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