[mirror_ubuntu-bionic-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;
	raw_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)
		/* Another active reader was observed, so wakeup is not
		 * likely to succeed. Save the atomic op.
		 */
		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_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->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,
			 unsigned int flags, signed long adjustment)
{
	struct rwsem_waiter waiter;
	struct task_struct *tsk = current;
	signed long count;

	set_task_state(tsk, TASK_UNINTERRUPTIBLE);

	/* set up my own style of waitqueue */
	raw_spin_lock_irq(&sem->wait_lock);
	waiter.task = tsk;
	waiter.flags = flags;
	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.
	 *
	 * Alternatively, if we're called from a failed down_write(), 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)
		sem = __rwsem_do_wake(sem, RWSEM_WAKE_NO_ACTIVE);
	else if (count > RWSEM_WAITING_BIAS &&
		 adjustment == -RWSEM_ACTIVE_WRITE_BIAS)
		sem = __rwsem_do_wake(sem, RWSEM_WAKE_READ_OWNED);

	raw_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
 */
struct rw_semaphore __sched *rwsem_down_read_failed(struct rw_semaphore *sem)
{
	return rwsem_down_failed_common(sem, RWSEM_WAITING_FOR_READ,
					-RWSEM_ACTIVE_READ_BIAS);
}

/*
 * wait for the write lock to be granted
 */
struct rw_semaphore __sched *rwsem_down_write_failed(struct rw_semaphore *sem)
{
	return rwsem_down_failed_common(sem, RWSEM_WAITING_FOR_WRITE,
					-RWSEM_ACTIVE_WRITE_BIAS);
}

/*
 * 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
	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
4ea2176d IM	24	sem->count = RWSEM_UNLOCKED_VALUE;
ddb6c9b5	25	raw_spin_lock_init(&sem->wait_lock);
4ea2176d IM	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)
424acaae ML	70	/* Another active reader was observed, so wakeup is not
	71	* likely to succeed. Save the atomic op.
	72	*/
345af7bf	73	goto out;
1da177e4	74
345af7bf ML	75	/* There's a writer at the front of the queue - try to grant it the
	76	* write lock. However, we only wake this writer if we can transition
	77	* the active part of the count from 0 -> 1
1da177e4	78	*/
fd41b334 ML	79	adjustment = RWSEM_ACTIVE_WRITE_BIAS;
	80	if (waiter->list.next == &sem->wait_list)
	81	adjustment -= RWSEM_WAITING_BIAS;
	82
345af7bf	83	try_again_write:
fd41b334	84	oldcount = rwsem_atomic_update(adjustment, sem) - adjustment;
1da177e4	85	if (oldcount & RWSEM_ACTIVE_MASK)
345af7bf ML	86	/* Someone grabbed the sem already */
345af7bf ML	87	goto undo_write;
1da177e4 LT	88
	89	/* We must be careful not to touch 'waiter' after we set ->task = NULL.
	90	* It is an allocated on the waiter's stack and may become invalid at
	91	* any time after that point (due to a wakeup from another source).
	92	*/
	93	list_del(&waiter->list);
	94	tsk = waiter->task;
d59dd462	95	smp_mb();
1da177e4 LT	96	waiter->task = NULL;
	97	wake_up_process(tsk);
	98	put_task_struct(tsk);
	99	goto out;
	100
345af7bf	101	readers_only:
70bdc6e0 ML	102	/* If we come here from up_xxxx(), another thread might have reached
	103	* rwsem_down_failed_common() before we acquired the spinlock and
	104	* woken up a waiter, making it now active. We prefer to check for
	105	* this first in order to not spend too much time with the spinlock
	106	* held if we're not going to be able to wake up readers in the end.
	107	*
	108	* Note that we do not need to update the rwsem count: any writer
	109	* trying to acquire rwsem will run rwsem_down_write_failed() due
	110	* to the waiting threads and block trying to acquire the spinlock.
	111	*
	112	* We use a dummy atomic update in order to acquire the cache line
	113	* exclusively since we expect to succeed and run the final rwsem
	114	* count adjustment pretty soon.
	115	*/
	116	if (wake_type == RWSEM_WAKE_ANY &&
424acaae ML	117	rwsem_atomic_update(0, sem) < RWSEM_WAITING_BIAS)
424acaae ML	118	/* Someone grabbed the sem for write already */
70bdc6e0	119	goto out;
1da177e4	120
345af7bf ML	121	/* Grant an infinite number of read locks to the readers at the front
	122	* of the queue. Note we increment the 'active part' of the count by
	123	* the number of readers before waking any processes up.
1da177e4	124	*/
1da177e4 LT	125	woken = 0;
	126	do {
	127	woken++;
	128
	129	if (waiter->list.next == &sem->wait_list)
	130	break;
	131
	132	waiter = list_entry(waiter->list.next,
	133	struct rwsem_waiter, list);
	134
	135	} while (waiter->flags & RWSEM_WAITING_FOR_READ);
	136
fd41b334 ML	137	adjustment = woken * RWSEM_ACTIVE_READ_BIAS;
	138	if (waiter->flags & RWSEM_WAITING_FOR_READ)
	139	/* hit end of list above */
	140	adjustment -= RWSEM_WAITING_BIAS;
1da177e4	141
fd41b334	142	rwsem_atomic_add(adjustment, sem);
1da177e4 LT	143
1da177e4 LT	144	next = sem->wait_list.next;
fd41b334	145	for (loop = woken; loop > 0; loop--) {
1da177e4 LT	146	waiter = list_entry(next, struct rwsem_waiter, list);
	147	next = waiter->list.next;
	148	tsk = waiter->task;
d59dd462	149	smp_mb();
1da177e4 LT	150	waiter->task = NULL;
	151	wake_up_process(tsk);
	152	put_task_struct(tsk);
	153	}
	154
	155	sem->wait_list.next = next;
	156	next->prev = &sem->wait_list;
	157
	158	out:
1da177e4 LT	159	return sem;
1da177e4 LT	160
91af7081 ML	161	/* undo the change to the active count, but check for a transition
91af7081 ML	162	* 1->0 */
345af7bf	163	undo_write:
fd41b334	164	if (rwsem_atomic_update(-adjustment, sem) & RWSEM_ACTIVE_MASK)
345af7bf ML	165	goto out;
345af7bf ML	166	goto try_again_write;
1da177e4 LT	167	}
	168
	169	/*
	170	* wait for a lock to be granted
	171	*/
c7af77b5	172	static struct rw_semaphore __sched *
1da177e4	173	rwsem_down_failed_common(struct rw_semaphore *sem,
a8618a0e	174	unsigned int flags, signed long adjustment)
1da177e4	175	{
a8618a0e	176	struct rwsem_waiter waiter;
1da177e4 LT	177	struct task_struct *tsk = current;
	178	signed long count;
	179
	180	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
	181
	182	/* set up my own style of waitqueue */
ddb6c9b5	183	raw_spin_lock_irq(&sem->wait_lock);
a8618a0e ML	184	waiter.task = tsk;
a8618a0e ML	185	waiter.flags = flags;
1da177e4 LT	186	get_task_struct(tsk);
1da177e4 LT	187
fd41b334 ML	188	if (list_empty(&sem->wait_list))
fd41b334 ML	189	adjustment += RWSEM_WAITING_BIAS;
a8618a0e	190	list_add_tail(&waiter.list, &sem->wait_list);
1da177e4	191
70bdc6e0	192	/* we're now waiting on the lock, but no longer actively locking */
1da177e4 LT	193	count = rwsem_atomic_update(adjustment, sem);
1da177e4 LT	194
424acaae ML	195	/* If there are no active locks, wake the front queued process(es) up.
	196	*
	197	* Alternatively, if we're called from a failed down_write(), there
	198	* were already threads queued before us and there are no active
	199	* writers, the lock must be read owned; so we try to wake any read
	200	* locks that were queued ahead of us. */
	201	if (count == RWSEM_WAITING_BIAS)
70bdc6e0	202	sem = __rwsem_do_wake(sem, RWSEM_WAKE_NO_ACTIVE);
424acaae ML	203	else if (count > RWSEM_WAITING_BIAS &&
	204	adjustment == -RWSEM_ACTIVE_WRITE_BIAS)
	205	sem = __rwsem_do_wake(sem, RWSEM_WAKE_READ_OWNED);
1da177e4	206
ddb6c9b5	207	raw_spin_unlock_irq(&sem->wait_lock);
1da177e4 LT	208
	209	/* wait to be given the lock */
	210	for (;;) {
a8618a0e	211	if (!waiter.task)
1da177e4 LT	212	break;
	213	schedule();
	214	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
	215	}
	216
	217	tsk->state = TASK_RUNNING;
	218
	219	return sem;
	220	}
	221
	222	/*
	223	* wait for the read lock to be granted
	224	*/
d1233754	225	struct rw_semaphore __sched rwsem_down_read_failed(struct rw_semaphore sem)
1da177e4	226	{
a8618a0e ML	227	return rwsem_down_failed_common(sem, RWSEM_WAITING_FOR_READ,
a8618a0e ML	228	-RWSEM_ACTIVE_READ_BIAS);
1da177e4 LT	229	}
	230
	231	/*
	232	* wait for the write lock to be granted
	233	*/
d1233754	234	struct rw_semaphore __sched rwsem_down_write_failed(struct rw_semaphore sem)
1da177e4	235	{
a8618a0e ML	236	return rwsem_down_failed_common(sem, RWSEM_WAITING_FOR_WRITE,
a8618a0e ML	237	-RWSEM_ACTIVE_WRITE_BIAS);
1da177e4 LT	238	}
	239
	240	/*
	241	* handle waking up a waiter on the semaphore
	242	* - up_read/up_write has decremented the active part of count if we come here
	243	*/
d1233754	244	struct rw_semaphore rwsem_wake(struct rw_semaphore sem)
1da177e4 LT	245	{
	246	unsigned long flags;
	247
ddb6c9b5	248	raw_spin_lock_irqsave(&sem->wait_lock, flags);
1da177e4 LT	249
	250	/* do nothing if list empty */
	251	if (!list_empty(&sem->wait_list))
70bdc6e0	252	sem = __rwsem_do_wake(sem, RWSEM_WAKE_ANY);
1da177e4	253
ddb6c9b5	254	raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
1da177e4	255
1da177e4 LT	256	return sem;
	257	}
	258
	259	/*
	260	* downgrade a write lock into a read lock
	261	* - caller incremented waiting part of count and discovered it still negative
	262	* - just wake up any readers at the front of the queue
	263	*/
d1233754	264	struct rw_semaphore rwsem_downgrade_wake(struct rw_semaphore sem)
1da177e4 LT	265	{
	266	unsigned long flags;
	267
ddb6c9b5	268	raw_spin_lock_irqsave(&sem->wait_lock, flags);
1da177e4 LT	269
	270	/* do nothing if list empty */
	271	if (!list_empty(&sem->wait_list))
70bdc6e0	272	sem = __rwsem_do_wake(sem, RWSEM_WAKE_READ_OWNED);
1da177e4	273
ddb6c9b5	274	raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
1da177e4	275
1da177e4 LT	276	return sem;
	277	}
	278
	279	EXPORT_SYMBOL(rwsem_down_read_failed);
	280	EXPORT_SYMBOL(rwsem_down_write_failed);
	281	EXPORT_SYMBOL(rwsem_wake);
	282	EXPORT_SYMBOL(rwsem_downgrade_wake);