xref: /openbmc/linux/drivers/tty/tty_ldsem.c (revision 930beb5a)
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 
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  */
94 static inline int ldsem_cmpxchg(long *old, long new, struct ld_semaphore *sem)
95 {
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 	}
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
464