swap_state.c 10.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12
/*
 *  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>
13
#include <linux/swapops.h>
L
Linus Torvalds 已提交
14 15 16 17
#include <linux/init.h>
#include <linux/pagemap.h>
#include <linux/buffer_head.h>
#include <linux/backing-dev.h>
18
#include <linux/pagevec.h>
C
Christoph Lameter 已提交
19
#include <linux/migrate.h>
20
#include <linux/page_cgroup.h>
L
Linus Torvalds 已提交
21 22 23 24 25

#include <asm/pgtable.h>

/*
 * swapper_space is a fiction, retained to simplify the path through
26
 * vmscan's shrink_page_list, to make sync_page look nicer, and to allow
L
Linus Torvalds 已提交
27 28
 * future use of radix_tree tags in the swap cache.
 */
29
static const struct address_space_operations swap_aops = {
L
Linus Torvalds 已提交
30 31 32
	.writepage	= swap_writepage,
	.sync_page	= block_sync_page,
	.set_page_dirty	= __set_page_dirty_nobuffers,
33
	.migratepage	= migrate_page,
L
Linus Torvalds 已提交
34 35 36
};

static struct backing_dev_info swap_backing_dev_info = {
37
	.name		= "swap",
38
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK | BDI_CAP_SWAP_BACKED,
L
Linus Torvalds 已提交
39 40 41 42 43
	.unplug_io_fn	= swap_unplug_io_fn,
};

struct address_space swapper_space = {
	.page_tree	= RADIX_TREE_INIT(GFP_ATOMIC|__GFP_NOWARN),
N
Nick Piggin 已提交
44
	.tree_lock	= __SPIN_LOCK_UNLOCKED(swapper_space.tree_lock),
L
Linus Torvalds 已提交
45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
	.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)
{
61 62
	printk("%lu pages in swap cache\n", total_swapcache_pages);
	printk("Swap cache stats: add %lu, delete %lu, find %lu/%lu\n",
L
Linus Torvalds 已提交
63
		swap_cache_info.add_total, swap_cache_info.del_total,
64
		swap_cache_info.find_success, swap_cache_info.find_total);
H
Hugh Dickins 已提交
65
	printk("Free swap  = %ldkB\n", nr_swap_pages << (PAGE_SHIFT - 10));
L
Linus Torvalds 已提交
66 67 68 69
	printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
}

/*
N
Nick Piggin 已提交
70
 * add_to_swap_cache resembles add_to_page_cache_locked on swapper_space,
L
Linus Torvalds 已提交
71 72
 * but sets SwapCache flag and private instead of mapping and index.
 */
73
int add_to_swap_cache(struct page *page, swp_entry_t entry, gfp_t gfp_mask)
L
Linus Torvalds 已提交
74 75 76
{
	int error;

77 78 79 80
	VM_BUG_ON(!PageLocked(page));
	VM_BUG_ON(PageSwapCache(page));
	VM_BUG_ON(!PageSwapBacked(page));

B
Balbir Singh 已提交
81 82
	error = radix_tree_preload(gfp_mask);
	if (!error) {
N
Nick Piggin 已提交
83 84 85 86
		page_cache_get(page);
		SetPageSwapCache(page);
		set_page_private(page, entry.val);

N
Nick Piggin 已提交
87
		spin_lock_irq(&swapper_space.tree_lock);
L
Linus Torvalds 已提交
88 89
		error = radix_tree_insert(&swapper_space.page_tree,
						entry.val, page);
N
Nick Piggin 已提交
90
		if (likely(!error)) {
L
Linus Torvalds 已提交
91
			total_swapcache_pages++;
92
			__inc_zone_page_state(page, NR_FILE_PAGES);
93
			INC_CACHE_INFO(add_total);
L
Linus Torvalds 已提交
94
		}
N
Nick Piggin 已提交
95
		spin_unlock_irq(&swapper_space.tree_lock);
L
Linus Torvalds 已提交
96
		radix_tree_preload_end();
N
Nick Piggin 已提交
97 98 99 100 101 102

		if (unlikely(error)) {
			set_page_private(page, 0UL);
			ClearPageSwapCache(page);
			page_cache_release(page);
		}
H
Hugh Dickins 已提交
103
	}
L
Linus Torvalds 已提交
104 105 106 107 108 109 110 111 112
	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)
{
113 114 115
	VM_BUG_ON(!PageLocked(page));
	VM_BUG_ON(!PageSwapCache(page));
	VM_BUG_ON(PageWriteback(page));
L
Linus Torvalds 已提交
116

H
Hugh Dickins 已提交
117 118
	radix_tree_delete(&swapper_space.page_tree, page_private(page));
	set_page_private(page, 0);
L
Linus Torvalds 已提交
119 120
	ClearPageSwapCache(page);
	total_swapcache_pages--;
121
	__dec_zone_page_state(page, NR_FILE_PAGES);
L
Linus Torvalds 已提交
122 123 124 125 126 127 128 129 130 131
	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. 
 */
132
int add_to_swap(struct page *page)
L
Linus Torvalds 已提交
133 134 135 136
{
	swp_entry_t entry;
	int err;

137 138
	VM_BUG_ON(!PageLocked(page));
	VM_BUG_ON(!PageUptodate(page));
L
Linus Torvalds 已提交
139 140 141 142 143 144

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

N
Nick Piggin 已提交
145 146 147 148
		/*
		 * Radix-tree node allocations from PF_MEMALLOC contexts could
		 * completely exhaust the page allocator. __GFP_NOMEMALLOC
		 * stops emergency reserves from being allocated.
L
Linus Torvalds 已提交
149
		 *
N
Nick Piggin 已提交
150 151
		 * TODO: this could cause a theoretical memory reclaim
		 * deadlock in the swap out path.
L
Linus Torvalds 已提交
152 153 154 155
		 */
		/*
		 * Add it to the swap cache and mark it dirty
		 */
H
Hugh Dickins 已提交
156
		err = add_to_swap_cache(page, entry,
157
				__GFP_HIGH|__GFP_NOMEMALLOC|__GFP_NOWARN);
L
Linus Torvalds 已提交
158 159 160 161 162 163 164

		switch (err) {
		case 0:				/* Success */
			SetPageDirty(page);
			return 1;
		case -EEXIST:
			/* Raced with "speculative" read_swap_cache_async */
165
			swapcache_free(entry, NULL);
L
Linus Torvalds 已提交
166 167 168
			continue;
		default:
			/* -ENOMEM radix-tree allocation failure */
169
			swapcache_free(entry, NULL);
L
Linus Torvalds 已提交
170 171 172 173 174 175 176 177 178 179 180 181 182 183 184
			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;

H
Hugh Dickins 已提交
185
	entry.val = page_private(page);
L
Linus Torvalds 已提交
186

N
Nick Piggin 已提交
187
	spin_lock_irq(&swapper_space.tree_lock);
L
Linus Torvalds 已提交
188
	__delete_from_swap_cache(page);
N
Nick Piggin 已提交
189
	spin_unlock_irq(&swapper_space.tree_lock);
L
Linus Torvalds 已提交
190

191
	swapcache_free(entry, page);
L
Linus Torvalds 已提交
192 193 194 195 196 197 198
	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
199 200
 * here because we are going to recheck again inside
 * try_to_free_swap() _with_ the lock.
L
Linus Torvalds 已提交
201 202 203 204
 * 					- Marcelo
 */
static inline void free_swap_cache(struct page *page)
{
205 206
	if (PageSwapCache(page) && !page_mapped(page) && trylock_page(page)) {
		try_to_free_swap(page);
L
Linus Torvalds 已提交
207 208 209 210 211 212
		unlock_page(page);
	}
}

/* 
 * Perform a free_page(), also freeing any swap cache associated with
213
 * this page if it is the last user of the page.
L
Linus Torvalds 已提交
214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230
 */
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) {
231
		int todo = min(nr, PAGEVEC_SIZE);
L
Linus Torvalds 已提交
232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266
		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.
 */
267
struct page *read_swap_cache_async(swp_entry_t entry, gfp_t gfp_mask,
L
Linus Torvalds 已提交
268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286
			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) {
287
			new_page = alloc_page_vma(gfp_mask, vma, addr);
L
Linus Torvalds 已提交
288 289 290 291
			if (!new_page)
				break;		/* Out of memory */
		}

H
Hugh Dickins 已提交
292 293 294
		/*
		 * Swap entry may have been freed since our caller observed it.
		 */
295 296 297 298
		err = swapcache_prepare(entry);
		if (err == -EEXIST) /* seems racy */
			continue;
		if (err)           /* swp entry is obsolete ? */
H
Hugh Dickins 已提交
299 300
			break;

L
Linus Torvalds 已提交
301 302
		/*
		 * Associate the page with swap entry in the swap cache.
H
Hugh Dickins 已提交
303 304 305 306
		 * 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 add_to_swap or shmem_writepage
		 * re-using the just freed swap entry for an existing page.
L
Linus Torvalds 已提交
307 308
		 * May fail (-ENOMEM) if radix-tree node allocation failed.
		 */
309
		__set_page_locked(new_page);
R
Rik van Riel 已提交
310
		SetPageSwapBacked(new_page);
H
Hugh Dickins 已提交
311
		err = add_to_swap_cache(new_page, entry, gfp_mask & GFP_KERNEL);
N
Nick Piggin 已提交
312
		if (likely(!err)) {
L
Linus Torvalds 已提交
313 314 315
			/*
			 * Initiate read into locked page and return.
			 */
316
			lru_cache_add_anon(new_page);
317
			swap_readpage(new_page);
L
Linus Torvalds 已提交
318 319
			return new_page;
		}
R
Rik van Riel 已提交
320
		ClearPageSwapBacked(new_page);
321
		__clear_page_locked(new_page);
322
		swapcache_free(entry, NULL);
H
Hugh Dickins 已提交
323
	} while (err != -ENOMEM);
L
Linus Torvalds 已提交
324 325 326 327 328

	if (new_page)
		page_cache_release(new_page);
	return found_page;
}
329 330 331 332

/**
 * swapin_readahead - swap in pages in hope we need them soon
 * @entry: swap entry of this memory
333
 * @gfp_mask: memory allocation flags
334 335 336 337 338 339 340 341 342 343 344 345 346 347 348
 * @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.
 */
349
struct page *swapin_readahead(swp_entry_t entry, gfp_t gfp_mask,
350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367
			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),
368
						gfp_mask, vma, addr);
369 370 371 372 373
		if (!page)
			break;
		page_cache_release(page);
	}
	lru_add_drain();	/* Push any new pages onto the LRU now */
374
	return read_swap_cache_async(entry, gfp_mask, vma, addr);
375
}