swap_state.c 13.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9
/*
 *  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>
10
#include <linux/gfp.h>
L
Linus Torvalds 已提交
11 12
#include <linux/kernel_stat.h>
#include <linux/swap.h>
13
#include <linux/swapops.h>
L
Linus Torvalds 已提交
14 15 16
#include <linux/init.h>
#include <linux/pagemap.h>
#include <linux/backing-dev.h>
17
#include <linux/blkdev.h>
18
#include <linux/pagevec.h>
C
Christoph Lameter 已提交
19
#include <linux/migrate.h>
L
Linus Torvalds 已提交
20 21 22 23 24

#include <asm/pgtable.h>

/*
 * swapper_space is a fiction, retained to simplify the path through
J
Jens Axboe 已提交
25
 * vmscan's shrink_page_list.
L
Linus Torvalds 已提交
26
 */
27
static const struct address_space_operations swap_aops = {
L
Linus Torvalds 已提交
28
	.writepage	= swap_writepage,
29
	.set_page_dirty	= swap_set_page_dirty,
30
#ifdef CONFIG_MIGRATION
31
	.migratepage	= migrate_page,
32
#endif
L
Linus Torvalds 已提交
33 34
};

35 36 37
struct address_space swapper_spaces[MAX_SWAPFILES] = {
	[0 ... MAX_SWAPFILES - 1] = {
		.page_tree	= RADIX_TREE_INIT(GFP_ATOMIC|__GFP_NOWARN),
38
		.i_mmap_writable = ATOMIC_INIT(0),
39 40
		.a_ops		= &swap_aops,
	}
L
Linus Torvalds 已提交
41 42 43 44 45 46 47 48 49 50 51
};

#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;

52 53 54 55 56 57 58 59 60 61
unsigned long total_swapcache_pages(void)
{
	int i;
	unsigned long ret = 0;

	for (i = 0; i < MAX_SWAPFILES; i++)
		ret += swapper_spaces[i].nrpages;
	return ret;
}

62 63
static atomic_t swapin_readahead_hits = ATOMIC_INIT(4);

L
Linus Torvalds 已提交
64 65
void show_swap_cache_info(void)
{
66
	printk("%lu pages in swap cache\n", total_swapcache_pages());
67
	printk("Swap cache stats: add %lu, delete %lu, find %lu/%lu\n",
L
Linus Torvalds 已提交
68
		swap_cache_info.add_total, swap_cache_info.del_total,
69
		swap_cache_info.find_success, swap_cache_info.find_total);
70 71
	printk("Free swap  = %ldkB\n",
		get_nr_swap_pages() << (PAGE_SHIFT - 10));
L
Linus Torvalds 已提交
72 73 74 75
	printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
}

/*
76
 * __add_to_swap_cache resembles add_to_page_cache_locked on swapper_space,
L
Linus Torvalds 已提交
77 78
 * but sets SwapCache flag and private instead of mapping and index.
 */
79
int __add_to_swap_cache(struct page *page, swp_entry_t entry)
L
Linus Torvalds 已提交
80 81
{
	int error;
82
	struct address_space *address_space;
L
Linus Torvalds 已提交
83

84 85 86
	VM_BUG_ON_PAGE(!PageLocked(page), page);
	VM_BUG_ON_PAGE(PageSwapCache(page), page);
	VM_BUG_ON_PAGE(!PageSwapBacked(page), page);
87

88 89 90 91
	page_cache_get(page);
	SetPageSwapCache(page);
	set_page_private(page, entry.val);

92 93 94 95
	address_space = swap_address_space(entry);
	spin_lock_irq(&address_space->tree_lock);
	error = radix_tree_insert(&address_space->page_tree,
					entry.val, page);
96
	if (likely(!error)) {
97
		address_space->nrpages++;
98 99 100
		__inc_zone_page_state(page, NR_FILE_PAGES);
		INC_CACHE_INFO(add_total);
	}
101
	spin_unlock_irq(&address_space->tree_lock);
102 103

	if (unlikely(error)) {
104 105 106 107 108 109
		/*
		 * 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);
110 111 112 113 114 115 116 117 118 119 120 121 122
		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;

123
	error = radix_tree_maybe_preload(gfp_mask);
B
Balbir Singh 已提交
124
	if (!error) {
125
		error = __add_to_swap_cache(page, entry);
L
Linus Torvalds 已提交
126
		radix_tree_preload_end();
H
Hugh Dickins 已提交
127
	}
L
Linus Torvalds 已提交
128 129 130 131 132 133 134 135 136
	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)
{
137 138 139
	swp_entry_t entry;
	struct address_space *address_space;

140 141 142
	VM_BUG_ON_PAGE(!PageLocked(page), page);
	VM_BUG_ON_PAGE(!PageSwapCache(page), page);
	VM_BUG_ON_PAGE(PageWriteback(page), page);
L
Linus Torvalds 已提交
143

144 145 146
	entry.val = page_private(page);
	address_space = swap_address_space(entry);
	radix_tree_delete(&address_space->page_tree, page_private(page));
H
Hugh Dickins 已提交
147
	set_page_private(page, 0);
L
Linus Torvalds 已提交
148
	ClearPageSwapCache(page);
149
	address_space->nrpages--;
150
	__dec_zone_page_state(page, NR_FILE_PAGES);
L
Linus Torvalds 已提交
151 152 153 154 155 156 157 158 159 160
	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. 
 */
161
int add_to_swap(struct page *page, struct list_head *list)
L
Linus Torvalds 已提交
162 163 164 165
{
	swp_entry_t entry;
	int err;

166 167
	VM_BUG_ON_PAGE(!PageLocked(page), page);
	VM_BUG_ON_PAGE(!PageUptodate(page), page);
L
Linus Torvalds 已提交
168

169 170 171 172
	entry = get_swap_page();
	if (!entry.val)
		return 0;

173 174 175 176 177
	if (mem_cgroup_try_charge_swap(page, entry)) {
		swapcache_free(entry);
		return 0;
	}

A
Andrea Arcangeli 已提交
178
	if (unlikely(PageTransHuge(page)))
179
		if (unlikely(split_huge_page_to_list(page, list))) {
180
			swapcache_free(entry);
A
Andrea Arcangeli 已提交
181 182 183
			return 0;
		}

184 185 186 187 188 189 190 191 192
	/*
	 * 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.
	 */
	/*
M
Minchan Kim 已提交
193
	 * Add it to the swap cache.
194 195 196 197
	 */
	err = add_to_swap_cache(page, entry,
			__GFP_HIGH|__GFP_NOMEMALLOC|__GFP_NOWARN);

M
Minchan Kim 已提交
198
	if (!err) {
199 200
		return 1;
	} else {	/* -ENOMEM radix-tree allocation failure */
N
Nick Piggin 已提交
201
		/*
202 203
		 * add_to_swap_cache() doesn't return -EEXIST, so we can safely
		 * clear SWAP_HAS_CACHE flag.
L
Linus Torvalds 已提交
204
		 */
205
		swapcache_free(entry);
206
		return 0;
L
Linus Torvalds 已提交
207 208 209 210 211 212 213 214 215 216 217 218
	}
}

/*
 * 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;
219
	struct address_space *address_space;
L
Linus Torvalds 已提交
220

H
Hugh Dickins 已提交
221
	entry.val = page_private(page);
L
Linus Torvalds 已提交
222

223 224
	address_space = swap_address_space(entry);
	spin_lock_irq(&address_space->tree_lock);
L
Linus Torvalds 已提交
225
	__delete_from_swap_cache(page);
226
	spin_unlock_irq(&address_space->tree_lock);
L
Linus Torvalds 已提交
227

228
	swapcache_free(entry);
L
Linus Torvalds 已提交
229 230 231 232 233 234 235
	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
236 237
 * here because we are going to recheck again inside
 * try_to_free_swap() _with_ the lock.
L
Linus Torvalds 已提交
238 239 240 241
 * 					- Marcelo
 */
static inline void free_swap_cache(struct page *page)
{
242 243
	if (PageSwapCache(page) && !page_mapped(page) && trylock_page(page)) {
		try_to_free_swap(page);
L
Linus Torvalds 已提交
244 245 246 247 248 249
		unlock_page(page);
	}
}

/* 
 * Perform a free_page(), also freeing any swap cache associated with
250
 * this page if it is the last user of the page.
L
Linus Torvalds 已提交
251 252 253 254 255 256 257 258 259 260 261 262 263 264
 */
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;
265
	int i;
L
Linus Torvalds 已提交
266 267

	lru_add_drain();
268 269 270
	for (i = 0; i < nr; i++)
		free_swap_cache(pagep[i]);
	release_pages(pagep, nr, false);
L
Linus Torvalds 已提交
271 272 273 274 275 276 277 278 279 280 281 282
}

/*
 * 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;

283
	page = find_get_page(swap_address_space(entry), entry.val);
L
Linus Torvalds 已提交
284

285
	if (page) {
L
Linus Torvalds 已提交
286
		INC_CACHE_INFO(find_success);
287 288 289
		if (TestClearPageReadahead(page))
			atomic_inc(&swapin_readahead_hits);
	}
L
Linus Torvalds 已提交
290 291 292 293 294

	INC_CACHE_INFO(find_total);
	return page;
}

295 296 297
struct page *__read_swap_cache_async(swp_entry_t entry, gfp_t gfp_mask,
			struct vm_area_struct *vma, unsigned long addr,
			bool *new_page_allocated)
L
Linus Torvalds 已提交
298 299
{
	struct page *found_page, *new_page = NULL;
300
	struct address_space *swapper_space = swap_address_space(entry);
L
Linus Torvalds 已提交
301
	int err;
302
	*new_page_allocated = false;
L
Linus Torvalds 已提交
303 304 305 306 307 308 309

	do {
		/*
		 * First check the swap cache.  Since this is normally
		 * called after lookup_swap_cache() failed, re-calling
		 * that would confuse statistics.
		 */
310
		found_page = find_get_page(swapper_space, entry.val);
L
Linus Torvalds 已提交
311 312 313 314 315 316 317
		if (found_page)
			break;

		/*
		 * Get a new page to read into from swap.
		 */
		if (!new_page) {
318
			new_page = alloc_page_vma(gfp_mask, vma, addr);
L
Linus Torvalds 已提交
319 320 321 322
			if (!new_page)
				break;		/* Out of memory */
		}

323 324 325
		/*
		 * call radix_tree_preload() while we can wait.
		 */
326
		err = radix_tree_maybe_preload(gfp_mask & GFP_KERNEL);
327 328 329
		if (err)
			break;

H
Hugh Dickins 已提交
330 331 332
		/*
		 * Swap entry may have been freed since our caller observed it.
		 */
333
		err = swapcache_prepare(entry);
334
		if (err == -EEXIST) {
335
			radix_tree_preload_end();
336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351
			/*
			 * We might race against get_swap_page() and stumble
			 * across a SWAP_HAS_CACHE swap_map entry whose page
			 * has not been brought into the swapcache yet, while
			 * the other end is scheduled away waiting on discard
			 * I/O completion at scan_swap_map().
			 *
			 * In order to avoid turning this transitory state
			 * into a permanent loop around this -EEXIST case
			 * if !CONFIG_PREEMPT and the I/O completion happens
			 * to be waiting on the CPU waitqueue where we are now
			 * busy looping, we just conditionally invoke the
			 * scheduler here, if there are some more important
			 * tasks to run.
			 */
			cond_resched();
352
			continue;
353 354 355
		}
		if (err) {		/* swp entry is obsolete ? */
			radix_tree_preload_end();
H
Hugh Dickins 已提交
356
			break;
357
		}
H
Hugh Dickins 已提交
358

359
		/* May fail (-ENOMEM) if radix-tree node allocation failed. */
360
		__SetPageLocked(new_page);
R
Rik van Riel 已提交
361
		SetPageSwapBacked(new_page);
362
		err = __add_to_swap_cache(new_page, entry);
N
Nick Piggin 已提交
363
		if (likely(!err)) {
364
			radix_tree_preload_end();
L
Linus Torvalds 已提交
365 366 367
			/*
			 * Initiate read into locked page and return.
			 */
368
			lru_cache_add_anon(new_page);
369
			*new_page_allocated = true;
L
Linus Torvalds 已提交
370 371
			return new_page;
		}
372
		radix_tree_preload_end();
R
Rik van Riel 已提交
373
		ClearPageSwapBacked(new_page);
374
		__ClearPageLocked(new_page);
375 376 377 378
		/*
		 * add_to_swap_cache() doesn't return -EEXIST, so we can safely
		 * clear SWAP_HAS_CACHE flag.
		 */
379
		swapcache_free(entry);
H
Hugh Dickins 已提交
380
	} while (err != -ENOMEM);
L
Linus Torvalds 已提交
381 382 383 384 385

	if (new_page)
		page_cache_release(new_page);
	return found_page;
}
386

387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405
/*
 * 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)
{
	bool page_was_allocated;
	struct page *retpage = __read_swap_cache_async(entry, gfp_mask,
			vma, addr, &page_was_allocated);

	if (page_was_allocated)
		swap_readpage(retpage);

	return retpage;
}

406 407 408 409 410 411
static unsigned long swapin_nr_pages(unsigned long offset)
{
	static unsigned long prev_offset;
	unsigned int pages, max_pages, last_ra;
	static atomic_t last_readahead_pages;

412
	max_pages = 1 << READ_ONCE(page_cluster);
413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449
	if (max_pages <= 1)
		return 1;

	/*
	 * This heuristic has been found to work well on both sequential and
	 * random loads, swapping to hard disk or to SSD: please don't ask
	 * what the "+ 2" means, it just happens to work well, that's all.
	 */
	pages = atomic_xchg(&swapin_readahead_hits, 0) + 2;
	if (pages == 2) {
		/*
		 * We can have no readahead hits to judge by: but must not get
		 * stuck here forever, so check for an adjacent offset instead
		 * (and don't even bother to check whether swap type is same).
		 */
		if (offset != prev_offset + 1 && offset != prev_offset - 1)
			pages = 1;
		prev_offset = offset;
	} else {
		unsigned int roundup = 4;
		while (roundup < pages)
			roundup <<= 1;
		pages = roundup;
	}

	if (pages > max_pages)
		pages = max_pages;

	/* Don't shrink readahead too fast */
	last_ra = atomic_read(&last_readahead_pages) / 2;
	if (pages < last_ra)
		pages = last_ra;
	atomic_set(&last_readahead_pages, pages);

	return pages;
}

450 451 452
/**
 * swapin_readahead - swap in pages in hope we need them soon
 * @entry: swap entry of this memory
453
 * @gfp_mask: memory allocation flags
454 455 456 457 458 459 460 461 462 463 464 465 466 467 468
 * @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.
 */
469
struct page *swapin_readahead(swp_entry_t entry, gfp_t gfp_mask,
470 471 472
			struct vm_area_struct *vma, unsigned long addr)
{
	struct page *page;
473 474
	unsigned long entry_offset = swp_offset(entry);
	unsigned long offset = entry_offset;
475
	unsigned long start_offset, end_offset;
476
	unsigned long mask;
477
	struct blk_plug plug;
478

479 480 481 482
	mask = swapin_nr_pages(offset) - 1;
	if (!mask)
		goto skip;

483 484 485 486 487 488
	/* 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++;

489
	blk_start_plug(&plug);
490
	for (offset = start_offset; offset <= end_offset ; offset++) {
491 492
		/* Ok, do the async read-ahead now */
		page = read_swap_cache_async(swp_entry(swp_type(entry), offset),
493
						gfp_mask, vma, addr);
494
		if (!page)
495
			continue;
496 497
		if (offset != entry_offset)
			SetPageReadahead(page);
498 499
		page_cache_release(page);
	}
500 501
	blk_finish_plug(&plug);

502
	lru_add_drain();	/* Push any new pages onto the LRU now */
503
skip:
504
	return read_swap_cache_async(entry, gfp_mask, vma, addr);
505
}