[net-next-2.6.git] / mm / mempool.c

/*
 *  linux/mm/mempool.c
 *
 *  memory buffer pool support. Such pools are mostly used
 *  for guaranteed, deadlock-free memory allocations during
 *  extreme VM load.
 *
 *  started by Ingo Molnar, Copyright (C) 2001
 */

#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/module.h>
#include <linux/mempool.h>
#include <linux/blkdev.h>
#include <linux/writeback.h>

static void add_element(mempool_t *pool, void *element)
{
	BUG_ON(pool->curr_nr >= pool->min_nr);
	pool->elements[pool->curr_nr++] = element;
}

static void *remove_element(mempool_t *pool)
{
	BUG_ON(pool->curr_nr <= 0);
	return pool->elements[--pool->curr_nr];
}

static void free_pool(mempool_t *pool)
{
	while (pool->curr_nr) {
		void *element = remove_element(pool);
		pool->free(element, pool->pool_data);
	}
	kfree(pool->elements);
	kfree(pool);
}

/**
 * mempool_create - create a memory pool
 * @min_nr:    the minimum number of elements guaranteed to be
 *             allocated for this pool.
 * @alloc_fn:  user-defined element-allocation function.
 * @free_fn:   user-defined element-freeing function.
 * @pool_data: optional private data available to the user-defined functions.
 *
 * this function creates and allocates a guaranteed size, preallocated
 * memory pool. The pool can be used from the mempool_alloc() and mempool_free()
 * functions. This function might sleep. Both the alloc_fn() and the free_fn()
 * functions might sleep - as long as the mempool_alloc() function is not called
 * from IRQ contexts.
 */
mempool_t *mempool_create(int min_nr, mempool_alloc_t *alloc_fn,
				mempool_free_t *free_fn, void *pool_data)
{
	return  mempool_create_node(min_nr,alloc_fn,free_fn, pool_data,-1);
}
EXPORT_SYMBOL(mempool_create);

mempool_t *mempool_create_node(int min_nr, mempool_alloc_t *alloc_fn,
			mempool_free_t *free_fn, void *pool_data, int node_id)
{
	mempool_t *pool;
	pool = kmalloc_node(sizeof(*pool), GFP_KERNEL | __GFP_ZERO, node_id);
	if (!pool)
		return NULL;
	pool->elements = kmalloc_node(min_nr * sizeof(void *),
					GFP_KERNEL, node_id);
	if (!pool->elements) {
		kfree(pool);
		return NULL;
	}
	spin_lock_init(&pool->lock);
	pool->min_nr = min_nr;
	pool->pool_data = pool_data;
	init_waitqueue_head(&pool->wait);
	pool->alloc = alloc_fn;
	pool->free = free_fn;

	/*
	 * First pre-allocate the guaranteed number of buffers.
	 */
	while (pool->curr_nr < pool->min_nr) {
		void *element;

		element = pool->alloc(GFP_KERNEL, pool->pool_data);
		if (unlikely(!element)) {
			free_pool(pool);
			return NULL;
		}
		add_element(pool, element);
	}
	return pool;
}
EXPORT_SYMBOL(mempool_create_node);

/**
 * mempool_resize - resize an existing memory pool
 * @pool:       pointer to the memory pool which was allocated via
 *              mempool_create().
 * @new_min_nr: the new minimum number of elements guaranteed to be
 *              allocated for this pool.
 * @gfp_mask:   the usual allocation bitmask.
 *
 * This function shrinks/grows the pool. In the case of growing,
 * it cannot be guaranteed that the pool will be grown to the new
 * size immediately, but new mempool_free() calls will refill it.
 *
 * Note, the caller must guarantee that no mempool_destroy is called
 * while this function is running. mempool_alloc() & mempool_free()
 * might be called (eg. from IRQ contexts) while this function executes.
 */
int mempool_resize(mempool_t *pool, int new_min_nr, gfp_t gfp_mask)
{
	void *element;
	void **new_elements;
	unsigned long flags;

	BUG_ON(new_min_nr <= 0);

	spin_lock_irqsave(&pool->lock, flags);
	if (new_min_nr <= pool->min_nr) {
		while (new_min_nr < pool->curr_nr) {
			element = remove_element(pool);
			spin_unlock_irqrestore(&pool->lock, flags);
			pool->free(element, pool->pool_data);
			spin_lock_irqsave(&pool->lock, flags);
		}
		pool->min_nr = new_min_nr;
		goto out_unlock;
	}
	spin_unlock_irqrestore(&pool->lock, flags);

	/* Grow the pool */
	new_elements = kmalloc(new_min_nr * sizeof(*new_elements), gfp_mask);
	if (!new_elements)
		return -ENOMEM;

	spin_lock_irqsave(&pool->lock, flags);
	if (unlikely(new_min_nr <= pool->min_nr)) {
		/* Raced, other resize will do our work */
		spin_unlock_irqrestore(&pool->lock, flags);
		kfree(new_elements);
		goto out;
	}
	memcpy(new_elements, pool->elements,
			pool->curr_nr * sizeof(*new_elements));
	kfree(pool->elements);
	pool->elements = new_elements;
	pool->min_nr = new_min_nr;

	while (pool->curr_nr < pool->min_nr) {
		spin_unlock_irqrestore(&pool->lock, flags);
		element = pool->alloc(gfp_mask, pool->pool_data);
		if (!element)
			goto out;
		spin_lock_irqsave(&pool->lock, flags);
		if (pool->curr_nr < pool->min_nr) {
			add_element(pool, element);
		} else {
			spin_unlock_irqrestore(&pool->lock, flags);
			pool->free(element, pool->pool_data);	/* Raced */
			goto out;
		}
	}
out_unlock:
	spin_unlock_irqrestore(&pool->lock, flags);
out:
	return 0;
}
EXPORT_SYMBOL(mempool_resize);

/**
 * mempool_destroy - deallocate a memory pool
 * @pool:      pointer to the memory pool which was allocated via
 *             mempool_create().
 *
 * this function only sleeps if the free_fn() function sleeps. The caller
 * has to guarantee that all elements have been returned to the pool (ie:
 * freed) prior to calling mempool_destroy().
 */
void mempool_destroy(mempool_t *pool)
{
	/* Check for outstanding elements */
	BUG_ON(pool->curr_nr != pool->min_nr);
	free_pool(pool);
}
EXPORT_SYMBOL(mempool_destroy);

/**
 * mempool_alloc - allocate an element from a specific memory pool
 * @pool:      pointer to the memory pool which was allocated via
 *             mempool_create().
 * @gfp_mask:  the usual allocation bitmask.
 *
 * this function only sleeps if the alloc_fn() function sleeps or
 * returns NULL. Note that due to preallocation, this function
 * *never* fails when called from process contexts. (it might
 * fail if called from an IRQ context.)
 */
void * mempool_alloc(mempool_t *pool, gfp_t gfp_mask)
{
	void *element;
	unsigned long flags;
	wait_queue_t wait;
	gfp_t gfp_temp;

	might_sleep_if(gfp_mask & __GFP_WAIT);

	gfp_mask |= __GFP_NOMEMALLOC;	/* don't allocate emergency reserves */
	gfp_mask |= __GFP_NORETRY;	/* don't loop in __alloc_pages */
	gfp_mask |= __GFP_NOWARN;	/* failures are OK */

	gfp_temp = gfp_mask & ~(__GFP_WAIT|__GFP_IO);

repeat_alloc:

	element = pool->alloc(gfp_temp, pool->pool_data);
	if (likely(element != NULL))
		return element;

	spin_lock_irqsave(&pool->lock, flags);
	if (likely(pool->curr_nr)) {
		element = remove_element(pool);
		spin_unlock_irqrestore(&pool->lock, flags);
		return element;
	}
	spin_unlock_irqrestore(&pool->lock, flags);

	/* We must not sleep in the GFP_ATOMIC case */
	if (!(gfp_mask & __GFP_WAIT))
		return NULL;

	/* Now start performing page reclaim */
	gfp_temp = gfp_mask;
	init_wait(&wait);
	prepare_to_wait(&pool->wait, &wait, TASK_UNINTERRUPTIBLE);
	smp_mb();
	if (!pool->curr_nr) {
		/*
		 * FIXME: this should be io_schedule().  The timeout is there
		 * as a workaround for some DM problems in 2.6.18.
		 */
		io_schedule_timeout(5*HZ);
	}
	finish_wait(&pool->wait, &wait);

	goto repeat_alloc;
}
EXPORT_SYMBOL(mempool_alloc);

/**
 * mempool_free - return an element to the pool.
 * @element:   pool element pointer.
 * @pool:      pointer to the memory pool which was allocated via
 *             mempool_create().
 *
 * this function only sleeps if the free_fn() function sleeps.
 */
void mempool_free(void *element, mempool_t *pool)
{
	unsigned long flags;

	if (unlikely(element == NULL))
		return;

	smp_mb();
	if (pool->curr_nr < pool->min_nr) {
		spin_lock_irqsave(&pool->lock, flags);
		if (pool->curr_nr < pool->min_nr) {
			add_element(pool, element);
			spin_unlock_irqrestore(&pool->lock, flags);
			wake_up(&pool->wait);
			return;
		}
		spin_unlock_irqrestore(&pool->lock, flags);
	}
	pool->free(element, pool->pool_data);
}
EXPORT_SYMBOL(mempool_free);

/*
 * A commonly used alloc and free fn.
 */
void *mempool_alloc_slab(gfp_t gfp_mask, void *pool_data)
{
	struct kmem_cache *mem = pool_data;
	return kmem_cache_alloc(mem, gfp_mask);
}
EXPORT_SYMBOL(mempool_alloc_slab);

void mempool_free_slab(void *element, void *pool_data)
{
	struct kmem_cache *mem = pool_data;
	kmem_cache_free(mem, element);
}
EXPORT_SYMBOL(mempool_free_slab);

/*
 * A commonly used alloc and free fn that kmalloc/kfrees the amount of memory
 * specified by pool_data
 */
void *mempool_kmalloc(gfp_t gfp_mask, void *pool_data)
{
	size_t size = (size_t)pool_data;
	return kmalloc(size, gfp_mask);
}
EXPORT_SYMBOL(mempool_kmalloc);

void mempool_kfree(void *element, void *pool_data)
{
	kfree(element);
}
EXPORT_SYMBOL(mempool_kfree);

/*
 * A simple mempool-backed page allocator that allocates pages
 * of the order specified by pool_data.
 */
void *mempool_alloc_pages(gfp_t gfp_mask, void *pool_data)
{
	int order = (int)(long)pool_data;
	return alloc_pages(gfp_mask, order);
}
EXPORT_SYMBOL(mempool_alloc_pages);

void mempool_free_pages(void *element, void *pool_data)
{
	int order = (int)(long)pool_data;
	__free_pages(element, order);
}
EXPORT_SYMBOL(mempool_free_pages);
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/mm/mempool.c
	3	*
	4	* memory buffer pool support. Such pools are mostly used
	5	* for guaranteed, deadlock-free memory allocations during
	6	* extreme VM load.
	7	*
	8	* started by Ingo Molnar, Copyright (C) 2001
	9	*/
	10
	11	#include <linux/mm.h>
	12	#include <linux/slab.h>
	13	#include <linux/module.h>
	14	#include <linux/mempool.h>
	15	#include <linux/blkdev.h>
	16	#include <linux/writeback.h>
	17
	18	static void add_element(mempool_t pool, void element)
	19	{
	20	BUG_ON(pool->curr_nr >= pool->min_nr);
	21	pool->elements[pool->curr_nr++] = element;
	22	}
	23
	24	static void remove_element(mempool_t pool)
	25	{
	26	BUG_ON(pool->curr_nr <= 0);
	27	return pool->elements[--pool->curr_nr];
	28	}
	29
	30	static void free_pool(mempool_t *pool)
	31	{
	32	while (pool->curr_nr) {
	33	void *element = remove_element(pool);
	34	pool->free(element, pool->pool_data);
	35	}
	36	kfree(pool->elements);
	37	kfree(pool);
	38	}
	39
	40	/**
	41	* mempool_create - create a memory pool
	42	* @min_nr: the minimum number of elements guaranteed to be
	43	* allocated for this pool.
	44	* @alloc_fn: user-defined element-allocation function.
	45	* @free_fn: user-defined element-freeing function.
	46	* @pool_data: optional private data available to the user-defined functions.
	47	*
	48	* this function creates and allocates a guaranteed size, preallocated
72fd4a35	49	* memory pool. The pool can be used from the mempool_alloc() and mempool_free()
1da177e4	50	* functions. This function might sleep. Both the alloc_fn() and the free_fn()
72fd4a35	51	* functions might sleep - as long as the mempool_alloc() function is not called
1da177e4 LT	52	* from IRQ contexts.
1da177e4 LT	53	*/
1946089a	54	mempool_t mempool_create(int min_nr, mempool_alloc_t alloc_fn,
1da177e4 LT	55	mempool_free_t free_fn, void pool_data)
1da177e4 LT	56	{
1946089a CL	57	return mempool_create_node(min_nr,alloc_fn,free_fn, pool_data,-1);
	58	}
	59	EXPORT_SYMBOL(mempool_create);
1da177e4	60
1946089a CL	61	mempool_t mempool_create_node(int min_nr, mempool_alloc_t alloc_fn,
	62	mempool_free_t free_fn, void pool_data, int node_id)
	63	{
	64	mempool_t *pool;
94f6030c	65	pool = kmalloc_node(sizeof(*pool), GFP_KERNEL \| __GFP_ZERO, node_id);
1da177e4 LT	66	if (!pool)
1da177e4 LT	67	return NULL;
1946089a CL	68	pool->elements = kmalloc_node(min_nr * sizeof(void *),
1946089a CL	69	GFP_KERNEL, node_id);
1da177e4 LT	70	if (!pool->elements) {
	71	kfree(pool);
	72	return NULL;
	73	}
	74	spin_lock_init(&pool->lock);
	75	pool->min_nr = min_nr;
	76	pool->pool_data = pool_data;
	77	init_waitqueue_head(&pool->wait);
	78	pool->alloc = alloc_fn;
	79	pool->free = free_fn;
	80
	81	/*
	82	* First pre-allocate the guaranteed number of buffers.
	83	*/
	84	while (pool->curr_nr < pool->min_nr) {
	85	void *element;
	86
	87	element = pool->alloc(GFP_KERNEL, pool->pool_data);
	88	if (unlikely(!element)) {
	89	free_pool(pool);
	90	return NULL;
	91	}
	92	add_element(pool, element);
	93	}
	94	return pool;
	95	}
1946089a	96	EXPORT_SYMBOL(mempool_create_node);
1da177e4 LT	97
	98	/**
	99	* mempool_resize - resize an existing memory pool
	100	* @pool: pointer to the memory pool which was allocated via
	101	* mempool_create().
	102	* @new_min_nr: the new minimum number of elements guaranteed to be
	103	* allocated for this pool.
	104	* @gfp_mask: the usual allocation bitmask.
	105	*
	106	* This function shrinks/grows the pool. In the case of growing,
	107	* it cannot be guaranteed that the pool will be grown to the new
	108	* size immediately, but new mempool_free() calls will refill it.
	109	*
	110	* Note, the caller must guarantee that no mempool_destroy is called
	111	* while this function is running. mempool_alloc() & mempool_free()
	112	* might be called (eg. from IRQ contexts) while this function executes.
	113	*/
dd0fc66f	114	int mempool_resize(mempool_t *pool, int new_min_nr, gfp_t gfp_mask)
1da177e4 LT	115	{
	116	void *element;
	117	void **new_elements;
	118	unsigned long flags;
	119
	120	BUG_ON(new_min_nr <= 0);
	121
	122	spin_lock_irqsave(&pool->lock, flags);
	123	if (new_min_nr <= pool->min_nr) {
	124	while (new_min_nr < pool->curr_nr) {
	125	element = remove_element(pool);
	126	spin_unlock_irqrestore(&pool->lock, flags);
	127	pool->free(element, pool->pool_data);
	128	spin_lock_irqsave(&pool->lock, flags);
	129	}
	130	pool->min_nr = new_min_nr;
	131	goto out_unlock;
	132	}
	133	spin_unlock_irqrestore(&pool->lock, flags);
	134
	135	/* Grow the pool */
	136	new_elements = kmalloc(new_min_nr * sizeof(*new_elements), gfp_mask);
	137	if (!new_elements)
	138	return -ENOMEM;
	139
	140	spin_lock_irqsave(&pool->lock, flags);
	141	if (unlikely(new_min_nr <= pool->min_nr)) {
	142	/* Raced, other resize will do our work */
	143	spin_unlock_irqrestore(&pool->lock, flags);
	144	kfree(new_elements);
	145	goto out;
	146	}
	147	memcpy(new_elements, pool->elements,
	148	pool->curr_nr * sizeof(*new_elements));
	149	kfree(pool->elements);
	150	pool->elements = new_elements;
	151	pool->min_nr = new_min_nr;
	152
	153	while (pool->curr_nr < pool->min_nr) {
	154	spin_unlock_irqrestore(&pool->lock, flags);
	155	element = pool->alloc(gfp_mask, pool->pool_data);
	156	if (!element)
	157	goto out;
	158	spin_lock_irqsave(&pool->lock, flags);
	159	if (pool->curr_nr < pool->min_nr) {
	160	add_element(pool, element);
	161	} else {
	162	spin_unlock_irqrestore(&pool->lock, flags);
	163	pool->free(element, pool->pool_data); /* Raced */
	164	goto out;
	165	}
	166	}
	167	out_unlock:
	168	spin_unlock_irqrestore(&pool->lock, flags);
	169	out:
	170	return 0;
	171	}
	172	EXPORT_SYMBOL(mempool_resize);
	173
	174	/**
	175	* mempool_destroy - deallocate a memory pool
	176	* @pool: pointer to the memory pool which was allocated via
	177	* mempool_create().
	178	*
179	* this function only sleeps if the free_fn() function sleeps. The caller
180	* has to guarantee that all elements have been returned to the pool (ie:
181	* freed) prior to calling mempool_destroy().
182	*/
183	void mempool_destroy(mempool_t *pool)
184	{
f02e1faf ES	185	/* Check for outstanding elements */
f02e1faf ES	186	BUG_ON(pool->curr_nr != pool->min_nr);
1da177e4 LT	187	free_pool(pool);
	188	}
	189	EXPORT_SYMBOL(mempool_destroy);
	190
	191	/**
	192	* mempool_alloc - allocate an element from a specific memory pool
	193	* @pool: pointer to the memory pool which was allocated via
	194	* mempool_create().
	195	* @gfp_mask: the usual allocation bitmask.
	196	*
72fd4a35	197	* this function only sleeps if the alloc_fn() function sleeps or
1da177e4 LT	198	* returns NULL. Note that due to preallocation, this function
	199	* never fails when called from process contexts. (it might
	200	* fail if called from an IRQ context.)
	201	*/
dd0fc66f	202	void * mempool_alloc(mempool_t *pool, gfp_t gfp_mask)
1da177e4 LT	203	{
	204	void *element;
	205	unsigned long flags;
01890a4c	206	wait_queue_t wait;
6daa0e28	207	gfp_t gfp_temp;
20a77776 NP	208
20a77776 NP	209	might_sleep_if(gfp_mask & __GFP_WAIT);
b84a35be NP	210
	211	gfp_mask \|= __GFP_NOMEMALLOC; /* don't allocate emergency reserves */
	212	gfp_mask \|= __GFP_NORETRY; /* don't loop in __alloc_pages */
	213	gfp_mask \|= __GFP_NOWARN; /* failures are OK */
1da177e4	214
20a77776 NP	215	gfp_temp = gfp_mask & ~(__GFP_WAIT\|__GFP_IO);
20a77776 NP	216
1da177e4	217	repeat_alloc:
20a77776 NP	218
20a77776 NP	219	element = pool->alloc(gfp_temp, pool->pool_data);
1da177e4 LT	220	if (likely(element != NULL))
	221	return element;
	222
1da177e4 LT	223	spin_lock_irqsave(&pool->lock, flags);
	224	if (likely(pool->curr_nr)) {
	225	element = remove_element(pool);
	226	spin_unlock_irqrestore(&pool->lock, flags);
	227	return element;
	228	}
	229	spin_unlock_irqrestore(&pool->lock, flags);
	230
	231	/* We must not sleep in the GFP_ATOMIC case */
	232	if (!(gfp_mask & __GFP_WAIT))
	233	return NULL;
	234
20a77776 NP	235	/* Now start performing page reclaim */
20a77776 NP	236	gfp_temp = gfp_mask;
01890a4c	237	init_wait(&wait);
1da177e4	238	prepare_to_wait(&pool->wait, &wait, TASK_UNINTERRUPTIBLE);
d59dd462	239	smp_mb();
0b1d647a PM	240	if (!pool->curr_nr) {
	241	/*
	242	* FIXME: this should be io_schedule(). The timeout is there
	243	* as a workaround for some DM problems in 2.6.18.
	244	*/
	245	io_schedule_timeout(5*HZ);
	246	}
1da177e4 LT	247	finish_wait(&pool->wait, &wait);
	248
	249	goto repeat_alloc;
	250	}
	251	EXPORT_SYMBOL(mempool_alloc);
	252
	253	/**
	254	* mempool_free - return an element to the pool.
	255	* @element: pool element pointer.
	256	* @pool: pointer to the memory pool which was allocated via
	257	* mempool_create().
	258	*
	259	* this function only sleeps if the free_fn() function sleeps.
	260	*/
	261	void mempool_free(void element, mempool_t pool)
	262	{
	263	unsigned long flags;
	264
c80e7a82 RR	265	if (unlikely(element == NULL))
	266	return;
	267
d59dd462	268	smp_mb();
1da177e4 LT	269	if (pool->curr_nr < pool->min_nr) {
	270	spin_lock_irqsave(&pool->lock, flags);
	271	if (pool->curr_nr < pool->min_nr) {
	272	add_element(pool, element);
	273	spin_unlock_irqrestore(&pool->lock, flags);
	274	wake_up(&pool->wait);
	275	return;
	276	}
	277	spin_unlock_irqrestore(&pool->lock, flags);
	278	}
	279	pool->free(element, pool->pool_data);
	280	}
	281	EXPORT_SYMBOL(mempool_free);
	282
	283	/*
	284	* A commonly used alloc and free fn.
	285	*/
dd0fc66f	286	void mempool_alloc_slab(gfp_t gfp_mask, void pool_data)
1da177e4	287	{
fcc234f8	288	struct kmem_cache *mem = pool_data;
1da177e4 LT	289	return kmem_cache_alloc(mem, gfp_mask);
	290	}
	291	EXPORT_SYMBOL(mempool_alloc_slab);
	292
	293	void mempool_free_slab(void element, void pool_data)
	294	{
fcc234f8	295	struct kmem_cache *mem = pool_data;
1da177e4 LT	296	kmem_cache_free(mem, element);
	297	}
	298	EXPORT_SYMBOL(mempool_free_slab);
6e0678f3	299
53184082 MD	300	/*
53184082 MD	301	* A commonly used alloc and free fn that kmalloc/kfrees the amount of memory
183ff22b	302	* specified by pool_data
53184082 MD	303	*/
	304	void mempool_kmalloc(gfp_t gfp_mask, void pool_data)
	305	{
5e2f89b5	306	size_t size = (size_t)pool_data;
53184082 MD	307	return kmalloc(size, gfp_mask);
	308	}
	309	EXPORT_SYMBOL(mempool_kmalloc);
	310
	311	void mempool_kfree(void element, void pool_data)
	312	{
	313	kfree(element);
	314	}
	315	EXPORT_SYMBOL(mempool_kfree);
	316
6e0678f3 MD	317	/*
	318	* A simple mempool-backed page allocator that allocates pages
	319	* of the order specified by pool_data.
	320	*/
	321	void mempool_alloc_pages(gfp_t gfp_mask, void pool_data)
	322	{
	323	int order = (int)(long)pool_data;
	324	return alloc_pages(gfp_mask, order);
	325	}
	326	EXPORT_SYMBOL(mempool_alloc_pages);
	327
	328	void mempool_free_pages(void element, void pool_data)
	329	{
	330	int order = (int)(long)pool_data;
	331	__free_pages(element, order);
	332	}
	333	EXPORT_SYMBOL(mempool_free_pages);