[deliverable/linux.git] / lib / rwsem.c

/* rwsem.c: R/W semaphores: contention handling functions
 *
 * Written by David Howells (dhowells@redhat.com).
 * Derived from arch/i386/kernel/semaphore.c
 *
 * Writer lock-stealing by Alex Shi <alex.shi@intel.com>
 */
#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;
};

/* Wake types for __rwsem_do_wake().  Note that RWSEM_WAKE_NO_ACTIVE and
 * RWSEM_WAKE_READ_OWNED imply that the spinlock must have been kept held
 * since the rwsem value was observed.
 */
#define RWSEM_WAKE_ANY        0 /* Wake whatever's at head of wait list */
#define RWSEM_WAKE_NO_ACTIVE  1 /* rwsem was observed with no active thread */
#define RWSEM_WAKE_READ_OWNED 2 /* rwsem was observed to be read owned */

/*
 * 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, int wake_type)
{
	struct rwsem_waiter *waiter;
	struct task_struct *tsk;
	struct list_head *next;
	signed long 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_READ_OWNED)
			/* 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;
	}

	/* If we come here from up_xxxx(), another thread might have reached
	 * rwsem_down_failed_common() before we acquired the spinlock and
	 * woken up a waiter, making it now active.  We prefer to check for
	 * this first in order to not spend too much time with the spinlock
	 * held if we're not going to be able to wake up readers in the end.
	 *
	 * Note that we do not need to update the rwsem count: any writer
	 * trying to acquire rwsem will run rwsem_down_write_failed() due
	 * to the waiting threads and block trying to acquire the spinlock.
	 *
	 * We use a dummy atomic update in order to acquire the cache line
	 * exclusively since we expect to succeed and run the final rwsem
	 * count adjustment pretty soon.
	 */
	if (wake_type == RWSEM_WAKE_ANY &&
	    rwsem_atomic_update(0, sem) < RWSEM_WAITING_BIAS)
		/* Someone grabbed the sem for write already */
		goto out;

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

	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)
{
	signed long adjustment = -RWSEM_ACTIVE_READ_BIAS;
	struct rwsem_waiter waiter;
	struct task_struct *tsk = current;
	signed long count;

	/* 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 (count == RWSEM_WAITING_BIAS)
		sem = __rwsem_do_wake(sem, RWSEM_WAKE_NO_ACTIVE);

	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)
{
	signed long adjustment = -RWSEM_ACTIVE_WRITE_BIAS;
	struct rwsem_waiter waiter;
	struct task_struct *tsk = current;
	signed long count;

	/* 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_READ_OWNED);

	/* 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 (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 (dhowells@redhat.com).
	4	* Derived from arch/i386/kernel/semaphore.c
ce6711f3 AS	5	*
ce6711f3 AS	6	* Writer lock-stealing by Alex Shi <alex.shi@intel.com>
1da177e4 LT	7	*/
	8	#include <linux/rwsem.h>
	9	#include <linux/sched.h>
	10	#include <linux/init.h>
8bc3bcc9	11	#include <linux/export.h>
1da177e4	12
4ea2176d IM	13	/*
	14	* Initialize an rwsem:
	15	*/
	16	void __init_rwsem(struct rw_semaphore sem, const char name,
	17	struct lock_class_key *key)
	18	{
	19	#ifdef CONFIG_DEBUG_LOCK_ALLOC
	20	/*
	21	* Make sure we are not reinitializing a held semaphore:
	22	*/
	23	debug_check_no_locks_freed((void )sem, sizeof(sem));
4dfbb9d8	24	lockdep_init_map(&sem->dep_map, name, key, 0);
4ea2176d IM	25	#endif
4ea2176d IM	26	sem->count = RWSEM_UNLOCKED_VALUE;
ddb6c9b5	27	raw_spin_lock_init(&sem->wait_lock);
4ea2176d IM	28	INIT_LIST_HEAD(&sem->wait_list);
	29	}
	30
	31	EXPORT_SYMBOL(__init_rwsem);
	32
e2d57f78 ML	33	enum rwsem_waiter_type {
	34	RWSEM_WAITING_FOR_WRITE,
	35	RWSEM_WAITING_FOR_READ
	36	};
	37
1da177e4 LT	38	struct rwsem_waiter {
	39	struct list_head list;
	40	struct task_struct *task;
e2d57f78	41	enum rwsem_waiter_type type;
1da177e4 LT	42	};
1da177e4 LT	43
70bdc6e0 ML	44	/* Wake types for __rwsem_do_wake(). Note that RWSEM_WAKE_NO_ACTIVE and
	45	* RWSEM_WAKE_READ_OWNED imply that the spinlock must have been kept held
	46	* since the rwsem value was observed.
	47	*/
	48	#define RWSEM_WAKE_ANY 0 /* Wake whatever's at head of wait list */
	49	#define RWSEM_WAKE_NO_ACTIVE 1 /* rwsem was observed with no active thread */
	50	#define RWSEM_WAKE_READ_OWNED 2 /* rwsem was observed to be read owned */
	51
1da177e4 LT	52	/*
	53	* handle the lock release when processes blocked on it that can now run
	54	* - if we come here from up_xxxx(), then:
	55	* - the 'active part' of count (&0x0000ffff) reached 0 (but may have changed)
	56	* - the 'waiting part' of count (&0xffff0000) is -ve (and will still be so)
345af7bf	57	* - there must be someone on the queue
1da177e4 LT	58	* - the spinlock must be held by the caller
	59	* - woken process blocks are discarded from the list after having task zeroed
	60	* - writers are only woken if downgrading is false
	61	*/
70bdc6e0 ML	62	static struct rw_semaphore *
70bdc6e0 ML	63	__rwsem_do_wake(struct rw_semaphore *sem, int wake_type)
1da177e4 LT	64	{
	65	struct rwsem_waiter *waiter;
	66	struct task_struct *tsk;
	67	struct list_head *next;
ce6711f3	68	signed long woken, loop, adjustment;
1da177e4	69
345af7bf	70	waiter = list_entry(sem->wait_list.next, struct rwsem_waiter, list);
8cf5322c ML	71	if (waiter->type == RWSEM_WAITING_FOR_WRITE) {
	72	if (wake_type != RWSEM_WAKE_READ_OWNED)
	73	/* Wake writer at the front of the queue, but do not
	74	* grant it the lock yet as we want other writers
	75	* to be able to steal it. Readers, on the other hand,
	76	* will block as they will notice the queued writer.
	77	*/
	78	wake_up_process(waiter->task);
345af7bf	79	goto out;
8cf5322c	80	}
1da177e4	81
70bdc6e0 ML	82	/* If we come here from up_xxxx(), another thread might have reached
	83	* rwsem_down_failed_common() before we acquired the spinlock and
	84	* woken up a waiter, making it now active. We prefer to check for
	85	* this first in order to not spend too much time with the spinlock
	86	* held if we're not going to be able to wake up readers in the end.
	87	*
	88	* Note that we do not need to update the rwsem count: any writer
	89	* trying to acquire rwsem will run rwsem_down_write_failed() due
	90	* to the waiting threads and block trying to acquire the spinlock.
	91	*
	92	* We use a dummy atomic update in order to acquire the cache line
	93	* exclusively since we expect to succeed and run the final rwsem
	94	* count adjustment pretty soon.
	95	*/
	96	if (wake_type == RWSEM_WAKE_ANY &&
424acaae ML	97	rwsem_atomic_update(0, sem) < RWSEM_WAITING_BIAS)
424acaae ML	98	/* Someone grabbed the sem for write already */
70bdc6e0	99	goto out;
1da177e4	100
345af7bf ML	101	/* Grant an infinite number of read locks to the readers at the front
	102	* of the queue. Note we increment the 'active part' of the count by
	103	* the number of readers before waking any processes up.
1da177e4	104	*/
1da177e4 LT	105	woken = 0;
	106	do {
	107	woken++;
	108
	109	if (waiter->list.next == &sem->wait_list)
	110	break;
	111
	112	waiter = list_entry(waiter->list.next,
	113	struct rwsem_waiter, list);
	114
e2d57f78	115	} while (waiter->type != RWSEM_WAITING_FOR_WRITE);
1da177e4	116
fd41b334	117	adjustment = woken * RWSEM_ACTIVE_READ_BIAS;
e2d57f78	118	if (waiter->type != RWSEM_WAITING_FOR_WRITE)
fd41b334 ML	119	/* hit end of list above */
fd41b334 ML	120	adjustment -= RWSEM_WAITING_BIAS;
1da177e4	121
fd41b334	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	{
1e78277c	148	signed long adjustment = -RWSEM_ACTIVE_READ_BIAS;
a8618a0e	149	struct rwsem_waiter waiter;
1da177e4 LT	150	struct task_struct *tsk = current;
	151	signed long count;
	152
1da177e4	153	/* set up my own style of waitqueue */
a8618a0e	154	waiter.task = tsk;
da16922c	155	waiter.type = RWSEM_WAITING_FOR_READ;
1da177e4 LT	156	get_task_struct(tsk);
1da177e4 LT	157
f7dd1cee	158	raw_spin_lock_irq(&sem->wait_lock);
fd41b334 ML	159	if (list_empty(&sem->wait_list))
fd41b334 ML	160	adjustment += RWSEM_WAITING_BIAS;
a8618a0e	161	list_add_tail(&waiter.list, &sem->wait_list);
1da177e4	162
70bdc6e0	163	/* we're now waiting on the lock, but no longer actively locking */
1da177e4 LT	164	count = rwsem_atomic_update(adjustment, sem);
1da177e4 LT	165
da16922c	166	/* If there are no active locks, wake the front queued process(es). */
424acaae	167	if (count == RWSEM_WAITING_BIAS)
70bdc6e0	168	sem = __rwsem_do_wake(sem, RWSEM_WAKE_NO_ACTIVE);
1da177e4	169
ddb6c9b5	170	raw_spin_unlock_irq(&sem->wait_lock);
1da177e4 LT	171
1da177e4 LT	172	/* wait to be given the lock */
f7dd1cee ML	173	while (true) {
f7dd1cee ML	174	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
a8618a0e	175	if (!waiter.task)
1da177e4 LT	176	break;
1da177e4 LT	177	schedule();
1da177e4 LT	178	}
	179
	180	tsk->state = TASK_RUNNING;
	181
	182	return sem;
	183	}
	184
1da177e4	185	/*
023fe4f7	186	* wait until we successfully acquire the write lock
1da177e4	187	*/
d1233754	188	struct rw_semaphore __sched rwsem_down_write_failed(struct rw_semaphore sem)
1da177e4	189	{
1e78277c ML	190	signed long adjustment = -RWSEM_ACTIVE_WRITE_BIAS;
	191	struct rwsem_waiter waiter;
	192	struct task_struct *tsk = current;
	193	signed long count;
	194
	195	/* set up my own style of waitqueue */
	196	waiter.task = tsk;
023fe4f7	197	waiter.type = RWSEM_WAITING_FOR_WRITE;
1e78277c ML	198
	199	raw_spin_lock_irq(&sem->wait_lock);
	200	if (list_empty(&sem->wait_list))
	201	adjustment += RWSEM_WAITING_BIAS;
	202	list_add_tail(&waiter.list, &sem->wait_list);
	203
	204	/* we're now waiting on the lock, but no longer actively locking */
	205	count = rwsem_atomic_update(adjustment, sem);
	206
ed00f643 ML	207	/* If there were already threads queued before us and there are no
	208	* active writers, the lock must be read owned; so we try to wake
	209	* any read locks that were queued ahead of us. */
	210	if (count > RWSEM_WAITING_BIAS &&
	211	adjustment == -RWSEM_ACTIVE_WRITE_BIAS)
1e78277c ML	212	sem = __rwsem_do_wake(sem, RWSEM_WAKE_READ_OWNED);
1e78277c ML	213
023fe4f7	214	/* wait until we successfully acquire the lock */
a7d2c573	215	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
1e78277c	216	while (true) {
9b0fc9c0 ML	217	if (!(count & RWSEM_ACTIVE_MASK)) {
	218	/* Try acquiring the write lock. */
	219	count = RWSEM_ACTIVE_WRITE_BIAS;
	220	if (!list_is_singular(&sem->wait_list))
	221	count += RWSEM_WAITING_BIAS;
	222	if (cmpxchg(&sem->count, RWSEM_WAITING_BIAS, count) ==
5ede972d	223	RWSEM_WAITING_BIAS)
9b0fc9c0 ML	224	break;
9b0fc9c0 ML	225	}
1e78277c	226
1e78277c	227	raw_spin_unlock_irq(&sem->wait_lock);
a7d2c573 ML	228
	229	/* Block until there are no active lockers. */
	230	do {
	231	schedule();
	232	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
9b0fc9c0	233	} while ((count = sem->count) & RWSEM_ACTIVE_MASK);
a7d2c573	234
023fe4f7	235	raw_spin_lock_irq(&sem->wait_lock);
1e78277c ML	236	}
1e78277c ML	237
023fe4f7 ML	238	list_del(&waiter.list);
023fe4f7 ML	239	raw_spin_unlock_irq(&sem->wait_lock);
1e78277c ML	240	tsk->state = TASK_RUNNING;
	241
	242	return sem;
1da177e4 LT	243	}
	244
	245	/*
	246	* handle waking up a waiter on the semaphore
	247	* - up_read/up_write has decremented the active part of count if we come here
	248	*/
d1233754	249	struct rw_semaphore rwsem_wake(struct rw_semaphore sem)
1da177e4 LT	250	{
	251	unsigned long flags;
	252
ddb6c9b5	253	raw_spin_lock_irqsave(&sem->wait_lock, flags);
1da177e4 LT	254
	255	/* do nothing if list empty */
	256	if (!list_empty(&sem->wait_list))
70bdc6e0	257	sem = __rwsem_do_wake(sem, RWSEM_WAKE_ANY);
1da177e4	258
ddb6c9b5	259	raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
1da177e4	260
1da177e4 LT	261	return sem;
	262	}
	263
	264	/*
	265	* downgrade a write lock into a read lock
	266	* - caller incremented waiting part of count and discovered it still negative
	267	* - just wake up any readers at the front of the queue
	268	*/
d1233754	269	struct rw_semaphore rwsem_downgrade_wake(struct rw_semaphore sem)
1da177e4 LT	270	{
	271	unsigned long flags;
	272
ddb6c9b5	273	raw_spin_lock_irqsave(&sem->wait_lock, flags);
1da177e4 LT	274
	275	/* do nothing if list empty */
	276	if (!list_empty(&sem->wait_list))
70bdc6e0	277	sem = __rwsem_do_wake(sem, RWSEM_WAKE_READ_OWNED);
1da177e4	278
ddb6c9b5	279	raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
1da177e4	280
1da177e4 LT	281	return sem;
	282	}
	283
	284	EXPORT_SYMBOL(rwsem_down_read_failed);
	285	EXPORT_SYMBOL(rwsem_down_write_failed);
	286	EXPORT_SYMBOL(rwsem_wake);
	287	EXPORT_SYMBOL(rwsem_downgrade_wake);