[mirror_ubuntu-hirsute-kernel.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
 */
#include <linux/rwsem.h>
#include <linux/sched.h>
#include <linux/init.h>
#include <linux/module.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;
	spin_lock_init(&sem->wait_lock);
	INIT_LIST_HEAD(&sem->wait_list);
}

EXPORT_SYMBOL(__init_rwsem);

struct rwsem_waiter {
	struct list_head list;
	struct task_struct *task;
	unsigned int flags;
#define RWSEM_WAITING_FOR_READ	0x00000001
#define RWSEM_WAITING_FOR_WRITE	0x00000002
};

/* 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 oldcount, woken, loop, adjustment;

	waiter = list_entry(sem->wait_list.next, struct rwsem_waiter, list);
	if (!(waiter->flags & RWSEM_WAITING_FOR_WRITE))
		goto readers_only;

	if (wake_type == RWSEM_WAKE_READ_OWNED)
		goto out;

	/* There's a writer at the front of the queue - try to grant it the
	 * write lock.  However, we only wake this writer if we can transition
	 * the active part of the count from 0 -> 1
	 */
	adjustment = RWSEM_ACTIVE_WRITE_BIAS;
	if (waiter->list.next == &sem->wait_list)
		adjustment -= RWSEM_WAITING_BIAS;

 try_again_write:
	oldcount = rwsem_atomic_update(adjustment, sem) - adjustment;
	if (oldcount & RWSEM_ACTIVE_MASK)
		/* Someone grabbed the sem already */
		goto undo_write;

	/* We must be careful not to touch 'waiter' after we set ->task = NULL.
	 * It is an allocated on the waiter's stack and may become invalid at
	 * any time after that point (due to a wakeup from another source).
	 */
	list_del(&waiter->list);
	tsk = waiter->task;
	smp_mb();
	waiter->task = NULL;
	wake_up_process(tsk);
	put_task_struct(tsk);
	goto out;

 readers_only:
	/* 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_ACTIVE_MASK))
		/* Someone grabbed the sem 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->flags & RWSEM_WAITING_FOR_READ);

	adjustment = woken * RWSEM_ACTIVE_READ_BIAS;
	if (waiter->flags & RWSEM_WAITING_FOR_READ)
		/* hit end of list above */
		adjustment -= RWSEM_WAITING_BIAS;

	rwsem_atomic_add(adjustment, sem);

	next = sem->wait_list.next;
	for (loop = woken; loop > 0; loop--) {
		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);
	}

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

 out:
	return sem;

	/* undo the change to the active count, but check for a transition
	 * 1->0 */
 undo_write:
	if (rwsem_atomic_update(-adjustment, sem) & RWSEM_ACTIVE_MASK)
		goto out;
	goto try_again_write;
}

/*
 * wait for a lock to be granted
 */
static struct rw_semaphore __sched *
rwsem_down_failed_common(struct rw_semaphore *sem,
			struct rwsem_waiter *waiter, signed long adjustment)
{
	struct task_struct *tsk = current;
	signed long count;

	set_task_state(tsk, TASK_UNINTERRUPTIBLE);

	/* set up my own style of waitqueue */
	spin_lock_irq(&sem->wait_lock);
	waiter->task = tsk;
	get_task_struct(tsk);

	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) up */
	if (!(count & RWSEM_ACTIVE_MASK))
		sem = __rwsem_do_wake(sem, RWSEM_WAKE_NO_ACTIVE);

	spin_unlock_irq(&sem->wait_lock);

	/* wait to be given the lock */
	for (;;) {
		if (!waiter->task)
			break;
		schedule();
		set_task_state(tsk, TASK_UNINTERRUPTIBLE);
	}

	tsk->state = TASK_RUNNING;

	return sem;
}

/*
 * wait for the read lock to be granted
 */
asmregparm struct rw_semaphore __sched *
rwsem_down_read_failed(struct rw_semaphore *sem)
{
	struct rwsem_waiter waiter;

	waiter.flags = RWSEM_WAITING_FOR_READ;
	rwsem_down_failed_common(sem, &waiter, -RWSEM_ACTIVE_READ_BIAS);
	return sem;
}

/*
 * wait for the write lock to be granted
 */
asmregparm struct rw_semaphore __sched *
rwsem_down_write_failed(struct rw_semaphore *sem)
{
	struct rwsem_waiter waiter;

	waiter.flags = RWSEM_WAITING_FOR_WRITE;
	rwsem_down_failed_common(sem, &waiter, -RWSEM_ACTIVE_WRITE_BIAS);

	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
 */
asmregparm struct rw_semaphore *rwsem_wake(struct rw_semaphore *sem)
{
	unsigned long flags;

	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);

	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
 */
asmregparm struct rw_semaphore *rwsem_downgrade_wake(struct rw_semaphore *sem)
{
	unsigned long flags;

	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);

	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
	5	*/
	6	#include <linux/rwsem.h>
	7	#include <linux/sched.h>
	8	#include <linux/init.h>
	9	#include <linux/module.h>
	10
4ea2176d IM	11	/*
	12	* Initialize an rwsem:
	13	*/
	14	void __init_rwsem(struct rw_semaphore sem, const char name,
	15	struct lock_class_key *key)
	16	{
	17	#ifdef CONFIG_DEBUG_LOCK_ALLOC
	18	/*
	19	* Make sure we are not reinitializing a held semaphore:
	20	*/
	21	debug_check_no_locks_freed((void )sem, sizeof(sem));
4dfbb9d8	22	lockdep_init_map(&sem->dep_map, name, key, 0);
4ea2176d IM	23	#endif
	24	sem->count = RWSEM_UNLOCKED_VALUE;
	25	spin_lock_init(&sem->wait_lock);
	26	INIT_LIST_HEAD(&sem->wait_list);
	27	}
	28
	29	EXPORT_SYMBOL(__init_rwsem);
	30
1da177e4 LT	31	struct rwsem_waiter {
	32	struct list_head list;
	33	struct task_struct *task;
	34	unsigned int flags;
	35	#define RWSEM_WAITING_FOR_READ 0x00000001
	36	#define RWSEM_WAITING_FOR_WRITE 0x00000002
	37	};
	38
70bdc6e0 ML	39	/* Wake types for __rwsem_do_wake(). Note that RWSEM_WAKE_NO_ACTIVE and
	40	* RWSEM_WAKE_READ_OWNED imply that the spinlock must have been kept held
	41	* since the rwsem value was observed.
	42	*/
	43	#define RWSEM_WAKE_ANY 0 /* Wake whatever's at head of wait list */
	44	#define RWSEM_WAKE_NO_ACTIVE 1 /* rwsem was observed with no active thread */
	45	#define RWSEM_WAKE_READ_OWNED 2 /* rwsem was observed to be read owned */
	46
1da177e4 LT	47	/*
	48	* handle the lock release when processes blocked on it that can now run
	49	* - if we come here from up_xxxx(), then:
	50	* - the 'active part' of count (&0x0000ffff) reached 0 (but may have changed)
	51	* - the 'waiting part' of count (&0xffff0000) is -ve (and will still be so)
345af7bf	52	* - there must be someone on the queue
1da177e4 LT	53	* - the spinlock must be held by the caller
	54	* - woken process blocks are discarded from the list after having task zeroed
	55	* - writers are only woken if downgrading is false
	56	*/
70bdc6e0 ML	57	static struct rw_semaphore *
70bdc6e0 ML	58	__rwsem_do_wake(struct rw_semaphore *sem, int wake_type)
1da177e4 LT	59	{
	60	struct rwsem_waiter *waiter;
	61	struct task_struct *tsk;
	62	struct list_head *next;
fd41b334	63	signed long oldcount, woken, loop, adjustment;
1da177e4	64
345af7bf ML	65	waiter = list_entry(sem->wait_list.next, struct rwsem_waiter, list);
	66	if (!(waiter->flags & RWSEM_WAITING_FOR_WRITE))
	67	goto readers_only;
	68
70bdc6e0	69	if (wake_type == RWSEM_WAKE_READ_OWNED)
345af7bf	70	goto out;
1da177e4	71
345af7bf ML	72	/* There's a writer at the front of the queue - try to grant it the
	73	* write lock. However, we only wake this writer if we can transition
	74	* the active part of the count from 0 -> 1
1da177e4	75	*/
fd41b334 ML	76	adjustment = RWSEM_ACTIVE_WRITE_BIAS;
	77	if (waiter->list.next == &sem->wait_list)
	78	adjustment -= RWSEM_WAITING_BIAS;
	79
345af7bf	80	try_again_write:
fd41b334	81	oldcount = rwsem_atomic_update(adjustment, sem) - adjustment;
1da177e4	82	if (oldcount & RWSEM_ACTIVE_MASK)
345af7bf ML	83	/* Someone grabbed the sem already */
345af7bf ML	84	goto undo_write;
1da177e4 LT	85
	86	/* We must be careful not to touch 'waiter' after we set ->task = NULL.
	87	* It is an allocated on the waiter's stack and may become invalid at
	88	* any time after that point (due to a wakeup from another source).
	89	*/
	90	list_del(&waiter->list);
	91	tsk = waiter->task;
d59dd462	92	smp_mb();
1da177e4 LT	93	waiter->task = NULL;
	94	wake_up_process(tsk);
	95	put_task_struct(tsk);
	96	goto out;
	97
345af7bf	98	readers_only:
70bdc6e0 ML	99	/* If we come here from up_xxxx(), another thread might have reached
	100	* rwsem_down_failed_common() before we acquired the spinlock and
	101	* woken up a waiter, making it now active. We prefer to check for
	102	* this first in order to not spend too much time with the spinlock
	103	* held if we're not going to be able to wake up readers in the end.
	104	*
	105	* Note that we do not need to update the rwsem count: any writer
	106	* trying to acquire rwsem will run rwsem_down_write_failed() due
	107	* to the waiting threads and block trying to acquire the spinlock.
	108	*
	109	* We use a dummy atomic update in order to acquire the cache line
	110	* exclusively since we expect to succeed and run the final rwsem
	111	* count adjustment pretty soon.
	112	*/
	113	if (wake_type == RWSEM_WAKE_ANY &&
	114	(rwsem_atomic_update(0, sem) & RWSEM_ACTIVE_MASK))
345af7bf	115	/* Someone grabbed the sem already */
70bdc6e0	116	goto out;
1da177e4	117
345af7bf ML	118	/* Grant an infinite number of read locks to the readers at the front
	119	* of the queue. Note we increment the 'active part' of the count by
	120	* the number of readers before waking any processes up.
1da177e4	121	*/
1da177e4 LT	122	woken = 0;
	123	do {
	124	woken++;
	125
	126	if (waiter->list.next == &sem->wait_list)
	127	break;
	128
	129	waiter = list_entry(waiter->list.next,
	130	struct rwsem_waiter, list);
	131
	132	} while (waiter->flags & RWSEM_WAITING_FOR_READ);
	133
fd41b334 ML	134	adjustment = woken * RWSEM_ACTIVE_READ_BIAS;
	135	if (waiter->flags & RWSEM_WAITING_FOR_READ)
	136	/* hit end of list above */
	137	adjustment -= RWSEM_WAITING_BIAS;
1da177e4	138
fd41b334	139	rwsem_atomic_add(adjustment, sem);
1da177e4 LT	140
1da177e4 LT	141	next = sem->wait_list.next;
fd41b334	142	for (loop = woken; loop > 0; loop--) {
1da177e4 LT	143	waiter = list_entry(next, struct rwsem_waiter, list);
	144	next = waiter->list.next;
	145	tsk = waiter->task;
d59dd462	146	smp_mb();
1da177e4 LT	147	waiter->task = NULL;
	148	wake_up_process(tsk);
	149	put_task_struct(tsk);
	150	}
	151
	152	sem->wait_list.next = next;
	153	next->prev = &sem->wait_list;
	154
	155	out:
1da177e4 LT	156	return sem;
1da177e4 LT	157
91af7081 ML	158	/* undo the change to the active count, but check for a transition
91af7081 ML	159	* 1->0 */
345af7bf	160	undo_write:
fd41b334	161	if (rwsem_atomic_update(-adjustment, sem) & RWSEM_ACTIVE_MASK)
345af7bf ML	162	goto out;
345af7bf ML	163	goto try_again_write;
1da177e4 LT	164	}
	165
	166	/*
	167	* wait for a lock to be granted
	168	*/
c7af77b5	169	static struct rw_semaphore __sched *
1da177e4 LT	170	rwsem_down_failed_common(struct rw_semaphore *sem,
	171	struct rwsem_waiter *waiter, signed long adjustment)
	172	{
	173	struct task_struct *tsk = current;
	174	signed long count;
	175
	176	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
	177
	178	/* set up my own style of waitqueue */
	179	spin_lock_irq(&sem->wait_lock);
	180	waiter->task = tsk;
	181	get_task_struct(tsk);
	182
fd41b334 ML	183	if (list_empty(&sem->wait_list))
fd41b334 ML	184	adjustment += RWSEM_WAITING_BIAS;
1da177e4 LT	185	list_add_tail(&waiter->list, &sem->wait_list);
1da177e4 LT	186
70bdc6e0	187	/* we're now waiting on the lock, but no longer actively locking */
1da177e4 LT	188	count = rwsem_atomic_update(adjustment, sem);
	189
	190	/* if there are no active locks, wake the front queued process(es) up */
	191	if (!(count & RWSEM_ACTIVE_MASK))
70bdc6e0	192	sem = __rwsem_do_wake(sem, RWSEM_WAKE_NO_ACTIVE);
1da177e4 LT	193
	194	spin_unlock_irq(&sem->wait_lock);
	195
	196	/* wait to be given the lock */
	197	for (;;) {
	198	if (!waiter->task)
	199	break;
	200	schedule();
	201	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
	202	}
	203
	204	tsk->state = TASK_RUNNING;
	205
	206	return sem;
	207	}
	208
	209	/*
	210	* wait for the read lock to be granted
	211	*/
d50efc6c	212	asmregparm struct rw_semaphore __sched *
1da177e4 LT	213	rwsem_down_read_failed(struct rw_semaphore *sem)
	214	{
	215	struct rwsem_waiter waiter;
	216
1da177e4	217	waiter.flags = RWSEM_WAITING_FOR_READ;
fd41b334	218	rwsem_down_failed_common(sem, &waiter, -RWSEM_ACTIVE_READ_BIAS);
1da177e4 LT	219	return sem;
	220	}
	221
	222	/*
	223	* wait for the write lock to be granted
	224	*/
d50efc6c	225	asmregparm struct rw_semaphore __sched *
1da177e4 LT	226	rwsem_down_write_failed(struct rw_semaphore *sem)
	227	{
	228	struct rwsem_waiter waiter;
	229
1da177e4	230	waiter.flags = RWSEM_WAITING_FOR_WRITE;
fd41b334	231	rwsem_down_failed_common(sem, &waiter, -RWSEM_ACTIVE_WRITE_BIAS);
1da177e4	232
1da177e4 LT	233	return sem;
	234	}
	235
	236	/*
	237	* handle waking up a waiter on the semaphore
	238	* - up_read/up_write has decremented the active part of count if we come here
	239	*/
d50efc6c	240	asmregparm struct rw_semaphore rwsem_wake(struct rw_semaphore sem)
1da177e4 LT	241	{
	242	unsigned long flags;
	243
1da177e4 LT	244	spin_lock_irqsave(&sem->wait_lock, flags);
	245
	246	/* do nothing if list empty */
	247	if (!list_empty(&sem->wait_list))
70bdc6e0	248	sem = __rwsem_do_wake(sem, RWSEM_WAKE_ANY);
1da177e4 LT	249
	250	spin_unlock_irqrestore(&sem->wait_lock, flags);
	251
1da177e4 LT	252	return sem;
	253	}
	254
	255	/*
	256	* downgrade a write lock into a read lock
	257	* - caller incremented waiting part of count and discovered it still negative
	258	* - just wake up any readers at the front of the queue
	259	*/
d50efc6c	260	asmregparm struct rw_semaphore rwsem_downgrade_wake(struct rw_semaphore sem)
1da177e4 LT	261	{
	262	unsigned long flags;
	263
1da177e4 LT	264	spin_lock_irqsave(&sem->wait_lock, flags);
	265
	266	/* do nothing if list empty */
	267	if (!list_empty(&sem->wait_list))
70bdc6e0	268	sem = __rwsem_do_wake(sem, RWSEM_WAKE_READ_OWNED);
1da177e4 LT	269
	270	spin_unlock_irqrestore(&sem->wait_lock, flags);
	271
1da177e4 LT	272	return sem;
	273	}
	274
	275	EXPORT_SYMBOL(rwsem_down_read_failed);
	276	EXPORT_SYMBOL(rwsem_down_write_failed);
	277	EXPORT_SYMBOL(rwsem_wake);
	278	EXPORT_SYMBOL(rwsem_downgrade_wake);