[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,
};
EXPORT_SYMBOL(swapper_space);

#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, int 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);
			page->private = 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));

	radix_tree_delete(&swapper_space.page_tree, page->private);
	page->private = 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;

	BUG_ON(!PageSwapCache(page));
	BUG_ON(!PageLocked(page));
	BUG_ON(PageWriteback(page));
	BUG_ON(PagePrivate(page));
  
	entry.val = page->private;

	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. Can not do a lock_page,
 * as we are holding the page_table_lock spinlock.
 */
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	};
	43	EXPORT_SYMBOL(swapper_space);
	44
	45	#define INC_CACHE_INFO(x) do { swap_cache_info.x++; } while (0)
	46
	47	static struct {
	48	unsigned long add_total;
	49	unsigned long del_total;
	50	unsigned long find_success;
	51	unsigned long find_total;
	52	unsigned long noent_race;
	53	unsigned long exist_race;
	54	} swap_cache_info;
	55
	56	void show_swap_cache_info(void)
	57	{
	58	printk("Swap cache: add %lu, delete %lu, find %lu/%lu, race %lu+%lu\n",
	59	swap_cache_info.add_total, swap_cache_info.del_total,
	60	swap_cache_info.find_success, swap_cache_info.find_total,
	61	swap_cache_info.noent_race, swap_cache_info.exist_race);
	62	printk("Free swap = %lukB\n", nr_swap_pages << (PAGE_SHIFT - 10));
	63	printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
	64	}
65
66	/*
67	* __add_to_swap_cache resembles add_to_page_cache on swapper_space,
68	* but sets SwapCache flag and private instead of mapping and index.
69	*/
70	static int __add_to_swap_cache(struct page *page,
71	swp_entry_t entry, int gfp_mask)
72	{
73	int error;
74
75	BUG_ON(PageSwapCache(page));
76	BUG_ON(PagePrivate(page));
77	error = radix_tree_preload(gfp_mask);
78	if (!error) {
79	write_lock_irq(&swapper_space.tree_lock);
80	error = radix_tree_insert(&swapper_space.page_tree,
81	entry.val, page);
82	if (!error) {
83	page_cache_get(page);
84	SetPageLocked(page);
85	SetPageSwapCache(page);
86	page->private = entry.val;
87	total_swapcache_pages++;
88	pagecache_acct(1);
89	}
90	write_unlock_irq(&swapper_space.tree_lock);
91	radix_tree_preload_end();
92	}
93	return error;
94	}
95
96	static int add_to_swap_cache(struct page *page, swp_entry_t entry)
97	{
98	int error;
99
100	if (!swap_duplicate(entry)) {
101	INC_CACHE_INFO(noent_race);
102	return -ENOENT;
103	}
104	error = __add_to_swap_cache(page, entry, GFP_KERNEL);
105	/*
106	* Anon pages are already on the LRU, we don't run lru_cache_add here.
107	*/
108	if (error) {
109	swap_free(entry);
110	if (error == -EEXIST)
111	INC_CACHE_INFO(exist_race);
112	return error;
113	}
114	INC_CACHE_INFO(add_total);
115	return 0;
116	}
117
118	/*
119	* This must be called only on pages that have
120	* been verified to be in the swap cache.
121	*/
122	void __delete_from_swap_cache(struct page *page)
123	{
124	BUG_ON(!PageLocked(page));
125	BUG_ON(!PageSwapCache(page));
126	BUG_ON(PageWriteback(page));
127
128	radix_tree_delete(&swapper_space.page_tree, page->private);
129	page->private = 0;
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
	199	BUG_ON(!PageSwapCache(page));
	200	BUG_ON(!PageLocked(page));
	201	BUG_ON(PageWriteback(page));
	202	BUG_ON(PagePrivate(page));
	203
	204	entry.val = page->private;
	205
	206	write_lock_irq(&swapper_space.tree_lock);
	207	__delete_from_swap_cache(page);
	208	write_unlock_irq(&swapper_space.tree_lock);
	209
	210	swap_free(entry);
	211	page_cache_release(page);
	212	}
	213
	214	/*
	215	* Strange swizzling function only for use by shmem_writepage
	216	*/
	217	int move_to_swap_cache(struct page *page, swp_entry_t entry)
	218	{
	219	int err = __add_to_swap_cache(page, entry, GFP_ATOMIC);
	220	if (!err) {
	221	remove_from_page_cache(page);
	222	page_cache_release(page); /* pagecache ref */
	223	if (!swap_duplicate(entry))
	224	BUG();
	225	SetPageDirty(page);
	226	INC_CACHE_INFO(add_total);
	227	} else if (err == -EEXIST)
	228	INC_CACHE_INFO(exist_race);
	229	return err;
	230	}
	231
	232	/*
233	* Strange swizzling function for shmem_getpage (and shmem_unuse)
234	*/
235	int move_from_swap_cache(struct page *page, unsigned long index,
236	struct address_space *mapping)
237	{
238	int err = add_to_page_cache(page, mapping, index, GFP_ATOMIC);
239	if (!err) {
240	delete_from_swap_cache(page);
241	/* shift page from clean_pages to dirty_pages list */
242	ClearPageDirty(page);
243	set_page_dirty(page);
244	}
245	return err;
246	}
247
248	/*
249	* If we are the only user, then try to free up the swap cache.
250	*
251	* Its ok to check for PageSwapCache without the page lock
252	* here because we are going to recheck again inside
253	* exclusive_swap_page() _with_ the lock.
254	* - Marcelo
255	*/
256	static inline void free_swap_cache(struct page *page)
257	{
258	if (PageSwapCache(page) && !TestSetPageLocked(page)) {
259	remove_exclusive_swap_page(page);
260	unlock_page(page);
261	}
262	}
263
264	/*
265	* Perform a free_page(), also freeing any swap cache associated with
266	* this page if it is the last user of the page. Can not do a lock_page,
267	* as we are holding the page_table_lock spinlock.
268	*/
269	void free_page_and_swap_cache(struct page *page)
270	{
271	free_swap_cache(page);
272	page_cache_release(page);
273	}
274
275	/*
276	* Passed an array of pages, drop them all from swapcache and then release
277	* them. They are removed from the LRU and freed if this is their last use.
278	*/
279	void free_pages_and_swap_cache(struct page **pages, int nr)
280	{
281	int chunk = 16;
282	struct page **pagep = pages;
283
284	lru_add_drain();
285	while (nr) {
286	int todo = min(chunk, nr);
287	int i;
288
289	for (i = 0; i < todo; i++)
290	free_swap_cache(pagep[i]);
291	release_pages(pagep, todo, 0);
292	pagep += todo;
293	nr -= todo;
294	}
295	}
296
297	/*
298	* Lookup a swap entry in the swap cache. A found page will be returned
299	* unlocked and with its refcount incremented - we rely on the kernel
300	* lock getting page table operations atomic even if we drop the page
301	* lock before returning.
302	*/
303	struct page * lookup_swap_cache(swp_entry_t entry)
304	{
305	struct page *page;
306
307	page = find_get_page(&swapper_space, entry.val);
308
309	if (page)
310	INC_CACHE_INFO(find_success);
311
312	INC_CACHE_INFO(find_total);
313	return page;
314	}
315
316	/*
317	* Locate a page of swap in physical memory, reserving swap cache space
318	* and reading the disk if it is not already cached.
319	* A failure return means that either the page allocation failed or that
320	* the swap entry is no longer in use.
321	*/
322	struct page *read_swap_cache_async(swp_entry_t entry,
323	struct vm_area_struct *vma, unsigned long addr)
324	{
325	struct page found_page, new_page = NULL;
326	int err;
327
328	do {
329	/*
330	* First check the swap cache. Since this is normally
331	* called after lookup_swap_cache() failed, re-calling
332	* that would confuse statistics.
333	*/
334	found_page = find_get_page(&swapper_space, entry.val);
335	if (found_page)
336	break;
337
338	/*
339	* Get a new page to read into from swap.
340	*/
341	if (!new_page) {
342	new_page = alloc_page_vma(GFP_HIGHUSER, vma, addr);
343	if (!new_page)
344	break; /* Out of memory */
345	}
346
347	/*
348	* Associate the page with swap entry in the swap cache.
349	* May fail (-ENOENT) if swap entry has been freed since
350	* our caller observed it. May fail (-EEXIST) if there
351	* is already a page associated with this entry in the
352	* swap cache: added by a racing read_swap_cache_async,
353	* or by try_to_swap_out (or shmem_writepage) re-using
354	* the just freed swap entry for an existing page.
355	* May fail (-ENOMEM) if radix-tree node allocation failed.
356	*/
357	err = add_to_swap_cache(new_page, entry);
358	if (!err) {
359	/*
360	* Initiate read into locked page and return.
361	*/
362	lru_cache_add_active(new_page);
363	swap_readpage(NULL, new_page);
364	return new_page;
365	}
366	} while (err != -ENOENT && err != -ENOMEM);
367
368	if (new_page)
369	page_cache_release(new_page);
370	return found_page;
371	}