struct list_head        node;
 73 
 74         struct delayed_work     poll_work;
 75         struct work_struct      uevent_work;
 76         struct work_struct      sync_work;
 77         char                    name[];
 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 int 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         return 0;
153 }
154 
155 const char *rfkill_get_led_trigger_name(struct rfkill *rfkill)
156 {
157         return rfkill->led_trigger.name;
158 }
159 EXPORT_SYMBOL(rfkill_get_led_trigger_name);
160 
161 void rfkill_set_led_trigger_name(struct rfkill *rfkill, const char *name)
162 {
163         BUG_ON(!rfkill);
164 
165         rfkill->ledtrigname = name;
166 }
167 EXPORT_SYMBOL(rfkill_set_led_trigger_name);
168 
169 static int rfkill_led_trigger_register(struct rfkill *rfkill)
170 {
171         rfkill->led_trigger.name = rfkill->ledtrigname
172                                         ? : dev_name(&rfkill->dev);
173         rfkill->led_trigger.activate = rfkill_led_trigger_activate;
174         return led_trigger_register(&rfkill->led_trigger);
175 }
176 
177 static void rfkill_led_trigger_unregister(struct rfkill *rfkill)
178 {
179         led_trigger_unregister(&rfkill->led_trigger);
180 }
181 
182 static struct led_trigger rfkill_any_led_trigger;
183 static struct led_trigger rfkill_none_led_trigger;
184 static struct work_struct rfkill_global_led_trigger_work;
185 
186 static void rfkill_global_led_trigger_worker(struct work_struct *work)
187 {
188         enum led_brightness brightness = LED_OFF;
189         struct rfkill *rfkill;
190 
191         mutex_lock(&rfkill_global_mutex);
192         list_for_each_entry(rfkill, &rfkill_list, node) {
193                 if (!(rfkill->state & RFKILL_BLOCK_ANY)) {
194                         brightness = LED_FULL;
195                         break;
196                 }
197         }
198         mutex_unlock(&rfkill_global_mutex);
199 
200         led_trigger_event(&rfkill_any_led_trigger, brightness);
201         led_trigger_event(&rfkill_none_led_trigger,
202                           brightness == LED_OFF ? LED_FULL : LED_OFF);
203 }
204 
205 static void rfkill_global_led_trigger_event(void)
206 {
207         schedule_work(&rfkill_global_led_trigger_work);
208 }
209 
210 static int rfkill_global_led_trigger_register(void)
211 {
212         int ret;
213 
214         INIT_WORK(&rfkill_global_led_trigger_work,
215                         rfkill_global_led_trigger_worker);
216 
217         rfkill_any_led_trigger.name = "rfkill-any";
218         ret = led_trigger_register(&rfkill_any_led_trigger);
219         if (ret)
220                 return ret;
221 
222         rfkill_none_led_trigger.name = "rfkill-none";
223         ret = led_trigger_register(&rfkill_none_led_trigger);
224         if (ret)
225                 led_trigger_unregister(&rfkill_any_led_trigger);
226         else
227                 /* Delay activation until all global triggers are registered */
228                 rfkill_global_led_trigger_event();
229 
230         return ret;
231 }
232 
233 static void rfkill_global_led_trigger_unregister(void)
234 {
235         led_trigger_unregister(&rfkill_none_led_trigger);
236         led_trigger_unregister(&rfkill_any_led_trigger);
237         cancel_work_sync(&rfkill_global_led_trigger_work);
238 }
239 #else
240 static void rfkill_led_trigger_event(struct rfkill *rfkill)
241 {
242 }
243 
244 static inline int rfkill_led_trigger_register(struct rfkill *rfkill)
245 {
246         return 0;
247 }
248 
249 static inline void rfkill_led_trigger_unregister(struct rfkill *rfkill)
250 {
251 }
252 
253 static void rfkill_global_led_trigger_event(void)
254 {
255 }
256 
257 static int rfkill_global_led_trigger_register(void)
258 {
259         return 0;
260 }
261 
262 static void rfkill_global_led_trigger_unregister(void)
263 {
264 }
265 #endif /* CONFIG_RFKILL_LEDS */
266 
267 static void rfkill_fill_event(struct rfkill_event *ev, struct rfkill *rfkill,
268                               enum rfkill_operation op)
269 {
270         unsigned long flags;
271 
272         ev->idx = rfkill->idx;
273         ev->type = rfkill->type;
274         ev->op = op;
275 
276         spin_lock_irqsave(&rfkill->lock, flags);
277         ev->hard = !!(rfkill->state & RFKILL_BLOCK_HW);
278         ev->soft = !!(rfkill->state & (RFKILL_BLOCK_SW |
279                                         RFKILL_BLOCK_SW_PREV));
280         spin_unlock_irqrestore(&rfkill->lock, flags);
281 }
282 
283 static void rfkill_send_events(struct rfkill *rfkill, enum rfkill_operation op)
284 {
285         struct rfkill_data *data;
286         struct rfkill_int_event *ev;
287 
288         list_for_each_entry(data, &rfkill_fds, list) {
289                 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
290                 if (!ev)
291                         continue;
292                 rfkill_fill_event(&ev->ev, rfkill, op);
293                 mutex_lock(&data->mtx);
294                 list_add_tail(&ev->list, &data->events);
295                 mutex_unlock(&data->mtx);
296                 wake_up_interruptible(&data->read_wait);
297         }
298 }
299 
300 static void rfkill_event(struct rfkill *rfkill)
301 {
302         if (!rfkill->registered)
303                 return;
304 
305         kobject_uevent(&rfkill->dev.kobj, KOBJ_CHANGE);
306 
307         /* also send event to /dev/rfkill */
308         rfkill_send_events(rfkill, RFKILL_OP_CHANGE);
309 }
310 
311 /**
312  * rfkill_set_block - wrapper for set_block method
313  *
314  * @rfkill: the rfkill struct to use
315  * @blocked: the new software state
316  *
317  * Calls the set_block method (when applicable) and handles notifications
318  * etc. as well.
319  */
320 static void rfkill_set_block(struct rfkill *rfkill, bool blocked)
321 {
322         unsigned long flags;
323         bool prev, curr;
324         int err;
325 
326         if (unlikely(rfkill->dev.power.power_state.event & PM_EVENT_SLEEP))
327                 return;
328 
329         /*
330          * Some platforms (...!) generate input events which affect the
331          * _hard_ kill state -- whenever something tries to change the
332          * current software state query the hardware state too.
333          */
334         if (rfkill->ops->query)
335                 rfkill->ops->query(rfkill, rfkill->data);
336 
337         spin_lock_irqsave(&rfkill->lock, flags);
338         prev = rfkill->state & RFKILL_BLOCK_SW;
339 
340         if (prev)
341                 rfkill->state |= RFKILL_BLOCK_SW_PREV;
342         else
343                 rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
344 
345         if (blocked)
346                 rfkill->state |= RFKILL_BLOCK_SW;
347         else
348                 rfkill->state &= ~RFKILL_BLOCK_SW;
349 
350         rfkill->state |= RFKILL_BLOCK_SW_SETCALL;
351         spin_unlock_irqrestore(&rfkill->lock, flags);
352 
353         err = rfkill->ops->set_block(rfkill->data, blocked);
354 
355         spin_lock_irqsave(&rfkill->lock, flags);
356         if (err) {
357                 /*
358                  * Failed -- reset status to _PREV, which may be different
359                  * from what we have set _PREV to earlier in this function
360                  * if rfkill_set_sw_state was invoked.
361                  */
362                 if (rfkill->state & RFKILL_BLOCK_SW_PREV)
363                         rfkill->state |= RFKILL_BLOCK_SW;
364                 else
365                         rfkill->state &= ~RFKILL_BLOCK_SW;
366         }
367         rfkill->state &= ~RFKILL_BLOCK_SW_SETCALL;
368         rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
369         curr = rfkill->state & RFKILL_BLOCK_SW;
370         spin_unlock_irqrestore(&rfkill->lock, flags);
371 
372         rfkill_led_trigger_event(rfkill);
373         rfkill_global_led_trigger_event();
374 
375         if (prev != curr)
376                 rfkill_event(rfkill);
377 }
378 
379 static void rfkill_update_global_state(enum rfkill_type type, bool blocked)
380 {
381         int i;
382 
383         if (type != RFKILL_TYPE_ALL) {
384                 rfkill_global_states[type].cur = blocked;
385                 return;
386         }
387 
388         for (i = 0; i < NUM_RFKILL_TYPES; i++)
389                 rfkill_global_states[i].cur = blocked;
390 }
391 
392 #ifdef CONFIG_RFKILL_INPUT
393 static atomic_t rfkill_input_disabled = ATOMIC_INIT(0);
394 
395 /**
396  * __rfkill_switch_all - Toggle state of all switches of given type
397  * @type: type of interfaces to be affected
398  * @blocked: the new state
399  *
400  * This function sets the state of all switches of given type,
401  * unless a specific switch is suspended.
402  *
403  * Caller must have acquired rfkill_global_mutex.
404  */
405 static void __rfkill_switch_all(const enum rfkill_type type, bool blocked)
406 {
407         struct rfkill *rfkill;
408 
409         rfkill_update_global_state(type, blocked);
410         list_for_each_entry(rfkill, &rfkill_list, node) {
411                 if (rfkill->type != type && type != RFKILL_TYPE_ALL)
412                         continue;
413 
414                 rfkill_set_block(rfkill, blocked);
415         }
416 }
417 
418 /**
419  * rfkill_switch_all - Toggle state of all switches of given type
420  * @type: type of interfaces to be affected
421  * @blocked: the new state
422  *
423  * Acquires rfkill_global_mutex and calls __rfkill_switch_all(@type, @state).
424  * Please refer to __rfkill_switch_all() for details.
425  *
426  * Does nothing if the EPO lock is active.
427  */
428 void rfkill_switch_all(enum rfkill_type type, bool blocked)
429 {
430         if (atomic_read(&rfkill_input_disabled))
431                 return;
432 
433         mutex_lock(&rfkill_global_mutex);
434 
435         if (!rfkill_epo_lock_active)
436                 __rfkill_switch_all(type, blocked);
437 
438         mutex_unlock(&rfkill_global_mutex);
439 }
440 
441 /**
442  * rfkill_epo - emergency power off all transmitters
443  *
444  * This kicks all non-suspended rfkill devices to RFKILL_STATE_SOFT_BLOCKED,
445  * ignoring everything in its path but rfkill_global_mutex and rfkill->mutex.
446  *
447  * The global state before the EPO is saved and can be restored later
448  * using rfkill_restore_states().
449  */
450 void rfkill_epo(void)
451 {
452         struct rfkill *rfkill;
453         int i;
454 
455         if (atomic_read(&rfkill_input_disabled))
456                 return;
457 
458         mutex_lock(&rfkill_global_mutex);
459 
460         rfkill_epo_lock_active = true;
461         list_for_each_entry(rfkill, &rfkill_list, node)
462                 rfkill_set_block(rfkill, true);
463 
464         for (i = 0; i < NUM_RFKILL_TYPES; i++) {
465                 rfkill_global_states[i].sav = rfkill_global_states[i].cur;
466                 rfkill_global_states[i].cur = true;
467         }
468 
469         mutex_unlock(&rfkill_global_mutex);
470 }
471 
472 /**
473  * rfkill_restore_states - restore global states
474  *
475  * Restore (and sync switches to) the global state from the
476  * states in rfkill_default_states.  This can undo the effects of
477  * a call to rfkill_epo().
478  */
479 void rfkill_restore_states(void)
480 {
481         int i;
482 
483         if (atomic_read(&rfkill_input_disabled))
484                 return;
485 
486         mutex_lock(&rfkill_global_mutex);
487 
488         rfkill_epo_lock_active = false;
489         for (i = 0; i < NUM_RFKILL_TYPES; i++)
490                 __rfkill_switch_all(i, rfkill_global_states[i].sav);
491         mutex_unlock(&rfkill_global_mutex);
492 }
493 
494 /**
495  * rfkill_remove_epo_lock - unlock state changes
496  *
497  * Used by rfkill-input manually unlock state changes, when
498  * the EPO switch is deactivated.
499  */
500 void rfkill_remove_epo_lock(void)
501 {
502         if (atomic_read(&rfkill_input_disabled))
503                 return;
504 
505         mutex_lock(&rfkill_global_mutex);
506         rfkill_epo_lock_active = false;
507         mutex_unlock(&rfkill_global_mutex);
508 }
509 
510 /**
511  * rfkill_is_epo_lock_active - returns true EPO is active
512  *
513  * Returns 0 (false) if there is NOT an active EPO condition,
514  * and 1 (true) if there is an active EPO condition, which
515  * locks all radios in one of the BLOCKED states.
516  *
517  * Can be called in atomic context.
518  */
519 bool rfkill_is_epo_lock_active(void)
520 {
521         return rfkill_epo_lock_active;
522 }
523 
524 /**
525  * rfkill_get_global_sw_state - returns global state for a type
526  * @type: the type to get the global state of
527  *
528  * Returns the current global state for a given wireless
529  * device type.
530  */
531 bool rfkill_get_global_sw_state(const enum rfkill_type type)
532 {
533         return rfkill_global_states[type].cur;
534 }
535 #endif
536 
537 bool rfkill_set_hw_state(struct rfkill *rfkill, bool blocked)
538 {
539         unsigned long flags;
540         bool ret, prev;
541 
542         BUG_ON(!rfkill);
543 
544         spin_lock_irqsave(&rfkill->lock, flags);
545         prev = !!(rfkill->state & RFKILL_BLOCK_HW);
546         if (blocked)
547                 rfkill->state |= RFKILL_BLOCK_HW;
548         else
549                 rfkill->state &= ~RFKILL_BLOCK_HW;
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 }
561 EXPORT_SYMBOL(rfkill_set_hw_state);
562 
563 static 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 
577 bool 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 }
602 EXPORT_SYMBOL(rfkill_set_sw_state);
603 
604 void 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 }
616 EXPORT_SYMBOL(rfkill_init_sw_state);
617 
618 void 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 }
651 EXPORT_SYMBOL(rfkill_set_states);
652 
653 static 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 
665 enum 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 }
679 EXPORT_SYMBOL(rfkill_find_type);
680 
681 static 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 }
688 static DEVICE_ATTR_RO(name);
689 
690 static 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 }
697 static DEVICE_ATTR_RO(type);
698 
699 static 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 }
706 static DEVICE_ATTR_RO(index);
707 
708 static 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 }
715 static DEVICE_ATTR_RO(persistent);
716 
717 static 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 }
724 static DEVICE_ATTR_RO(hard);
725 
726 static 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 
734 static 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 }
757 static DEVICE_ATTR_RW(soft);
758 
759 static u8 user_state_from_blocked(unsigned long state)
760 {
761         if (state & RFKILL_BLOCK_HW)
762                 return RFKILL_USER_STATE_HARD_BLOCKED;
763         if (state & RFKILL_BLOCK_SW)
764                 return RFKILL_USER_STATE_SOFT_BLOCKED;
765 
766         return RFKILL_USER_STATE_UNBLOCKED;
767 }
768 
769 static ssize_t state_show(struct device *dev, struct device_attribute *attr,
770                           char *buf)
771 {
772         struct rfkill *rfkill = to_rfkill(dev);
773 
774         return sprintf(buf, "%d\n", user_state_from_blocked(rfkill->state));
775 }
776 
777 static ssize_t state_store(struct device *dev, struct device_attribute *attr,
778                            const char *buf, size_t count)
779 {
780         struct rfkill *rfkill = to_rfkill(dev);
781         unsigned long state;
782         int err;
783 
784         if (!capable(CAP_NET_ADMIN))
785                 return -EPERM;
786 
787         err = kstrtoul(buf, 0, &state);
788         if (err)
789                 return err;
790 
791         if (state != RFKILL_USER_STATE_SOFT_BLOCKED &&
792             state != RFKILL_USER_STATE_UNBLOCKED)
793                 return -EINVAL;
794 
795         mutex_lock(&rfkill_global_mutex);
796         rfkill_set_block(rfkill, state == RFKILL_USER_STATE_SOFT_BLOCKED);
797         mutex_unlock(&rfkill_global_mutex);
798 
799         return count;
800 }
801 static DEVICE_ATTR_RW(state);
802 
803 static struct attribute *rfkill_dev_attrs[] = {
804         &dev_attr_name.attr,
805         &dev_attr_type.attr,
806         &dev_attr_index.attr,
807         &dev_attr_persistent.attr,
808         &dev_attr_state.attr,
809         &dev_attr_soft.attr,
810         &dev_attr_hard.attr,
811         NULL,
812 };
813 ATTRIBUTE_GROUPS(rfkill_dev);
814 
815 static void rfkill_release(struct device *dev)
816 {
817         struct rfkill *rfkill = to_rfkill(dev);
818 
819         kfree(rfkill);
820 }
821 
822 static int rfkill_dev_uevent(struct device *dev, struct kobj_uevent_env *env)
823 {
824         struct rfkill *rfkill = to_rfkill(dev);
825         unsigned long flags;
826         u32 state;
827         int error;
828 
829         error = add_uevent_var(env, "RFKILL_NAME=%s", rfkill->name);
830         if (error)
831                 return error;
832         error = add_uevent_var(env, "RFKILL_TYPE=%s",
833                                rfkill_types[rfkill->type]);
834         if (error)
835                 return error;
836         spin_lock_irqsave(&rfkill->lock, flags);
837         state = rfkill->state;
838         spin_unlock_irqrestore(&rfkill->lock, flags);
839         error = add_uevent_var(env, "RFKILL_STATE=%d",
840                                user_state_from_blocked(state));
841         return error;
842 }
843 
844 void rfkill_pause_polling(struct rfkill *rfkill)
845 {
846         BUG_ON(!rfkill);
847 
848         if (!rfkill->ops->poll)
849                 return;
850 
851         rfkill->polling_paused = true;
852         cancel_delayed_work_sync(&rfkill->poll_work);
853 }
854 EXPORT_SYMBOL(rfkill_pause_polling);
855 
856 void rfkill_resume_polling(struct rfkill *rfkill)
857 {
858         BUG_ON(!rfkill);
859 
860         if (!rfkill->ops->poll)
861                 return;
862 
863         rfkill->polling_paused = false;
864 
865         if (rfkill->suspended)
866                 return;
867 
868         queue_delayed_work(system_power_efficient_wq,
869                            &rfkill->poll_work, 0);
870 }
871 EXPORT_SYMBOL(rfkill_resume_polling);
872 
873 #ifdef CONFIG_PM_SLEEP
874 static int rfkill_suspend(struct device *dev)
875 {
876         struct rfkill *rfkill = to_rfkill(dev);
877 
878         rfkill->suspended = true;
879         cancel_delayed_work_sync(&rfkill->poll_work);
880 
881         return 0;
882 }
883 
884 static int rfkill_resume(struct device *dev)
885 {
886         struct rfkill *rfkill = to_rfkill(dev);
887         bool cur;
888 
889         rfkill->suspended = false;
890 
891         if (!rfkill->persistent) {
892                 cur = !!(rfkill->state & RFKILL_BLOCK_SW);
893                 rfkill_set_block(rfkill, cur);
894         }
895 
896         if (rfkill->ops->poll && !rfkill->polling_paused)
897                 queue_delayed_work(system_power_efficient_wq,
898                                    &rfkill->poll_work, 0);
899 
900         return 0;
901 }
902 
903 static SIMPLE_DEV_PM_OPS(rfkill_pm_ops, rfkill_suspend, rfkill_resume);
904 #define RFKILL_PM_OPS (&rfkill_pm_ops)
905 #else
906 #define RFKILL_PM_OPS NULL
907 #endif
908 
909 static struct class rfkill_class = {
910         .name           = "rfkill",
911         .dev_release    = rfkill_release,
912         .dev_groups     = rfkill_dev_groups,
913         .dev_uevent     = rfkill_dev_uevent,
914         .pm             = RFKILL_PM_OPS,
915 };
916 
917 bool rfkill_blocked(struct rfkill *rfkill)
918 {
919         unsigned long flags;
920         u32 state;
921 
922         spin_lock_irqsave(&rfkill->lock, flags);
923         state = rfkill->state;
924         spin_unlock_irqrestore(&rfkill->lock, flags);
925 
926         return !!(state & RFKILL_BLOCK_ANY);
927 }
928 EXPORT_SYMBOL(rfkill_blocked);
929 
930 
931 struct rfkill * __must_check rfkill_alloc(const char *name,
932                                           struct device *parent,
933                                           const enum rfkill_type type,
934                                           const struct rfkill_ops *ops,
935                                           void *ops_data)
936 {
937         struct rfkill *rfkill;
938         struct device *dev;
939 
940         if (WARN_ON(!ops))
941                 return NULL;
942 
943         if (WARN_ON(!ops->set_block))
944                 return NULL;
945 
946         if (WARN_ON(!name))
947                 return NULL;
948 
949         if (WARN_ON(type == RFKILL_TYPE_ALL || type >= NUM_RFKILL_TYPES))
950                 return NULL;
951 
952         rfkill = kzalloc(sizeof(*rfkill) + strlen(name) + 1, GFP_KERNEL);
953         if (!rfkill)
954                 return NULL;
955 
956         spin_lock_init(&rfkill->lock);
957         INIT_LIST_HEAD(&rfkill->node);
958         rfkill->type = type;
959         strcpy(rfkill->name, name);
960         rfkill->ops = ops;
961         rfkill->data = ops_data;
962 
963         dev = &rfkill->dev;
964         dev->class = &rfkill_class;
965         dev->parent = parent;
966         device_initialize(dev);
967 
968         return rfkill;
969 }
970 EXPORT_SYMBOL(rfkill_alloc);
971 
972 static void rfkill_poll(struct work_struct *work)
973 {
974         struct rfkill *rfkill;
975 
976         rfkill = container_of(work, struct rfkill, poll_work.work);
977 
978         /*
979          * Poll hardware state -- driver will use one of the
980          * rfkill_set{,_hw,_sw}_state functions and use its
981          * return value to update the current status.
982          */
983         rfkill->ops->poll(rfkill, rfkill->data);
984 
985         queue_delayed_work(system_power_efficient_wq,
986                 &rfkill->poll_work,
987                 round_jiffies_relative(POLL_INTERVAL));
988 }
989 
990 static void rfkill_uevent_work(struct work_struct *work)
991 {
992         struct rfkill *rfkill;
993 
994         rfkill = container_of(work, struct rfkill, uevent_work);
995 
996         mutex_lock(&rfkill_global_mutex);
997         rfkill_event(rfkill);
998         mutex_unlock(&rfkill_global_mutex);
999 }
1000 
1001 static void rfkill_sync_work(struct work_struct *work)
1002 {
1003         struct rfkill *rfkill;
1004         bool cur;
1005 
1006         rfkill = container_of(work, struct rfkill, sync_work);
1007 
1008         mutex_lock(&rfkill_global_mutex);
1009         cur = rfkill_global_states[rfkill->type].cur;
1010         rfkill_set_block(rfkill, cur);
1011         mutex_unlock(&rfkill_global_mutex);
1012 }
1013 
1014 int __must_check rfkill_register(struct rfkill *rfkill)
1015 {
1016         static unsigned long rfkill_no;
1017         struct device *dev = &rfkill->dev;
1018         int error;
1019 
1020         BUG_ON(!rfkill);
1021 
1022         mutex_lock(&rfkill_global_mutex);
1023 
1024         if (rfkill->registered) {
1025                 error = -EALREADY;
1026                 goto unlock;
1027         }
1028 
1029         rfkill->idx = rfkill_no;
1030         dev_set_name(dev, "rfkill%lu", rfkill_no);
1031         rfkill_no++;
1032 
1033         list_add_tail(&rfkill->node, &rfkill_list);
1034 
1035         error = device_add(dev);
1036         if (error)
1037                 goto remove;
1038 
1039         error = rfkill_led_trigger_register(rfkill);
1040         if (error)
1041                 goto devdel;
1042 
1043         rfkill->registered = true;
1044 
1045         INIT_DELAYED_WORK(&rfkill->poll_work, rfkill_poll);
1046         INIT_WORK(&rfkill->uevent_work, rfkill_uevent_work);
1047         INIT_WORK(&rfkill->sync_work, rfkill_sync_work);
1048 
1049         if (rfkill->ops->poll)
1050                 queue_delayed_work(system_power_efficient_wq,
1051                         &rfkill->poll_work,
1052                         round_jiffies_relative(POLL_INTERVAL));
1053 
1054         if (!rfkill->persistent || rfkill_epo_lock_active) {
1055                 schedule_work(&rfkill->sync_work);
1056         } else {
1057 #ifdef CONFIG_RFKILL_INPUT
1058                 bool soft_blocked = !!(rfkill->state & RFKILL_BLOCK_SW);
1059 
1060                 if (!atomic_read(&rfkill_input_disabled))
1061                         __rfkill_switch_all(rfkill->type, soft_blocked);
1062 #endif
1063         }
1064 
1065         rfkill_global_led_trigger_event();
1066         rfkill_send_events(rfkill, RFKILL_OP_ADD);
1067 
1068         mutex_unlock(&rfkill_global_mutex);
1069         return 0;
1070 
1071  devdel:
1072         device_del(&rfkill->dev);
1073  remove:
1074         list_del_init(&rfkill->node);
1075  unlock:
1076         mutex_unlock(&rfkill_global_mutex);
1077         return error;
1078 }
1079 EXPORT_SYMBOL(rfkill_register);
1080 
1081 void rfkill_unregister(struct rfkill *rfkill)
1082 {
1083         BUG_ON(!rfkill);
1084 
1085         if (rfkill->ops->poll)
1086                 cancel_delayed_work_sync(&rfkill->poll_work);
1087 
1088         cancel_work_sync(&rfkill->uevent_work);
1089         cancel_work_sync(&rfkill->sync_work);
1090 
1091         rfkill->registered = false;
1092 
1093         device_del(&rfkill->dev);
1094 
1095         mutex_lock(&rfkill_global_mutex);
1096         rfkill_send_events(rfkill, RFKILL_OP_DEL);
1097         list_del_init(&rfkill->node);
1098         rfkill_global_led_trigger_event();
1099         mutex_unlock(&rfkill_global_mutex);
1100 
1101         rfkill_led_trigger_unregister(rfkill);
1102 }
1103 EXPORT_SYMBOL(rfkill_unregister);
1104 
1105 void rfkill_destroy(struct rfkill *rfkill)
1106 {
1107         if (rfkill)
1108                 put_device(&rfkill->dev);
1109 }
1110 EXPORT_SYMBOL(rfkill_destroy);
1111 
1112 static int rfkill_fop_open(struct inode *inode, struct file *file)
1113 {
1114         struct rfkill_data *data;
1115         struct rfkill *rfkill;
1116         struct rfkill_int_event *ev, *tmp;
1117 
1118         data = kzalloc(sizeof(*data), GFP_KERNEL);
1119         if (!data)
1120                 return -ENOMEM;
1121 
1122         INIT_LIST_HEAD(&data->events);
1123         mutex_init(&data->mtx);
1124         init_waitqueue_head(&data->read_wait);
1125 
1126         mutex_lock(&rfkill_global_mutex);
1127         mutex_lock(&data->mtx);
1128         /*
1129          * start getting events from elsewhere but hold mtx to get
1130          * startup events added first
1131          */
1132 
1133         list_for_each_entry(rfkill, &rfkill_list, node) {
1134                 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1135                 if (!ev)
1136                         goto free;
1137                 rfkill_fill_event(&ev->ev, rfkill, RFKILL_OP_ADD);
1138                 list_add_tail(&ev->list, &data->events);
1139         }
1140         list_add(&data->list, &rfkill_fds);
1141         mutex_unlock(&data->mtx);
1142         mutex_unlock(&rfkill_global_mutex);
1143 
1144         file->private_data = data;
1145 
1146         return nonseekable_open(inode, file);
1147 
1148  free:
1149         mutex_unlock(&data->mtx);
1150         mutex_unlock(&rfkill_global_mutex);
1151         mutex_destroy(&data->mtx);
1152         list_for_each_entry_safe(ev, tmp, &data->events, list)
1153                 kfree(ev);
1154         kfree(data);
1155         return -ENOMEM;
1156 }
1157 
1158 static __poll_t rfkill_fop_poll(struct file *file, poll_table *wait)
1159 {
1160         struct rfkill_data *data = file->private_data;
1161         __poll_t res = EPOLLOUT | EPOLLWRNORM;
1162 
1163         poll_wait(file, &data->read_wait, wait);
1164 
1165         mutex_lock(&data->mtx);
1166         if (!list_empty(&data->events))
1167                 res = EPOLLIN | EPOLLRDNORM;
1168         mutex_unlock(&data->mtx);
1169 
1170         return res;
1171 }
1172 
1173 static ssize_t rfkill_fop_read(struct file *file, char __user *buf,
1174                                size_t count, loff_t *pos)
1175 {
1176         struct rfkill_data *data = file->private_data;
1177         struct rfkill_int_event *ev;
1178         unsigned long sz;
1179         int ret;
1180 
1181         mutex_lock(&data->mtx);
1182 
1183         while (list_empty(&data->events)) {
1184                 if (file->f_flags & O_NONBLOCK) {
1185                         ret = -EAGAIN;
1186                         goto out;
1187                 }
1188                 mutex_unlock(&data->mtx);
1189                 /* since we re-check and it just compares pointers,
1190                  * using !list_empty() without locking isn't a problem
1191                  */
1192                 ret = wait_event_interruptible(data->read_wait,
1193                                                !list_empty(&data->events));
1194                 mutex_lock(&data->mtx);
1195 
1196                 if (ret)
1197                         goto out;
1198         }
1199 
1200         ev = list_first_entry(&data->events, struct rfkill_int_event,
1201                                 list);
1202 
1203         sz = min_t(unsigned long, sizeof(ev->ev), count);
1204         ret = sz;
1205         if (copy_to_user(buf, &ev->ev, sz))
1206                 ret = -EFAULT;
1207 
1208         list_del(&ev->list);
1209         kfree(ev);
1210  out:
1211         mutex_unlock(&data->mtx);
1212         return ret;
1213 }
1214 
1215 static ssize_t rfkill_fop_write(struct file *file, const char __user *buf,
1216                                 size_t count, loff_t *pos)
1217 {
1218         struct rfkill *rfkill;
1219         struct rfkill_event ev;
1220         int ret;
1221 
1222         /* we don't need the 'hard' variable but accept it */
1223         if (count < RFKILL_EVENT_SIZE_V1 - 1)
1224                 return -EINVAL;
1225 
1226         /*
1227          * Copy as much data as we can accept into our 'ev' buffer,
1228          * but tell userspace how much we've copied so it can determine
1229          * our API version even in a write() call, if it cares.
1230          */
1231         count = min(count, sizeof(ev));
1232         if (copy_from_user(&ev, buf, count))
1233                 return -EFAULT;
1234 
1235         if (ev.type >= NUM_RFKILL_TYPES)
1236                 return -EINVAL;
1237 
1238         mutex_lock(&rfkill_global_mutex);
1239 
1240         switch (ev.op) {
1241         case RFKILL_OP_CHANGE_ALL:
1242                 rfkill_update_global_state(ev.type, ev.soft);
1243                 list_for_each_entry(rfkill, &rfkill_list, node)
1244                         if (rfkill->type == ev.type ||
1245                             ev.type == RFKILL_TYPE_ALL)
1246                                 rfkill_set_block(rfkill, ev.soft);
1247                 ret = 0;
1248                 break;
1249         case RFKILL_OP_CHANGE:
1250                 list_for_each_entry(rfkill, &rfkill_list, node)
1251                         if (rfkill->idx == ev.idx &&
1252                             (rfkill->type == ev.type ||
1253                              ev.type == RFKILL_TYPE_ALL))
1254                                 rfkill_set_block(rfkill, ev.soft);
1255                 ret = 0;
1256                 break;
1257         default:
1258                 ret = -EINVAL;
1259                 break;
1260         }
1261 
1262         mutex_unlock(&rfkill_global_mutex);
1263 
1264         return ret ?: count;
1265 }
1266 
1267 static int rfkill_fop_release(struct inode *inode, struct file *file)
1268 {
1269         struct rfkill_data *data = file->private_data;
1270         struct rfkill_int_event *ev, *tmp;
1271 
1272         mutex_lock(&rfkill_global_mutex);
1273         list_del(&data->list);
1274         mutex_unlock(&rfkill_global_mutex);
1275 
1276         mutex_destroy(&data->mtx);
1277         list_for_each_entry_safe(ev, tmp, &data->events, list)
1278                 kfree(ev);
1279 
1280 #ifdef CONFIG_RFKILL_INPUT
1281         if (data->input_handler)
1282                 if (atomic_dec_return(&rfkill_input_disabled) == 0)
1283                         printk(KERN_DEBUG "rfkill: input handler enabled\n");
1284 #endif
1285 
1286         kfree(data);
1287 
1288         return 0;
1289 }
1290 
1291 #ifdef CONFIG_RFKILL_INPUT
1292 static long rfkill_fop_ioctl(struct file *file, unsigned int cmd,
1293                              unsigned long arg)
1294 {
1295         struct rfkill_data *data = file->private_data;
1296 
1297         if (_IOC_TYPE(cmd) != RFKILL_IOC_MAGIC)
1298                 return -ENOSYS;
1299 
1300         if (_IOC_NR(cmd) != RFKILL_IOC_NOINPUT)
1301                 return -ENOSYS;
1302 
1303         mutex_lock(&data->mtx);
1304 
1305         if (!data->input_handler) {
1306                 if (atomic_inc_return(&rfkill_input_disabled) == 1)
1307                         printk(KERN_DEBUG "rfkill: input handler disabled\n");
1308                 data->input_handler = true;
1309         }
1310 
1311         mutex_unlock(&data->mtx);
1312 
1313         return 0;
1314 }
1315 #endif
1316 
1317 static const struct file_operations rfkill_fops = {
1318         .owner          = THIS_MODULE,
1319         .open           = rfkill_fop_open,
1320         .read           = rfkill_fop_read,
1321         .write          = rfkill_fop_write,
1322         .poll           = rfkill_fop_poll,
1323         .release        = rfkill_fop_release,
1324 #ifdef CONFIG_RFKILL_INPUT
1325         .unlocked_ioctl = rfkill_fop_ioctl,
1326         .compat_ioctl   = rfkill_fop_ioctl,
1327 #endif
1328         .llseek         = no_llseek,
1329 };
1330 
1331 static struct miscdevice rfkill_miscdev = {
1332         .name   = "rfkill",
1333         .fops   = &rfkill_fops,
1334         .minor  = MISC_DYNAMIC_MINOR,
1335 };
1336 
1337 static int __init rfkill_init(void)
1338 {
1339         int error;
1340 
1341         rfkill_update_global_state(RFKILL_TYPE_ALL, !rfkill_default_state);
1342 
1343         error = class_register(&rfkill_class);
1344         if (error)
1345                 goto error_class;
1346 
1347         error = misc_register(&rfkill_miscdev);
1348         if (error)
1349                 goto error_misc;
1350 
1351         error = rfkill_global_led_trigger_register();
1352         if (error)
1353                 goto error_led_trigger;
1354 
1355 #ifdef CONFIG_RFKILL_INPUT
1356         error = rfkill_handler_init();
1357         if (error)
1358                 goto error_input;
1359 #endif
1360 
1361         return 0;
1362 
1363 #ifdef CONFIG_RFKILL_INPUT
1364 error_input:
1365         rfkill_global_led_trigger_unregister();
1366 #endif
1367 error_led_trigger:
1368         misc_deregister(&rfkill_miscdev);
1369 error_misc:
1370         class_unregister(&rfkill_class);
1371 error_class:
1372         return error;
1373 }
1374 subsys_initcall(rfkill_init);
1375 
1376 static void __exit rfkill_exit(void)
1377 {
1378 #ifdef CONFIG_RFKILL_INPUT
1379         rfkill_handler_exit();
1380 #endif
1381         rfkill_global_led_trigger_unregister();
1382         misc_deregister(&rfkill_miscdev);
1383         class_unregister(&rfkill_class);
1384 }
1385 module_exit(rfkill_exit);
1386 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp