[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/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/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;

	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)
{
	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>
5a0e3ad6	11	#include <linux/gfp.h>
1da177e4 LT	12	#include <linux/kernel_stat.h>
1da177e4 LT	13	#include <linux/swap.h>
46017e95	14	#include <linux/swapops.h>
1da177e4 LT	15	#include <linux/init.h>
	16	#include <linux/pagemap.h>
	17	#include <linux/buffer_head.h>
	18	#include <linux/backing-dev.h>
c484d410	19	#include <linux/pagevec.h>
b20a3503	20	#include <linux/migrate.h>
8c7c6e34	21	#include <linux/page_cgroup.h>
1da177e4 LT	22
	23	#include <asm/pgtable.h>
	24
	25	/*
	26	* swapper_space is a fiction, retained to simplify the path through
2706a1b8	27	* vmscan's shrink_page_list, to make sync_page look nicer, and to allow
1da177e4 LT	28	* future use of radix_tree tags in the swap cache.
1da177e4 LT	29	*/
f5e54d6e	30	static const struct address_space_operations swap_aops = {
1da177e4 LT	31	.writepage = swap_writepage,
	32	.sync_page = block_sync_page,
	33	.set_page_dirty = __set_page_dirty_nobuffers,
e965f963	34	.migratepage = migrate_page,
1da177e4 LT	35	};
	36
	37	static struct backing_dev_info swap_backing_dev_info = {
d993831f	38	.name = "swap",
4f98a2fe	39	.capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK \| BDI_CAP_SWAP_BACKED,
1da177e4 LT	40	.unplug_io_fn = swap_unplug_io_fn,
	41	};
	42
	43	struct address_space swapper_space = {
	44	.page_tree = RADIX_TREE_INIT(GFP_ATOMIC\|__GFP_NOWARN),
19fd6231	45	.tree_lock = __SPIN_LOCK_UNLOCKED(swapper_space.tree_lock),
1da177e4 LT	46	.a_ops = &swap_aops,
	47	.i_mmap_nonlinear = LIST_HEAD_INIT(swapper_space.i_mmap_nonlinear),
	48	.backing_dev_info = &swap_backing_dev_info,
	49	};
1da177e4 LT	50
	51	#define INC_CACHE_INFO(x) do { swap_cache_info.x++; } while (0)
	52
	53	static struct {
	54	unsigned long add_total;
	55	unsigned long del_total;
	56	unsigned long find_success;
	57	unsigned long find_total;
1da177e4 LT	58	} swap_cache_info;
	59
	60	void show_swap_cache_info(void)
	61	{
2c97b7fc JW	62	printk("%lu pages in swap cache\n", total_swapcache_pages);
2c97b7fc JW	63	printk("Swap cache stats: add %lu, delete %lu, find %lu/%lu\n",
1da177e4	64	swap_cache_info.add_total, swap_cache_info.del_total,
bb63be0a	65	swap_cache_info.find_success, swap_cache_info.find_total);
07279cdf	66	printk("Free swap = %ldkB\n", nr_swap_pages << (PAGE_SHIFT - 10));
1da177e4 LT	67	printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
	68	}
	69
	70	/*
31a56396	71	* __add_to_swap_cache resembles add_to_page_cache_locked on swapper_space,
1da177e4 LT	72	* but sets SwapCache flag and private instead of mapping and index.
1da177e4 LT	73	*/
31a56396	74	static int __add_to_swap_cache(struct page *page, swp_entry_t entry)
1da177e4 LT	75	{
	76	int error;
	77
51726b12 HD	78	VM_BUG_ON(!PageLocked(page));
	79	VM_BUG_ON(PageSwapCache(page));
	80	VM_BUG_ON(!PageSwapBacked(page));
	81
31a56396 DN	82	page_cache_get(page);
	83	SetPageSwapCache(page);
	84	set_page_private(page, entry.val);
	85
	86	spin_lock_irq(&swapper_space.tree_lock);
	87	error = radix_tree_insert(&swapper_space.page_tree, entry.val, page);
	88	if (likely(!error)) {
	89	total_swapcache_pages++;
	90	__inc_zone_page_state(page, NR_FILE_PAGES);
	91	INC_CACHE_INFO(add_total);
	92	}
	93	spin_unlock_irq(&swapper_space.tree_lock);
	94
	95	if (unlikely(error)) {
2ca4532a DN	96	/*
	97	* Only the context which have set SWAP_HAS_CACHE flag
	98	* would call add_to_swap_cache().
	99	* So add_to_swap_cache() doesn't returns -EEXIST.
	100	*/
	101	VM_BUG_ON(error == -EEXIST);
31a56396 DN	102	set_page_private(page, 0UL);
	103	ClearPageSwapCache(page);
	104	page_cache_release(page);
	105	}
	106
	107	return error;
	108	}
	109
	110
	111	int add_to_swap_cache(struct page *page, swp_entry_t entry, gfp_t gfp_mask)
	112	{
	113	int error;
	114
35c754d7 BS	115	error = radix_tree_preload(gfp_mask);
35c754d7 BS	116	if (!error) {
31a56396	117	error = __add_to_swap_cache(page, entry);
1da177e4	118	radix_tree_preload_end();
fa1de900	119	}
1da177e4 LT	120	return error;
	121	}
	122
1da177e4 LT	123	/*
	124	* This must be called only on pages that have
	125	* been verified to be in the swap cache.
	126	*/
	127	void __delete_from_swap_cache(struct page *page)
	128	{
51726b12 HD	129	VM_BUG_ON(!PageLocked(page));
	130	VM_BUG_ON(!PageSwapCache(page));
	131	VM_BUG_ON(PageWriteback(page));
1da177e4	132
4c21e2f2 HD	133	radix_tree_delete(&swapper_space.page_tree, page_private(page));
4c21e2f2 HD	134	set_page_private(page, 0);
1da177e4 LT	135	ClearPageSwapCache(page);
1da177e4 LT	136	total_swapcache_pages--;
347ce434	137	__dec_zone_page_state(page, NR_FILE_PAGES);
1da177e4 LT	138	INC_CACHE_INFO(del_total);
	139	}
	140
	141	/**
	142	* add_to_swap - allocate swap space for a page
	143	* @page: page we want to move to swap
	144	*
	145	* Allocate swap space for the page and add the page to the
	146	* swap cache. Caller needs to hold the page lock.
	147	*/
ac47b003	148	int add_to_swap(struct page *page)
1da177e4 LT	149	{
1da177e4 LT	150	swp_entry_t entry;
1da177e4 LT	151	int err;
1da177e4 LT	152
51726b12 HD	153	VM_BUG_ON(!PageLocked(page));
51726b12 HD	154	VM_BUG_ON(!PageUptodate(page));
1da177e4	155
2ca4532a DN	156	entry = get_swap_page();
	157	if (!entry.val)
	158	return 0;
	159
3f04f62f AA	160	if (unlikely(PageTransHuge(page)))
	161	if (unlikely(split_huge_page(page))) {
	162	swapcache_free(entry, NULL);
	163	return 0;
	164	}
	165
2ca4532a DN	166	/*
	167	* Radix-tree node allocations from PF_MEMALLOC contexts could
	168	* completely exhaust the page allocator. __GFP_NOMEMALLOC
	169	* stops emergency reserves from being allocated.
	170	*
	171	* TODO: this could cause a theoretical memory reclaim
	172	* deadlock in the swap out path.
	173	*/
	174	/*
	175	* Add it to the swap cache and mark it dirty
	176	*/
	177	err = add_to_swap_cache(page, entry,
	178	__GFP_HIGH\|__GFP_NOMEMALLOC\|__GFP_NOWARN);
	179
	180	if (!err) { /* Success */
	181	SetPageDirty(page);
	182	return 1;
	183	} else { /* -ENOMEM radix-tree allocation failure */
bd53b714	184	/*
2ca4532a DN	185	* add_to_swap_cache() doesn't return -EEXIST, so we can safely
2ca4532a DN	186	* clear SWAP_HAS_CACHE flag.
1da177e4	187	*/
2ca4532a DN	188	swapcache_free(entry, NULL);
2ca4532a DN	189	return 0;
1da177e4 LT	190	}
	191	}
	192
	193	/*
	194	* This must be called only on pages that have
	195	* been verified to be in the swap cache and locked.
	196	* It will never put the page into the free list,
	197	* the caller has a reference on the page.
	198	*/
	199	void delete_from_swap_cache(struct page *page)
	200	{
	201	swp_entry_t entry;
	202
4c21e2f2	203	entry.val = page_private(page);
1da177e4	204
19fd6231	205	spin_lock_irq(&swapper_space.tree_lock);
1da177e4	206	__delete_from_swap_cache(page);
19fd6231	207	spin_unlock_irq(&swapper_space.tree_lock);
1da177e4	208
cb4b86ba	209	swapcache_free(entry, page);
1da177e4 LT	210	page_cache_release(page);
	211	}
	212
1da177e4 LT	213	/*
	214	* If we are the only user, then try to free up the swap cache.
	215	*
	216	* Its ok to check for PageSwapCache without the page lock
a2c43eed HD	217	* here because we are going to recheck again inside
a2c43eed HD	218	* try_to_free_swap() _with_ the lock.
1da177e4 LT	219	* - Marcelo
	220	*/
	221	static inline void free_swap_cache(struct page *page)
	222	{
a2c43eed HD	223	if (PageSwapCache(page) && !page_mapped(page) && trylock_page(page)) {
a2c43eed HD	224	try_to_free_swap(page);
1da177e4 LT	225	unlock_page(page);
	226	}
	227	}
	228
	229	/*
	230	* Perform a free_page(), also freeing any swap cache associated with
b8072f09	231	* this page if it is the last user of the page.
1da177e4 LT	232	*/
	233	void free_page_and_swap_cache(struct page *page)
	234	{
	235	free_swap_cache(page);
	236	page_cache_release(page);
	237	}
	238
	239	/*
	240	* Passed an array of pages, drop them all from swapcache and then release
	241	* them. They are removed from the LRU and freed if this is their last use.
	242	*/
	243	void free_pages_and_swap_cache(struct page **pages, int nr)
	244	{
1da177e4 LT	245	struct page **pagep = pages;
	246
	247	lru_add_drain();
	248	while (nr) {
c484d410	249	int todo = min(nr, PAGEVEC_SIZE);
1da177e4 LT	250	int i;
	251
	252	for (i = 0; i < todo; i++)
	253	free_swap_cache(pagep[i]);
	254	release_pages(pagep, todo, 0);
	255	pagep += todo;
	256	nr -= todo;
	257	}
	258	}
	259
	260	/*
	261	* Lookup a swap entry in the swap cache. A found page will be returned
	262	* unlocked and with its refcount incremented - we rely on the kernel
	263	* lock getting page table operations atomic even if we drop the page
	264	* lock before returning.
	265	*/
	266	struct page * lookup_swap_cache(swp_entry_t entry)
	267	{
	268	struct page *page;
	269
	270	page = find_get_page(&swapper_space, entry.val);
	271
	272	if (page)
	273	INC_CACHE_INFO(find_success);
	274
	275	INC_CACHE_INFO(find_total);
	276	return page;
	277	}
	278
	279	/*
	280	* Locate a page of swap in physical memory, reserving swap cache space
	281	* and reading the disk if it is not already cached.
	282	* A failure return means that either the page allocation failed or that
	283	* the swap entry is no longer in use.
	284	*/
02098fea	285	struct page *read_swap_cache_async(swp_entry_t entry, gfp_t gfp_mask,
1da177e4 LT	286	struct vm_area_struct *vma, unsigned long addr)
	287	{
	288	struct page found_page, new_page = NULL;
	289	int err;
	290
	291	do {
	292	/*
	293	* First check the swap cache. Since this is normally
	294	* called after lookup_swap_cache() failed, re-calling
	295	* that would confuse statistics.
	296	*/
	297	found_page = find_get_page(&swapper_space, entry.val);
	298	if (found_page)
	299	break;
	300
	301	/*
	302	* Get a new page to read into from swap.
	303	*/
	304	if (!new_page) {
02098fea	305	new_page = alloc_page_vma(gfp_mask, vma, addr);
1da177e4 LT	306	if (!new_page)
	307	break; /* Out of memory */
	308	}
	309
31a56396 DN	310	/*
	311	* call radix_tree_preload() while we can wait.
	312	*/
	313	err = radix_tree_preload(gfp_mask & GFP_KERNEL);
	314	if (err)
	315	break;
	316
f000944d HD	317	/*
	318	* Swap entry may have been freed since our caller observed it.
	319	*/
355cfa73	320	err = swapcache_prepare(entry);
31a56396 DN	321	if (err == -EEXIST) { /* seems racy */
31a56396 DN	322	radix_tree_preload_end();
355cfa73	323	continue;
31a56396 DN	324	}
	325	if (err) { /* swp entry is obsolete ? */
	326	radix_tree_preload_end();
f000944d	327	break;
31a56396	328	}
f000944d	329
2ca4532a	330	/* May fail (-ENOMEM) if radix-tree node allocation failed. */
f45840b5	331	__set_page_locked(new_page);
b2e18538	332	SetPageSwapBacked(new_page);
31a56396	333	err = __add_to_swap_cache(new_page, entry);
529ae9aa	334	if (likely(!err)) {
31a56396	335	radix_tree_preload_end();
1da177e4 LT	336	/*
	337	* Initiate read into locked page and return.
	338	*/
c5fdae46	339	lru_cache_add_anon(new_page);
aca8bf32	340	swap_readpage(new_page);
1da177e4 LT	341	return new_page;
1da177e4 LT	342	}
31a56396	343	radix_tree_preload_end();
b2e18538	344	ClearPageSwapBacked(new_page);
f45840b5	345	__clear_page_locked(new_page);
2ca4532a DN	346	/*
	347	* add_to_swap_cache() doesn't return -EEXIST, so we can safely
	348	* clear SWAP_HAS_CACHE flag.
	349	*/
cb4b86ba	350	swapcache_free(entry, NULL);
f000944d	351	} while (err != -ENOMEM);
1da177e4 LT	352
	353	if (new_page)
	354	page_cache_release(new_page);
	355	return found_page;
	356	}
46017e95 HD	357
	358	/**
	359	* swapin_readahead - swap in pages in hope we need them soon
	360	* @entry: swap entry of this memory
7682486b	361	* @gfp_mask: memory allocation flags
46017e95 HD	362	* @vma: user vma this address belongs to
	363	* @addr: target address for mempolicy
	364	*
	365	* Returns the struct page for entry and addr, after queueing swapin.
	366	*
	367	* Primitive swap readahead code. We simply read an aligned block of
	368	* (1 << page_cluster) entries in the swap area. This method is chosen
	369	* because it doesn't cost us any seek time. We also make sure to queue
	370	* the 'original' request together with the readahead ones...
	371	*
	372	* This has been extended to use the NUMA policies from the mm triggering
	373	* the readahead.
	374	*
	375	* Caller must hold down_read on the vma->vm_mm if vma is not NULL.
	376	*/
02098fea	377	struct page *swapin_readahead(swp_entry_t entry, gfp_t gfp_mask,
46017e95 HD	378	struct vm_area_struct *vma, unsigned long addr)
	379	{
	380	int nr_pages;
	381	struct page *page;
	382	unsigned long offset;
	383	unsigned long end_offset;
	384
	385	/*
	386	* Get starting offset for readaround, and number of pages to read.
	387	* Adjust starting address by readbehind (for NUMA interleave case)?
	388	* No, it's very unlikely that swap layout would follow vma layout,
	389	* more likely that neighbouring swap pages came from the same node:
	390	* so use the same "addr" to choose the same node for each swap read.
	391	*/
	392	nr_pages = valid_swaphandles(entry, &offset);
	393	for (end_offset = offset + nr_pages; offset < end_offset; offset++) {
	394	/* Ok, do the async read-ahead now */
	395	page = read_swap_cache_async(swp_entry(swp_type(entry), offset),
02098fea	396	gfp_mask, vma, addr);
46017e95 HD	397	if (!page)
	398	break;
	399	page_cache_release(page);
	400	}
	401	lru_add_drain(); /* Push any new pages onto the LRU now */
02098fea	402	return read_swap_cache_async(entry, gfp_mask, vma, addr);
46017e95	403	}