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

/*
 *  linux/mm/swap.c
 *
 *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 */

/*
 * This file contains the default values for the operation of the
 * Linux VM subsystem. Fine-tuning documentation can be found in
 * Documentation/sysctl/vm.txt.
 * Started 18.12.91
 * Swap aging added 23.2.95, Stephen Tweedie.
 * Buffermem limits added 12.3.98, Rik van Riel.
 */

#include <linux/mm.h>
#include <linux/sched.h>
#include <linux/kernel_stat.h>
#include <linux/swap.h>
#include <linux/mman.h>
#include <linux/pagemap.h>
#include <linux/pagevec.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/mm_inline.h>
#include <linux/buffer_head.h>	/* for try_to_release_page() */
#include <linux/percpu_counter.h>
#include <linux/percpu.h>
#include <linux/cpu.h>
#include <linux/notifier.h>
#include <linux/backing-dev.h>

/* How many pages do we try to swap or page in/out together? */
int page_cluster;

static DEFINE_PER_CPU(struct pagevec, lru_add_pvecs) = { 0, };
static DEFINE_PER_CPU(struct pagevec, lru_add_active_pvecs) = { 0, };
static DEFINE_PER_CPU(struct pagevec, lru_rotate_pvecs) = { 0, };

/*
 * This path almost never happens for VM activity - pages are normally
 * freed via pagevecs.  But it gets used by networking.
 */
static void __page_cache_release(struct page *page)
{
	if (PageLRU(page)) {
		unsigned long flags;
		struct zone *zone = page_zone(page);

		spin_lock_irqsave(&zone->lru_lock, flags);
		VM_BUG_ON(!PageLRU(page));
		__ClearPageLRU(page);
		del_page_from_lru(zone, page);
		spin_unlock_irqrestore(&zone->lru_lock, flags);
	}
	free_hot_page(page);
}

static void put_compound_page(struct page *page)
{
	page = compound_head(page);
	if (put_page_testzero(page)) {
		compound_page_dtor *dtor;

		dtor = get_compound_page_dtor(page);
		(*dtor)(page);
	}
}

void put_page(struct page *page)
{
	if (unlikely(PageCompound(page)))
		put_compound_page(page);
	else if (put_page_testzero(page))
		__page_cache_release(page);
}
EXPORT_SYMBOL(put_page);

/**
 * put_pages_list(): release a list of pages
 *
 * Release a list of pages which are strung together on page.lru.  Currently
 * used by read_cache_pages() and related error recovery code.
 *
 * @pages: list of pages threaded on page->lru
 */
void put_pages_list(struct list_head *pages)
{
	while (!list_empty(pages)) {
		struct page *victim;

		victim = list_entry(pages->prev, struct page, lru);
		list_del(&victim->lru);
		page_cache_release(victim);
	}
}
EXPORT_SYMBOL(put_pages_list);

/*
 * pagevec_move_tail() must be called with IRQ disabled.
 * Otherwise this may cause nasty races.
 */
static void pagevec_move_tail(struct pagevec *pvec)
{
	int i;
	int pgmoved = 0;
	struct zone *zone = NULL;

	for (i = 0; i < pagevec_count(pvec); i++) {
		struct page *page = pvec->pages[i];
		struct zone *pagezone = page_zone(page);

		if (pagezone != zone) {
			if (zone)
				spin_unlock(&zone->lru_lock);
			zone = pagezone;
			spin_lock(&zone->lru_lock);
		}
		if (PageLRU(page) && !PageActive(page)) {
			list_move_tail(&page->lru, &zone->inactive_list);
			pgmoved++;
		}
	}
	if (zone)
		spin_unlock(&zone->lru_lock);
	__count_vm_events(PGROTATED, pgmoved);
	release_pages(pvec->pages, pvec->nr, pvec->cold);
	pagevec_reinit(pvec);
}

/*
 * Writeback is about to end against a page which has been marked for immediate
 * reclaim.  If it still appears to be reclaimable, move it to the tail of the
 * inactive list.
 *
 * Returns zero if it cleared PG_writeback.
 */
int rotate_reclaimable_page(struct page *page)
{
	struct pagevec *pvec;
	unsigned long flags;

	if (PageLocked(page))
		return 1;
	if (PageDirty(page))
		return 1;
	if (PageActive(page))
		return 1;
	if (!PageLRU(page))
		return 1;

	page_cache_get(page);
	local_irq_save(flags);
	pvec = &__get_cpu_var(lru_rotate_pvecs);
	if (!pagevec_add(pvec, page))
		pagevec_move_tail(pvec);
	local_irq_restore(flags);

	if (!test_clear_page_writeback(page))
		BUG();

	return 0;
}

/*
 * FIXME: speed this up?
 */
void activate_page(struct page *page)
{
	struct zone *zone = page_zone(page);

	spin_lock_irq(&zone->lru_lock);
	if (PageLRU(page) && !PageActive(page)) {
		del_page_from_inactive_list(zone, page);
		SetPageActive(page);
		add_page_to_active_list(zone, page);
		__count_vm_event(PGACTIVATE);
	}
	spin_unlock_irq(&zone->lru_lock);
}

/*
 * Mark a page as having seen activity.
 *
 * inactive,unreferenced	->	inactive,referenced
 * inactive,referenced		->	active,unreferenced
 * active,unreferenced		->	active,referenced
 */
void mark_page_accessed(struct page *page)
{
	if (!PageActive(page) && PageReferenced(page) && PageLRU(page)) {
		activate_page(page);
		ClearPageReferenced(page);
	} else if (!PageReferenced(page)) {
		SetPageReferenced(page);
	}
}

EXPORT_SYMBOL(mark_page_accessed);

/**
 * lru_cache_add: add a page to the page lists
 * @page: the page to add
 */
void lru_cache_add(struct page *page)
{
	struct pagevec *pvec = &get_cpu_var(lru_add_pvecs);

	page_cache_get(page);
	if (!pagevec_add(pvec, page))
		__pagevec_lru_add(pvec);
	put_cpu_var(lru_add_pvecs);
}

void lru_cache_add_active(struct page *page)
{
	struct pagevec *pvec = &get_cpu_var(lru_add_active_pvecs);

	page_cache_get(page);
	if (!pagevec_add(pvec, page))
		__pagevec_lru_add_active(pvec);
	put_cpu_var(lru_add_active_pvecs);
}

/*
 * Drain pages out of the cpu's pagevecs.
 * Either "cpu" is the current CPU, and preemption has already been
 * disabled; or "cpu" is being hot-unplugged, and is already dead.
 */
static void drain_cpu_pagevecs(int cpu)
{
	struct pagevec *pvec;

	pvec = &per_cpu(lru_add_pvecs, cpu);
	if (pagevec_count(pvec))
		__pagevec_lru_add(pvec);

	pvec = &per_cpu(lru_add_active_pvecs, cpu);
	if (pagevec_count(pvec))
		__pagevec_lru_add_active(pvec);

	pvec = &per_cpu(lru_rotate_pvecs, cpu);
	if (pagevec_count(pvec)) {
		unsigned long flags;

		/* No harm done if a racing interrupt already did this */
		local_irq_save(flags);
		pagevec_move_tail(pvec);
		local_irq_restore(flags);
	}
}

void lru_add_drain(void)
{
	drain_cpu_pagevecs(get_cpu());
	put_cpu();
}

#ifdef CONFIG_NUMA
static void lru_add_drain_per_cpu(struct work_struct *dummy)
{
	lru_add_drain();
}

/*
 * Returns 0 for success
 */
int lru_add_drain_all(void)
{
	return schedule_on_each_cpu(lru_add_drain_per_cpu);
}

#else

/*
 * Returns 0 for success
 */
int lru_add_drain_all(void)
{
	lru_add_drain();
	return 0;
}
#endif

/*
 * Batched page_cache_release().  Decrement the reference count on all the
 * passed pages.  If it fell to zero then remove the page from the LRU and
 * free it.
 *
 * Avoid taking zone->lru_lock if possible, but if it is taken, retain it
 * for the remainder of the operation.
 *
 * The locking in this function is against shrink_cache(): we recheck the
 * page count inside the lock to see whether shrink_cache grabbed the page
 * via the LRU.  If it did, give up: shrink_cache will free it.
 */
void release_pages(struct page **pages, int nr, int cold)
{
	int i;
	struct pagevec pages_to_free;
	struct zone *zone = NULL;
	unsigned long uninitialized_var(flags);

	pagevec_init(&pages_to_free, cold);
	for (i = 0; i < nr; i++) {
		struct page *page = pages[i];

		if (unlikely(PageCompound(page))) {
			if (zone) {
				spin_unlock_irqrestore(&zone->lru_lock, flags);
				zone = NULL;
			}
			put_compound_page(page);
			continue;
		}

		if (!put_page_testzero(page))
			continue;

		if (PageLRU(page)) {
			struct zone *pagezone = page_zone(page);
			if (pagezone != zone) {
				if (zone)
					spin_unlock_irqrestore(&zone->lru_lock,
									flags);
				zone = pagezone;
				spin_lock_irqsave(&zone->lru_lock, flags);
			}
			VM_BUG_ON(!PageLRU(page));
			__ClearPageLRU(page);
			del_page_from_lru(zone, page);
		}

		if (!pagevec_add(&pages_to_free, page)) {
			if (zone) {
				spin_unlock_irqrestore(&zone->lru_lock, flags);
				zone = NULL;
			}
			__pagevec_free(&pages_to_free);
			pagevec_reinit(&pages_to_free);
  		}
	}
	if (zone)
		spin_unlock_irqrestore(&zone->lru_lock, flags);

	pagevec_free(&pages_to_free);
}

/*
 * The pages which we're about to release may be in the deferred lru-addition
 * queues.  That would prevent them from really being freed right now.  That's
 * OK from a correctness point of view but is inefficient - those pages may be
 * cache-warm and we want to give them back to the page allocator ASAP.
 *
 * So __pagevec_release() will drain those queues here.  __pagevec_lru_add()
 * and __pagevec_lru_add_active() call release_pages() directly to avoid
 * mutual recursion.
 */
void __pagevec_release(struct pagevec *pvec)
{
	lru_add_drain();
	release_pages(pvec->pages, pagevec_count(pvec), pvec->cold);
	pagevec_reinit(pvec);
}

EXPORT_SYMBOL(__pagevec_release);

/*
 * pagevec_release() for pages which are known to not be on the LRU
 *
 * This function reinitialises the caller's pagevec.
 */
void __pagevec_release_nonlru(struct pagevec *pvec)
{
	int i;
	struct pagevec pages_to_free;

	pagevec_init(&pages_to_free, pvec->cold);
	for (i = 0; i < pagevec_count(pvec); i++) {
		struct page *page = pvec->pages[i];

		VM_BUG_ON(PageLRU(page));
		if (put_page_testzero(page))
			pagevec_add(&pages_to_free, page);
	}
	pagevec_free(&pages_to_free);
	pagevec_reinit(pvec);
}

/*
 * Add the passed pages to the LRU, then drop the caller's refcount
 * on them.  Reinitialises the caller's pagevec.
 */
void __pagevec_lru_add(struct pagevec *pvec)
{
	int i;
	struct zone *zone = NULL;

	for (i = 0; i < pagevec_count(pvec); i++) {
		struct page *page = pvec->pages[i];
		struct zone *pagezone = page_zone(page);

		if (pagezone != zone) {
			if (zone)
				spin_unlock_irq(&zone->lru_lock);
			zone = pagezone;
			spin_lock_irq(&zone->lru_lock);
		}
		VM_BUG_ON(PageLRU(page));
		SetPageLRU(page);
		add_page_to_inactive_list(zone, page);
	}
	if (zone)
		spin_unlock_irq(&zone->lru_lock);
	release_pages(pvec->pages, pvec->nr, pvec->cold);
	pagevec_reinit(pvec);
}

EXPORT_SYMBOL(__pagevec_lru_add);

void __pagevec_lru_add_active(struct pagevec *pvec)
{
	int i;
	struct zone *zone = NULL;

	for (i = 0; i < pagevec_count(pvec); i++) {
		struct page *page = pvec->pages[i];
		struct zone *pagezone = page_zone(page);

		if (pagezone != zone) {
			if (zone)
				spin_unlock_irq(&zone->lru_lock);
			zone = pagezone;
			spin_lock_irq(&zone->lru_lock);
		}
		VM_BUG_ON(PageLRU(page));
		SetPageLRU(page);
		VM_BUG_ON(PageActive(page));
		SetPageActive(page);
		add_page_to_active_list(zone, page);
	}
	if (zone)
		spin_unlock_irq(&zone->lru_lock);
	release_pages(pvec->pages, pvec->nr, pvec->cold);
	pagevec_reinit(pvec);
}

/*
 * Try to drop buffers from the pages in a pagevec
 */
void pagevec_strip(struct pagevec *pvec)
{
	int i;

	for (i = 0; i < pagevec_count(pvec); i++) {
		struct page *page = pvec->pages[i];

		if (PagePrivate(page) && !TestSetPageLocked(page)) {
			if (PagePrivate(page))
				try_to_release_page(page, 0);
			unlock_page(page);
		}
	}
}

/**
 * pagevec_lookup - gang pagecache lookup
 * @pvec:	Where the resulting pages are placed
 * @mapping:	The address_space to search
 * @start:	The starting page index
 * @nr_pages:	The maximum number of pages
 *
 * pagevec_lookup() will search for and return a group of up to @nr_pages pages
 * in the mapping.  The pages are placed in @pvec.  pagevec_lookup() takes a
 * reference against the pages in @pvec.
 *
 * The search returns a group of mapping-contiguous pages with ascending
 * indexes.  There may be holes in the indices due to not-present pages.
 *
 * pagevec_lookup() returns the number of pages which were found.
 */
unsigned pagevec_lookup(struct pagevec *pvec, struct address_space *mapping,
		pgoff_t start, unsigned nr_pages)
{
	pvec->nr = find_get_pages(mapping, start, nr_pages, pvec->pages);
	return pagevec_count(pvec);
}

EXPORT_SYMBOL(pagevec_lookup);

unsigned pagevec_lookup_tag(struct pagevec *pvec, struct address_space *mapping,
		pgoff_t *index, int tag, unsigned nr_pages)
{
	pvec->nr = find_get_pages_tag(mapping, index, tag,
					nr_pages, pvec->pages);
	return pagevec_count(pvec);
}

EXPORT_SYMBOL(pagevec_lookup_tag);

#ifdef CONFIG_SMP
/*
 * We tolerate a little inaccuracy to avoid ping-ponging the counter between
 * CPUs
 */
#define ACCT_THRESHOLD	max(16, NR_CPUS * 2)

static DEFINE_PER_CPU(long, committed_space) = 0;

void vm_acct_memory(long pages)
{
	long *local;

	preempt_disable();
	local = &__get_cpu_var(committed_space);
	*local += pages;
	if (*local > ACCT_THRESHOLD || *local < -ACCT_THRESHOLD) {
		atomic_add(*local, &vm_committed_space);
		*local = 0;
	}
	preempt_enable();
}

#ifdef CONFIG_HOTPLUG_CPU

/* Drop the CPU's cached committed space back into the central pool. */
static int cpu_swap_callback(struct notifier_block *nfb,
			     unsigned long action,
			     void *hcpu)
{
	long *committed;

	committed = &per_cpu(committed_space, (long)hcpu);
	if (action == CPU_DEAD || action == CPU_DEAD_FROZEN) {
		atomic_add(*committed, &vm_committed_space);
		*committed = 0;
		drain_cpu_pagevecs((long)hcpu);
	}
	return NOTIFY_OK;
}
#endif /* CONFIG_HOTPLUG_CPU */
#endif /* CONFIG_SMP */

/*
 * Perform any setup for the swap system
 */
void __init swap_setup(void)
{
	unsigned long megs = num_physpages >> (20 - PAGE_SHIFT);

#ifdef CONFIG_SWAP
	bdi_init(swapper_space.backing_dev_info);
#endif

	/* Use a smaller cluster for small-memory machines */
	if (megs < 16)
		page_cluster = 2;
	else
		page_cluster = 3;
	/*
	 * Right now other parts of the system means that we
	 * _really_ don't want to cluster much more
	 */
#ifdef CONFIG_HOTPLUG_CPU
	hotcpu_notifier(cpu_swap_callback, 0);
#endif
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/mm/swap.c
	3	*
	4	* Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
	5	*/
	6
	7	/*
183ff22b	8	* This file contains the default values for the operation of the
1da177e4 LT	9	* Linux VM subsystem. Fine-tuning documentation can be found in
	10	* Documentation/sysctl/vm.txt.
	11	* Started 18.12.91
	12	* Swap aging added 23.2.95, Stephen Tweedie.
	13	* Buffermem limits added 12.3.98, Rik van Riel.
	14	*/
	15
	16	#include <linux/mm.h>
	17	#include <linux/sched.h>
	18	#include <linux/kernel_stat.h>
	19	#include <linux/swap.h>
	20	#include <linux/mman.h>
	21	#include <linux/pagemap.h>
	22	#include <linux/pagevec.h>
	23	#include <linux/init.h>
	24	#include <linux/module.h>
	25	#include <linux/mm_inline.h>
	26	#include <linux/buffer_head.h> /* for try_to_release_page() */
1da177e4 LT	27	#include <linux/percpu_counter.h>
	28	#include <linux/percpu.h>
	29	#include <linux/cpu.h>
	30	#include <linux/notifier.h>
e0bf68dd	31	#include <linux/backing-dev.h>
1da177e4 LT	32
	33	/* How many pages do we try to swap or page in/out together? */
	34	int page_cluster;
	35
902aaed0 HH	36	static DEFINE_PER_CPU(struct pagevec, lru_add_pvecs) = { 0, };
	37	static DEFINE_PER_CPU(struct pagevec, lru_add_active_pvecs) = { 0, };
	38	static DEFINE_PER_CPU(struct pagevec, lru_rotate_pvecs) = { 0, };
	39
b221385b AB	40	/*
	41	* This path almost never happens for VM activity - pages are normally
	42	* freed via pagevecs. But it gets used by networking.
	43	*/
920c7a5d	44	static void __page_cache_release(struct page *page)
b221385b AB	45	{
	46	if (PageLRU(page)) {
	47	unsigned long flags;
	48	struct zone *zone = page_zone(page);
	49
	50	spin_lock_irqsave(&zone->lru_lock, flags);
	51	VM_BUG_ON(!PageLRU(page));
	52	__ClearPageLRU(page);
	53	del_page_from_lru(zone, page);
	54	spin_unlock_irqrestore(&zone->lru_lock, flags);
	55	}
	56	free_hot_page(page);
	57	}
	58
8519fb30	59	static void put_compound_page(struct page *page)
1da177e4	60	{
d85f3385	61	page = compound_head(page);
8519fb30	62	if (put_page_testzero(page)) {
33f2ef89	63	compound_page_dtor *dtor;
1da177e4	64
33f2ef89	65	dtor = get_compound_page_dtor(page);
8519fb30	66	(*dtor)(page);
1da177e4	67	}
8519fb30 NP	68	}
	69
	70	void put_page(struct page *page)
	71	{
	72	if (unlikely(PageCompound(page)))
	73	put_compound_page(page);
	74	else if (put_page_testzero(page))
1da177e4 LT	75	__page_cache_release(page);
	76	}
	77	EXPORT_SYMBOL(put_page);
1da177e4	78
1d7ea732 AZ	79	/**
	80	* put_pages_list(): release a list of pages
	81	*
	82	* Release a list of pages which are strung together on page.lru. Currently
	83	* used by read_cache_pages() and related error recovery code.
	84	*
	85	* @pages: list of pages threaded on page->lru
	86	*/
	87	void put_pages_list(struct list_head *pages)
	88	{
	89	while (!list_empty(pages)) {
	90	struct page *victim;
	91
	92	victim = list_entry(pages->prev, struct page, lru);
	93	list_del(&victim->lru);
	94	page_cache_release(victim);
	95	}
	96	}
	97	EXPORT_SYMBOL(put_pages_list);
	98
902aaed0 HH	99	/*
	100	* pagevec_move_tail() must be called with IRQ disabled.
	101	* Otherwise this may cause nasty races.
	102	*/
	103	static void pagevec_move_tail(struct pagevec *pvec)
	104	{
	105	int i;
	106	int pgmoved = 0;
	107	struct zone *zone = NULL;
	108
	109	for (i = 0; i < pagevec_count(pvec); i++) {
	110	struct page *page = pvec->pages[i];
	111	struct zone *pagezone = page_zone(page);
	112
	113	if (pagezone != zone) {
	114	if (zone)
	115	spin_unlock(&zone->lru_lock);
	116	zone = pagezone;
	117	spin_lock(&zone->lru_lock);
	118	}
	119	if (PageLRU(page) && !PageActive(page)) {
	120	list_move_tail(&page->lru, &zone->inactive_list);
	121	pgmoved++;
	122	}
	123	}
	124	if (zone)
	125	spin_unlock(&zone->lru_lock);
	126	__count_vm_events(PGROTATED, pgmoved);
	127	release_pages(pvec->pages, pvec->nr, pvec->cold);
	128	pagevec_reinit(pvec);
	129	}
	130
1da177e4 LT	131	/*
	132	* Writeback is about to end against a page which has been marked for immediate
	133	* reclaim. If it still appears to be reclaimable, move it to the tail of the
902aaed0	134	* inactive list.
1da177e4 LT	135	*
	136	* Returns zero if it cleared PG_writeback.
	137	*/
	138	int rotate_reclaimable_page(struct page *page)
	139	{
902aaed0	140	struct pagevec *pvec;
1da177e4 LT	141	unsigned long flags;
	142
	143	if (PageLocked(page))
	144	return 1;
	145	if (PageDirty(page))
	146	return 1;
	147	if (PageActive(page))
	148	return 1;
	149	if (!PageLRU(page))
	150	return 1;
	151
902aaed0 HH	152	page_cache_get(page);
	153	local_irq_save(flags);
	154	pvec = &__get_cpu_var(lru_rotate_pvecs);
	155	if (!pagevec_add(pvec, page))
	156	pagevec_move_tail(pvec);
	157	local_irq_restore(flags);
	158
1da177e4 LT	159	if (!test_clear_page_writeback(page))
1da177e4 LT	160	BUG();
902aaed0	161
1da177e4 LT	162	return 0;
	163	}
	164
	165	/*
	166	* FIXME: speed this up?
	167	*/
920c7a5d	168	void activate_page(struct page *page)
1da177e4 LT	169	{
	170	struct zone *zone = page_zone(page);
	171
	172	spin_lock_irq(&zone->lru_lock);
	173	if (PageLRU(page) && !PageActive(page)) {
	174	del_page_from_inactive_list(zone, page);
	175	SetPageActive(page);
	176	add_page_to_active_list(zone, page);
f8891e5e	177	__count_vm_event(PGACTIVATE);
1da177e4 LT	178	}
	179	spin_unlock_irq(&zone->lru_lock);
	180	}
	181
	182	/*
	183	* Mark a page as having seen activity.
	184	*
	185	* inactive,unreferenced -> inactive,referenced
	186	* inactive,referenced -> active,unreferenced
	187	* active,unreferenced -> active,referenced
	188	*/
920c7a5d	189	void mark_page_accessed(struct page *page)
1da177e4 LT	190	{
	191	if (!PageActive(page) && PageReferenced(page) && PageLRU(page)) {
	192	activate_page(page);
	193	ClearPageReferenced(page);
	194	} else if (!PageReferenced(page)) {
	195	SetPageReferenced(page);
	196	}
	197	}
	198
	199	EXPORT_SYMBOL(mark_page_accessed);
	200
	201	/**
	202	* lru_cache_add: add a page to the page lists
	203	* @page: the page to add
	204	*/
920c7a5d	205	void lru_cache_add(struct page *page)
1da177e4 LT	206	{
	207	struct pagevec *pvec = &get_cpu_var(lru_add_pvecs);
	208
	209	page_cache_get(page);
	210	if (!pagevec_add(pvec, page))
	211	__pagevec_lru_add(pvec);
	212	put_cpu_var(lru_add_pvecs);
	213	}
	214
920c7a5d	215	void lru_cache_add_active(struct page *page)
1da177e4 LT	216	{
	217	struct pagevec *pvec = &get_cpu_var(lru_add_active_pvecs);
	218
	219	page_cache_get(page);
	220	if (!pagevec_add(pvec, page))
	221	__pagevec_lru_add_active(pvec);
	222	put_cpu_var(lru_add_active_pvecs);
	223	}
	224
902aaed0 HH	225	/*
	226	* Drain pages out of the cpu's pagevecs.
	227	* Either "cpu" is the current CPU, and preemption has already been
	228	* disabled; or "cpu" is being hot-unplugged, and is already dead.
	229	*/
	230	static void drain_cpu_pagevecs(int cpu)
1da177e4	231	{
902aaed0	232	struct pagevec *pvec;
1da177e4	233
902aaed0	234	pvec = &per_cpu(lru_add_pvecs, cpu);
1da177e4 LT	235	if (pagevec_count(pvec))
1da177e4 LT	236	__pagevec_lru_add(pvec);
902aaed0	237
80bfed90	238	pvec = &per_cpu(lru_add_active_pvecs, cpu);
1da177e4 LT	239	if (pagevec_count(pvec))
1da177e4 LT	240	__pagevec_lru_add_active(pvec);
902aaed0 HH	241
	242	pvec = &per_cpu(lru_rotate_pvecs, cpu);
	243	if (pagevec_count(pvec)) {
	244	unsigned long flags;
	245
	246	/* No harm done if a racing interrupt already did this */
	247	local_irq_save(flags);
	248	pagevec_move_tail(pvec);
	249	local_irq_restore(flags);
	250	}
80bfed90 AM	251	}
	252
	253	void lru_add_drain(void)
	254	{
902aaed0	255	drain_cpu_pagevecs(get_cpu());
80bfed90	256	put_cpu();
1da177e4 LT	257	}
1da177e4 LT	258
053837fc	259	#ifdef CONFIG_NUMA
c4028958	260	static void lru_add_drain_per_cpu(struct work_struct *dummy)
053837fc NP	261	{
	262	lru_add_drain();
	263	}
	264
	265	/*
	266	* Returns 0 for success
	267	*/
	268	int lru_add_drain_all(void)
	269	{
c4028958	270	return schedule_on_each_cpu(lru_add_drain_per_cpu);
053837fc NP	271	}
	272
	273	#else
	274
	275	/*
	276	* Returns 0 for success
	277	*/
	278	int lru_add_drain_all(void)
	279	{
	280	lru_add_drain();
	281	return 0;
	282	}
	283	#endif
	284
1da177e4 LT	285	/*
	286	* Batched page_cache_release(). Decrement the reference count on all the
	287	* passed pages. If it fell to zero then remove the page from the LRU and
	288	* free it.
	289	*
	290	* Avoid taking zone->lru_lock if possible, but if it is taken, retain it
	291	* for the remainder of the operation.
	292	*
	293	* The locking in this function is against shrink_cache(): we recheck the
	294	* page count inside the lock to see whether shrink_cache grabbed the page
	295	* via the LRU. If it did, give up: shrink_cache will free it.
	296	*/
	297	void release_pages(struct page **pages, int nr, int cold)
	298	{
	299	int i;
	300	struct pagevec pages_to_free;
	301	struct zone *zone = NULL;
902aaed0	302	unsigned long uninitialized_var(flags);
1da177e4 LT	303
	304	pagevec_init(&pages_to_free, cold);
	305	for (i = 0; i < nr; i++) {
	306	struct page *page = pages[i];
1da177e4	307
8519fb30 NP	308	if (unlikely(PageCompound(page))) {
8519fb30 NP	309	if (zone) {
902aaed0	310	spin_unlock_irqrestore(&zone->lru_lock, flags);
8519fb30 NP	311	zone = NULL;
	312	}
	313	put_compound_page(page);
	314	continue;
	315	}
	316
b5810039	317	if (!put_page_testzero(page))
1da177e4 LT	318	continue;
1da177e4 LT	319
46453a6e NP	320	if (PageLRU(page)) {
	321	struct zone *pagezone = page_zone(page);
	322	if (pagezone != zone) {
	323	if (zone)
902aaed0 HH	324	spin_unlock_irqrestore(&zone->lru_lock,
902aaed0 HH	325	flags);
46453a6e	326	zone = pagezone;
902aaed0	327	spin_lock_irqsave(&zone->lru_lock, flags);
46453a6e	328	}
725d704e	329	VM_BUG_ON(!PageLRU(page));
67453911	330	__ClearPageLRU(page);
1da177e4	331	del_page_from_lru(zone, page);
46453a6e NP	332	}
	333
	334	if (!pagevec_add(&pages_to_free, page)) {
	335	if (zone) {
902aaed0	336	spin_unlock_irqrestore(&zone->lru_lock, flags);
46453a6e	337	zone = NULL;
1da177e4	338	}
46453a6e NP	339	__pagevec_free(&pages_to_free);
	340	pagevec_reinit(&pages_to_free);
	341	}
1da177e4 LT	342	}
1da177e4 LT	343	if (zone)
902aaed0	344	spin_unlock_irqrestore(&zone->lru_lock, flags);
1da177e4 LT	345
	346	pagevec_free(&pages_to_free);
	347	}
	348
	349	/*
	350	* The pages which we're about to release may be in the deferred lru-addition
	351	* queues. That would prevent them from really being freed right now. That's
	352	* OK from a correctness point of view but is inefficient - those pages may be
	353	* cache-warm and we want to give them back to the page allocator ASAP.
	354	*
	355	* So __pagevec_release() will drain those queues here. __pagevec_lru_add()
	356	* and __pagevec_lru_add_active() call release_pages() directly to avoid
	357	* mutual recursion.
	358	*/
	359	void __pagevec_release(struct pagevec *pvec)
	360	{
	361	lru_add_drain();
	362	release_pages(pvec->pages, pagevec_count(pvec), pvec->cold);
	363	pagevec_reinit(pvec);
	364	}
	365
7f285701 SF	366	EXPORT_SYMBOL(__pagevec_release);
7f285701 SF	367
1da177e4 LT	368	/*
	369	* pagevec_release() for pages which are known to not be on the LRU
	370	*
	371	* This function reinitialises the caller's pagevec.
	372	*/
	373	void __pagevec_release_nonlru(struct pagevec *pvec)
	374	{
	375	int i;
	376	struct pagevec pages_to_free;
	377
	378	pagevec_init(&pages_to_free, pvec->cold);
1da177e4 LT	379	for (i = 0; i < pagevec_count(pvec); i++) {
	380	struct page *page = pvec->pages[i];
	381
725d704e	382	VM_BUG_ON(PageLRU(page));
1da177e4 LT	383	if (put_page_testzero(page))
	384	pagevec_add(&pages_to_free, page);
	385	}
	386	pagevec_free(&pages_to_free);
	387	pagevec_reinit(pvec);
	388	}
	389
	390	/*
	391	* Add the passed pages to the LRU, then drop the caller's refcount
	392	* on them. Reinitialises the caller's pagevec.
	393	*/
	394	void __pagevec_lru_add(struct pagevec *pvec)
	395	{
	396	int i;
	397	struct zone *zone = NULL;
	398
	399	for (i = 0; i < pagevec_count(pvec); i++) {
	400	struct page *page = pvec->pages[i];
	401	struct zone *pagezone = page_zone(page);
	402
	403	if (pagezone != zone) {
	404	if (zone)
	405	spin_unlock_irq(&zone->lru_lock);
	406	zone = pagezone;
	407	spin_lock_irq(&zone->lru_lock);
	408	}
725d704e	409	VM_BUG_ON(PageLRU(page));
8d438f96	410	SetPageLRU(page);
1da177e4 LT	411	add_page_to_inactive_list(zone, page);
	412	}
	413	if (zone)
	414	spin_unlock_irq(&zone->lru_lock);
	415	release_pages(pvec->pages, pvec->nr, pvec->cold);
	416	pagevec_reinit(pvec);
	417	}
	418
	419	EXPORT_SYMBOL(__pagevec_lru_add);
	420
	421	void __pagevec_lru_add_active(struct pagevec *pvec)
	422	{
	423	int i;
	424	struct zone *zone = NULL;
	425
	426	for (i = 0; i < pagevec_count(pvec); i++) {
	427	struct page *page = pvec->pages[i];
	428	struct zone *pagezone = page_zone(page);
	429
	430	if (pagezone != zone) {
	431	if (zone)
	432	spin_unlock_irq(&zone->lru_lock);
	433	zone = pagezone;
	434	spin_lock_irq(&zone->lru_lock);
	435	}
725d704e	436	VM_BUG_ON(PageLRU(page));
8d438f96	437	SetPageLRU(page);
725d704e	438	VM_BUG_ON(PageActive(page));
4c84cacf	439	SetPageActive(page);
1da177e4 LT	440	add_page_to_active_list(zone, page);
	441	}
	442	if (zone)
	443	spin_unlock_irq(&zone->lru_lock);
	444	release_pages(pvec->pages, pvec->nr, pvec->cold);
	445	pagevec_reinit(pvec);
	446	}
	447
	448	/*
	449	* Try to drop buffers from the pages in a pagevec
	450	*/
	451	void pagevec_strip(struct pagevec *pvec)
	452	{
	453	int i;
	454
	455	for (i = 0; i < pagevec_count(pvec); i++) {
	456	struct page *page = pvec->pages[i];
	457
	458	if (PagePrivate(page) && !TestSetPageLocked(page)) {
5b40dc78 CL	459	if (PagePrivate(page))
5b40dc78 CL	460	try_to_release_page(page, 0);
1da177e4 LT	461	unlock_page(page);
	462	}
	463	}
	464	}
	465
	466	/**
	467	* pagevec_lookup - gang pagecache lookup
	468	* @pvec: Where the resulting pages are placed
	469	* @mapping: The address_space to search
	470	* @start: The starting page index
	471	* @nr_pages: The maximum number of pages
	472	*
	473	* pagevec_lookup() will search for and return a group of up to @nr_pages pages
	474	* in the mapping. The pages are placed in @pvec. pagevec_lookup() takes a
	475	* reference against the pages in @pvec.
	476	*
	477	* The search returns a group of mapping-contiguous pages with ascending
	478	* indexes. There may be holes in the indices due to not-present pages.
	479	*
	480	* pagevec_lookup() returns the number of pages which were found.
	481	*/
	482	unsigned pagevec_lookup(struct pagevec pvec, struct address_space mapping,
	483	pgoff_t start, unsigned nr_pages)
	484	{
	485	pvec->nr = find_get_pages(mapping, start, nr_pages, pvec->pages);
	486	return pagevec_count(pvec);
	487	}
	488
78539fdf CH	489	EXPORT_SYMBOL(pagevec_lookup);
78539fdf CH	490
1da177e4 LT	491	unsigned pagevec_lookup_tag(struct pagevec pvec, struct address_space mapping,
	492	pgoff_t *index, int tag, unsigned nr_pages)
	493	{
	494	pvec->nr = find_get_pages_tag(mapping, index, tag,
	495	nr_pages, pvec->pages);
	496	return pagevec_count(pvec);
	497	}
	498
7f285701	499	EXPORT_SYMBOL(pagevec_lookup_tag);
1da177e4 LT	500
	501	#ifdef CONFIG_SMP
	502	/*
	503	* We tolerate a little inaccuracy to avoid ping-ponging the counter between
	504	* CPUs
	505	*/
	506	#define ACCT_THRESHOLD max(16, NR_CPUS * 2)
	507
	508	static DEFINE_PER_CPU(long, committed_space) = 0;
	509
	510	void vm_acct_memory(long pages)
	511	{
	512	long *local;
	513
	514	preempt_disable();
	515	local = &__get_cpu_var(committed_space);
	516	*local += pages;
	517	if (local > ACCT_THRESHOLD \|\| local < -ACCT_THRESHOLD) {
	518	atomic_add(*local, &vm_committed_space);
	519	*local = 0;
	520	}
	521	preempt_enable();
	522	}
1da177e4 LT	523
1da177e4 LT	524	#ifdef CONFIG_HOTPLUG_CPU
1da177e4 LT	525
	526	/* Drop the CPU's cached committed space back into the central pool. */
	527	static int cpu_swap_callback(struct notifier_block *nfb,
	528	unsigned long action,
	529	void *hcpu)
	530	{
	531	long *committed;
	532
	533	committed = &per_cpu(committed_space, (long)hcpu);
8bb78442	534	if (action == CPU_DEAD \|\| action == CPU_DEAD_FROZEN) {
1da177e4 LT	535	atomic_add(*committed, &vm_committed_space);
1da177e4 LT	536	*committed = 0;
902aaed0	537	drain_cpu_pagevecs((long)hcpu);
1da177e4 LT	538	}
	539	return NOTIFY_OK;
	540	}
	541	#endif /* CONFIG_HOTPLUG_CPU */
	542	#endif /* CONFIG_SMP */
	543
1da177e4 LT	544	/*
	545	* Perform any setup for the swap system
	546	*/
	547	void __init swap_setup(void)
	548	{
	549	unsigned long megs = num_physpages >> (20 - PAGE_SHIFT);
	550
e0bf68dd PZ	551	#ifdef CONFIG_SWAP
	552	bdi_init(swapper_space.backing_dev_info);
	553	#endif
	554
1da177e4 LT	555	/* Use a smaller cluster for small-memory machines */
	556	if (megs < 16)
	557	page_cluster = 2;
	558	else
	559	page_cluster = 3;
	560	/*
	561	* Right now other parts of the system means that we
	562	* _really_ don't want to cluster much more
	563	*/
02316067	564	#ifdef CONFIG_HOTPLUG_CPU
1da177e4	565	hotcpu_notifier(cpu_swap_callback, 0);
02316067	566	#endif
1da177e4	567	}