linux/net/rfkill/core.c
<<
>>
Prefs
   1/*
   2 * Copyright (C) 2006 - 2007 Ivo van Doorn
   3 * Copyright (C) 2007 Dmitry Torokhov
   4 * Copyright 2009 Johannes Berg <johannes@sipsolutions.net>
   5 *
   6 * This program is free software; you can redistribute it and/or modify
   7 * it under the terms of the GNU General Public License as published by
   8 * the Free Software Foundation; either version 2 of the License, or
   9 * (at your option) any later version.
  10 *
  11 * This program is distributed in the hope that it will be useful,
  12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  14 * GNU General Public License for more details.
  15 *
  16 * You should have received a copy of the GNU General Public License
  17 * along with this program; if not, write to the
  18 * Free Software Foundation, Inc.,
  19 * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
  20 */
  21
  22#include <linux/kernel.h>
  23#include <linux/module.h>
  24#include <linux/init.h>
  25#include <linux/workqueue.h>
  26#include <linux/capability.h>
  27#include <linux/list.h>
  28#include <linux/mutex.h>
  29#include <linux/rfkill.h>
  30#include <linux/sched.h>
  31#include <linux/spinlock.h>
  32#include <linux/miscdevice.h>
  33#include <linux/wait.h>
  34#include <linux/poll.h>
  35#include <linux/fs.h>
  36#include <linux/slab.h>
  37
  38#include "rfkill.h"
  39
  40#define POLL_INTERVAL           (5 * HZ)
  41
  42#define RFKILL_BLOCK_HW         BIT(0)
  43#define RFKILL_BLOCK_SW         BIT(1)
  44#define RFKILL_BLOCK_SW_PREV    BIT(2)
  45#define RFKILL_BLOCK_ANY        (RFKILL_BLOCK_HW |\
  46                                 RFKILL_BLOCK_SW |\
  47                                 RFKILL_BLOCK_SW_PREV)
  48#define RFKILL_BLOCK_SW_SETCALL BIT(31)
  49
  50struct rfkill {
  51        spinlock_t              lock;
  52
  53        const char              *name;
  54        enum rfkill_type        type;
  55
  56        unsigned long           state;
  57
  58        u32                     idx;
  59
  60        bool                    registered;
  61        bool                    persistent;
  62
  63        const struct rfkill_ops *ops;
  64        void                    *data;
  65
  66#ifdef CONFIG_RFKILL_LEDS
  67        struct led_trigger      led_trigger;
  68        const char              *ledtrigname;
  69#endif
  70
  71        struct device           dev;
  72        struct list_head        node;
  73
  74        struct delayed_work     poll_work;
  75        struct work_struct      uevent_work;
  76        struct work_struct      sync_work;
  77};
  78#define to_rfkill(d)    container_of(d, struct rfkill, dev)
  79
  80struct rfkill_int_event {
  81        struct list_head        list;
  82        struct rfkill_event     ev;
  83};
  84
  85struct rfkill_data {
  86        struct list_head        list;
  87        struct list_head        events;
  88        struct mutex            mtx;
  89        wait_queue_head_t       read_wait;
  90        bool                    input_handler;
  91};
  92
  93
  94MODULE_AUTHOR("Ivo van Doorn <IvDoorn@gmail.com>");
  95MODULE_AUTHOR("Johannes Berg <johannes@sipsolutions.net>");
  96MODULE_DESCRIPTION("RF switch support");
  97MODULE_LICENSE("GPL");
  98
  99
 100/*
 101 * The locking here should be made much smarter, we currently have
 102 * a bit of a stupid situation because drivers might want to register
 103 * the rfkill struct under their own lock, and take this lock during
 104 * rfkill method calls -- which will cause an AB-BA deadlock situation.
 105 *
 106 * To fix that, we need to rework this code here to be mostly lock-free
 107 * and only use the mutex for list manipulations, not to protect the
 108 * various other global variables. Then we can avoid holding the mutex
 109 * around driver operations, and all is happy.
 110 */
 111static LIST_HEAD(rfkill_list);  /* list of registered rf switches */
 112static DEFINE_MUTEX(rfkill_global_mutex);
 113static LIST_HEAD(rfkill_fds);   /* list of open fds of /dev/rfkill */
 114
 115static unsigned int rfkill_default_state = 1;
 116module_param_named(default_state, rfkill_default_state, uint, 0444);
 117MODULE_PARM_DESC(default_state,
 118                 "Default initial state for all radio types, 0 = radio off");
 119
 120static struct {
 121        bool cur, sav;
 122} rfkill_global_states[NUM_RFKILL_TYPES];
 123
 124static bool rfkill_epo_lock_active;
 125
 126
 127#ifdef CONFIG_RFKILL_LEDS
 128static void rfkill_led_trigger_event(struct rfkill *rfkill)
 129{
 130        struct led_trigger *trigger;
 131
 132        if (!rfkill->registered)
 133                return;
 134
 135        trigger = &rfkill->led_trigger;
 136
 137        if (rfkill->state & RFKILL_BLOCK_ANY)
 138                led_trigger_event(trigger, LED_OFF);
 139        else
 140                led_trigger_event(trigger, LED_FULL);
 141}
 142
 143static void rfkill_led_trigger_activate(struct led_classdev *led)
 144{
 145        struct rfkill *rfkill;
 146
 147        rfkill = container_of(led->trigger, struct rfkill, led_trigger);
 148
 149        rfkill_led_trigger_event(rfkill);
 150}
 151
 152static int rfkill_led_trigger_register(struct rfkill *rfkill)
 153{
 154        rfkill->led_trigger.name = rfkill->ledtrigname
 155                                        ? : dev_name(&rfkill->dev);
 156        rfkill->led_trigger.activate = rfkill_led_trigger_activate;
 157        return led_trigger_register(&rfkill->led_trigger);
 158}
 159
 160static void rfkill_led_trigger_unregister(struct rfkill *rfkill)
 161{
 162        led_trigger_unregister(&rfkill->led_trigger);
 163}
 164#else
 165static void rfkill_led_trigger_event(struct rfkill *rfkill)
 166{
 167}
 168
 169static inline int rfkill_led_trigger_register(struct rfkill *rfkill)
 170{
 171        return 0;
 172}
 173
 174static inline void rfkill_led_trigger_unregister(struct rfkill *rfkill)
 175{
 176}
 177#endif /* CONFIG_RFKILL_LEDS */
 178
 179static void rfkill_fill_event(struct rfkill_event *ev, struct rfkill *rfkill,
 180                              enum rfkill_operation op)
 181{
 182        unsigned long flags;
 183
 184        ev->idx = rfkill->idx;
 185        ev->type = rfkill->type;
 186        ev->op = op;
 187
 188        spin_lock_irqsave(&rfkill->lock, flags);
 189        ev->hard = !!(rfkill->state & RFKILL_BLOCK_HW);
 190        ev->soft = !!(rfkill->state & (RFKILL_BLOCK_SW |
 191                                        RFKILL_BLOCK_SW_PREV));
 192        spin_unlock_irqrestore(&rfkill->lock, flags);
 193}
 194
 195static void rfkill_send_events(struct rfkill *rfkill, enum rfkill_operation op)
 196{
 197        struct rfkill_data *data;
 198        struct rfkill_int_event *ev;
 199
 200        list_for_each_entry(data, &rfkill_fds, list) {
 201                ev = kzalloc(sizeof(*ev), GFP_KERNEL);
 202                if (!ev)
 203                        continue;
 204                rfkill_fill_event(&ev->ev, rfkill, op);
 205                mutex_lock(&data->mtx);
 206                list_add_tail(&ev->list, &data->events);
 207                mutex_unlock(&data->mtx);
 208                wake_up_interruptible(&data->read_wait);
 209        }
 210}
 211
 212static void rfkill_event(struct rfkill *rfkill)
 213{
 214        if (!rfkill->registered)
 215                return;
 216
 217        kobject_uevent(&rfkill->dev.kobj, KOBJ_CHANGE);
 218
 219        /* also send event to /dev/rfkill */
 220        rfkill_send_events(rfkill, RFKILL_OP_CHANGE);
 221}
 222
 223static bool __rfkill_set_hw_state(struct rfkill *rfkill,
 224                                  bool blocked, bool *change)
 225{
 226        unsigned long flags;
 227        bool prev, any;
 228
 229        BUG_ON(!rfkill);
 230
 231        spin_lock_irqsave(&rfkill->lock, flags);
 232        prev = !!(rfkill->state & RFKILL_BLOCK_HW);
 233        if (blocked)
 234                rfkill->state |= RFKILL_BLOCK_HW;
 235        else
 236                rfkill->state &= ~RFKILL_BLOCK_HW;
 237        *change = prev != blocked;
 238        any = rfkill->state & RFKILL_BLOCK_ANY;
 239        spin_unlock_irqrestore(&rfkill->lock, flags);
 240
 241        rfkill_led_trigger_event(rfkill);
 242
 243        return any;
 244}
 245
 246/**
 247 * rfkill_set_block - wrapper for set_block method
 248 *
 249 * @rfkill: the rfkill struct to use
 250 * @blocked: the new software state
 251 *
 252 * Calls the set_block method (when applicable) and handles notifications
 253 * etc. as well.
 254 */
 255static void rfkill_set_block(struct rfkill *rfkill, bool blocked)
 256{
 257        unsigned long flags;
 258        int err;
 259
 260        if (unlikely(rfkill->dev.power.power_state.event & PM_EVENT_SLEEP))
 261                return;
 262
 263        /*
 264         * Some platforms (...!) generate input events which affect the
 265         * _hard_ kill state -- whenever something tries to change the
 266         * current software state query the hardware state too.
 267         */
 268        if (rfkill->ops->query)
 269                rfkill->ops->query(rfkill, rfkill->data);
 270
 271        spin_lock_irqsave(&rfkill->lock, flags);
 272        if (rfkill->state & RFKILL_BLOCK_SW)
 273                rfkill->state |= RFKILL_BLOCK_SW_PREV;
 274        else
 275                rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
 276
 277        if (blocked)
 278                rfkill->state |= RFKILL_BLOCK_SW;
 279        else
 280                rfkill->state &= ~RFKILL_BLOCK_SW;
 281
 282        rfkill->state |= RFKILL_BLOCK_SW_SETCALL;
 283        spin_unlock_irqrestore(&rfkill->lock, flags);
 284
 285        err = rfkill->ops->set_block(rfkill->data, blocked);
 286
 287        spin_lock_irqsave(&rfkill->lock, flags);
 288        if (err) {
 289                /*
 290                 * Failed -- reset status to _prev, this may be different
 291                 * from what set set _PREV to earlier in this function
 292                 * if rfkill_set_sw_state was invoked.
 293                 */
 294                if (rfkill->state & RFKILL_BLOCK_SW_PREV)
 295                        rfkill->state |= RFKILL_BLOCK_SW;
 296                else
 297                        rfkill->state &= ~RFKILL_BLOCK_SW;
 298        }
 299        rfkill->state &= ~RFKILL_BLOCK_SW_SETCALL;
 300        rfkill->state &= ~RFKILL_BLOCK_SW_PREV;
 301        spin_unlock_irqrestore(&rfkill->lock, flags);
 302
 303        rfkill_led_trigger_event(rfkill);
 304        rfkill_event(rfkill);
 305}
 306
 307#ifdef CONFIG_RFKILL_INPUT
 308static atomic_t rfkill_input_disabled = ATOMIC_INIT(0);
 309
 310/**
 311 * __rfkill_switch_all - Toggle state of all switches of given type
 312 * @type: type of interfaces to be affected
 313 * @state: the new state
 314 *
 315 * This function sets the state of all switches of given type,
 316 * unless a specific switch is claimed by userspace (in which case,
 317 * that switch is left alone) or suspended.
 318 *
 319 * Caller must have acquired rfkill_global_mutex.
 320 */
 321static void __rfkill_switch_all(const enum rfkill_type type, bool blocked)
 322{
 323        struct rfkill *rfkill;
 324
 325        rfkill_global_states[type].cur = blocked;
 326        list_for_each_entry(rfkill, &rfkill_list, node) {
 327                if (rfkill->type != type)
 328                        continue;
 329
 330                rfkill_set_block(rfkill, blocked);
 331        }
 332}
 333
 334/**
 335 * rfkill_switch_all - Toggle state of all switches of given type
 336 * @type: type of interfaces to be affected
 337 * @state: the new state
 338 *
 339 * Acquires rfkill_global_mutex and calls __rfkill_switch_all(@type, @state).
 340 * Please refer to __rfkill_switch_all() for details.
 341 *
 342 * Does nothing if the EPO lock is active.
 343 */
 344void rfkill_switch_all(enum rfkill_type type, bool blocked)
 345{
 346        if (atomic_read(&rfkill_input_disabled))
 347                return;
 348
 349        mutex_lock(&rfkill_global_mutex);
 350
 351        if (!rfkill_epo_lock_active)
 352                __rfkill_switch_all(type, blocked);
 353
 354        mutex_unlock(&rfkill_global_mutex);
 355}
 356
 357/**
 358 * rfkill_epo - emergency power off all transmitters
 359 *
 360 * This kicks all non-suspended rfkill devices to RFKILL_STATE_SOFT_BLOCKED,
 361 * ignoring everything in its path but rfkill_global_mutex and rfkill->mutex.
 362 *
 363 * The global state before the EPO is saved and can be restored later
 364 * using rfkill_restore_states().
 365 */
 366void rfkill_epo(void)
 367{
 368        struct rfkill *rfkill;
 369        int i;
 370
 371        if (atomic_read(&rfkill_input_disabled))
 372                return;
 373
 374        mutex_lock(&rfkill_global_mutex);
 375
 376        rfkill_epo_lock_active = true;
 377        list_for_each_entry(rfkill, &rfkill_list, node)
 378                rfkill_set_block(rfkill, true);
 379
 380        for (i = 0; i < NUM_RFKILL_TYPES; i++) {
 381                rfkill_global_states[i].sav = rfkill_global_states[i].cur;
 382                rfkill_global_states[i].cur = true;
 383        }
 384
 385        mutex_unlock(&rfkill_global_mutex);
 386}
 387
 388/**
 389 * rfkill_restore_states - restore global states
 390 *
 391 * Restore (and sync switches to) the global state from the
 392 * states in rfkill_default_states.  This can undo the effects of
 393 * a call to rfkill_epo().
 394 */
 395void rfkill_restore_states(void)
 396{
 397        int i;
 398
 399        if (atomic_read(&rfkill_input_disabled))
 400                return;
 401
 402        mutex_lock(&rfkill_global_mutex);
 403
 404        rfkill_epo_lock_active = false;
 405        for (i = 0; i < NUM_RFKILL_TYPES; i++)
 406                __rfkill_switch_all(i, rfkill_global_states[i].sav);
 407        mutex_unlock(&rfkill_global_mutex);
 408}
 409
 410/**
 411 * rfkill_remove_epo_lock - unlock state changes
 412 *
 413 * Used by rfkill-input manually unlock state changes, when
 414 * the EPO switch is deactivated.
 415 */
 416void rfkill_remove_epo_lock(void)
 417{
 418        if (atomic_read(&rfkill_input_disabled))
 419                return;
 420
 421        mutex_lock(&rfkill_global_mutex);
 422        rfkill_epo_lock_active = false;
 423        mutex_unlock(&rfkill_global_mutex);
 424}
 425
 426/**
 427 * rfkill_is_epo_lock_active - returns true EPO is active
 428 *
 429 * Returns 0 (false) if there is NOT an active EPO contidion,
 430 * and 1 (true) if there is an active EPO contition, which
 431 * locks all radios in one of the BLOCKED states.
 432 *
 433 * Can be called in atomic context.
 434 */
 435bool rfkill_is_epo_lock_active(void)
 436{
 437        return rfkill_epo_lock_active;
 438}
 439
 440/**
 441 * rfkill_get_global_sw_state - returns global state for a type
 442 * @type: the type to get the global state of
 443 *
 444 * Returns the current global state for a given wireless
 445 * device type.
 446 */
 447bool rfkill_get_global_sw_state(const enum rfkill_type type)
 448{
 449        return rfkill_global_states[type].cur;
 450}
 451#endif
 452
 453
 454bool rfkill_set_hw_state(struct rfkill *rfkill, bool blocked)
 455{
 456        bool ret, change;
 457
 458        ret = __rfkill_set_hw_state(rfkill, blocked, &change);
 459
 460        if (!rfkill->registered)
 461                return ret;
 462
 463        if (change)
 464                schedule_work(&rfkill->uevent_work);
 465
 466        return ret;
 467}
 468EXPORT_SYMBOL(rfkill_set_hw_state);
 469
 470static void __rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
 471{
 472        u32 bit = RFKILL_BLOCK_SW;
 473
 474        /* if in a ops->set_block right now, use other bit */
 475        if (rfkill->state & RFKILL_BLOCK_SW_SETCALL)
 476                bit = RFKILL_BLOCK_SW_PREV;
 477
 478        if (blocked)
 479                rfkill->state |= bit;
 480        else
 481                rfkill->state &= ~bit;
 482}
 483
 484bool rfkill_set_sw_state(struct rfkill *rfkill, bool blocked)
 485{
 486        unsigned long flags;
 487        bool prev, hwblock;
 488
 489        BUG_ON(!rfkill);
 490
 491        spin_lock_irqsave(&rfkill->lock, flags);
 492        prev = !!(rfkill->state & RFKILL_BLOCK_SW);
 493        __rfkill_set_sw_state(rfkill, blocked);
 494        hwblock = !!(rfkill->state & RFKILL_BLOCK_HW);
 495        blocked = blocked || hwblock;
 496        spin_unlock_irqrestore(&rfkill->lock, flags);
 497
 498        if (!rfkill->registered)
 499                return blocked;
 500
 501        if (prev != blocked && !hwblock)
 502                schedule_work(&rfkill->uevent_work);
 503
 504        rfkill_led_trigger_event(rfkill);
 505
 506        return blocked;
 507}
 508EXPORT_SYMBOL(rfkill_set_sw_state);
 509
 510void rfkill_init_sw_state(struct rfkill *rfkill, bool blocked)
 511{
 512        unsigned long flags;
 513
 514        BUG_ON(!rfkill);
 515        BUG_ON(rfkill->registered);
 516
 517        spin_lock_irqsave(&rfkill->lock, flags);
 518        __rfkill_set_sw_state(rfkill, blocked);
 519        rfkill->persistent = true;
 520        spin_unlock_irqrestore(&rfkill->lock, flags);
 521}
 522EXPORT_SYMBOL(rfkill_init_sw_state);
 523
 524void rfkill_set_states(struct rfkill *rfkill, bool sw, bool hw)
 525{
 526        unsigned long flags;
 527        bool swprev, hwprev;
 528
 529        BUG_ON(!rfkill);
 530
 531        spin_lock_irqsave(&rfkill->lock, flags);
 532
 533        /*
 534         * No need to care about prev/setblock ... this is for uevent only
 535         * and that will get triggered by rfkill_set_block anyway.
 536         */
 537        swprev = !!(rfkill->state & RFKILL_BLOCK_SW);
 538        hwprev = !!(rfkill->state & RFKILL_BLOCK_HW);
 539        __rfkill_set_sw_state(rfkill, sw);
 540        if (hw)
 541                rfkill->state |= RFKILL_BLOCK_HW;
 542        else
 543                rfkill->state &= ~RFKILL_BLOCK_HW;
 544
 545        spin_unlock_irqrestore(&rfkill->lock, flags);
 546
 547        if (!rfkill->registered) {
 548                rfkill->persistent = true;
 549        } else {
 550                if (swprev != sw || hwprev != hw)
 551                        schedule_work(&rfkill->uevent_work);
 552
 553                rfkill_led_trigger_event(rfkill);
 554        }
 555}
 556EXPORT_SYMBOL(rfkill_set_states);
 557
 558static ssize_t rfkill_name_show(struct device *dev,
 559                                struct device_attribute *attr,
 560                                char *buf)
 561{
 562        struct rfkill *rfkill = to_rfkill(dev);
 563
 564        return sprintf(buf, "%s\n", rfkill->name);
 565}
 566
 567static const char *rfkill_get_type_str(enum rfkill_type type)
 568{
 569        BUILD_BUG_ON(NUM_RFKILL_TYPES != RFKILL_TYPE_FM + 1);
 570
 571        switch (type) {
 572        case RFKILL_TYPE_WLAN:
 573                return "wlan";
 574        case RFKILL_TYPE_BLUETOOTH:
 575                return "bluetooth";
 576        case RFKILL_TYPE_UWB:
 577                return "ultrawideband";
 578        case RFKILL_TYPE_WIMAX:
 579                return "wimax";
 580        case RFKILL_TYPE_WWAN:
 581                return "wwan";
 582        case RFKILL_TYPE_GPS:
 583                return "gps";
 584        case RFKILL_TYPE_FM:
 585                return "fm";
 586        default:
 587                BUG();
 588        }
 589}
 590
 591static ssize_t rfkill_type_show(struct device *dev,
 592                                struct device_attribute *attr,
 593                                char *buf)
 594{
 595        struct rfkill *rfkill = to_rfkill(dev);
 596
 597        return sprintf(buf, "%s\n", rfkill_get_type_str(rfkill->type));
 598}
 599
 600static ssize_t rfkill_idx_show(struct device *dev,
 601                               struct device_attribute *attr,
 602                               char *buf)
 603{
 604        struct rfkill *rfkill = to_rfkill(dev);
 605
 606        return sprintf(buf, "%d\n", rfkill->idx);
 607}
 608
 609static ssize_t rfkill_persistent_show(struct device *dev,
 610                               struct device_attribute *attr,
 611                               char *buf)
 612{
 613        struct rfkill *rfkill = to_rfkill(dev);
 614
 615        return sprintf(buf, "%d\n", rfkill->persistent);
 616}
 617
 618static ssize_t rfkill_hard_show(struct device *dev,
 619                                 struct device_attribute *attr,
 620                                 char *buf)
 621{
 622        struct rfkill *rfkill = to_rfkill(dev);
 623
 624        return sprintf(buf, "%d\n", (rfkill->state & RFKILL_BLOCK_HW) ? 1 : 0 );
 625}
 626
 627static ssize_t rfkill_soft_show(struct device *dev,
 628                                 struct device_attribute *attr,
 629                                 char *buf)
 630{
 631        struct rfkill *rfkill = to_rfkill(dev);
 632
 633        return sprintf(buf, "%d\n", (rfkill->state & RFKILL_BLOCK_SW) ? 1 : 0 );
 634}
 635
 636static ssize_t rfkill_soft_store(struct device *dev,
 637                                  struct device_attribute *attr,
 638                                  const char *buf, size_t count)
 639{
 640        struct rfkill *rfkill = to_rfkill(dev);
 641        unsigned long state;
 642        int err;
 643
 644        if (!capable(CAP_NET_ADMIN))
 645                return -EPERM;
 646
 647        err = strict_strtoul(buf, 0, &state);
 648        if (err)
 649                return err;
 650
 651        if (state > 1 )
 652                return -EINVAL;
 653
 654        mutex_lock(&rfkill_global_mutex);
 655        rfkill_set_block(rfkill, state);
 656        mutex_unlock(&rfkill_global_mutex);
 657
 658        return err ?: count;
 659}
 660
 661static u8 user_state_from_blocked(unsigned long state)
 662{
 663        if (state & RFKILL_BLOCK_HW)
 664                return RFKILL_USER_STATE_HARD_BLOCKED;
 665        if (state & RFKILL_BLOCK_SW)
 666                return RFKILL_USER_STATE_SOFT_BLOCKED;
 667
 668        return RFKILL_USER_STATE_UNBLOCKED;
 669}
 670
 671static ssize_t rfkill_state_show(struct device *dev,
 672                                 struct device_attribute *attr,
 673                                 char *buf)
 674{
 675        struct rfkill *rfkill = to_rfkill(dev);
 676
 677        return sprintf(buf, "%d\n", user_state_from_blocked(rfkill->state));
 678}
 679
 680static ssize_t rfkill_state_store(struct device *dev,
 681                                  struct device_attribute *attr,
 682                                  const char *buf, size_t count)
 683{
 684        struct rfkill *rfkill = to_rfkill(dev);
 685        unsigned long state;
 686        int err;
 687
 688        if (!capable(CAP_NET_ADMIN))
 689                return -EPERM;
 690
 691        err = strict_strtoul(buf, 0, &state);
 692        if (err)
 693                return err;
 694
 695        if (state != RFKILL_USER_STATE_SOFT_BLOCKED &&
 696            state != RFKILL_USER_STATE_UNBLOCKED)
 697                return -EINVAL;
 698
 699        mutex_lock(&rfkill_global_mutex);
 700        rfkill_set_block(rfkill, state == RFKILL_USER_STATE_SOFT_BLOCKED);
 701        mutex_unlock(&rfkill_global_mutex);
 702
 703        return err ?: count;
 704}
 705
 706static ssize_t rfkill_claim_show(struct device *dev,
 707                                 struct device_attribute *attr,
 708                                 char *buf)
 709{
 710        return sprintf(buf, "%d\n", 0);
 711}
 712
 713static ssize_t rfkill_claim_store(struct device *dev,
 714                                  struct device_attribute *attr,
 715                                  const char *buf, size_t count)
 716{
 717        return -EOPNOTSUPP;
 718}
 719
 720static struct device_attribute rfkill_dev_attrs[] = {
 721        __ATTR(name, S_IRUGO, rfkill_name_show, NULL),
 722        __ATTR(type, S_IRUGO, rfkill_type_show, NULL),
 723        __ATTR(index, S_IRUGO, rfkill_idx_show, NULL),
 724        __ATTR(persistent, S_IRUGO, rfkill_persistent_show, NULL),
 725        __ATTR(state, S_IRUGO|S_IWUSR, rfkill_state_show, rfkill_state_store),
 726        __ATTR(claim, S_IRUGO|S_IWUSR, rfkill_claim_show, rfkill_claim_store),
 727        __ATTR(soft, S_IRUGO|S_IWUSR, rfkill_soft_show, rfkill_soft_store),
 728        __ATTR(hard, S_IRUGO, rfkill_hard_show, NULL),
 729        __ATTR_NULL
 730};
 731
 732static void rfkill_release(struct device *dev)
 733{
 734        struct rfkill *rfkill = to_rfkill(dev);
 735
 736        kfree(rfkill);
 737}
 738
 739static int rfkill_dev_uevent(struct device *dev, struct kobj_uevent_env *env)
 740{
 741        struct rfkill *rfkill = to_rfkill(dev);
 742        unsigned long flags;
 743        u32 state;
 744        int error;
 745
 746        error = add_uevent_var(env, "RFKILL_NAME=%s", rfkill->name);
 747        if (error)
 748                return error;
 749        error = add_uevent_var(env, "RFKILL_TYPE=%s",
 750                               rfkill_get_type_str(rfkill->type));
 751        if (error)
 752                return error;
 753        spin_lock_irqsave(&rfkill->lock, flags);
 754        state = rfkill->state;
 755        spin_unlock_irqrestore(&rfkill->lock, flags);
 756        error = add_uevent_var(env, "RFKILL_STATE=%d",
 757                               user_state_from_blocked(state));
 758        return error;
 759}
 760
 761void rfkill_pause_polling(struct rfkill *rfkill)
 762{
 763        BUG_ON(!rfkill);
 764
 765        if (!rfkill->ops->poll)
 766                return;
 767
 768        cancel_delayed_work_sync(&rfkill->poll_work);
 769}
 770EXPORT_SYMBOL(rfkill_pause_polling);
 771
 772void rfkill_resume_polling(struct rfkill *rfkill)
 773{
 774        BUG_ON(!rfkill);
 775
 776        if (!rfkill->ops->poll)
 777                return;
 778
 779        schedule_work(&rfkill->poll_work.work);
 780}
 781EXPORT_SYMBOL(rfkill_resume_polling);
 782
 783static int rfkill_suspend(struct device *dev, pm_message_t state)
 784{
 785        struct rfkill *rfkill = to_rfkill(dev);
 786
 787        rfkill_pause_polling(rfkill);
 788
 789        return 0;
 790}
 791
 792static int rfkill_resume(struct device *dev)
 793{
 794        struct rfkill *rfkill = to_rfkill(dev);
 795        bool cur;
 796
 797        if (!rfkill->persistent) {
 798                cur = !!(rfkill->state & RFKILL_BLOCK_SW);
 799                rfkill_set_block(rfkill, cur);
 800        }
 801
 802        rfkill_resume_polling(rfkill);
 803
 804        return 0;
 805}
 806
 807static struct class rfkill_class = {
 808        .name           = "rfkill",
 809        .dev_release    = rfkill_release,
 810        .dev_attrs      = rfkill_dev_attrs,
 811        .dev_uevent     = rfkill_dev_uevent,
 812        .suspend        = rfkill_suspend,
 813        .resume         = rfkill_resume,
 814};
 815
 816bool rfkill_blocked(struct rfkill *rfkill)
 817{
 818        unsigned long flags;
 819        u32 state;
 820
 821        spin_lock_irqsave(&rfkill->lock, flags);
 822        state = rfkill->state;
 823        spin_unlock_irqrestore(&rfkill->lock, flags);
 824
 825        return !!(state & RFKILL_BLOCK_ANY);
 826}
 827EXPORT_SYMBOL(rfkill_blocked);
 828
 829
 830struct rfkill * __must_check rfkill_alloc(const char *name,
 831                                          struct device *parent,
 832                                          const enum rfkill_type type,
 833                                          const struct rfkill_ops *ops,
 834                                          void *ops_data)
 835{
 836        struct rfkill *rfkill;
 837        struct device *dev;
 838
 839        if (WARN_ON(!ops))
 840                return NULL;
 841
 842        if (WARN_ON(!ops->set_block))
 843                return NULL;
 844
 845        if (WARN_ON(!name))
 846                return NULL;
 847
 848        if (WARN_ON(type == RFKILL_TYPE_ALL || type >= NUM_RFKILL_TYPES))
 849                return NULL;
 850
 851        rfkill = kzalloc(sizeof(*rfkill), GFP_KERNEL);
 852        if (!rfkill)
 853                return NULL;
 854
 855        spin_lock_init(&rfkill->lock);
 856        INIT_LIST_HEAD(&rfkill->node);
 857        rfkill->type = type;
 858        rfkill->name = name;
 859        rfkill->ops = ops;
 860        rfkill->data = ops_data;
 861
 862        dev = &rfkill->dev;
 863        dev->class = &rfkill_class;
 864        dev->parent = parent;
 865        device_initialize(dev);
 866
 867        return rfkill;
 868}
 869EXPORT_SYMBOL(rfkill_alloc);
 870
 871static void rfkill_poll(struct work_struct *work)
 872{
 873        struct rfkill *rfkill;
 874
 875        rfkill = container_of(work, struct rfkill, poll_work.work);
 876
 877        /*
 878         * Poll hardware state -- driver will use one of the
 879         * rfkill_set{,_hw,_sw}_state functions and use its
 880         * return value to update the current status.
 881         */
 882        rfkill->ops->poll(rfkill, rfkill->data);
 883
 884        schedule_delayed_work(&rfkill->poll_work,
 885                round_jiffies_relative(POLL_INTERVAL));
 886}
 887
 888static void rfkill_uevent_work(struct work_struct *work)
 889{
 890        struct rfkill *rfkill;
 891
 892        rfkill = container_of(work, struct rfkill, uevent_work);
 893
 894        mutex_lock(&rfkill_global_mutex);
 895        rfkill_event(rfkill);
 896        mutex_unlock(&rfkill_global_mutex);
 897}
 898
 899static void rfkill_sync_work(struct work_struct *work)
 900{
 901        struct rfkill *rfkill;
 902        bool cur;
 903
 904        rfkill = container_of(work, struct rfkill, sync_work);
 905
 906        mutex_lock(&rfkill_global_mutex);
 907        cur = rfkill_global_states[rfkill->type].cur;
 908        rfkill_set_block(rfkill, cur);
 909        mutex_unlock(&rfkill_global_mutex);
 910}
 911
 912int __must_check rfkill_register(struct rfkill *rfkill)
 913{
 914        static unsigned long rfkill_no;
 915        struct device *dev = &rfkill->dev;
 916        int error;
 917
 918        BUG_ON(!rfkill);
 919
 920        mutex_lock(&rfkill_global_mutex);
 921
 922        if (rfkill->registered) {
 923                error = -EALREADY;
 924                goto unlock;
 925        }
 926
 927        rfkill->idx = rfkill_no;
 928        dev_set_name(dev, "rfkill%lu", rfkill_no);
 929        rfkill_no++;
 930
 931        list_add_tail(&rfkill->node, &rfkill_list);
 932
 933        error = device_add(dev);
 934        if (error)
 935                goto remove;
 936
 937        error = rfkill_led_trigger_register(rfkill);
 938        if (error)
 939                goto devdel;
 940
 941        rfkill->registered = true;
 942
 943        INIT_DELAYED_WORK(&rfkill->poll_work, rfkill_poll);
 944        INIT_WORK(&rfkill->uevent_work, rfkill_uevent_work);
 945        INIT_WORK(&rfkill->sync_work, rfkill_sync_work);
 946
 947        if (rfkill->ops->poll)
 948                schedule_delayed_work(&rfkill->poll_work,
 949                        round_jiffies_relative(POLL_INTERVAL));
 950
 951        if (!rfkill->persistent || rfkill_epo_lock_active) {
 952                schedule_work(&rfkill->sync_work);
 953        } else {
 954#ifdef CONFIG_RFKILL_INPUT
 955                bool soft_blocked = !!(rfkill->state & RFKILL_BLOCK_SW);
 956
 957                if (!atomic_read(&rfkill_input_disabled))
 958                        __rfkill_switch_all(rfkill->type, soft_blocked);
 959#endif
 960        }
 961
 962        rfkill_send_events(rfkill, RFKILL_OP_ADD);
 963
 964        mutex_unlock(&rfkill_global_mutex);
 965        return 0;
 966
 967 devdel:
 968        device_del(&rfkill->dev);
 969 remove:
 970        list_del_init(&rfkill->node);
 971 unlock:
 972        mutex_unlock(&rfkill_global_mutex);
 973        return error;
 974}
 975EXPORT_SYMBOL(rfkill_register);
 976
 977void rfkill_unregister(struct rfkill *rfkill)
 978{
 979        BUG_ON(!rfkill);
 980
 981        if (rfkill->ops->poll)
 982                cancel_delayed_work_sync(&rfkill->poll_work);
 983
 984        cancel_work_sync(&rfkill->uevent_work);
 985        cancel_work_sync(&rfkill->sync_work);
 986
 987        rfkill->registered = false;
 988
 989        device_del(&rfkill->dev);
 990
 991        mutex_lock(&rfkill_global_mutex);
 992        rfkill_send_events(rfkill, RFKILL_OP_DEL);
 993        list_del_init(&rfkill->node);
 994        mutex_unlock(&rfkill_global_mutex);
 995
 996        rfkill_led_trigger_unregister(rfkill);
 997}
 998EXPORT_SYMBOL(rfkill_unregister);
 999
1000void rfkill_destroy(struct rfkill *rfkill)
1001{
1002        if (rfkill)
1003                put_device(&rfkill->dev);
1004}
1005EXPORT_SYMBOL(rfkill_destroy);
1006
1007static int rfkill_fop_open(struct inode *inode, struct file *file)
1008{
1009        struct rfkill_data *data;
1010        struct rfkill *rfkill;
1011        struct rfkill_int_event *ev, *tmp;
1012
1013        data = kzalloc(sizeof(*data), GFP_KERNEL);
1014        if (!data)
1015                return -ENOMEM;
1016
1017        INIT_LIST_HEAD(&data->events);
1018        mutex_init(&data->mtx);
1019        init_waitqueue_head(&data->read_wait);
1020
1021        mutex_lock(&rfkill_global_mutex);
1022        mutex_lock(&data->mtx);
1023        /*
1024         * start getting events from elsewhere but hold mtx to get
1025         * startup events added first
1026         */
1027
1028        list_for_each_entry(rfkill, &rfkill_list, node) {
1029                ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1030                if (!ev)
1031                        goto free;
1032                rfkill_fill_event(&ev->ev, rfkill, RFKILL_OP_ADD);
1033                list_add_tail(&ev->list, &data->events);
1034        }
1035        list_add(&data->list, &rfkill_fds);
1036        mutex_unlock(&data->mtx);
1037        mutex_unlock(&rfkill_global_mutex);
1038
1039        file->private_data = data;
1040
1041        return nonseekable_open(inode, file);
1042
1043 free:
1044        mutex_unlock(&data->mtx);
1045        mutex_unlock(&rfkill_global_mutex);
1046        mutex_destroy(&data->mtx);
1047        list_for_each_entry_safe(ev, tmp, &data->events, list)
1048                kfree(ev);
1049        kfree(data);
1050        return -ENOMEM;
1051}
1052
1053static unsigned int rfkill_fop_poll(struct file *file, poll_table *wait)
1054{
1055        struct rfkill_data *data = file->private_data;
1056        unsigned int res = POLLOUT | POLLWRNORM;
1057
1058        poll_wait(file, &data->read_wait, wait);
1059
1060        mutex_lock(&data->mtx);
1061        if (!list_empty(&data->events))
1062                res = POLLIN | POLLRDNORM;
1063        mutex_unlock(&data->mtx);
1064
1065        return res;
1066}
1067
1068static bool rfkill_readable(struct rfkill_data *data)
1069{
1070        bool r;
1071
1072        mutex_lock(&data->mtx);
1073        r = !list_empty(&data->events);
1074        mutex_unlock(&data->mtx);
1075
1076        return r;
1077}
1078
1079static ssize_t rfkill_fop_read(struct file *file, char __user *buf,
1080                               size_t count, loff_t *pos)
1081{
1082        struct rfkill_data *data = file->private_data;
1083        struct rfkill_int_event *ev;
1084        unsigned long sz;
1085        int ret;
1086
1087        mutex_lock(&data->mtx);
1088
1089        while (list_empty(&data->events)) {
1090                if (file->f_flags & O_NONBLOCK) {
1091                        ret = -EAGAIN;
1092                        goto out;
1093                }
1094                mutex_unlock(&data->mtx);
1095                ret = wait_event_interruptible(data->read_wait,
1096                                               rfkill_readable(data));
1097                mutex_lock(&data->mtx);
1098
1099                if (ret)
1100                        goto out;
1101        }
1102
1103        ev = list_first_entry(&data->events, struct rfkill_int_event,
1104                                list);
1105
1106        sz = min_t(unsigned long, sizeof(ev->ev), count);
1107        ret = sz;
1108        if (copy_to_user(buf, &ev->ev, sz))
1109                ret = -EFAULT;
1110
1111        list_del(&ev->list);
1112        kfree(ev);
1113 out:
1114        mutex_unlock(&data->mtx);
1115        return ret;
1116}
1117
1118static ssize_t rfkill_fop_write(struct file *file, const char __user *buf,
1119                                size_t count, loff_t *pos)
1120{
1121        struct rfkill *rfkill;
1122        struct rfkill_event ev;
1123
1124        /* we don't need the 'hard' variable but accept it */
1125        if (count < RFKILL_EVENT_SIZE_V1 - 1)
1126                return -EINVAL;
1127
1128        /*
1129         * Copy as much data as we can accept into our 'ev' buffer,
1130         * but tell userspace how much we've copied so it can determine
1131         * our API version even in a write() call, if it cares.
1132         */
1133        count = min(count, sizeof(ev));
1134        if (copy_from_user(&ev, buf, count))
1135                return -EFAULT;
1136
1137        if (ev.op != RFKILL_OP_CHANGE && ev.op != RFKILL_OP_CHANGE_ALL)
1138                return -EINVAL;
1139
1140        if (ev.type >= NUM_RFKILL_TYPES)
1141                return -EINVAL;
1142
1143        mutex_lock(&rfkill_global_mutex);
1144
1145        if (ev.op == RFKILL_OP_CHANGE_ALL) {
1146                if (ev.type == RFKILL_TYPE_ALL) {
1147                        enum rfkill_type i;
1148                        for (i = 0; i < NUM_RFKILL_TYPES; i++)
1149                                rfkill_global_states[i].cur = ev.soft;
1150                } else {
1151                        rfkill_global_states[ev.type].cur = ev.soft;
1152                }
1153        }
1154
1155        list_for_each_entry(rfkill, &rfkill_list, node) {
1156                if (rfkill->idx != ev.idx && ev.op != RFKILL_OP_CHANGE_ALL)
1157                        continue;
1158
1159                if (rfkill->type != ev.type && ev.type != RFKILL_TYPE_ALL)
1160                        continue;
1161
1162                rfkill_set_block(rfkill, ev.soft);
1163        }
1164        mutex_unlock(&rfkill_global_mutex);
1165
1166        return count;
1167}
1168
1169static int rfkill_fop_release(struct inode *inode, struct file *file)
1170{
1171        struct rfkill_data *data = file->private_data;
1172        struct rfkill_int_event *ev, *tmp;
1173
1174        mutex_lock(&rfkill_global_mutex);
1175        list_del(&data->list);
1176        mutex_unlock(&rfkill_global_mutex);
1177
1178        mutex_destroy(&data->mtx);
1179        list_for_each_entry_safe(ev, tmp, &data->events, list)
1180                kfree(ev);
1181
1182#ifdef CONFIG_RFKILL_INPUT
1183        if (data->input_handler)
1184                if (atomic_dec_return(&rfkill_input_disabled) == 0)
1185                        printk(KERN_DEBUG "rfkill: input handler enabled\n");
1186#endif
1187
1188        kfree(data);
1189
1190        return 0;
1191}
1192
1193#ifdef CONFIG_RFKILL_INPUT
1194static long rfkill_fop_ioctl(struct file *file, unsigned int cmd,
1195                             unsigned long arg)
1196{
1197        struct rfkill_data *data = file->private_data;
1198
1199        if (_IOC_TYPE(cmd) != RFKILL_IOC_MAGIC)
1200                return -ENOSYS;
1201
1202        if (_IOC_NR(cmd) != RFKILL_IOC_NOINPUT)
1203                return -ENOSYS;
1204
1205        mutex_lock(&data->mtx);
1206
1207        if (!data->input_handler) {
1208                if (atomic_inc_return(&rfkill_input_disabled) == 1)
1209                        printk(KERN_DEBUG "rfkill: input handler disabled\n");
1210                data->input_handler = true;
1211        }
1212
1213        mutex_unlock(&data->mtx);
1214
1215        return 0;
1216}
1217#endif
1218
1219static const struct file_operations rfkill_fops = {
1220        .owner          = THIS_MODULE,
1221        .open           = rfkill_fop_open,
1222        .read           = rfkill_fop_read,
1223        .write          = rfkill_fop_write,
1224        .poll           = rfkill_fop_poll,
1225        .release        = rfkill_fop_release,
1226#ifdef CONFIG_RFKILL_INPUT
1227        .unlocked_ioctl = rfkill_fop_ioctl,
1228        .compat_ioctl   = rfkill_fop_ioctl,
1229#endif
1230        .llseek         = no_llseek,
1231};
1232
1233static struct miscdevice rfkill_miscdev = {
1234        .name   = "rfkill",
1235        .fops   = &rfkill_fops,
1236        .minor  = MISC_DYNAMIC_MINOR,
1237};
1238
1239static int __init rfkill_init(void)
1240{
1241        int error;
1242        int i;
1243
1244        for (i = 0; i < NUM_RFKILL_TYPES; i++)
1245                rfkill_global_states[i].cur = !rfkill_default_state;
1246
1247        error = class_register(&rfkill_class);
1248        if (error)
1249                goto out;
1250
1251        error = misc_register(&rfkill_miscdev);
1252        if (error) {
1253                class_unregister(&rfkill_class);
1254                goto out;
1255        }
1256
1257#ifdef CONFIG_RFKILL_INPUT
1258        error = rfkill_handler_init();
1259        if (error) {
1260                misc_deregister(&rfkill_miscdev);
1261                class_unregister(&rfkill_class);
1262                goto out;
1263        }
1264#endif
1265
1266 out:
1267        return error;
1268}
1269subsys_initcall(rfkill_init);
1270
1271static void __exit rfkill_exit(void)
1272{
1273#ifdef CONFIG_RFKILL_INPUT
1274        rfkill_handler_exit();
1275#endif
1276        misc_deregister(&rfkill_miscdev);
1277        class_unregister(&rfkill_class);
1278}
1279module_exit(rfkill_exit);
1280