[linux.git] / lib / rwsem.c

/* rwsem.c: R/W semaphores: contention handling functions
 *
 * Written by David Howells ([email protected]).
 * Derived from arch/i386/kernel/semaphore.c
 *
 * Writer lock-stealing by Alex Shi <[email protected]>
 * and Michel Lespinasse <[email protected]>
 */
#include <linux/rwsem.h>
#include <linux/sched.h>
#include <linux/init.h>
#include <linux/export.h>

/*
 * Initialize an rwsem:
 */
void __init_rwsem(struct rw_semaphore *sem, const char *name,
		  struct lock_class_key *key)
{
#ifdef CONFIG_DEBUG_LOCK_ALLOC
	/*
	 * Make sure we are not reinitializing a held semaphore:
	 */
	debug_check_no_locks_freed((void *)sem, sizeof(*sem));
	lockdep_init_map(&sem->dep_map, name, key, 0);
#endif
	sem->count = RWSEM_UNLOCKED_VALUE;
	raw_spin_lock_init(&sem->wait_lock);
	INIT_LIST_HEAD(&sem->wait_list);
}

EXPORT_SYMBOL(__init_rwsem);

enum rwsem_waiter_type {
	RWSEM_WAITING_FOR_WRITE,
	RWSEM_WAITING_FOR_READ
};

struct rwsem_waiter {
	struct list_head list;
	struct task_struct *task;
	enum rwsem_waiter_type type;
};

enum rwsem_wake_type {
	RWSEM_WAKE_ANY,		/* Wake whatever's at head of wait list */
	RWSEM_WAKE_READERS,	/* Wake readers only */
	RWSEM_WAKE_READ_OWNED	/* Waker thread holds the read lock */
};

/*
 * handle the lock release when processes blocked on it that can now run
 * - if we come here from up_xxxx(), then:
 *   - the 'active part' of count (&0x0000ffff) reached 0 (but may have changed)
 *   - the 'waiting part' of count (&0xffff0000) is -ve (and will still be so)
 * - there must be someone on the queue
 * - the spinlock must be held by the caller
 * - woken process blocks are discarded from the list after having task zeroed
 * - writers are only woken if downgrading is false
 */
static struct rw_semaphore *
__rwsem_do_wake(struct rw_semaphore *sem, enum rwsem_wake_type wake_type)
{
	struct rwsem_waiter *waiter;
	struct task_struct *tsk;
	struct list_head *next;
	long oldcount, woken, loop, adjustment;

	waiter = list_entry(sem->wait_list.next, struct rwsem_waiter, list);
	if (waiter->type == RWSEM_WAITING_FOR_WRITE) {
		if (wake_type == RWSEM_WAKE_ANY)
			/* Wake writer at the front of the queue, but do not
			 * grant it the lock yet as we want other writers
			 * to be able to steal it.  Readers, on the other hand,
			 * will block as they will notice the queued writer.
			 */
			wake_up_process(waiter->task);
		goto out;
	}

	/* Writers might steal the lock before we grant it to the next reader.
	 * We prefer to do the first reader grant before counting readers
	 * so we can bail out early if a writer stole the lock.
	 */
	adjustment = 0;
	if (wake_type != RWSEM_WAKE_READ_OWNED) {
		adjustment = RWSEM_ACTIVE_READ_BIAS;
 try_reader_grant:
		oldcount = rwsem_atomic_update(adjustment, sem) - adjustment;
		if (unlikely(oldcount < RWSEM_WAITING_BIAS)) {
			/* A writer stole the lock. Undo our reader grant. */
			if (rwsem_atomic_update(-adjustment, sem) &
						RWSEM_ACTIVE_MASK)
				goto out;
			/* Last active locker left. Retry waking readers. */
			goto try_reader_grant;
		}
	}

	/* Grant an infinite number of read locks to the readers at the front
	 * of the queue.  Note we increment the 'active part' of the count by
	 * the number of readers before waking any processes up.
	 */
	woken = 0;
	do {
		woken++;

		if (waiter->list.next == &sem->wait_list)
			break;

		waiter = list_entry(waiter->list.next,
					struct rwsem_waiter, list);

	} while (waiter->type != RWSEM_WAITING_FOR_WRITE);

	adjustment = woken * RWSEM_ACTIVE_READ_BIAS - adjustment;
	if (waiter->type != RWSEM_WAITING_FOR_WRITE)
		/* hit end of list above */
		adjustment -= RWSEM_WAITING_BIAS;

	if (adjustment)
		rwsem_atomic_add(adjustment, sem);

	next = sem->wait_list.next;
	loop = woken;
	do {
		waiter = list_entry(next, struct rwsem_waiter, list);
		next = waiter->list.next;
		tsk = waiter->task;
		smp_mb();
		waiter->task = NULL;
		wake_up_process(tsk);
		put_task_struct(tsk);
	} while (--loop);

	sem->wait_list.next = next;
	next->prev = &sem->wait_list;

 out:
	return sem;
}

/*
 * wait for the read lock to be granted
 */
struct rw_semaphore __sched *rwsem_down_read_failed(struct rw_semaphore *sem)
{
	long count, adjustment = -RWSEM_ACTIVE_READ_BIAS;
	struct rwsem_waiter waiter;
	struct task_struct *tsk = current;

	/* set up my own style of waitqueue */
	waiter.task = tsk;
	waiter.type = RWSEM_WAITING_FOR_READ;
	get_task_struct(tsk);

	raw_spin_lock_irq(&sem->wait_lock);
	if (list_empty(&sem->wait_list))
		adjustment += RWSEM_WAITING_BIAS;
	list_add_tail(&waiter.list, &sem->wait_list);

	/* we're now waiting on the lock, but no longer actively locking */
	count = rwsem_atomic_update(adjustment, sem);

	/* If there are no active locks, wake the front queued process(es).
	 *
	 * If there are no writers and we are first in the queue,
	 * wake our own waiter to join the existing active readers !
	 */
	if (count == RWSEM_WAITING_BIAS ||
	    (count > RWSEM_WAITING_BIAS &&
	     adjustment != -RWSEM_ACTIVE_READ_BIAS))
		sem = __rwsem_do_wake(sem, RWSEM_WAKE_ANY);

	raw_spin_unlock_irq(&sem->wait_lock);

	/* wait to be given the lock */
	while (true) {
		set_task_state(tsk, TASK_UNINTERRUPTIBLE);
		if (!waiter.task)
			break;
		schedule();
	}

	tsk->state = TASK_RUNNING;

	return sem;
}

/*
 * wait until we successfully acquire the write lock
 */
struct rw_semaphore __sched *rwsem_down_write_failed(struct rw_semaphore *sem)
{
	long count, adjustment = -RWSEM_ACTIVE_WRITE_BIAS;
	struct rwsem_waiter waiter;
	struct task_struct *tsk = current;

	/* set up my own style of waitqueue */
	waiter.task = tsk;
	waiter.type = RWSEM_WAITING_FOR_WRITE;

	raw_spin_lock_irq(&sem->wait_lock);
	if (list_empty(&sem->wait_list))
		adjustment += RWSEM_WAITING_BIAS;
	list_add_tail(&waiter.list, &sem->wait_list);

	/* we're now waiting on the lock, but no longer actively locking */
	count = rwsem_atomic_update(adjustment, sem);

	/* If there were already threads queued before us and there are no
	 * active writers, the lock must be read owned; so we try to wake
	 * any read locks that were queued ahead of us. */
	if (count > RWSEM_WAITING_BIAS &&
	    adjustment == -RWSEM_ACTIVE_WRITE_BIAS)
		sem = __rwsem_do_wake(sem, RWSEM_WAKE_READERS);

	/* wait until we successfully acquire the lock */
	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
	while (true) {
		if (!(count & RWSEM_ACTIVE_MASK)) {
			/* Try acquiring the write lock. */
			count = RWSEM_ACTIVE_WRITE_BIAS;
			if (!list_is_singular(&sem->wait_list))
				count += RWSEM_WAITING_BIAS;

			if (sem->count == RWSEM_WAITING_BIAS &&
			    cmpxchg(&sem->count, RWSEM_WAITING_BIAS, count) ==
							RWSEM_WAITING_BIAS)
				break;
		}

		raw_spin_unlock_irq(&sem->wait_lock);

		/* Block until there are no active lockers. */
		do {
			schedule();
			set_task_state(tsk, TASK_UNINTERRUPTIBLE);
		} while ((count = sem->count) & RWSEM_ACTIVE_MASK);

		raw_spin_lock_irq(&sem->wait_lock);
	}

	list_del(&waiter.list);
	raw_spin_unlock_irq(&sem->wait_lock);
	tsk->state = TASK_RUNNING;

	return sem;
}

/*
 * handle waking up a waiter on the semaphore
 * - up_read/up_write has decremented the active part of count if we come here
 */
struct rw_semaphore *rwsem_wake(struct rw_semaphore *sem)
{
	unsigned long flags;

	raw_spin_lock_irqsave(&sem->wait_lock, flags);

	/* do nothing if list empty */
	if (!list_empty(&sem->wait_list))
		sem = __rwsem_do_wake(sem, RWSEM_WAKE_ANY);

	raw_spin_unlock_irqrestore(&sem->wait_lock, flags);

	return sem;
}

/*
 * downgrade a write lock into a read lock
 * - caller incremented waiting part of count and discovered it still negative
 * - just wake up any readers at the front of the queue
 */
struct rw_semaphore *rwsem_downgrade_wake(struct rw_semaphore *sem)
{
	unsigned long flags;

	raw_spin_lock_irqsave(&sem->wait_lock, flags);

	/* do nothing if list empty */
	if (!list_empty(&sem->wait_list))
		sem = __rwsem_do_wake(sem, RWSEM_WAKE_READ_OWNED);

	raw_spin_unlock_irqrestore(&sem->wait_lock, flags);

	return sem;
}

EXPORT_SYMBOL(rwsem_down_read_failed);
EXPORT_SYMBOL(rwsem_down_write_failed);
EXPORT_SYMBOL(rwsem_wake);
EXPORT_SYMBOL(rwsem_downgrade_wake);
Commit	Line	Data
1da177e4 LT	1	/* rwsem.c: R/W semaphores: contention handling functions
	2	*
	3	* Written by David Howells ([email protected]).
	4	* Derived from arch/i386/kernel/semaphore.c
ce6711f3 AS	5	*
ce6711f3 AS	6	* Writer lock-stealing by Alex Shi <[email protected]>
fe6e674c	7	* and Michel Lespinasse <[email protected]>
1da177e4 LT	8	*/
	9	#include <linux/rwsem.h>
	10	#include <linux/sched.h>
	11	#include <linux/init.h>
8bc3bcc9	12	#include <linux/export.h>
1da177e4	13
4ea2176d IM	14	/*
	15	* Initialize an rwsem:
	16	*/
	17	void __init_rwsem(struct rw_semaphore sem, const char name,
	18	struct lock_class_key *key)
	19	{
	20	#ifdef CONFIG_DEBUG_LOCK_ALLOC
	21	/*
	22	* Make sure we are not reinitializing a held semaphore:
	23	*/
	24	debug_check_no_locks_freed((void )sem, sizeof(sem));
4dfbb9d8	25	lockdep_init_map(&sem->dep_map, name, key, 0);
4ea2176d IM	26	#endif
4ea2176d IM	27	sem->count = RWSEM_UNLOCKED_VALUE;
ddb6c9b5	28	raw_spin_lock_init(&sem->wait_lock);
4ea2176d IM	29	INIT_LIST_HEAD(&sem->wait_list);
	30	}
	31
	32	EXPORT_SYMBOL(__init_rwsem);
	33
e2d57f78 ML	34	enum rwsem_waiter_type {
	35	RWSEM_WAITING_FOR_WRITE,
	36	RWSEM_WAITING_FOR_READ
	37	};
	38
1da177e4 LT	39	struct rwsem_waiter {
	40	struct list_head list;
	41	struct task_struct *task;
e2d57f78	42	enum rwsem_waiter_type type;
1da177e4 LT	43	};
1da177e4 LT	44
fe6e674c ML	45	enum rwsem_wake_type {
	46	RWSEM_WAKE_ANY, /* Wake whatever's at head of wait list */
	47	RWSEM_WAKE_READERS, /* Wake readers only */
	48	RWSEM_WAKE_READ_OWNED /* Waker thread holds the read lock */
	49	};
70bdc6e0	50
1da177e4 LT	51	/*
	52	* handle the lock release when processes blocked on it that can now run
	53	* - if we come here from up_xxxx(), then:
	54	* - the 'active part' of count (&0x0000ffff) reached 0 (but may have changed)
	55	* - the 'waiting part' of count (&0xffff0000) is -ve (and will still be so)
345af7bf	56	* - there must be someone on the queue
1da177e4 LT	57	* - the spinlock must be held by the caller
	58	* - woken process blocks are discarded from the list after having task zeroed
	59	* - writers are only woken if downgrading is false
	60	*/
70bdc6e0	61	static struct rw_semaphore *
fe6e674c	62	__rwsem_do_wake(struct rw_semaphore *sem, enum rwsem_wake_type wake_type)
1da177e4 LT	63	{
	64	struct rwsem_waiter *waiter;
	65	struct task_struct *tsk;
	66	struct list_head *next;
b5f54181	67	long oldcount, woken, loop, adjustment;
1da177e4	68
345af7bf	69	waiter = list_entry(sem->wait_list.next, struct rwsem_waiter, list);
8cf5322c	70	if (waiter->type == RWSEM_WAITING_FOR_WRITE) {
fe6e674c	71	if (wake_type == RWSEM_WAKE_ANY)
8cf5322c ML	72	/* Wake writer at the front of the queue, but do not
	73	* grant it the lock yet as we want other writers
	74	* to be able to steal it. Readers, on the other hand,
	75	* will block as they will notice the queued writer.
	76	*/
	77	wake_up_process(waiter->task);
345af7bf	78	goto out;
8cf5322c	79	}
1da177e4	80
fe6e674c ML	81	/* Writers might steal the lock before we grant it to the next reader.
	82	* We prefer to do the first reader grant before counting readers
	83	* so we can bail out early if a writer stole the lock.
70bdc6e0	84	*/
fe6e674c ML	85	adjustment = 0;
	86	if (wake_type != RWSEM_WAKE_READ_OWNED) {
	87	adjustment = RWSEM_ACTIVE_READ_BIAS;
	88	try_reader_grant:
	89	oldcount = rwsem_atomic_update(adjustment, sem) - adjustment;
	90	if (unlikely(oldcount < RWSEM_WAITING_BIAS)) {
	91	/* A writer stole the lock. Undo our reader grant. */
	92	if (rwsem_atomic_update(-adjustment, sem) &
	93	RWSEM_ACTIVE_MASK)
	94	goto out;
	95	/* Last active locker left. Retry waking readers. */
	96	goto try_reader_grant;
	97	}
	98	}
1da177e4	99
345af7bf ML	100	/* Grant an infinite number of read locks to the readers at the front
	101	* of the queue. Note we increment the 'active part' of the count by
	102	* the number of readers before waking any processes up.
1da177e4	103	*/
1da177e4 LT	104	woken = 0;
	105	do {
	106	woken++;
	107
	108	if (waiter->list.next == &sem->wait_list)
	109	break;
	110
	111	waiter = list_entry(waiter->list.next,
	112	struct rwsem_waiter, list);
	113
e2d57f78	114	} while (waiter->type != RWSEM_WAITING_FOR_WRITE);
1da177e4	115
fe6e674c	116	adjustment = woken * RWSEM_ACTIVE_READ_BIAS - adjustment;
e2d57f78	117	if (waiter->type != RWSEM_WAITING_FOR_WRITE)
fd41b334 ML	118	/* hit end of list above */
fd41b334 ML	119	adjustment -= RWSEM_WAITING_BIAS;
1da177e4	120
fe6e674c ML	121	if (adjustment)
fe6e674c ML	122	rwsem_atomic_add(adjustment, sem);
1da177e4 LT	123
1da177e4 LT	124	next = sem->wait_list.next;
8cf5322c ML	125	loop = woken;
8cf5322c ML	126	do {
1da177e4 LT	127	waiter = list_entry(next, struct rwsem_waiter, list);
	128	next = waiter->list.next;
	129	tsk = waiter->task;
d59dd462	130	smp_mb();
1da177e4 LT	131	waiter->task = NULL;
	132	wake_up_process(tsk);
	133	put_task_struct(tsk);
8cf5322c	134	} while (--loop);
1da177e4 LT	135
	136	sem->wait_list.next = next;
	137	next->prev = &sem->wait_list;
	138
	139	out:
1da177e4	140	return sem;
ce6711f3 AS	141	}
ce6711f3 AS	142
1da177e4	143	/*
1e78277c	144	* wait for the read lock to be granted
1da177e4	145	*/
1e78277c	146	struct rw_semaphore __sched rwsem_down_read_failed(struct rw_semaphore sem)
1da177e4	147	{
b5f54181	148	long count, adjustment = -RWSEM_ACTIVE_READ_BIAS;
a8618a0e	149	struct rwsem_waiter waiter;
1da177e4	150	struct task_struct *tsk = current;
1da177e4	151
1da177e4	152	/* set up my own style of waitqueue */
a8618a0e	153	waiter.task = tsk;
da16922c	154	waiter.type = RWSEM_WAITING_FOR_READ;
1da177e4 LT	155	get_task_struct(tsk);
1da177e4 LT	156
f7dd1cee	157	raw_spin_lock_irq(&sem->wait_lock);
fd41b334 ML	158	if (list_empty(&sem->wait_list))
fd41b334 ML	159	adjustment += RWSEM_WAITING_BIAS;
a8618a0e	160	list_add_tail(&waiter.list, &sem->wait_list);
1da177e4	161
70bdc6e0	162	/* we're now waiting on the lock, but no longer actively locking */
1da177e4 LT	163	count = rwsem_atomic_update(adjustment, sem);
1da177e4 LT	164
25c39325 ML	165	/* If there are no active locks, wake the front queued process(es).
	166	*
	167	* If there are no writers and we are first in the queue,
	168	* wake our own waiter to join the existing active readers !
	169	*/
	170	if (count == RWSEM_WAITING_BIAS \|\|
	171	(count > RWSEM_WAITING_BIAS &&
	172	adjustment != -RWSEM_ACTIVE_READ_BIAS))
fe6e674c	173	sem = __rwsem_do_wake(sem, RWSEM_WAKE_ANY);
1da177e4	174
ddb6c9b5	175	raw_spin_unlock_irq(&sem->wait_lock);
1da177e4 LT	176
1da177e4 LT	177	/* wait to be given the lock */
f7dd1cee ML	178	while (true) {
f7dd1cee ML	179	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
a8618a0e	180	if (!waiter.task)
1da177e4 LT	181	break;
1da177e4 LT	182	schedule();
1da177e4 LT	183	}
	184
	185	tsk->state = TASK_RUNNING;
	186
	187	return sem;
	188	}
	189
1da177e4	190	/*
023fe4f7	191	* wait until we successfully acquire the write lock
1da177e4	192	*/
d1233754	193	struct rw_semaphore __sched rwsem_down_write_failed(struct rw_semaphore sem)
1da177e4	194	{
b5f54181	195	long count, adjustment = -RWSEM_ACTIVE_WRITE_BIAS;
1e78277c ML	196	struct rwsem_waiter waiter;
1e78277c ML	197	struct task_struct *tsk = current;
1e78277c ML	198
	199	/* set up my own style of waitqueue */
	200	waiter.task = tsk;
023fe4f7	201	waiter.type = RWSEM_WAITING_FOR_WRITE;
1e78277c ML	202
	203	raw_spin_lock_irq(&sem->wait_lock);
	204	if (list_empty(&sem->wait_list))
	205	adjustment += RWSEM_WAITING_BIAS;
	206	list_add_tail(&waiter.list, &sem->wait_list);
	207
	208	/* we're now waiting on the lock, but no longer actively locking */
	209	count = rwsem_atomic_update(adjustment, sem);
	210
ed00f643 ML	211	/* If there were already threads queued before us and there are no
	212	* active writers, the lock must be read owned; so we try to wake
	213	* any read locks that were queued ahead of us. */
	214	if (count > RWSEM_WAITING_BIAS &&
	215	adjustment == -RWSEM_ACTIVE_WRITE_BIAS)
fe6e674c	216	sem = __rwsem_do_wake(sem, RWSEM_WAKE_READERS);
1e78277c	217
023fe4f7	218	/* wait until we successfully acquire the lock */
a7d2c573	219	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
1e78277c	220	while (true) {
9b0fc9c0 ML	221	if (!(count & RWSEM_ACTIVE_MASK)) {
	222	/* Try acquiring the write lock. */
	223	count = RWSEM_ACTIVE_WRITE_BIAS;
	224	if (!list_is_singular(&sem->wait_list))
	225	count += RWSEM_WAITING_BIAS;
9607a85b DB	226
	227	if (sem->count == RWSEM_WAITING_BIAS &&
	228	cmpxchg(&sem->count, RWSEM_WAITING_BIAS, count) ==
5ede972d	229	RWSEM_WAITING_BIAS)
9b0fc9c0 ML	230	break;
9b0fc9c0 ML	231	}
1e78277c	232
1e78277c	233	raw_spin_unlock_irq(&sem->wait_lock);
a7d2c573 ML	234
	235	/* Block until there are no active lockers. */
	236	do {
	237	schedule();
	238	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
9b0fc9c0	239	} while ((count = sem->count) & RWSEM_ACTIVE_MASK);
a7d2c573	240
023fe4f7	241	raw_spin_lock_irq(&sem->wait_lock);
1e78277c ML	242	}
1e78277c ML	243
023fe4f7 ML	244	list_del(&waiter.list);
023fe4f7 ML	245	raw_spin_unlock_irq(&sem->wait_lock);
1e78277c ML	246	tsk->state = TASK_RUNNING;
	247
	248	return sem;
1da177e4 LT	249	}
	250
	251	/*
	252	* handle waking up a waiter on the semaphore
	253	* - up_read/up_write has decremented the active part of count if we come here
	254	*/
d1233754	255	struct rw_semaphore rwsem_wake(struct rw_semaphore sem)
1da177e4 LT	256	{
	257	unsigned long flags;
	258
ddb6c9b5	259	raw_spin_lock_irqsave(&sem->wait_lock, flags);
1da177e4 LT	260
	261	/* do nothing if list empty */
	262	if (!list_empty(&sem->wait_list))
70bdc6e0	263	sem = __rwsem_do_wake(sem, RWSEM_WAKE_ANY);
1da177e4	264
ddb6c9b5	265	raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
1da177e4	266
1da177e4 LT	267	return sem;
	268	}
	269
	270	/*
	271	* downgrade a write lock into a read lock
	272	* - caller incremented waiting part of count and discovered it still negative
	273	* - just wake up any readers at the front of the queue
	274	*/
d1233754	275	struct rw_semaphore rwsem_downgrade_wake(struct rw_semaphore sem)
1da177e4 LT	276	{
	277	unsigned long flags;
	278
ddb6c9b5	279	raw_spin_lock_irqsave(&sem->wait_lock, flags);
1da177e4 LT	280
	281	/* do nothing if list empty */
	282	if (!list_empty(&sem->wait_list))
70bdc6e0	283	sem = __rwsem_do_wake(sem, RWSEM_WAKE_READ_OWNED);
1da177e4	284
ddb6c9b5	285	raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
1da177e4	286
1da177e4 LT	287	return sem;
	288	}
	289
	290	EXPORT_SYMBOL(rwsem_down_read_failed);
	291	EXPORT_SYMBOL(rwsem_down_write_failed);
	292	EXPORT_SYMBOL(rwsem_wake);
	293	EXPORT_SYMBOL(rwsem_downgrade_wake);