[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/module.h>
#include <linux/mm.h>
#include <linux/kernel_stat.h>
#include <linux/swap.h>
#include <linux/swapops.h>
#include <linux/init.h>
#include <linux/pagemap.h>
#include <linux/buffer_head.h>
#include <linux/backing-dev.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, to make sync_page look nicer, and to allow
 * future use of radix_tree tags in the swap cache.
 */
static const struct address_space_operations swap_aops = {
	.writepage	= swap_writepage,
	.sync_page	= block_sync_page,
	.set_page_dirty	= __set_page_dirty_nobuffers,
	.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,
	.unplug_io_fn	= swap_unplug_io_fn,
};

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;

	/*
	 * 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)
{
	int nr_pages;
	struct page *page;
	unsigned long offset;
	unsigned long end_offset;

	/*
	 * Get starting offset for readaround, and number of pages to read.
	 * Adjust starting address by readbehind (for NUMA interleave case)?
	 * No, it's very unlikely that swap layout would follow vma layout,
	 * more likely that neighbouring swap pages came from the same node:
	 * so use the same "addr" to choose the same node for each swap read.
	 */
	nr_pages = valid_swaphandles(entry, &offset);
	for (end_offset = offset + nr_pages; 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)
			break;
		page_cache_release(page);
	}
	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	*/
	9	#include <linux/module.h>
	10	#include <linux/mm.h>
	11	#include <linux/kernel_stat.h>
	12	#include <linux/swap.h>
46017e95	13	#include <linux/swapops.h>
1da177e4 LT	14	#include <linux/init.h>
	15	#include <linux/pagemap.h>
	16	#include <linux/buffer_head.h>
	17	#include <linux/backing-dev.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
2706a1b8	26	* vmscan's shrink_page_list, to make sync_page look nicer, and to allow
1da177e4 LT	27	* future use of radix_tree tags in the swap cache.
1da177e4 LT	28	*/
f5e54d6e	29	static const struct address_space_operations swap_aops = {
1da177e4 LT	30	.writepage = swap_writepage,
	31	.sync_page = block_sync_page,
	32	.set_page_dirty = __set_page_dirty_nobuffers,
e965f963	33	.migratepage = migrate_page,
1da177e4 LT	34	};
	35
	36	static struct backing_dev_info swap_backing_dev_info = {
d993831f	37	.name = "swap",
4f98a2fe	38	.capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK \| BDI_CAP_SWAP_BACKED,
1da177e4 LT	39	.unplug_io_fn = swap_unplug_io_fn,
	40	};
	41
	42	struct address_space swapper_space = {
	43	.page_tree = RADIX_TREE_INIT(GFP_ATOMIC\|__GFP_NOWARN),
19fd6231	44	.tree_lock = __SPIN_LOCK_UNLOCKED(swapper_space.tree_lock),
1da177e4 LT	45	.a_ops = &swap_aops,
	46	.i_mmap_nonlinear = LIST_HEAD_INIT(swapper_space.i_mmap_nonlinear),
	47	.backing_dev_info = &swap_backing_dev_info,
	48	};
1da177e4 LT	49
	50	#define INC_CACHE_INFO(x) do { swap_cache_info.x++; } while (0)
	51
	52	static struct {
	53	unsigned long add_total;
	54	unsigned long del_total;
	55	unsigned long find_success;
	56	unsigned long find_total;
1da177e4 LT	57	} swap_cache_info;
	58
	59	void show_swap_cache_info(void)
	60	{
2c97b7fc JW	61	printk("%lu pages in swap cache\n", total_swapcache_pages);
2c97b7fc JW	62	printk("Swap cache stats: add %lu, delete %lu, find %lu/%lu\n",
1da177e4	63	swap_cache_info.add_total, swap_cache_info.del_total,
bb63be0a	64	swap_cache_info.find_success, swap_cache_info.find_total);
07279cdf	65	printk("Free swap = %ldkB\n", nr_swap_pages << (PAGE_SHIFT - 10));
1da177e4 LT	66	printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
	67	}
	68
	69	/*
31a56396	70	* __add_to_swap_cache resembles add_to_page_cache_locked on swapper_space,
1da177e4 LT	71	* but sets SwapCache flag and private instead of mapping and index.
1da177e4 LT	72	*/
31a56396	73	static int __add_to_swap_cache(struct page *page, swp_entry_t entry)
1da177e4 LT	74	{
	75	int error;
	76
51726b12 HD	77	VM_BUG_ON(!PageLocked(page));
	78	VM_BUG_ON(PageSwapCache(page));
	79	VM_BUG_ON(!PageSwapBacked(page));
	80
31a56396 DN	81	page_cache_get(page);
	82	SetPageSwapCache(page);
	83	set_page_private(page, entry.val);
	84
	85	spin_lock_irq(&swapper_space.tree_lock);
	86	error = radix_tree_insert(&swapper_space.page_tree, entry.val, page);
	87	if (likely(!error)) {
	88	total_swapcache_pages++;
	89	__inc_zone_page_state(page, NR_FILE_PAGES);
	90	INC_CACHE_INFO(add_total);
	91	}
	92	spin_unlock_irq(&swapper_space.tree_lock);
	93
	94	if (unlikely(error)) {
2ca4532a DN	95	/*
	96	* Only the context which have set SWAP_HAS_CACHE flag
	97	* would call add_to_swap_cache().
	98	* So add_to_swap_cache() doesn't returns -EEXIST.
	99	*/
	100	VM_BUG_ON(error == -EEXIST);
31a56396 DN	101	set_page_private(page, 0UL);
	102	ClearPageSwapCache(page);
	103	page_cache_release(page);
	104	}
	105
	106	return error;
	107	}
	108
	109
	110	int add_to_swap_cache(struct page *page, swp_entry_t entry, gfp_t gfp_mask)
	111	{
	112	int error;
	113
35c754d7 BS	114	error = radix_tree_preload(gfp_mask);
35c754d7 BS	115	if (!error) {
31a56396	116	error = __add_to_swap_cache(page, entry);
1da177e4	117	radix_tree_preload_end();
fa1de900	118	}
1da177e4 LT	119	return error;
	120	}
	121
1da177e4 LT	122	/*
	123	* This must be called only on pages that have
	124	* been verified to be in the swap cache.
	125	*/
	126	void __delete_from_swap_cache(struct page *page)
	127	{
51726b12 HD	128	VM_BUG_ON(!PageLocked(page));
	129	VM_BUG_ON(!PageSwapCache(page));
	130	VM_BUG_ON(PageWriteback(page));
1da177e4	131
4c21e2f2 HD	132	radix_tree_delete(&swapper_space.page_tree, page_private(page));
4c21e2f2 HD	133	set_page_private(page, 0);
1da177e4 LT	134	ClearPageSwapCache(page);
1da177e4 LT	135	total_swapcache_pages--;
347ce434	136	__dec_zone_page_state(page, NR_FILE_PAGES);
1da177e4 LT	137	INC_CACHE_INFO(del_total);
	138	}
	139
	140	/**
	141	* add_to_swap - allocate swap space for a page
	142	* @page: page we want to move to swap
	143	*
	144	* Allocate swap space for the page and add the page to the
	145	* swap cache. Caller needs to hold the page lock.
	146	*/
ac47b003	147	int add_to_swap(struct page *page)
1da177e4 LT	148	{
1da177e4 LT	149	swp_entry_t entry;
1da177e4 LT	150	int err;
1da177e4 LT	151
51726b12 HD	152	VM_BUG_ON(!PageLocked(page));
51726b12 HD	153	VM_BUG_ON(!PageUptodate(page));
1da177e4	154
2ca4532a DN	155	entry = get_swap_page();
	156	if (!entry.val)
	157	return 0;
	158
	159	/*
	160	* Radix-tree node allocations from PF_MEMALLOC contexts could
	161	* completely exhaust the page allocator. __GFP_NOMEMALLOC
	162	* stops emergency reserves from being allocated.
	163	*
	164	* TODO: this could cause a theoretical memory reclaim
	165	* deadlock in the swap out path.
	166	*/
	167	/*
	168	* Add it to the swap cache and mark it dirty
	169	*/
	170	err = add_to_swap_cache(page, entry,
	171	__GFP_HIGH\|__GFP_NOMEMALLOC\|__GFP_NOWARN);
	172
	173	if (!err) { /* Success */
	174	SetPageDirty(page);
	175	return 1;
	176	} else { /* -ENOMEM radix-tree allocation failure */
bd53b714	177	/*
2ca4532a DN	178	* add_to_swap_cache() doesn't return -EEXIST, so we can safely
2ca4532a DN	179	* clear SWAP_HAS_CACHE flag.
1da177e4	180	*/
2ca4532a DN	181	swapcache_free(entry, NULL);
2ca4532a DN	182	return 0;
1da177e4 LT	183	}
	184	}
	185
	186	/*
	187	* This must be called only on pages that have
	188	* been verified to be in the swap cache and locked.
	189	* It will never put the page into the free list,
	190	* the caller has a reference on the page.
	191	*/
	192	void delete_from_swap_cache(struct page *page)
	193	{
	194	swp_entry_t entry;
	195
4c21e2f2	196	entry.val = page_private(page);
1da177e4	197
19fd6231	198	spin_lock_irq(&swapper_space.tree_lock);
1da177e4	199	__delete_from_swap_cache(page);
19fd6231	200	spin_unlock_irq(&swapper_space.tree_lock);
1da177e4	201
cb4b86ba	202	swapcache_free(entry, page);
1da177e4 LT	203	page_cache_release(page);
	204	}
	205
1da177e4 LT	206	/*
	207	* If we are the only user, then try to free up the swap cache.
	208	*
	209	* Its ok to check for PageSwapCache without the page lock
a2c43eed HD	210	* here because we are going to recheck again inside
a2c43eed HD	211	* try_to_free_swap() _with_ the lock.
1da177e4 LT	212	* - Marcelo
	213	*/
	214	static inline void free_swap_cache(struct page *page)
	215	{
a2c43eed HD	216	if (PageSwapCache(page) && !page_mapped(page) && trylock_page(page)) {
a2c43eed HD	217	try_to_free_swap(page);
1da177e4 LT	218	unlock_page(page);
	219	}
	220	}
	221
	222	/*
	223	* Perform a free_page(), also freeing any swap cache associated with
b8072f09	224	* this page if it is the last user of the page.
1da177e4 LT	225	*/
	226	void free_page_and_swap_cache(struct page *page)
	227	{
	228	free_swap_cache(page);
	229	page_cache_release(page);
	230	}
	231
	232	/*
	233	* Passed an array of pages, drop them all from swapcache and then release
	234	* them. They are removed from the LRU and freed if this is their last use.
	235	*/
	236	void free_pages_and_swap_cache(struct page **pages, int nr)
	237	{
1da177e4 LT	238	struct page **pagep = pages;
	239
	240	lru_add_drain();
	241	while (nr) {
c484d410	242	int todo = min(nr, PAGEVEC_SIZE);
1da177e4 LT	243	int i;
	244
	245	for (i = 0; i < todo; i++)
	246	free_swap_cache(pagep[i]);
	247	release_pages(pagep, todo, 0);
	248	pagep += todo;
	249	nr -= todo;
	250	}
	251	}
	252
	253	/*
	254	* Lookup a swap entry in the swap cache. A found page will be returned
	255	* unlocked and with its refcount incremented - we rely on the kernel
	256	* lock getting page table operations atomic even if we drop the page
	257	* lock before returning.
	258	*/
	259	struct page * lookup_swap_cache(swp_entry_t entry)
	260	{
	261	struct page *page;
	262
	263	page = find_get_page(&swapper_space, entry.val);
	264
	265	if (page)
	266	INC_CACHE_INFO(find_success);
	267
	268	INC_CACHE_INFO(find_total);
	269	return page;
	270	}
	271
	272	/*
	273	* Locate a page of swap in physical memory, reserving swap cache space
	274	* and reading the disk if it is not already cached.
	275	* A failure return means that either the page allocation failed or that
	276	* the swap entry is no longer in use.
	277	*/
02098fea	278	struct page *read_swap_cache_async(swp_entry_t entry, gfp_t gfp_mask,
1da177e4 LT	279	struct vm_area_struct *vma, unsigned long addr)
	280	{
	281	struct page found_page, new_page = NULL;
	282	int err;
	283
	284	do {
	285	/*
	286	* First check the swap cache. Since this is normally
	287	* called after lookup_swap_cache() failed, re-calling
	288	* that would confuse statistics.
	289	*/
	290	found_page = find_get_page(&swapper_space, entry.val);
	291	if (found_page)
	292	break;
	293
	294	/*
	295	* Get a new page to read into from swap.
	296	*/
	297	if (!new_page) {
02098fea	298	new_page = alloc_page_vma(gfp_mask, vma, addr);
1da177e4 LT	299	if (!new_page)
	300	break; /* Out of memory */
	301	}
	302
31a56396 DN	303	/*
	304	* call radix_tree_preload() while we can wait.
	305	*/
	306	err = radix_tree_preload(gfp_mask & GFP_KERNEL);
	307	if (err)
	308	break;
	309
f000944d HD	310	/*
	311	* Swap entry may have been freed since our caller observed it.
	312	*/
355cfa73	313	err = swapcache_prepare(entry);
31a56396 DN	314	if (err == -EEXIST) { /* seems racy */
31a56396 DN	315	radix_tree_preload_end();
355cfa73	316	continue;
31a56396 DN	317	}
	318	if (err) { /* swp entry is obsolete ? */
	319	radix_tree_preload_end();
f000944d	320	break;
31a56396	321	}
f000944d	322
2ca4532a	323	/* May fail (-ENOMEM) if radix-tree node allocation failed. */
f45840b5	324	__set_page_locked(new_page);
b2e18538	325	SetPageSwapBacked(new_page);
31a56396	326	err = __add_to_swap_cache(new_page, entry);
529ae9aa	327	if (likely(!err)) {
31a56396	328	radix_tree_preload_end();
1da177e4 LT	329	/*
	330	* Initiate read into locked page and return.
	331	*/
c5fdae46	332	lru_cache_add_anon(new_page);
aca8bf32	333	swap_readpage(new_page);
1da177e4 LT	334	return new_page;
1da177e4 LT	335	}
31a56396	336	radix_tree_preload_end();
b2e18538	337	ClearPageSwapBacked(new_page);
f45840b5	338	__clear_page_locked(new_page);
2ca4532a DN	339	/*
	340	* add_to_swap_cache() doesn't return -EEXIST, so we can safely
	341	* clear SWAP_HAS_CACHE flag.
	342	*/
cb4b86ba	343	swapcache_free(entry, NULL);
f000944d	344	} while (err != -ENOMEM);
1da177e4 LT	345
	346	if (new_page)
	347	page_cache_release(new_page);
	348	return found_page;
	349	}
46017e95 HD	350
	351	/**
	352	* swapin_readahead - swap in pages in hope we need them soon
	353	* @entry: swap entry of this memory
7682486b	354	* @gfp_mask: memory allocation flags
46017e95 HD	355	* @vma: user vma this address belongs to
	356	* @addr: target address for mempolicy
	357	*
	358	* Returns the struct page for entry and addr, after queueing swapin.
	359	*
	360	* Primitive swap readahead code. We simply read an aligned block of
	361	* (1 << page_cluster) entries in the swap area. This method is chosen
	362	* because it doesn't cost us any seek time. We also make sure to queue
	363	* the 'original' request together with the readahead ones...
	364	*
	365	* This has been extended to use the NUMA policies from the mm triggering
	366	* the readahead.
	367	*
	368	* Caller must hold down_read on the vma->vm_mm if vma is not NULL.
	369	*/
02098fea	370	struct page *swapin_readahead(swp_entry_t entry, gfp_t gfp_mask,
46017e95 HD	371	struct vm_area_struct *vma, unsigned long addr)
	372	{
	373	int nr_pages;
	374	struct page *page;
	375	unsigned long offset;
	376	unsigned long end_offset;
	377
	378	/*
	379	* Get starting offset for readaround, and number of pages to read.
	380	* Adjust starting address by readbehind (for NUMA interleave case)?
	381	* No, it's very unlikely that swap layout would follow vma layout,
	382	* more likely that neighbouring swap pages came from the same node:
	383	* so use the same "addr" to choose the same node for each swap read.
	384	*/
	385	nr_pages = valid_swaphandles(entry, &offset);
	386	for (end_offset = offset + nr_pages; offset < end_offset; offset++) {
	387	/* Ok, do the async read-ahead now */
	388	page = read_swap_cache_async(swp_entry(swp_type(entry), offset),
02098fea	389	gfp_mask, vma, addr);
46017e95 HD	390	if (!page)
	391	break;
	392	page_cache_release(page);
	393	}
	394	lru_add_drain(); /* Push any new pages onto the LRU now */
02098fea	395	return read_swap_cache_async(entry, gfp_mask, vma, addr);
46017e95	396	}