Peter Hurley | 4898e640 | 2013-04-16 06:15:50 -0400 | [diff] [blame] | 1 | /* |
| 2 | * Ldisc rw semaphore |
| 3 | * |
| 4 | * The ldisc semaphore is semantically a rw_semaphore but which enforces |
| 5 | * an alternate policy, namely: |
| 6 | * 1) Supports lock wait timeouts |
| 7 | * 2) Write waiter has priority |
| 8 | * 3) Downgrading is not supported |
| 9 | * |
| 10 | * Implementation notes: |
| 11 | * 1) Upper half of semaphore count is a wait count (differs from rwsem |
| 12 | * in that rwsem normalizes the upper half to the wait bias) |
| 13 | * 2) Lacks overflow checking |
| 14 | * |
| 15 | * The generic counting was copied and modified from include/asm-generic/rwsem.h |
| 16 | * by Paul Mackerras <paulus@samba.org>. |
| 17 | * |
| 18 | * The scheduling policy was copied and modified from lib/rwsem.c |
| 19 | * Written by David Howells (dhowells@redhat.com). |
| 20 | * |
| 21 | * This implementation incorporates the write lock stealing work of |
| 22 | * Michel Lespinasse <walken@google.com>. |
| 23 | * |
| 24 | * Copyright (C) 2013 Peter Hurley <peter@hurleysoftware.com> |
| 25 | * |
| 26 | * This file may be redistributed under the terms of the GNU General Public |
| 27 | * License v2. |
| 28 | */ |
| 29 | |
| 30 | #include <linux/list.h> |
| 31 | #include <linux/spinlock.h> |
| 32 | #include <linux/atomic.h> |
| 33 | #include <linux/tty.h> |
| 34 | #include <linux/sched.h> |
| 35 | |
| 36 | |
| 37 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
| 38 | # define __acq(l, s, t, r, c, n, i) \ |
| 39 | lock_acquire(&(l)->dep_map, s, t, r, c, n, i) |
| 40 | # define __rel(l, n, i) \ |
| 41 | lock_release(&(l)->dep_map, n, i) |
| 42 | # ifdef CONFIG_PROVE_LOCKING |
| 43 | # define lockdep_acquire(l, s, t, i) __acq(l, s, t, 0, 2, NULL, i) |
| 44 | # define lockdep_acquire_nest(l, s, t, n, i) __acq(l, s, t, 0, 2, n, i) |
| 45 | # define lockdep_acquire_read(l, s, t, i) __acq(l, s, t, 1, 2, NULL, i) |
| 46 | # define lockdep_release(l, n, i) __rel(l, n, i) |
| 47 | # else |
| 48 | # define lockdep_acquire(l, s, t, i) __acq(l, s, t, 0, 1, NULL, i) |
| 49 | # define lockdep_acquire_nest(l, s, t, n, i) __acq(l, s, t, 0, 1, n, i) |
| 50 | # define lockdep_acquire_read(l, s, t, i) __acq(l, s, t, 1, 1, NULL, i) |
| 51 | # define lockdep_release(l, n, i) __rel(l, n, i) |
| 52 | # endif |
| 53 | #else |
| 54 | # define lockdep_acquire(l, s, t, i) do { } while (0) |
| 55 | # define lockdep_acquire_nest(l, s, t, n, i) do { } while (0) |
| 56 | # define lockdep_acquire_read(l, s, t, i) do { } while (0) |
| 57 | # define lockdep_release(l, n, i) do { } while (0) |
| 58 | #endif |
| 59 | |
| 60 | #ifdef CONFIG_LOCK_STAT |
| 61 | # define lock_stat(_lock, stat) lock_##stat(&(_lock)->dep_map, _RET_IP_) |
| 62 | #else |
| 63 | # define lock_stat(_lock, stat) do { } while (0) |
| 64 | #endif |
| 65 | |
| 66 | |
| 67 | #if BITS_PER_LONG == 64 |
| 68 | # define LDSEM_ACTIVE_MASK 0xffffffffL |
| 69 | #else |
| 70 | # define LDSEM_ACTIVE_MASK 0x0000ffffL |
| 71 | #endif |
| 72 | |
| 73 | #define LDSEM_UNLOCKED 0L |
| 74 | #define LDSEM_ACTIVE_BIAS 1L |
| 75 | #define LDSEM_WAIT_BIAS (-LDSEM_ACTIVE_MASK-1) |
| 76 | #define LDSEM_READ_BIAS LDSEM_ACTIVE_BIAS |
| 77 | #define LDSEM_WRITE_BIAS (LDSEM_WAIT_BIAS + LDSEM_ACTIVE_BIAS) |
| 78 | |
| 79 | struct ldsem_waiter { |
| 80 | struct list_head list; |
| 81 | struct task_struct *task; |
| 82 | }; |
| 83 | |
| 84 | static inline long ldsem_atomic_update(long delta, struct ld_semaphore *sem) |
| 85 | { |
| 86 | return atomic_long_add_return(delta, (atomic_long_t *)&sem->count); |
| 87 | } |
| 88 | |
Peter Hurley | cf87277 | 2013-12-11 21:11:58 -0500 | [diff] [blame] | 89 | /* |
| 90 | * ldsem_cmpxchg() updates @*old with the last-known sem->count value. |
| 91 | * Returns 1 if count was successfully changed; @*old will have @new value. |
| 92 | * Returns 0 if count was not changed; @*old will have most recent sem->count |
| 93 | */ |
Peter Hurley | 4898e640 | 2013-04-16 06:15:50 -0400 | [diff] [blame] | 94 | static inline int ldsem_cmpxchg(long *old, long new, struct ld_semaphore *sem) |
| 95 | { |
Peter Hurley | cf87277 | 2013-12-11 21:11:58 -0500 | [diff] [blame] | 96 | long tmp = atomic_long_cmpxchg(&sem->count, *old, new); |
| 97 | if (tmp == *old) { |
| 98 | *old = new; |
| 99 | return 1; |
| 100 | } else { |
| 101 | *old = tmp; |
| 102 | return 0; |
| 103 | } |
Peter Hurley | 4898e640 | 2013-04-16 06:15:50 -0400 | [diff] [blame] | 104 | } |
| 105 | |
| 106 | /* |
| 107 | * Initialize an ldsem: |
| 108 | */ |
| 109 | void __init_ldsem(struct ld_semaphore *sem, const char *name, |
| 110 | struct lock_class_key *key) |
| 111 | { |
| 112 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
| 113 | /* |
| 114 | * Make sure we are not reinitializing a held semaphore: |
| 115 | */ |
| 116 | debug_check_no_locks_freed((void *)sem, sizeof(*sem)); |
| 117 | lockdep_init_map(&sem->dep_map, name, key, 0); |
| 118 | #endif |
| 119 | sem->count = LDSEM_UNLOCKED; |
| 120 | sem->wait_readers = 0; |
| 121 | raw_spin_lock_init(&sem->wait_lock); |
| 122 | INIT_LIST_HEAD(&sem->read_wait); |
| 123 | INIT_LIST_HEAD(&sem->write_wait); |
| 124 | } |
| 125 | |
| 126 | static void __ldsem_wake_readers(struct ld_semaphore *sem) |
| 127 | { |
| 128 | struct ldsem_waiter *waiter, *next; |
| 129 | struct task_struct *tsk; |
| 130 | long adjust, count; |
| 131 | |
| 132 | /* Try to grant read locks to all readers on the read wait list. |
| 133 | * Note the 'active part' of the count is incremented by |
| 134 | * the number of readers before waking any processes up. |
| 135 | */ |
| 136 | adjust = sem->wait_readers * (LDSEM_ACTIVE_BIAS - LDSEM_WAIT_BIAS); |
| 137 | count = ldsem_atomic_update(adjust, sem); |
| 138 | do { |
| 139 | if (count > 0) |
| 140 | break; |
| 141 | if (ldsem_cmpxchg(&count, count - adjust, sem)) |
| 142 | return; |
| 143 | } while (1); |
| 144 | |
| 145 | list_for_each_entry_safe(waiter, next, &sem->read_wait, list) { |
| 146 | tsk = waiter->task; |
| 147 | smp_mb(); |
| 148 | waiter->task = NULL; |
| 149 | wake_up_process(tsk); |
| 150 | put_task_struct(tsk); |
| 151 | } |
| 152 | INIT_LIST_HEAD(&sem->read_wait); |
| 153 | sem->wait_readers = 0; |
| 154 | } |
| 155 | |
| 156 | static inline int writer_trylock(struct ld_semaphore *sem) |
| 157 | { |
| 158 | /* only wake this writer if the active part of the count can be |
| 159 | * transitioned from 0 -> 1 |
| 160 | */ |
| 161 | long count = ldsem_atomic_update(LDSEM_ACTIVE_BIAS, sem); |
| 162 | do { |
| 163 | if ((count & LDSEM_ACTIVE_MASK) == LDSEM_ACTIVE_BIAS) |
| 164 | return 1; |
| 165 | if (ldsem_cmpxchg(&count, count - LDSEM_ACTIVE_BIAS, sem)) |
| 166 | return 0; |
| 167 | } while (1); |
| 168 | } |
| 169 | |
| 170 | static void __ldsem_wake_writer(struct ld_semaphore *sem) |
| 171 | { |
| 172 | struct ldsem_waiter *waiter; |
| 173 | |
| 174 | waiter = list_entry(sem->write_wait.next, struct ldsem_waiter, list); |
| 175 | wake_up_process(waiter->task); |
| 176 | } |
| 177 | |
| 178 | /* |
| 179 | * handle the lock release when processes blocked on it that can now run |
| 180 | * - if we come here from up_xxxx(), then: |
| 181 | * - the 'active part' of count (&0x0000ffff) reached 0 (but may have changed) |
| 182 | * - the 'waiting part' of count (&0xffff0000) is -ve (and will still be so) |
| 183 | * - the spinlock must be held by the caller |
| 184 | * - woken process blocks are discarded from the list after having task zeroed |
| 185 | */ |
| 186 | static void __ldsem_wake(struct ld_semaphore *sem) |
| 187 | { |
| 188 | if (!list_empty(&sem->write_wait)) |
| 189 | __ldsem_wake_writer(sem); |
| 190 | else if (!list_empty(&sem->read_wait)) |
| 191 | __ldsem_wake_readers(sem); |
| 192 | } |
| 193 | |
| 194 | static void ldsem_wake(struct ld_semaphore *sem) |
| 195 | { |
| 196 | unsigned long flags; |
| 197 | |
| 198 | raw_spin_lock_irqsave(&sem->wait_lock, flags); |
| 199 | __ldsem_wake(sem); |
| 200 | raw_spin_unlock_irqrestore(&sem->wait_lock, flags); |
| 201 | } |
| 202 | |
| 203 | /* |
| 204 | * wait for the read lock to be granted |
| 205 | */ |
| 206 | static struct ld_semaphore __sched * |
| 207 | down_read_failed(struct ld_semaphore *sem, long count, long timeout) |
| 208 | { |
| 209 | struct ldsem_waiter waiter; |
| 210 | struct task_struct *tsk = current; |
| 211 | long adjust = -LDSEM_ACTIVE_BIAS + LDSEM_WAIT_BIAS; |
| 212 | |
| 213 | /* set up my own style of waitqueue */ |
| 214 | raw_spin_lock_irq(&sem->wait_lock); |
| 215 | |
| 216 | /* Try to reverse the lock attempt but if the count has changed |
| 217 | * so that reversing fails, check if there are are no waiters, |
| 218 | * and early-out if not */ |
| 219 | do { |
| 220 | if (ldsem_cmpxchg(&count, count + adjust, sem)) |
| 221 | break; |
| 222 | if (count > 0) { |
| 223 | raw_spin_unlock_irq(&sem->wait_lock); |
| 224 | return sem; |
| 225 | } |
| 226 | } while (1); |
| 227 | |
| 228 | list_add_tail(&waiter.list, &sem->read_wait); |
| 229 | sem->wait_readers++; |
| 230 | |
| 231 | waiter.task = tsk; |
| 232 | get_task_struct(tsk); |
| 233 | |
| 234 | /* if there are no active locks, wake the new lock owner(s) */ |
| 235 | if ((count & LDSEM_ACTIVE_MASK) == 0) |
| 236 | __ldsem_wake(sem); |
| 237 | |
| 238 | raw_spin_unlock_irq(&sem->wait_lock); |
| 239 | |
| 240 | /* wait to be given the lock */ |
| 241 | for (;;) { |
| 242 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); |
| 243 | |
| 244 | if (!waiter.task) |
| 245 | break; |
| 246 | if (!timeout) |
| 247 | break; |
| 248 | timeout = schedule_timeout(timeout); |
| 249 | } |
| 250 | |
| 251 | __set_task_state(tsk, TASK_RUNNING); |
| 252 | |
| 253 | if (!timeout) { |
| 254 | /* lock timed out but check if this task was just |
| 255 | * granted lock ownership - if so, pretend there |
| 256 | * was no timeout; otherwise, cleanup lock wait */ |
| 257 | raw_spin_lock_irq(&sem->wait_lock); |
| 258 | if (waiter.task) { |
| 259 | ldsem_atomic_update(-LDSEM_WAIT_BIAS, sem); |
| 260 | list_del(&waiter.list); |
| 261 | raw_spin_unlock_irq(&sem->wait_lock); |
| 262 | put_task_struct(waiter.task); |
| 263 | return NULL; |
| 264 | } |
| 265 | raw_spin_unlock_irq(&sem->wait_lock); |
| 266 | } |
| 267 | |
| 268 | return sem; |
| 269 | } |
| 270 | |
| 271 | /* |
| 272 | * wait for the write lock to be granted |
| 273 | */ |
| 274 | static struct ld_semaphore __sched * |
| 275 | down_write_failed(struct ld_semaphore *sem, long count, long timeout) |
| 276 | { |
| 277 | struct ldsem_waiter waiter; |
| 278 | struct task_struct *tsk = current; |
| 279 | long adjust = -LDSEM_ACTIVE_BIAS; |
| 280 | int locked = 0; |
| 281 | |
| 282 | /* set up my own style of waitqueue */ |
| 283 | raw_spin_lock_irq(&sem->wait_lock); |
| 284 | |
| 285 | /* Try to reverse the lock attempt but if the count has changed |
| 286 | * so that reversing fails, check if the lock is now owned, |
| 287 | * and early-out if so */ |
| 288 | do { |
| 289 | if (ldsem_cmpxchg(&count, count + adjust, sem)) |
| 290 | break; |
| 291 | if ((count & LDSEM_ACTIVE_MASK) == LDSEM_ACTIVE_BIAS) { |
| 292 | raw_spin_unlock_irq(&sem->wait_lock); |
| 293 | return sem; |
| 294 | } |
| 295 | } while (1); |
| 296 | |
| 297 | list_add_tail(&waiter.list, &sem->write_wait); |
| 298 | |
| 299 | waiter.task = tsk; |
| 300 | |
| 301 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); |
| 302 | for (;;) { |
| 303 | if (!timeout) |
| 304 | break; |
| 305 | raw_spin_unlock_irq(&sem->wait_lock); |
| 306 | timeout = schedule_timeout(timeout); |
| 307 | raw_spin_lock_irq(&sem->wait_lock); |
| 308 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); |
| 309 | if ((locked = writer_trylock(sem))) |
| 310 | break; |
| 311 | } |
| 312 | |
| 313 | if (!locked) |
| 314 | ldsem_atomic_update(-LDSEM_WAIT_BIAS, sem); |
| 315 | list_del(&waiter.list); |
| 316 | raw_spin_unlock_irq(&sem->wait_lock); |
| 317 | |
| 318 | __set_task_state(tsk, TASK_RUNNING); |
| 319 | |
| 320 | /* lock wait may have timed out */ |
| 321 | if (!locked) |
| 322 | return NULL; |
| 323 | return sem; |
| 324 | } |
| 325 | |
| 326 | |
| 327 | |
| 328 | static inline int __ldsem_down_read_nested(struct ld_semaphore *sem, |
| 329 | int subclass, long timeout) |
| 330 | { |
| 331 | long count; |
| 332 | |
| 333 | lockdep_acquire_read(sem, subclass, 0, _RET_IP_); |
| 334 | |
| 335 | count = ldsem_atomic_update(LDSEM_READ_BIAS, sem); |
| 336 | if (count <= 0) { |
| 337 | lock_stat(sem, contended); |
| 338 | if (!down_read_failed(sem, count, timeout)) { |
| 339 | lockdep_release(sem, 1, _RET_IP_); |
| 340 | return 0; |
| 341 | } |
| 342 | } |
| 343 | lock_stat(sem, acquired); |
| 344 | return 1; |
| 345 | } |
| 346 | |
| 347 | static inline int __ldsem_down_write_nested(struct ld_semaphore *sem, |
| 348 | int subclass, long timeout) |
| 349 | { |
| 350 | long count; |
| 351 | |
| 352 | lockdep_acquire(sem, subclass, 0, _RET_IP_); |
| 353 | |
| 354 | count = ldsem_atomic_update(LDSEM_WRITE_BIAS, sem); |
| 355 | if ((count & LDSEM_ACTIVE_MASK) != LDSEM_ACTIVE_BIAS) { |
| 356 | lock_stat(sem, contended); |
| 357 | if (!down_write_failed(sem, count, timeout)) { |
| 358 | lockdep_release(sem, 1, _RET_IP_); |
| 359 | return 0; |
| 360 | } |
| 361 | } |
| 362 | lock_stat(sem, acquired); |
| 363 | return 1; |
| 364 | } |
| 365 | |
| 366 | |
| 367 | /* |
| 368 | * lock for reading -- returns 1 if successful, 0 if timed out |
| 369 | */ |
| 370 | int __sched ldsem_down_read(struct ld_semaphore *sem, long timeout) |
| 371 | { |
| 372 | might_sleep(); |
| 373 | return __ldsem_down_read_nested(sem, 0, timeout); |
| 374 | } |
| 375 | |
| 376 | /* |
| 377 | * trylock for reading -- returns 1 if successful, 0 if contention |
| 378 | */ |
| 379 | int ldsem_down_read_trylock(struct ld_semaphore *sem) |
| 380 | { |
| 381 | long count = sem->count; |
| 382 | |
| 383 | while (count >= 0) { |
| 384 | if (ldsem_cmpxchg(&count, count + LDSEM_READ_BIAS, sem)) { |
| 385 | lockdep_acquire_read(sem, 0, 1, _RET_IP_); |
| 386 | lock_stat(sem, acquired); |
| 387 | return 1; |
| 388 | } |
| 389 | } |
| 390 | return 0; |
| 391 | } |
| 392 | |
| 393 | /* |
| 394 | * lock for writing -- returns 1 if successful, 0 if timed out |
| 395 | */ |
| 396 | int __sched ldsem_down_write(struct ld_semaphore *sem, long timeout) |
| 397 | { |
| 398 | might_sleep(); |
| 399 | return __ldsem_down_write_nested(sem, 0, timeout); |
| 400 | } |
| 401 | |
| 402 | /* |
| 403 | * trylock for writing -- returns 1 if successful, 0 if contention |
| 404 | */ |
| 405 | int ldsem_down_write_trylock(struct ld_semaphore *sem) |
| 406 | { |
| 407 | long count = sem->count; |
| 408 | |
| 409 | while ((count & LDSEM_ACTIVE_MASK) == 0) { |
| 410 | if (ldsem_cmpxchg(&count, count + LDSEM_WRITE_BIAS, sem)) { |
| 411 | lockdep_acquire(sem, 0, 1, _RET_IP_); |
| 412 | lock_stat(sem, acquired); |
| 413 | return 1; |
| 414 | } |
| 415 | } |
| 416 | return 0; |
| 417 | } |
| 418 | |
| 419 | /* |
| 420 | * release a read lock |
| 421 | */ |
| 422 | void ldsem_up_read(struct ld_semaphore *sem) |
| 423 | { |
| 424 | long count; |
| 425 | |
| 426 | lockdep_release(sem, 1, _RET_IP_); |
| 427 | |
| 428 | count = ldsem_atomic_update(-LDSEM_READ_BIAS, sem); |
| 429 | if (count < 0 && (count & LDSEM_ACTIVE_MASK) == 0) |
| 430 | ldsem_wake(sem); |
| 431 | } |
| 432 | |
| 433 | /* |
| 434 | * release a write lock |
| 435 | */ |
| 436 | void ldsem_up_write(struct ld_semaphore *sem) |
| 437 | { |
| 438 | long count; |
| 439 | |
| 440 | lockdep_release(sem, 1, _RET_IP_); |
| 441 | |
| 442 | count = ldsem_atomic_update(-LDSEM_WRITE_BIAS, sem); |
| 443 | if (count < 0) |
| 444 | ldsem_wake(sem); |
| 445 | } |
| 446 | |
| 447 | |
| 448 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
| 449 | |
| 450 | int ldsem_down_read_nested(struct ld_semaphore *sem, int subclass, long timeout) |
| 451 | { |
| 452 | might_sleep(); |
| 453 | return __ldsem_down_read_nested(sem, subclass, timeout); |
| 454 | } |
| 455 | |
| 456 | int ldsem_down_write_nested(struct ld_semaphore *sem, int subclass, |
| 457 | long timeout) |
| 458 | { |
| 459 | might_sleep(); |
| 460 | return __ldsem_down_write_nested(sem, subclass, timeout); |
| 461 | } |
| 462 | |
| 463 | #endif |