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

/*
 *  linux/mm/swap_state.c
 *
 *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 *  Swap reorganised 29.12.95, Stephen Tweedie
 *
 *  Rewritten to use page cache, (C) 1998 Stephen Tweedie
 */
#include <linux/module.h>
#include <linux/mm.h>
#include <linux/kernel_stat.h>
#include <linux/swap.h>
#include <linux/init.h>
#include <linux/pagemap.h>
#include <linux/buffer_head.h>
#include <linux/backing-dev.h>

#include <asm/pgtable.h>

/*
 * swapper_space is a fiction, retained to simplify the path through
 * vmscan's shrink_list, to make sync_page look nicer, and to allow
 * future use of radix_tree tags in the swap cache.
 */
static struct address_space_operations swap_aops = {
	.writepage	= swap_writepage,
	.sync_page	= block_sync_page,
	.set_page_dirty	= __set_page_dirty_nobuffers,
};

static struct backing_dev_info swap_backing_dev_info = {
	.capabilities	= BDI_CAP_NO_ACCT_DIRTY | BDI_CAP_NO_WRITEBACK,
	.unplug_io_fn	= swap_unplug_io_fn,
};

struct address_space swapper_space = {
	.page_tree	= RADIX_TREE_INIT(GFP_ATOMIC|__GFP_NOWARN),
	.tree_lock	= RW_LOCK_UNLOCKED,
	.a_ops		= &swap_aops,
	.i_mmap_nonlinear = LIST_HEAD_INIT(swapper_space.i_mmap_nonlinear),
	.backing_dev_info = &swap_backing_dev_info,
};

#define INC_CACHE_INFO(x)	do { swap_cache_info.x++; } while (0)

static struct {
	unsigned long add_total;
	unsigned long del_total;
	unsigned long find_success;
	unsigned long find_total;
	unsigned long noent_race;
	unsigned long exist_race;
} swap_cache_info;

void show_swap_cache_info(void)
{
	printk("Swap cache: add %lu, delete %lu, find %lu/%lu, race %lu+%lu\n",
		swap_cache_info.add_total, swap_cache_info.del_total,
		swap_cache_info.find_success, swap_cache_info.find_total,
		swap_cache_info.noent_race, swap_cache_info.exist_race);
	printk("Free swap  = %lukB\n", nr_swap_pages << (PAGE_SHIFT - 10));
	printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
}

/*
 * __add_to_swap_cache resembles add_to_page_cache on swapper_space,
 * but sets SwapCache flag and private instead of mapping and index.
 */
static int __add_to_swap_cache(struct page *page, swp_entry_t entry,
			       gfp_t gfp_mask)
{
	int error;

	BUG_ON(PageSwapCache(page));
	BUG_ON(PagePrivate(page));
	error = radix_tree_preload(gfp_mask);
	if (!error) {
		write_lock_irq(&swapper_space.tree_lock);
		error = radix_tree_insert(&swapper_space.page_tree,
						entry.val, page);
		if (!error) {
			page_cache_get(page);
			SetPageLocked(page);
			SetPageSwapCache(page);
			set_page_private(page, entry.val);
			total_swapcache_pages++;
			pagecache_acct(1);
		}
		write_unlock_irq(&swapper_space.tree_lock);
		radix_tree_preload_end();
	}
	return error;
}

static int add_to_swap_cache(struct page *page, swp_entry_t entry)
{
	int error;

	if (!swap_duplicate(entry)) {
		INC_CACHE_INFO(noent_race);
		return -ENOENT;
	}
	error = __add_to_swap_cache(page, entry, GFP_KERNEL);
	/*
	 * Anon pages are already on the LRU, we don't run lru_cache_add here.
	 */
	if (error) {
		swap_free(entry);
		if (error == -EEXIST)
			INC_CACHE_INFO(exist_race);
		return error;
	}
	INC_CACHE_INFO(add_total);
	return 0;
}

/*
 * This must be called only on pages that have
 * been verified to be in the swap cache.
 */
void __delete_from_swap_cache(struct page *page)
{
	BUG_ON(!PageLocked(page));
	BUG_ON(!PageSwapCache(page));
	BUG_ON(PageWriteback(page));
	BUG_ON(PagePrivate(page));

	radix_tree_delete(&swapper_space.page_tree, page_private(page));
	set_page_private(page, 0);
	ClearPageSwapCache(page);
	total_swapcache_pages--;
	pagecache_acct(-1);
	INC_CACHE_INFO(del_total);
}

/**
 * add_to_swap - allocate swap space for a page
 * @page: page we want to move to swap
 *
 * Allocate swap space for the page and add the page to the
 * swap cache.  Caller needs to hold the page lock. 
 */
int add_to_swap(struct page * page)
{
	swp_entry_t entry;
	int err;

	if (!PageLocked(page))
		BUG();

	for (;;) {
		entry = get_swap_page();
		if (!entry.val)
			return 0;

		/*
		 * Radix-tree node allocations from PF_MEMALLOC contexts could
		 * completely exhaust the page allocator. __GFP_NOMEMALLOC
		 * stops emergency reserves from being allocated.
		 *
		 * TODO: this could cause a theoretical memory reclaim
		 * deadlock in the swap out path.
		 */
		/*
		 * Add it to the swap cache and mark it dirty
		 */
		err = __add_to_swap_cache(page, entry,
				GFP_ATOMIC|__GFP_NOMEMALLOC|__GFP_NOWARN);

		switch (err) {
		case 0:				/* Success */
			SetPageUptodate(page);
			SetPageDirty(page);
			INC_CACHE_INFO(add_total);
			return 1;
		case -EEXIST:
			/* Raced with "speculative" read_swap_cache_async */
			INC_CACHE_INFO(exist_race);
			swap_free(entry);
			continue;
		default:
			/* -ENOMEM radix-tree allocation failure */
			swap_free(entry);
			return 0;
		}
	}
}

/*
 * This must be called only on pages that have
 * been verified to be in the swap cache and locked.
 * It will never put the page into the free list,
 * the caller has a reference on the page.
 */
void delete_from_swap_cache(struct page *page)
{
	swp_entry_t entry;

	entry.val = page_private(page);

	write_lock_irq(&swapper_space.tree_lock);
	__delete_from_swap_cache(page);
	write_unlock_irq(&swapper_space.tree_lock);

	swap_free(entry);
	page_cache_release(page);
}

/*
 * Strange swizzling function only for use by shmem_writepage
 */
int move_to_swap_cache(struct page *page, swp_entry_t entry)
{
	int err = __add_to_swap_cache(page, entry, GFP_ATOMIC);
	if (!err) {
		remove_from_page_cache(page);
		page_cache_release(page);	/* pagecache ref */
		if (!swap_duplicate(entry))
			BUG();
		SetPageDirty(page);
		INC_CACHE_INFO(add_total);
	} else if (err == -EEXIST)
		INC_CACHE_INFO(exist_race);
	return err;
}

/*
 * Strange swizzling function for shmem_getpage (and shmem_unuse)
 */
int move_from_swap_cache(struct page *page, unsigned long index,
		struct address_space *mapping)
{
	int err = add_to_page_cache(page, mapping, index, GFP_ATOMIC);
	if (!err) {
		delete_from_swap_cache(page);
		/* shift page from clean_pages to dirty_pages list */
		ClearPageDirty(page);
		set_page_dirty(page);
	}
	return err;
}

/* 
 * If we are the only user, then try to free up the swap cache. 
 * 
 * Its ok to check for PageSwapCache without the page lock
 * here because we are going to recheck again inside 
 * exclusive_swap_page() _with_ the lock. 
 * 					- Marcelo
 */
static inline void free_swap_cache(struct page *page)
{
	if (PageSwapCache(page) && !TestSetPageLocked(page)) {
		remove_exclusive_swap_page(page);
		unlock_page(page);
	}
}

/* 
 * Perform a free_page(), also freeing any swap cache associated with
 * this page if it is the last user of the page.
 */
void free_page_and_swap_cache(struct page *page)
{
	free_swap_cache(page);
	page_cache_release(page);
}

/*
 * Passed an array of pages, drop them all from swapcache and then release
 * them.  They are removed from the LRU and freed if this is their last use.
 */
void free_pages_and_swap_cache(struct page **pages, int nr)
{
	int chunk = 16;
	struct page **pagep = pages;

	lru_add_drain();
	while (nr) {
		int todo = min(chunk, nr);
		int i;

		for (i = 0; i < todo; i++)
			free_swap_cache(pagep[i]);
		release_pages(pagep, todo, 0);
		pagep += todo;
		nr -= todo;
	}
}

/*
 * Lookup a swap entry in the swap cache. A found page will be returned
 * unlocked and with its refcount incremented - we rely on the kernel
 * lock getting page table operations atomic even if we drop the page
 * lock before returning.
 */
struct page * lookup_swap_cache(swp_entry_t entry)
{
	struct page *page;

	page = find_get_page(&swapper_space, entry.val);

	if (page)
		INC_CACHE_INFO(find_success);

	INC_CACHE_INFO(find_total);
	return page;
}

/* 
 * Locate a page of swap in physical memory, reserving swap cache space
 * and reading the disk if it is not already cached.
 * A failure return means that either the page allocation failed or that
 * the swap entry is no longer in use.
 */
struct page *read_swap_cache_async(swp_entry_t entry,
			struct vm_area_struct *vma, unsigned long addr)
{
	struct page *found_page, *new_page = NULL;
	int err;

	do {
		/*
		 * First check the swap cache.  Since this is normally
		 * called after lookup_swap_cache() failed, re-calling
		 * that would confuse statistics.
		 */
		found_page = find_get_page(&swapper_space, entry.val);
		if (found_page)
			break;

		/*
		 * Get a new page to read into from swap.
		 */
		if (!new_page) {
			new_page = alloc_page_vma(GFP_HIGHUSER, vma, addr);
			if (!new_page)
				break;		/* Out of memory */
		}

		/*
		 * Associate the page with swap entry in the swap cache.
		 * May fail (-ENOENT) if swap entry has been freed since
		 * our caller observed it.  May fail (-EEXIST) if there
		 * is already a page associated with this entry in the
		 * swap cache: added by a racing read_swap_cache_async,
		 * or by try_to_swap_out (or shmem_writepage) re-using
		 * the just freed swap entry for an existing page.
		 * May fail (-ENOMEM) if radix-tree node allocation failed.
		 */
		err = add_to_swap_cache(new_page, entry);
		if (!err) {
			/*
			 * Initiate read into locked page and return.
			 */
			lru_cache_add_active(new_page);
			swap_readpage(NULL, new_page);
			return new_page;
		}
	} while (err != -ENOENT && err != -ENOMEM);

	if (new_page)
		page_cache_release(new_page);
	return found_page;
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/mm/swap_state.c
	3	*
	4	* Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
	5	* Swap reorganised 29.12.95, Stephen Tweedie
	6	*
	7	* Rewritten to use page cache, (C) 1998 Stephen Tweedie
	8	*/
	9	#include <linux/module.h>
	10	#include <linux/mm.h>
	11	#include <linux/kernel_stat.h>
	12	#include <linux/swap.h>
	13	#include <linux/init.h>
	14	#include <linux/pagemap.h>
	15	#include <linux/buffer_head.h>
	16	#include <linux/backing-dev.h>
	17
	18	#include <asm/pgtable.h>
	19
	20	/*
	21	* swapper_space is a fiction, retained to simplify the path through
	22	* vmscan's shrink_list, to make sync_page look nicer, and to allow
	23	* future use of radix_tree tags in the swap cache.
	24	*/
	25	static struct address_space_operations swap_aops = {
	26	.writepage = swap_writepage,
	27	.sync_page = block_sync_page,
	28	.set_page_dirty = __set_page_dirty_nobuffers,
	29	};
	30
	31	static struct backing_dev_info swap_backing_dev_info = {
	32	.capabilities = BDI_CAP_NO_ACCT_DIRTY \| BDI_CAP_NO_WRITEBACK,
	33	.unplug_io_fn = swap_unplug_io_fn,
	34	};
	35
	36	struct address_space swapper_space = {
	37	.page_tree = RADIX_TREE_INIT(GFP_ATOMIC\|__GFP_NOWARN),
	38	.tree_lock = RW_LOCK_UNLOCKED,
	39	.a_ops = &swap_aops,
	40	.i_mmap_nonlinear = LIST_HEAD_INIT(swapper_space.i_mmap_nonlinear),
	41	.backing_dev_info = &swap_backing_dev_info,
	42	};
1da177e4 LT	43
	44	#define INC_CACHE_INFO(x) do { swap_cache_info.x++; } while (0)
	45
	46	static struct {
	47	unsigned long add_total;
	48	unsigned long del_total;
	49	unsigned long find_success;
	50	unsigned long find_total;
	51	unsigned long noent_race;
	52	unsigned long exist_race;
	53	} swap_cache_info;
	54
	55	void show_swap_cache_info(void)
	56	{
	57	printk("Swap cache: add %lu, delete %lu, find %lu/%lu, race %lu+%lu\n",
	58	swap_cache_info.add_total, swap_cache_info.del_total,
	59	swap_cache_info.find_success, swap_cache_info.find_total,
	60	swap_cache_info.noent_race, swap_cache_info.exist_race);
	61	printk("Free swap = %lukB\n", nr_swap_pages << (PAGE_SHIFT - 10));
	62	printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
	63	}
	64
	65	/*
	66	* __add_to_swap_cache resembles add_to_page_cache on swapper_space,
	67	* but sets SwapCache flag and private instead of mapping and index.
	68	*/
9de75d11	69	static int __add_to_swap_cache(struct page *page, swp_entry_t entry,
dd0fc66f	70	gfp_t gfp_mask)
1da177e4 LT	71	{
	72	int error;
	73
	74	BUG_ON(PageSwapCache(page));
	75	BUG_ON(PagePrivate(page));
	76	error = radix_tree_preload(gfp_mask);
	77	if (!error) {
	78	write_lock_irq(&swapper_space.tree_lock);
	79	error = radix_tree_insert(&swapper_space.page_tree,
	80	entry.val, page);
	81	if (!error) {
	82	page_cache_get(page);
	83	SetPageLocked(page);
	84	SetPageSwapCache(page);
4c21e2f2	85	set_page_private(page, entry.val);
1da177e4 LT	86	total_swapcache_pages++;
	87	pagecache_acct(1);
	88	}
	89	write_unlock_irq(&swapper_space.tree_lock);
	90	radix_tree_preload_end();
	91	}
	92	return error;
	93	}
	94
	95	static int add_to_swap_cache(struct page *page, swp_entry_t entry)
	96	{
	97	int error;
	98
	99	if (!swap_duplicate(entry)) {
	100	INC_CACHE_INFO(noent_race);
	101	return -ENOENT;
	102	}
	103	error = __add_to_swap_cache(page, entry, GFP_KERNEL);
	104	/*
	105	* Anon pages are already on the LRU, we don't run lru_cache_add here.
	106	*/
	107	if (error) {
	108	swap_free(entry);
	109	if (error == -EEXIST)
	110	INC_CACHE_INFO(exist_race);
	111	return error;
	112	}
	113	INC_CACHE_INFO(add_total);
	114	return 0;
	115	}
	116
	117	/*
	118	* This must be called only on pages that have
	119	* been verified to be in the swap cache.
	120	*/
	121	void __delete_from_swap_cache(struct page *page)
	122	{
	123	BUG_ON(!PageLocked(page));
	124	BUG_ON(!PageSwapCache(page));
	125	BUG_ON(PageWriteback(page));
3279ffd9	126	BUG_ON(PagePrivate(page));
1da177e4	127
4c21e2f2 HD	128	radix_tree_delete(&swapper_space.page_tree, page_private(page));
4c21e2f2 HD	129	set_page_private(page, 0);
1da177e4 LT	130	ClearPageSwapCache(page);
	131	total_swapcache_pages--;
	132	pagecache_acct(-1);
	133	INC_CACHE_INFO(del_total);
	134	}
	135
	136	/**
	137	* add_to_swap - allocate swap space for a page
	138	* @page: page we want to move to swap
	139	*
	140	* Allocate swap space for the page and add the page to the
	141	* swap cache. Caller needs to hold the page lock.
	142	*/
	143	int add_to_swap(struct page * page)
	144	{
	145	swp_entry_t entry;
1da177e4 LT	146	int err;
	147
	148	if (!PageLocked(page))
	149	BUG();
	150
	151	for (;;) {
	152	entry = get_swap_page();
	153	if (!entry.val)
	154	return 0;
	155
bd53b714 NP	156	/*
	157	* Radix-tree node allocations from PF_MEMALLOC contexts could
	158	* completely exhaust the page allocator. __GFP_NOMEMALLOC
	159	* stops emergency reserves from being allocated.
1da177e4	160	*
bd53b714 NP	161	* TODO: this could cause a theoretical memory reclaim
bd53b714 NP	162	* deadlock in the swap out path.
1da177e4	163	*/
1da177e4 LT	164	/*
	165	* Add it to the swap cache and mark it dirty
	166	*/
bd53b714 NP	167	err = __add_to_swap_cache(page, entry,
bd53b714 NP	168	GFP_ATOMIC\|__GFP_NOMEMALLOC\|__GFP_NOWARN);
1da177e4 LT	169
	170	switch (err) {
	171	case 0: /* Success */
	172	SetPageUptodate(page);
	173	SetPageDirty(page);
	174	INC_CACHE_INFO(add_total);
	175	return 1;
	176	case -EEXIST:
	177	/* Raced with "speculative" read_swap_cache_async */
	178	INC_CACHE_INFO(exist_race);
	179	swap_free(entry);
	180	continue;
	181	default:
	182	/* -ENOMEM radix-tree allocation failure */
	183	swap_free(entry);
	184	return 0;
	185	}
	186	}
	187	}
	188
	189	/*
	190	* This must be called only on pages that have
	191	* been verified to be in the swap cache and locked.
	192	* It will never put the page into the free list,
	193	* the caller has a reference on the page.
	194	*/
	195	void delete_from_swap_cache(struct page *page)
	196	{
	197	swp_entry_t entry;
	198
4c21e2f2	199	entry.val = page_private(page);
1da177e4 LT	200
	201	write_lock_irq(&swapper_space.tree_lock);
	202	__delete_from_swap_cache(page);
	203	write_unlock_irq(&swapper_space.tree_lock);
	204
	205	swap_free(entry);
	206	page_cache_release(page);
	207	}
	208
	209	/*
	210	* Strange swizzling function only for use by shmem_writepage
	211	*/
	212	int move_to_swap_cache(struct page *page, swp_entry_t entry)
	213	{
	214	int err = __add_to_swap_cache(page, entry, GFP_ATOMIC);
	215	if (!err) {
	216	remove_from_page_cache(page);
	217	page_cache_release(page); /* pagecache ref */
	218	if (!swap_duplicate(entry))
	219	BUG();
	220	SetPageDirty(page);
	221	INC_CACHE_INFO(add_total);
	222	} else if (err == -EEXIST)
	223	INC_CACHE_INFO(exist_race);
	224	return err;
	225	}
	226
	227	/*
	228	* Strange swizzling function for shmem_getpage (and shmem_unuse)
	229	*/
	230	int move_from_swap_cache(struct page *page, unsigned long index,
	231	struct address_space *mapping)
	232	{
	233	int err = add_to_page_cache(page, mapping, index, GFP_ATOMIC);
	234	if (!err) {
	235	delete_from_swap_cache(page);
	236	/* shift page from clean_pages to dirty_pages list */
	237	ClearPageDirty(page);
	238	set_page_dirty(page);
	239	}
	240	return err;
	241	}
	242
	243	/*
	244	* If we are the only user, then try to free up the swap cache.
	245	*
	246	* Its ok to check for PageSwapCache without the page lock
	247	* here because we are going to recheck again inside
	248	* exclusive_swap_page() _with_ the lock.
	249	* - Marcelo
	250	*/
	251	static inline void free_swap_cache(struct page *page)
	252	{
	253	if (PageSwapCache(page) && !TestSetPageLocked(page)) {
	254	remove_exclusive_swap_page(page);
	255	unlock_page(page);
	256	}
	257	}
	258
	259	/*
	260	* Perform a free_page(), also freeing any swap cache associated with
b8072f09	261	* this page if it is the last user of the page.
1da177e4 LT	262	*/
	263	void free_page_and_swap_cache(struct page *page)
	264	{
	265	free_swap_cache(page);
	266	page_cache_release(page);
	267	}
	268
	269	/*
	270	* Passed an array of pages, drop them all from swapcache and then release
	271	* them. They are removed from the LRU and freed if this is their last use.
	272	*/
	273	void free_pages_and_swap_cache(struct page **pages, int nr)
	274	{
	275	int chunk = 16;
	276	struct page **pagep = pages;
	277
	278	lru_add_drain();
	279	while (nr) {
	280	int todo = min(chunk, nr);
	281	int i;
	282
	283	for (i = 0; i < todo; i++)
	284	free_swap_cache(pagep[i]);
	285	release_pages(pagep, todo, 0);
	286	pagep += todo;
	287	nr -= todo;
	288	}
	289	}
	290
	291	/*
	292	* Lookup a swap entry in the swap cache. A found page will be returned
	293	* unlocked and with its refcount incremented - we rely on the kernel
	294	* lock getting page table operations atomic even if we drop the page
	295	* lock before returning.
	296	*/
	297	struct page * lookup_swap_cache(swp_entry_t entry)
	298	{
	299	struct page *page;
	300
	301	page = find_get_page(&swapper_space, entry.val);
	302
	303	if (page)
	304	INC_CACHE_INFO(find_success);
	305
	306	INC_CACHE_INFO(find_total);
	307	return page;
	308	}
	309
	310	/*
	311	* Locate a page of swap in physical memory, reserving swap cache space
	312	* and reading the disk if it is not already cached.
	313	* A failure return means that either the page allocation failed or that
	314	* the swap entry is no longer in use.
	315	*/
	316	struct page *read_swap_cache_async(swp_entry_t entry,
	317	struct vm_area_struct *vma, unsigned long addr)
	318	{
	319	struct page found_page, new_page = NULL;
	320	int err;
	321
	322	do {
	323	/*
	324	* First check the swap cache. Since this is normally
	325	* called after lookup_swap_cache() failed, re-calling
326	* that would confuse statistics.
327	*/
328	found_page = find_get_page(&swapper_space, entry.val);
329	if (found_page)
330	break;
331
332	/*
333	* Get a new page to read into from swap.
334	*/
335	if (!new_page) {
336	new_page = alloc_page_vma(GFP_HIGHUSER, vma, addr);
337	if (!new_page)
338	break; /* Out of memory */
339	}
340
341	/*
342	* Associate the page with swap entry in the swap cache.
343	* May fail (-ENOENT) if swap entry has been freed since
344	* our caller observed it. May fail (-EEXIST) if there
345	* is already a page associated with this entry in the
346	* swap cache: added by a racing read_swap_cache_async,
347	* or by try_to_swap_out (or shmem_writepage) re-using
348	* the just freed swap entry for an existing page.
349	* May fail (-ENOMEM) if radix-tree node allocation failed.
350	*/
351	err = add_to_swap_cache(new_page, entry);
352	if (!err) {
353	/*
354	* Initiate read into locked page and return.
355	*/
356	lru_cache_add_active(new_page);
357	swap_readpage(NULL, new_page);
358	return new_page;
359	}
360	} while (err != -ENOENT && err != -ENOMEM);
361
362	if (new_page)
363	page_cache_release(new_page);
364	return found_page;
365	}