| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* | 
|  | 2 | * Generic waiting primitives. | 
|  | 3 | * | 
|  | 4 | * (C) 2004 William Irwin, Oracle | 
|  | 5 | */ | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6 | #include <linux/init.h> | 
|  | 7 | #include <linux/module.h> | 
|  | 8 | #include <linux/sched.h> | 
|  | 9 | #include <linux/mm.h> | 
|  | 10 | #include <linux/wait.h> | 
|  | 11 | #include <linux/hash.h> | 
|  | 12 |  | 
| Ingo Molnar | 21d71f5 | 2006-07-10 04:45:32 -0700 | [diff] [blame] | 13 | void init_waitqueue_head(wait_queue_head_t *q) | 
|  | 14 | { | 
|  | 15 | spin_lock_init(&q->lock); | 
|  | 16 | INIT_LIST_HEAD(&q->task_list); | 
|  | 17 | } | 
| Ingo Molnar | eb4542b | 2006-07-03 00:25:07 -0700 | [diff] [blame] | 18 |  | 
| Ingo Molnar | 21d71f5 | 2006-07-10 04:45:32 -0700 | [diff] [blame] | 19 | EXPORT_SYMBOL(init_waitqueue_head); | 
| Ingo Molnar | eb4542b | 2006-07-03 00:25:07 -0700 | [diff] [blame] | 20 |  | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 21 | void add_wait_queue(wait_queue_head_t *q, wait_queue_t *wait) | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 22 | { | 
|  | 23 | unsigned long flags; | 
|  | 24 |  | 
|  | 25 | wait->flags &= ~WQ_FLAG_EXCLUSIVE; | 
|  | 26 | spin_lock_irqsave(&q->lock, flags); | 
|  | 27 | __add_wait_queue(q, wait); | 
|  | 28 | spin_unlock_irqrestore(&q->lock, flags); | 
|  | 29 | } | 
|  | 30 | EXPORT_SYMBOL(add_wait_queue); | 
|  | 31 |  | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 32 | void add_wait_queue_exclusive(wait_queue_head_t *q, wait_queue_t *wait) | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 33 | { | 
|  | 34 | unsigned long flags; | 
|  | 35 |  | 
|  | 36 | wait->flags |= WQ_FLAG_EXCLUSIVE; | 
|  | 37 | spin_lock_irqsave(&q->lock, flags); | 
|  | 38 | __add_wait_queue_tail(q, wait); | 
|  | 39 | spin_unlock_irqrestore(&q->lock, flags); | 
|  | 40 | } | 
|  | 41 | EXPORT_SYMBOL(add_wait_queue_exclusive); | 
|  | 42 |  | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 43 | void remove_wait_queue(wait_queue_head_t *q, wait_queue_t *wait) | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 44 | { | 
|  | 45 | unsigned long flags; | 
|  | 46 |  | 
|  | 47 | spin_lock_irqsave(&q->lock, flags); | 
|  | 48 | __remove_wait_queue(q, wait); | 
|  | 49 | spin_unlock_irqrestore(&q->lock, flags); | 
|  | 50 | } | 
|  | 51 | EXPORT_SYMBOL(remove_wait_queue); | 
|  | 52 |  | 
|  | 53 |  | 
|  | 54 | /* | 
|  | 55 | * Note: we use "set_current_state()" _after_ the wait-queue add, | 
|  | 56 | * because we need a memory barrier there on SMP, so that any | 
|  | 57 | * wake-function that tests for the wait-queue being active | 
|  | 58 | * will be guaranteed to see waitqueue addition _or_ subsequent | 
|  | 59 | * tests in this thread will see the wakeup having taken place. | 
|  | 60 | * | 
|  | 61 | * The spin_unlock() itself is semi-permeable and only protects | 
|  | 62 | * one way (it only protects stuff inside the critical region and | 
|  | 63 | * stops them from bleeding out - it would still allow subsequent | 
| Michael Opdenacker | 59c5159 | 2007-05-09 08:57:56 +0200 | [diff] [blame] | 64 | * loads to move into the critical region). | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 65 | */ | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 66 | void | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 67 | prepare_to_wait(wait_queue_head_t *q, wait_queue_t *wait, int state) | 
|  | 68 | { | 
|  | 69 | unsigned long flags; | 
|  | 70 |  | 
|  | 71 | wait->flags &= ~WQ_FLAG_EXCLUSIVE; | 
|  | 72 | spin_lock_irqsave(&q->lock, flags); | 
|  | 73 | if (list_empty(&wait->task_list)) | 
|  | 74 | __add_wait_queue(q, wait); | 
| Tejun Heo | a25d644 | 2008-10-15 22:01:38 -0700 | [diff] [blame] | 75 | set_current_state(state); | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 76 | spin_unlock_irqrestore(&q->lock, flags); | 
|  | 77 | } | 
|  | 78 | EXPORT_SYMBOL(prepare_to_wait); | 
|  | 79 |  | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 80 | void | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 81 | prepare_to_wait_exclusive(wait_queue_head_t *q, wait_queue_t *wait, int state) | 
|  | 82 | { | 
|  | 83 | unsigned long flags; | 
|  | 84 |  | 
|  | 85 | wait->flags |= WQ_FLAG_EXCLUSIVE; | 
|  | 86 | spin_lock_irqsave(&q->lock, flags); | 
|  | 87 | if (list_empty(&wait->task_list)) | 
|  | 88 | __add_wait_queue_tail(q, wait); | 
| Tejun Heo | a25d644 | 2008-10-15 22:01:38 -0700 | [diff] [blame] | 89 | set_current_state(state); | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 90 | spin_unlock_irqrestore(&q->lock, flags); | 
|  | 91 | } | 
|  | 92 | EXPORT_SYMBOL(prepare_to_wait_exclusive); | 
|  | 93 |  | 
| Johannes Weiner | 777c6c5 | 2009-02-04 15:12:14 -0800 | [diff] [blame] | 94 | /* | 
|  | 95 | * finish_wait - clean up after waiting in a queue | 
|  | 96 | * @q: waitqueue waited on | 
|  | 97 | * @wait: wait descriptor | 
|  | 98 | * | 
|  | 99 | * Sets current thread back to running state and removes | 
|  | 100 | * the wait descriptor from the given waitqueue if still | 
|  | 101 | * queued. | 
|  | 102 | */ | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 103 | void finish_wait(wait_queue_head_t *q, wait_queue_t *wait) | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 104 | { | 
|  | 105 | unsigned long flags; | 
|  | 106 |  | 
|  | 107 | __set_current_state(TASK_RUNNING); | 
|  | 108 | /* | 
|  | 109 | * We can check for list emptiness outside the lock | 
|  | 110 | * IFF: | 
|  | 111 | *  - we use the "careful" check that verifies both | 
|  | 112 | *    the next and prev pointers, so that there cannot | 
|  | 113 | *    be any half-pending updates in progress on other | 
|  | 114 | *    CPU's that we haven't seen yet (and that might | 
|  | 115 | *    still change the stack area. | 
|  | 116 | * and | 
|  | 117 | *  - all other users take the lock (ie we can only | 
|  | 118 | *    have _one_ other CPU that looks at or modifies | 
|  | 119 | *    the list). | 
|  | 120 | */ | 
|  | 121 | if (!list_empty_careful(&wait->task_list)) { | 
|  | 122 | spin_lock_irqsave(&q->lock, flags); | 
|  | 123 | list_del_init(&wait->task_list); | 
|  | 124 | spin_unlock_irqrestore(&q->lock, flags); | 
|  | 125 | } | 
|  | 126 | } | 
|  | 127 | EXPORT_SYMBOL(finish_wait); | 
|  | 128 |  | 
| Johannes Weiner | 777c6c5 | 2009-02-04 15:12:14 -0800 | [diff] [blame] | 129 | /* | 
|  | 130 | * abort_exclusive_wait - abort exclusive waiting in a queue | 
|  | 131 | * @q: waitqueue waited on | 
|  | 132 | * @wait: wait descriptor | 
|  | 133 | * @state: runstate of the waiter to be woken | 
|  | 134 | * @key: key to identify a wait bit queue or %NULL | 
|  | 135 | * | 
|  | 136 | * Sets current thread back to running state and removes | 
|  | 137 | * the wait descriptor from the given waitqueue if still | 
|  | 138 | * queued. | 
|  | 139 | * | 
|  | 140 | * Wakes up the next waiter if the caller is concurrently | 
|  | 141 | * woken up through the queue. | 
|  | 142 | * | 
|  | 143 | * This prevents waiter starvation where an exclusive waiter | 
|  | 144 | * aborts and is woken up concurrently and noone wakes up | 
|  | 145 | * the next waiter. | 
|  | 146 | */ | 
|  | 147 | void abort_exclusive_wait(wait_queue_head_t *q, wait_queue_t *wait, | 
|  | 148 | unsigned int mode, void *key) | 
|  | 149 | { | 
|  | 150 | unsigned long flags; | 
|  | 151 |  | 
|  | 152 | __set_current_state(TASK_RUNNING); | 
|  | 153 | spin_lock_irqsave(&q->lock, flags); | 
|  | 154 | if (!list_empty(&wait->task_list)) | 
|  | 155 | list_del_init(&wait->task_list); | 
|  | 156 | else if (waitqueue_active(q)) | 
| Johannes Weiner | 78ddb08 | 2009-04-14 16:53:05 +0200 | [diff] [blame] | 157 | __wake_up_locked_key(q, mode, key); | 
| Johannes Weiner | 777c6c5 | 2009-02-04 15:12:14 -0800 | [diff] [blame] | 158 | spin_unlock_irqrestore(&q->lock, flags); | 
|  | 159 | } | 
|  | 160 | EXPORT_SYMBOL(abort_exclusive_wait); | 
|  | 161 |  | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 162 | int autoremove_wake_function(wait_queue_t *wait, unsigned mode, int sync, void *key) | 
|  | 163 | { | 
|  | 164 | int ret = default_wake_function(wait, mode, sync, key); | 
|  | 165 |  | 
|  | 166 | if (ret) | 
|  | 167 | list_del_init(&wait->task_list); | 
|  | 168 | return ret; | 
|  | 169 | } | 
|  | 170 | EXPORT_SYMBOL(autoremove_wake_function); | 
|  | 171 |  | 
|  | 172 | int wake_bit_function(wait_queue_t *wait, unsigned mode, int sync, void *arg) | 
|  | 173 | { | 
|  | 174 | struct wait_bit_key *key = arg; | 
|  | 175 | struct wait_bit_queue *wait_bit | 
|  | 176 | = container_of(wait, struct wait_bit_queue, wait); | 
|  | 177 |  | 
|  | 178 | if (wait_bit->key.flags != key->flags || | 
|  | 179 | wait_bit->key.bit_nr != key->bit_nr || | 
|  | 180 | test_bit(key->bit_nr, key->flags)) | 
|  | 181 | return 0; | 
|  | 182 | else | 
|  | 183 | return autoremove_wake_function(wait, mode, sync, key); | 
|  | 184 | } | 
|  | 185 | EXPORT_SYMBOL(wake_bit_function); | 
|  | 186 |  | 
|  | 187 | /* | 
|  | 188 | * To allow interruptible waiting and asynchronous (i.e. nonblocking) | 
|  | 189 | * waiting, the actions of __wait_on_bit() and __wait_on_bit_lock() are | 
|  | 190 | * permitted return codes. Nonzero return codes halt waiting and return. | 
|  | 191 | */ | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 192 | int __sched | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 193 | __wait_on_bit(wait_queue_head_t *wq, struct wait_bit_queue *q, | 
|  | 194 | int (*action)(void *), unsigned mode) | 
|  | 195 | { | 
|  | 196 | int ret = 0; | 
|  | 197 |  | 
|  | 198 | do { | 
|  | 199 | prepare_to_wait(wq, &q->wait, mode); | 
|  | 200 | if (test_bit(q->key.bit_nr, q->key.flags)) | 
|  | 201 | ret = (*action)(q->key.flags); | 
|  | 202 | } while (test_bit(q->key.bit_nr, q->key.flags) && !ret); | 
|  | 203 | finish_wait(wq, &q->wait); | 
|  | 204 | return ret; | 
|  | 205 | } | 
|  | 206 | EXPORT_SYMBOL(__wait_on_bit); | 
|  | 207 |  | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 208 | int __sched out_of_line_wait_on_bit(void *word, int bit, | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 209 | int (*action)(void *), unsigned mode) | 
|  | 210 | { | 
|  | 211 | wait_queue_head_t *wq = bit_waitqueue(word, bit); | 
|  | 212 | DEFINE_WAIT_BIT(wait, word, bit); | 
|  | 213 |  | 
|  | 214 | return __wait_on_bit(wq, &wait, action, mode); | 
|  | 215 | } | 
|  | 216 | EXPORT_SYMBOL(out_of_line_wait_on_bit); | 
|  | 217 |  | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 218 | int __sched | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 219 | __wait_on_bit_lock(wait_queue_head_t *wq, struct wait_bit_queue *q, | 
|  | 220 | int (*action)(void *), unsigned mode) | 
|  | 221 | { | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 222 | do { | 
| Johannes Weiner | 777c6c5 | 2009-02-04 15:12:14 -0800 | [diff] [blame] | 223 | int ret; | 
|  | 224 |  | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 225 | prepare_to_wait_exclusive(wq, &q->wait, mode); | 
| Johannes Weiner | 777c6c5 | 2009-02-04 15:12:14 -0800 | [diff] [blame] | 226 | if (!test_bit(q->key.bit_nr, q->key.flags)) | 
|  | 227 | continue; | 
|  | 228 | ret = action(q->key.flags); | 
|  | 229 | if (!ret) | 
|  | 230 | continue; | 
|  | 231 | abort_exclusive_wait(wq, &q->wait, mode, &q->key); | 
|  | 232 | return ret; | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 233 | } while (test_and_set_bit(q->key.bit_nr, q->key.flags)); | 
|  | 234 | finish_wait(wq, &q->wait); | 
| Johannes Weiner | 777c6c5 | 2009-02-04 15:12:14 -0800 | [diff] [blame] | 235 | return 0; | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 236 | } | 
|  | 237 | EXPORT_SYMBOL(__wait_on_bit_lock); | 
|  | 238 |  | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 239 | int __sched out_of_line_wait_on_bit_lock(void *word, int bit, | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 240 | int (*action)(void *), unsigned mode) | 
|  | 241 | { | 
|  | 242 | wait_queue_head_t *wq = bit_waitqueue(word, bit); | 
|  | 243 | DEFINE_WAIT_BIT(wait, word, bit); | 
|  | 244 |  | 
|  | 245 | return __wait_on_bit_lock(wq, &wait, action, mode); | 
|  | 246 | } | 
|  | 247 | EXPORT_SYMBOL(out_of_line_wait_on_bit_lock); | 
|  | 248 |  | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 249 | void __wake_up_bit(wait_queue_head_t *wq, void *word, int bit) | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 250 | { | 
|  | 251 | struct wait_bit_key key = __WAIT_BIT_KEY_INITIALIZER(word, bit); | 
|  | 252 | if (waitqueue_active(wq)) | 
| Matthew Wilcox | e64d66c | 2007-12-06 17:34:36 -0500 | [diff] [blame] | 253 | __wake_up(wq, TASK_NORMAL, 1, &key); | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 254 | } | 
|  | 255 | EXPORT_SYMBOL(__wake_up_bit); | 
|  | 256 |  | 
|  | 257 | /** | 
|  | 258 | * wake_up_bit - wake up a waiter on a bit | 
|  | 259 | * @word: the word being waited on, a kernel virtual address | 
|  | 260 | * @bit: the bit of the word being waited on | 
|  | 261 | * | 
|  | 262 | * There is a standard hashed waitqueue table for generic use. This | 
|  | 263 | * is the part of the hashtable's accessor API that wakes up waiters | 
|  | 264 | * on a bit. For instance, if one were to have waiters on a bitflag, | 
|  | 265 | * one would call wake_up_bit() after clearing the bit. | 
|  | 266 | * | 
|  | 267 | * In order for this to function properly, as it uses waitqueue_active() | 
|  | 268 | * internally, some kind of memory barrier must be done prior to calling | 
|  | 269 | * this. Typically, this will be smp_mb__after_clear_bit(), but in some | 
|  | 270 | * cases where bitflags are manipulated non-atomically under a lock, one | 
|  | 271 | * may need to use a less regular barrier, such fs/inode.c's smp_mb(), | 
|  | 272 | * because spin_unlock() does not guarantee a memory barrier. | 
|  | 273 | */ | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 274 | void wake_up_bit(void *word, int bit) | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 275 | { | 
|  | 276 | __wake_up_bit(bit_waitqueue(word, bit), word, bit); | 
|  | 277 | } | 
|  | 278 | EXPORT_SYMBOL(wake_up_bit); | 
|  | 279 |  | 
| Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 280 | wait_queue_head_t *bit_waitqueue(void *word, int bit) | 
| Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 281 | { | 
|  | 282 | const int shift = BITS_PER_LONG == 32 ? 5 : 6; | 
|  | 283 | const struct zone *zone = page_zone(virt_to_page(word)); | 
|  | 284 | unsigned long val = (unsigned long)word << shift | bit; | 
|  | 285 |  | 
|  | 286 | return &zone->wait_table[hash_long(val, zone->wait_table_bits)]; | 
|  | 287 | } | 
|  | 288 | EXPORT_SYMBOL(bit_waitqueue); |