[deliverable/linux.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/mm.h>
#include <linux/gfp.h>
#include <linux/kernel_stat.h>
#include <linux/swap.h>
#include <linux/swapops.h>
#include <linux/init.h>
#include <linux/pagemap.h>
#include <linux/backing-dev.h>
#include <linux/blkdev.h>
#include <linux/pagevec.h>
#include <linux/migrate.h>
#include <linux/page_cgroup.h>

#include <asm/pgtable.h>

/*
 * swapper_space is a fiction, retained to simplify the path through
 * vmscan's shrink_page_list.
 */
static const struct address_space_operations swap_aops = {
	.writepage	= swap_writepage,
	.set_page_dirty	= swap_set_page_dirty,
	.migratepage	= migrate_page,
};

static struct backing_dev_info swap_backing_dev_info = {
	.name		= "swap",
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK | BDI_CAP_SWAP_BACKED,
};

struct address_space swapper_space = {
	.page_tree	= RADIX_TREE_INIT(GFP_ATOMIC|__GFP_NOWARN),
	.tree_lock	= __SPIN_LOCK_UNLOCKED(swapper_space.tree_lock),
	.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;
} swap_cache_info;

void show_swap_cache_info(void)
{
	printk("%lu pages in swap cache\n", total_swapcache_pages);
	printk("Swap cache stats: add %lu, delete %lu, find %lu/%lu\n",
		swap_cache_info.add_total, swap_cache_info.del_total,
		swap_cache_info.find_success, swap_cache_info.find_total);
	printk("Free swap  = %ldkB\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_locked 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)
{
	int error;

	VM_BUG_ON(!PageLocked(page));
	VM_BUG_ON(PageSwapCache(page));
	VM_BUG_ON(!PageSwapBacked(page));

	page_cache_get(page);
	SetPageSwapCache(page);
	set_page_private(page, entry.val);

	spin_lock_irq(&swapper_space.tree_lock);
	error = radix_tree_insert(&swapper_space.page_tree, entry.val, page);
	if (likely(!error)) {
		total_swapcache_pages++;
		__inc_zone_page_state(page, NR_FILE_PAGES);
		INC_CACHE_INFO(add_total);
	}
	spin_unlock_irq(&swapper_space.tree_lock);

	if (unlikely(error)) {
		/*
		 * Only the context which have set SWAP_HAS_CACHE flag
		 * would call add_to_swap_cache().
		 * So add_to_swap_cache() doesn't returns -EEXIST.
		 */
		VM_BUG_ON(error == -EEXIST);
		set_page_private(page, 0UL);
		ClearPageSwapCache(page);
		page_cache_release(page);
	}

	return error;
}


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

	error = radix_tree_preload(gfp_mask);
	if (!error) {
		error = __add_to_swap_cache(page, entry);
		radix_tree_preload_end();
	}
	return error;
}

/*
 * 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)
{
	VM_BUG_ON(!PageLocked(page));
	VM_BUG_ON(!PageSwapCache(page));
	VM_BUG_ON(PageWriteback(page));

	radix_tree_delete(&swapper_space.page_tree, page_private(page));
	set_page_private(page, 0);
	ClearPageSwapCache(page);
	total_swapcache_pages--;
	__dec_zone_page_state(page, NR_FILE_PAGES);
	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;

	VM_BUG_ON(!PageLocked(page));
	VM_BUG_ON(!PageUptodate(page));

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

	if (unlikely(PageTransHuge(page)))
		if (unlikely(split_huge_page(page))) {
			swapcache_free(entry, NULL);
			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_HIGH|__GFP_NOMEMALLOC|__GFP_NOWARN);

	if (!err) {	/* Success */
		SetPageDirty(page);
		return 1;
	} else {	/* -ENOMEM radix-tree allocation failure */
		/*
		 * add_to_swap_cache() doesn't return -EEXIST, so we can safely
		 * clear SWAP_HAS_CACHE flag.
		 */
		swapcache_free(entry, NULL);
		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);

	spin_lock_irq(&swapper_space.tree_lock);
	__delete_from_swap_cache(page);
	spin_unlock_irq(&swapper_space.tree_lock);

	swapcache_free(entry, page);
	page_cache_release(page);
}

/* 
 * 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
 * try_to_free_swap() _with_ the lock.
 * 					- Marcelo
 */
static inline void free_swap_cache(struct page *page)
{
	if (PageSwapCache(page) && !page_mapped(page) && trylock_page(page)) {
		try_to_free_swap(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)
{
	struct page **pagep = pages;

	lru_add_drain();
	while (nr) {
		int todo = min(nr, PAGEVEC_SIZE);
		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, gfp_t gfp_mask,
			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_mask, vma, addr);
			if (!new_page)
				break;		/* Out of memory */
		}

		/*
		 * call radix_tree_preload() while we can wait.
		 */
		err = radix_tree_preload(gfp_mask & GFP_KERNEL);
		if (err)
			break;

		/*
		 * Swap entry may have been freed since our caller observed it.
		 */
		err = swapcache_prepare(entry);
		if (err == -EEXIST) {	/* seems racy */
			radix_tree_preload_end();
			continue;
		}
		if (err) {		/* swp entry is obsolete ? */
			radix_tree_preload_end();
			break;
		}

		/* May fail (-ENOMEM) if radix-tree node allocation failed. */
		__set_page_locked(new_page);
		SetPageSwapBacked(new_page);
		err = __add_to_swap_cache(new_page, entry);
		if (likely(!err)) {
			radix_tree_preload_end();
			/*
			 * Initiate read into locked page and return.
			 */
			lru_cache_add_anon(new_page);
			swap_readpage(new_page);
			return new_page;
		}
		radix_tree_preload_end();
		ClearPageSwapBacked(new_page);
		__clear_page_locked(new_page);
		/*
		 * add_to_swap_cache() doesn't return -EEXIST, so we can safely
		 * clear SWAP_HAS_CACHE flag.
		 */
		swapcache_free(entry, NULL);
	} while (err != -ENOMEM);

	if (new_page)
		page_cache_release(new_page);
	return found_page;
}

/**
 * swapin_readahead - swap in pages in hope we need them soon
 * @entry: swap entry of this memory
 * @gfp_mask: memory allocation flags
 * @vma: user vma this address belongs to
 * @addr: target address for mempolicy
 *
 * Returns the struct page for entry and addr, after queueing swapin.
 *
 * Primitive swap readahead code. We simply read an aligned block of
 * (1 << page_cluster) entries in the swap area. This method is chosen
 * because it doesn't cost us any seek time.  We also make sure to queue
 * the 'original' request together with the readahead ones...
 *
 * This has been extended to use the NUMA policies from the mm triggering
 * the readahead.
 *
 * Caller must hold down_read on the vma->vm_mm if vma is not NULL.
 */
struct page *swapin_readahead(swp_entry_t entry, gfp_t gfp_mask,
			struct vm_area_struct *vma, unsigned long addr)
{
	struct page *page;
	unsigned long offset = swp_offset(entry);
	unsigned long start_offset, end_offset;
	unsigned long mask = (1UL << page_cluster) - 1;
	struct blk_plug plug;

	/* Read a page_cluster sized and aligned cluster around offset. */
	start_offset = offset & ~mask;
	end_offset = offset | mask;
	if (!start_offset)	/* First page is swap header. */
		start_offset++;

	blk_start_plug(&plug);
	for (offset = start_offset; offset <= end_offset ; offset++) {
		/* Ok, do the async read-ahead now */
		page = read_swap_cache_async(swp_entry(swp_type(entry), offset),
						gfp_mask, vma, addr);
		if (!page)
			continue;
		page_cache_release(page);
	}
	blk_finish_plug(&plug);

	lru_add_drain();	/* Push any new pages onto the LRU now */
	return read_swap_cache_async(entry, gfp_mask, vma, addr);
}
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	*/
1da177e4	9	#include <linux/mm.h>
5a0e3ad6	10	#include <linux/gfp.h>
1da177e4 LT	11	#include <linux/kernel_stat.h>
1da177e4 LT	12	#include <linux/swap.h>
46017e95	13	#include <linux/swapops.h>
1da177e4 LT	14	#include <linux/init.h>
1da177e4 LT	15	#include <linux/pagemap.h>
1da177e4	16	#include <linux/backing-dev.h>
3fb5c298	17	#include <linux/blkdev.h>
c484d410	18	#include <linux/pagevec.h>
b20a3503	19	#include <linux/migrate.h>
8c7c6e34	20	#include <linux/page_cgroup.h>
1da177e4 LT	21
	22	#include <asm/pgtable.h>
	23
	24	/*
	25	* swapper_space is a fiction, retained to simplify the path through
7eaceacc	26	* vmscan's shrink_page_list.
1da177e4	27	*/
f5e54d6e	28	static const struct address_space_operations swap_aops = {
1da177e4	29	.writepage = swap_writepage,
62c230bc	30	.set_page_dirty = swap_set_page_dirty,
e965f963	31	.migratepage = migrate_page,
1da177e4 LT	32	};
	33
	34	static struct backing_dev_info swap_backing_dev_info = {
d993831f	35	.name = "swap",
4f98a2fe	36	.capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK \| BDI_CAP_SWAP_BACKED,
1da177e4 LT	37	};
	38
	39	struct address_space swapper_space = {
	40	.page_tree = RADIX_TREE_INIT(GFP_ATOMIC\|__GFP_NOWARN),
19fd6231	41	.tree_lock = __SPIN_LOCK_UNLOCKED(swapper_space.tree_lock),
1da177e4 LT	42	.a_ops = &swap_aops,
	43	.i_mmap_nonlinear = LIST_HEAD_INIT(swapper_space.i_mmap_nonlinear),
	44	.backing_dev_info = &swap_backing_dev_info,
	45	};
1da177e4 LT	46
	47	#define INC_CACHE_INFO(x) do { swap_cache_info.x++; } while (0)
	48
	49	static struct {
	50	unsigned long add_total;
	51	unsigned long del_total;
	52	unsigned long find_success;
	53	unsigned long find_total;
1da177e4 LT	54	} swap_cache_info;
	55
	56	void show_swap_cache_info(void)
	57	{
2c97b7fc JW	58	printk("%lu pages in swap cache\n", total_swapcache_pages);
2c97b7fc JW	59	printk("Swap cache stats: add %lu, delete %lu, find %lu/%lu\n",
1da177e4	60	swap_cache_info.add_total, swap_cache_info.del_total,
bb63be0a	61	swap_cache_info.find_success, swap_cache_info.find_total);
07279cdf	62	printk("Free swap = %ldkB\n", nr_swap_pages << (PAGE_SHIFT - 10));
1da177e4 LT	63	printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
	64	}
	65
	66	/*
31a56396	67	* __add_to_swap_cache resembles add_to_page_cache_locked on swapper_space,
1da177e4 LT	68	* but sets SwapCache flag and private instead of mapping and index.
1da177e4 LT	69	*/
31a56396	70	static int __add_to_swap_cache(struct page *page, swp_entry_t entry)
1da177e4 LT	71	{
	72	int error;
	73
51726b12 HD	74	VM_BUG_ON(!PageLocked(page));
	75	VM_BUG_ON(PageSwapCache(page));
	76	VM_BUG_ON(!PageSwapBacked(page));
	77
31a56396 DN	78	page_cache_get(page);
	79	SetPageSwapCache(page);
	80	set_page_private(page, entry.val);
	81
	82	spin_lock_irq(&swapper_space.tree_lock);
	83	error = radix_tree_insert(&swapper_space.page_tree, entry.val, page);
	84	if (likely(!error)) {
	85	total_swapcache_pages++;
	86	__inc_zone_page_state(page, NR_FILE_PAGES);
	87	INC_CACHE_INFO(add_total);
	88	}
	89	spin_unlock_irq(&swapper_space.tree_lock);
	90
	91	if (unlikely(error)) {
2ca4532a DN	92	/*
	93	* Only the context which have set SWAP_HAS_CACHE flag
	94	* would call add_to_swap_cache().
	95	* So add_to_swap_cache() doesn't returns -EEXIST.
	96	*/
	97	VM_BUG_ON(error == -EEXIST);
31a56396 DN	98	set_page_private(page, 0UL);
	99	ClearPageSwapCache(page);
	100	page_cache_release(page);
	101	}
	102
	103	return error;
	104	}
	105
	106
	107	int add_to_swap_cache(struct page *page, swp_entry_t entry, gfp_t gfp_mask)
	108	{
	109	int error;
	110
35c754d7 BS	111	error = radix_tree_preload(gfp_mask);
35c754d7 BS	112	if (!error) {
31a56396	113	error = __add_to_swap_cache(page, entry);
1da177e4	114	radix_tree_preload_end();
fa1de900	115	}
1da177e4 LT	116	return error;
	117	}
	118
1da177e4 LT	119	/*
	120	* This must be called only on pages that have
	121	* been verified to be in the swap cache.
	122	*/
	123	void __delete_from_swap_cache(struct page *page)
	124	{
51726b12 HD	125	VM_BUG_ON(!PageLocked(page));
	126	VM_BUG_ON(!PageSwapCache(page));
	127	VM_BUG_ON(PageWriteback(page));
1da177e4	128
4c21e2f2 HD	129	radix_tree_delete(&swapper_space.page_tree, page_private(page));
4c21e2f2 HD	130	set_page_private(page, 0);
1da177e4 LT	131	ClearPageSwapCache(page);
1da177e4 LT	132	total_swapcache_pages--;
347ce434	133	__dec_zone_page_state(page, NR_FILE_PAGES);
1da177e4 LT	134	INC_CACHE_INFO(del_total);
	135	}
	136
	137	/**
	138	* add_to_swap - allocate swap space for a page
	139	* @page: page we want to move to swap
	140	*
	141	* Allocate swap space for the page and add the page to the
	142	* swap cache. Caller needs to hold the page lock.
	143	*/
ac47b003	144	int add_to_swap(struct page *page)
1da177e4 LT	145	{
1da177e4 LT	146	swp_entry_t entry;
1da177e4 LT	147	int err;
1da177e4 LT	148
51726b12 HD	149	VM_BUG_ON(!PageLocked(page));
51726b12 HD	150	VM_BUG_ON(!PageUptodate(page));
1da177e4	151
2ca4532a DN	152	entry = get_swap_page();
	153	if (!entry.val)
	154	return 0;
	155
3f04f62f AA	156	if (unlikely(PageTransHuge(page)))
	157	if (unlikely(split_huge_page(page))) {
	158	swapcache_free(entry, NULL);
	159	return 0;
	160	}
	161
2ca4532a DN	162	/*
	163	* Radix-tree node allocations from PF_MEMALLOC contexts could
	164	* completely exhaust the page allocator. __GFP_NOMEMALLOC
	165	* stops emergency reserves from being allocated.
	166	*
	167	* TODO: this could cause a theoretical memory reclaim
	168	* deadlock in the swap out path.
	169	*/
	170	/*
	171	* Add it to the swap cache and mark it dirty
	172	*/
	173	err = add_to_swap_cache(page, entry,
	174	__GFP_HIGH\|__GFP_NOMEMALLOC\|__GFP_NOWARN);
	175
	176	if (!err) { /* Success */
	177	SetPageDirty(page);
	178	return 1;
	179	} else { /* -ENOMEM radix-tree allocation failure */
bd53b714	180	/*
2ca4532a DN	181	* add_to_swap_cache() doesn't return -EEXIST, so we can safely
2ca4532a DN	182	* clear SWAP_HAS_CACHE flag.
1da177e4	183	*/
2ca4532a DN	184	swapcache_free(entry, NULL);
2ca4532a DN	185	return 0;
1da177e4 LT	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	200
19fd6231	201	spin_lock_irq(&swapper_space.tree_lock);
1da177e4	202	__delete_from_swap_cache(page);
19fd6231	203	spin_unlock_irq(&swapper_space.tree_lock);
1da177e4	204
cb4b86ba	205	swapcache_free(entry, page);
1da177e4 LT	206	page_cache_release(page);
	207	}
	208
1da177e4 LT	209	/*
	210	* If we are the only user, then try to free up the swap cache.
	211	*
	212	* Its ok to check for PageSwapCache without the page lock
a2c43eed HD	213	* here because we are going to recheck again inside
a2c43eed HD	214	* try_to_free_swap() _with_ the lock.
1da177e4 LT	215	* - Marcelo
	216	*/
	217	static inline void free_swap_cache(struct page *page)
	218	{
a2c43eed HD	219	if (PageSwapCache(page) && !page_mapped(page) && trylock_page(page)) {
a2c43eed HD	220	try_to_free_swap(page);
1da177e4 LT	221	unlock_page(page);
	222	}
	223	}
	224
	225	/*
	226	* Perform a free_page(), also freeing any swap cache associated with
b8072f09	227	* this page if it is the last user of the page.
1da177e4 LT	228	*/
	229	void free_page_and_swap_cache(struct page *page)
	230	{
	231	free_swap_cache(page);
	232	page_cache_release(page);
	233	}
	234
	235	/*
	236	* Passed an array of pages, drop them all from swapcache and then release
	237	* them. They are removed from the LRU and freed if this is their last use.
	238	*/
	239	void free_pages_and_swap_cache(struct page **pages, int nr)
	240	{
1da177e4 LT	241	struct page **pagep = pages;
	242
	243	lru_add_drain();
	244	while (nr) {
c484d410	245	int todo = min(nr, PAGEVEC_SIZE);
1da177e4 LT	246	int i;
	247
	248	for (i = 0; i < todo; i++)
	249	free_swap_cache(pagep[i]);
	250	release_pages(pagep, todo, 0);
	251	pagep += todo;
	252	nr -= todo;
	253	}
	254	}
	255
	256	/*
	257	* Lookup a swap entry in the swap cache. A found page will be returned
	258	* unlocked and with its refcount incremented - we rely on the kernel
	259	* lock getting page table operations atomic even if we drop the page
	260	* lock before returning.
	261	*/
	262	struct page * lookup_swap_cache(swp_entry_t entry)
	263	{
	264	struct page *page;
	265
	266	page = find_get_page(&swapper_space, entry.val);
	267
	268	if (page)
	269	INC_CACHE_INFO(find_success);
	270
	271	INC_CACHE_INFO(find_total);
	272	return page;
	273	}
	274
	275	/*
	276	* Locate a page of swap in physical memory, reserving swap cache space
	277	* and reading the disk if it is not already cached.
	278	* A failure return means that either the page allocation failed or that
	279	* the swap entry is no longer in use.
	280	*/
02098fea	281	struct page *read_swap_cache_async(swp_entry_t entry, gfp_t gfp_mask,
1da177e4 LT	282	struct vm_area_struct *vma, unsigned long addr)
	283	{
	284	struct page found_page, new_page = NULL;
	285	int err;
	286
	287	do {
	288	/*
	289	* First check the swap cache. Since this is normally
	290	* called after lookup_swap_cache() failed, re-calling
	291	* that would confuse statistics.
	292	*/
	293	found_page = find_get_page(&swapper_space, entry.val);
	294	if (found_page)
	295	break;
	296
	297	/*
	298	* Get a new page to read into from swap.
	299	*/
	300	if (!new_page) {
02098fea	301	new_page = alloc_page_vma(gfp_mask, vma, addr);
1da177e4 LT	302	if (!new_page)
	303	break; /* Out of memory */
	304	}
	305
31a56396 DN	306	/*
	307	* call radix_tree_preload() while we can wait.
	308	*/
	309	err = radix_tree_preload(gfp_mask & GFP_KERNEL);
	310	if (err)
	311	break;
	312
f000944d HD	313	/*
	314	* Swap entry may have been freed since our caller observed it.
	315	*/
355cfa73	316	err = swapcache_prepare(entry);
31a56396 DN	317	if (err == -EEXIST) { /* seems racy */
31a56396 DN	318	radix_tree_preload_end();
355cfa73	319	continue;
31a56396 DN	320	}
	321	if (err) { /* swp entry is obsolete ? */
	322	radix_tree_preload_end();
f000944d	323	break;
31a56396	324	}
f000944d	325
2ca4532a	326	/* May fail (-ENOMEM) if radix-tree node allocation failed. */
f45840b5	327	__set_page_locked(new_page);
b2e18538	328	SetPageSwapBacked(new_page);
31a56396	329	err = __add_to_swap_cache(new_page, entry);
529ae9aa	330	if (likely(!err)) {
31a56396	331	radix_tree_preload_end();
1da177e4 LT	332	/*
	333	* Initiate read into locked page and return.
	334	*/
c5fdae46	335	lru_cache_add_anon(new_page);
aca8bf32	336	swap_readpage(new_page);
1da177e4 LT	337	return new_page;
1da177e4 LT	338	}
31a56396	339	radix_tree_preload_end();
b2e18538	340	ClearPageSwapBacked(new_page);
f45840b5	341	__clear_page_locked(new_page);
2ca4532a DN	342	/*
	343	* add_to_swap_cache() doesn't return -EEXIST, so we can safely
	344	* clear SWAP_HAS_CACHE flag.
	345	*/
cb4b86ba	346	swapcache_free(entry, NULL);
f000944d	347	} while (err != -ENOMEM);
1da177e4 LT	348
	349	if (new_page)
	350	page_cache_release(new_page);
	351	return found_page;
	352	}
46017e95 HD	353
	354	/**
	355	* swapin_readahead - swap in pages in hope we need them soon
	356	* @entry: swap entry of this memory
7682486b	357	* @gfp_mask: memory allocation flags
46017e95 HD	358	* @vma: user vma this address belongs to
	359	* @addr: target address for mempolicy
	360	*
	361	* Returns the struct page for entry and addr, after queueing swapin.
	362	*
	363	* Primitive swap readahead code. We simply read an aligned block of
	364	* (1 << page_cluster) entries in the swap area. This method is chosen
	365	* because it doesn't cost us any seek time. We also make sure to queue
	366	* the 'original' request together with the readahead ones...
	367	*
	368	* This has been extended to use the NUMA policies from the mm triggering
	369	* the readahead.
	370	*
	371	* Caller must hold down_read on the vma->vm_mm if vma is not NULL.
	372	*/
02098fea	373	struct page *swapin_readahead(swp_entry_t entry, gfp_t gfp_mask,
46017e95 HD	374	struct vm_area_struct *vma, unsigned long addr)
46017e95 HD	375	{
46017e95	376	struct page *page;
67f96aa2 RR	377	unsigned long offset = swp_offset(entry);
	378	unsigned long start_offset, end_offset;
	379	unsigned long mask = (1UL << page_cluster) - 1;
3fb5c298	380	struct blk_plug plug;
46017e95	381
67f96aa2 RR	382	/* Read a page_cluster sized and aligned cluster around offset. */
	383	start_offset = offset & ~mask;
	384	end_offset = offset \| mask;
	385	if (!start_offset) /* First page is swap header. */
	386	start_offset++;
	387
3fb5c298	388	blk_start_plug(&plug);
67f96aa2	389	for (offset = start_offset; offset <= end_offset ; offset++) {
46017e95 HD	390	/* Ok, do the async read-ahead now */
46017e95 HD	391	page = read_swap_cache_async(swp_entry(swp_type(entry), offset),
02098fea	392	gfp_mask, vma, addr);
46017e95	393	if (!page)
67f96aa2	394	continue;
46017e95 HD	395	page_cache_release(page);
46017e95 HD	396	}
3fb5c298 CE	397	blk_finish_plug(&plug);
3fb5c298 CE	398
46017e95	399	lru_add_drain(); /* Push any new pages onto the LRU now */
02098fea	400	return read_swap_cache_async(entry, gfp_mask, vma, addr);
46017e95	401	}