swapfile.c 65.3 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
/*
 *  linux/mm/swapfile.c
 *
 *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 *  Swap reorganised 29.12.95, Stephen Tweedie
 */

#include <linux/mm.h>
#include <linux/hugetlb.h>
#include <linux/mman.h>
#include <linux/slab.h>
#include <linux/kernel_stat.h>
#include <linux/swap.h>
#include <linux/vmalloc.h>
#include <linux/pagemap.h>
#include <linux/namei.h>
17
#include <linux/shmem_fs.h>
L
Linus Torvalds 已提交
18
#include <linux/blkdev.h>
19
#include <linux/random.h>
L
Linus Torvalds 已提交
20 21 22 23
#include <linux/writeback.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/init.h>
H
Hugh Dickins 已提交
24
#include <linux/ksm.h>
L
Linus Torvalds 已提交
25 26 27
#include <linux/rmap.h>
#include <linux/security.h>
#include <linux/backing-dev.h>
I
Ingo Molnar 已提交
28
#include <linux/mutex.h>
29
#include <linux/capability.h>
L
Linus Torvalds 已提交
30
#include <linux/syscalls.h>
31
#include <linux/memcontrol.h>
K
Kay Sievers 已提交
32
#include <linux/poll.h>
33
#include <linux/oom.h>
L
Linus Torvalds 已提交
34 35 36 37

#include <asm/pgtable.h>
#include <asm/tlbflush.h>
#include <linux/swapops.h>
38
#include <linux/page_cgroup.h>
L
Linus Torvalds 已提交
39

H
Hugh Dickins 已提交
40 41 42
static bool swap_count_continued(struct swap_info_struct *, pgoff_t,
				 unsigned char);
static void free_swap_count_continuations(struct swap_info_struct *);
43
static sector_t map_swap_entry(swp_entry_t, struct block_device**);
H
Hugh Dickins 已提交
44

A
Adrian Bunk 已提交
45 46
static DEFINE_SPINLOCK(swap_lock);
static unsigned int nr_swapfiles;
47
long nr_swap_pages;
L
Linus Torvalds 已提交
48
long total_swap_pages;
49
static int least_priority;
L
Linus Torvalds 已提交
50 51 52 53 54 55

static const char Bad_file[] = "Bad swap file entry ";
static const char Unused_file[] = "Unused swap file entry ";
static const char Bad_offset[] = "Bad swap offset entry ";
static const char Unused_offset[] = "Unused swap offset entry ";

A
Adrian Bunk 已提交
56
static struct swap_list_t swap_list = {-1, -1};
L
Linus Torvalds 已提交
57

58
static struct swap_info_struct *swap_info[MAX_SWAPFILES];
L
Linus Torvalds 已提交
59

I
Ingo Molnar 已提交
60
static DEFINE_MUTEX(swapon_mutex);
L
Linus Torvalds 已提交
61

K
Kay Sievers 已提交
62 63 64 65
static DECLARE_WAIT_QUEUE_HEAD(proc_poll_wait);
/* Activity counter to indicate that a swapon or swapoff has occurred */
static atomic_t proc_poll_event = ATOMIC_INIT(0);

66
static inline unsigned char swap_count(unsigned char ent)
67
{
H
Hugh Dickins 已提交
68
	return ent & ~SWAP_HAS_CACHE;	/* may include SWAP_HAS_CONT flag */
69 70
}

71
/* returns 1 if swap entry is freed */
72 73 74
static int
__try_to_reclaim_swap(struct swap_info_struct *si, unsigned long offset)
{
75
	swp_entry_t entry = swp_entry(si->type, offset);
76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
	struct page *page;
	int ret = 0;

	page = find_get_page(&swapper_space, entry.val);
	if (!page)
		return 0;
	/*
	 * This function is called from scan_swap_map() and it's called
	 * by vmscan.c at reclaiming pages. So, we hold a lock on a page, here.
	 * We have to use trylock for avoiding deadlock. This is a special
	 * case and you should use try_to_free_swap() with explicit lock_page()
	 * in usual operations.
	 */
	if (trylock_page(page)) {
		ret = try_to_free_swap(page);
		unlock_page(page);
	}
	page_cache_release(page);
	return ret;
}
96

97 98 99 100 101 102 103
/*
 * swapon tell device that all the old swap contents can be discarded,
 * to allow the swap device to optimize its wear-levelling.
 */
static int discard_swap(struct swap_info_struct *si)
{
	struct swap_extent *se;
104 105
	sector_t start_block;
	sector_t nr_blocks;
106 107
	int err = 0;

108 109 110 111 112 113
	/* Do not discard the swap header page! */
	se = &si->first_swap_extent;
	start_block = (se->start_block + 1) << (PAGE_SHIFT - 9);
	nr_blocks = ((sector_t)se->nr_pages - 1) << (PAGE_SHIFT - 9);
	if (nr_blocks) {
		err = blkdev_issue_discard(si->bdev, start_block,
114
				nr_blocks, GFP_KERNEL, 0);
115 116 117 118
		if (err)
			return err;
		cond_resched();
	}
119

120 121 122
	list_for_each_entry(se, &si->first_swap_extent.list, list) {
		start_block = se->start_block << (PAGE_SHIFT - 9);
		nr_blocks = (sector_t)se->nr_pages << (PAGE_SHIFT - 9);
123 124

		err = blkdev_issue_discard(si->bdev, start_block,
125
				nr_blocks, GFP_KERNEL, 0);
126 127 128 129 130 131 132 133
		if (err)
			break;

		cond_resched();
	}
	return err;		/* That will often be -EOPNOTSUPP */
}

134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150
/*
 * swap allocation tell device that a cluster of swap can now be discarded,
 * to allow the swap device to optimize its wear-levelling.
 */
static void discard_swap_cluster(struct swap_info_struct *si,
				 pgoff_t start_page, pgoff_t nr_pages)
{
	struct swap_extent *se = si->curr_swap_extent;
	int found_extent = 0;

	while (nr_pages) {
		struct list_head *lh;

		if (se->start_page <= start_page &&
		    start_page < se->start_page + se->nr_pages) {
			pgoff_t offset = start_page - se->start_page;
			sector_t start_block = se->start_block + offset;
151
			sector_t nr_blocks = se->nr_pages - offset;
152 153 154 155 156 157 158 159 160 161 162 163

			if (nr_blocks > nr_pages)
				nr_blocks = nr_pages;
			start_page += nr_blocks;
			nr_pages -= nr_blocks;

			if (!found_extent++)
				si->curr_swap_extent = se;

			start_block <<= PAGE_SHIFT - 9;
			nr_blocks <<= PAGE_SHIFT - 9;
			if (blkdev_issue_discard(si->bdev, start_block,
164
				    nr_blocks, GFP_NOIO, 0))
165 166 167 168 169 170 171 172 173 174 175 176 177 178
				break;
		}

		lh = se->list.next;
		se = list_entry(lh, struct swap_extent, list);
	}
}

static int wait_for_discard(void *word)
{
	schedule();
	return 0;
}

179 180 181
#define SWAPFILE_CLUSTER	256
#define LATENCY_LIMIT		256

182 183
static unsigned long scan_swap_map(struct swap_info_struct *si,
				   unsigned char usage)
L
Linus Torvalds 已提交
184
{
185
	unsigned long offset;
186
	unsigned long scan_base;
187
	unsigned long last_in_cluster = 0;
188
	int latency_ration = LATENCY_LIMIT;
189
	int found_free_cluster = 0;
190

191
	/*
192 193 194 195 196 197 198
	 * We try to cluster swap pages by allocating them sequentially
	 * in swap.  Once we've allocated SWAPFILE_CLUSTER pages this
	 * way, however, we resort to first-free allocation, starting
	 * a new cluster.  This prevents us from scattering swap pages
	 * all over the entire swap partition, so that we reduce
	 * overall disk seek times between swap pages.  -- sct
	 * But we do now try to find an empty cluster.  -Andrea
199
	 * And we let swap pages go all over an SSD partition.  Hugh
200 201
	 */

202
	si->flags += SWP_SCANNING;
203
	scan_base = offset = si->cluster_next;
204 205 206 207 208 209

	if (unlikely(!si->cluster_nr--)) {
		if (si->pages - si->inuse_pages < SWAPFILE_CLUSTER) {
			si->cluster_nr = SWAPFILE_CLUSTER - 1;
			goto checks;
		}
210 211 212 213 214 215 216 217 218 219 220 221 222
		if (si->flags & SWP_DISCARDABLE) {
			/*
			 * Start range check on racing allocations, in case
			 * they overlap the cluster we eventually decide on
			 * (we scan without swap_lock to allow preemption).
			 * It's hardly conceivable that cluster_nr could be
			 * wrapped during our scan, but don't depend on it.
			 */
			if (si->lowest_alloc)
				goto checks;
			si->lowest_alloc = si->max;
			si->highest_alloc = 0;
		}
223
		spin_unlock(&swap_lock);
224

225 226 227 228 229 230 231 232 233 234
		/*
		 * If seek is expensive, start searching for new cluster from
		 * start of partition, to minimize the span of allocated swap.
		 * But if seek is cheap, search from our current position, so
		 * that swap is allocated from all over the partition: if the
		 * Flash Translation Layer only remaps within limited zones,
		 * we don't want to wear out the first zone too quickly.
		 */
		if (!(si->flags & SWP_SOLIDSTATE))
			scan_base = offset = si->lowest_bit;
235 236 237 238
		last_in_cluster = offset + SWAPFILE_CLUSTER - 1;

		/* Locate the first empty (unaligned) cluster */
		for (; last_in_cluster <= si->highest_bit; offset++) {
L
Linus Torvalds 已提交
239
			if (si->swap_map[offset])
240 241
				last_in_cluster = offset + SWAPFILE_CLUSTER;
			else if (offset == last_in_cluster) {
242
				spin_lock(&swap_lock);
243 244 245
				offset -= SWAPFILE_CLUSTER - 1;
				si->cluster_next = offset;
				si->cluster_nr = SWAPFILE_CLUSTER - 1;
246
				found_free_cluster = 1;
247
				goto checks;
L
Linus Torvalds 已提交
248
			}
249 250 251 252
			if (unlikely(--latency_ration < 0)) {
				cond_resched();
				latency_ration = LATENCY_LIMIT;
			}
253
		}
254 255

		offset = si->lowest_bit;
256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276
		last_in_cluster = offset + SWAPFILE_CLUSTER - 1;

		/* Locate the first empty (unaligned) cluster */
		for (; last_in_cluster < scan_base; offset++) {
			if (si->swap_map[offset])
				last_in_cluster = offset + SWAPFILE_CLUSTER;
			else if (offset == last_in_cluster) {
				spin_lock(&swap_lock);
				offset -= SWAPFILE_CLUSTER - 1;
				si->cluster_next = offset;
				si->cluster_nr = SWAPFILE_CLUSTER - 1;
				found_free_cluster = 1;
				goto checks;
			}
			if (unlikely(--latency_ration < 0)) {
				cond_resched();
				latency_ration = LATENCY_LIMIT;
			}
		}

		offset = scan_base;
277
		spin_lock(&swap_lock);
278
		si->cluster_nr = SWAPFILE_CLUSTER - 1;
279
		si->lowest_alloc = 0;
L
Linus Torvalds 已提交
280
	}
281

282 283
checks:
	if (!(si->flags & SWP_WRITEOK))
284
		goto no_page;
285 286
	if (!si->highest_bit)
		goto no_page;
287
	if (offset > si->highest_bit)
288
		scan_base = offset = si->lowest_bit;
289

290 291
	/* reuse swap entry of cache-only swap if not busy. */
	if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
292 293 294 295 296 297 298 299 300 301
		int swap_was_freed;
		spin_unlock(&swap_lock);
		swap_was_freed = __try_to_reclaim_swap(si, offset);
		spin_lock(&swap_lock);
		/* entry was freed successfully, try to use this again */
		if (swap_was_freed)
			goto checks;
		goto scan; /* check next one */
	}

302 303 304 305 306 307 308 309 310 311 312
	if (si->swap_map[offset])
		goto scan;

	if (offset == si->lowest_bit)
		si->lowest_bit++;
	if (offset == si->highest_bit)
		si->highest_bit--;
	si->inuse_pages++;
	if (si->inuse_pages == si->pages) {
		si->lowest_bit = si->max;
		si->highest_bit = 0;
L
Linus Torvalds 已提交
313
	}
H
Hugh Dickins 已提交
314
	si->swap_map[offset] = usage;
315 316
	si->cluster_next = offset + 1;
	si->flags -= SWP_SCANNING;
317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370

	if (si->lowest_alloc) {
		/*
		 * Only set when SWP_DISCARDABLE, and there's a scan
		 * for a free cluster in progress or just completed.
		 */
		if (found_free_cluster) {
			/*
			 * To optimize wear-levelling, discard the
			 * old data of the cluster, taking care not to
			 * discard any of its pages that have already
			 * been allocated by racing tasks (offset has
			 * already stepped over any at the beginning).
			 */
			if (offset < si->highest_alloc &&
			    si->lowest_alloc <= last_in_cluster)
				last_in_cluster = si->lowest_alloc - 1;
			si->flags |= SWP_DISCARDING;
			spin_unlock(&swap_lock);

			if (offset < last_in_cluster)
				discard_swap_cluster(si, offset,
					last_in_cluster - offset + 1);

			spin_lock(&swap_lock);
			si->lowest_alloc = 0;
			si->flags &= ~SWP_DISCARDING;

			smp_mb();	/* wake_up_bit advises this */
			wake_up_bit(&si->flags, ilog2(SWP_DISCARDING));

		} else if (si->flags & SWP_DISCARDING) {
			/*
			 * Delay using pages allocated by racing tasks
			 * until the whole discard has been issued. We
			 * could defer that delay until swap_writepage,
			 * but it's easier to keep this self-contained.
			 */
			spin_unlock(&swap_lock);
			wait_on_bit(&si->flags, ilog2(SWP_DISCARDING),
				wait_for_discard, TASK_UNINTERRUPTIBLE);
			spin_lock(&swap_lock);
		} else {
			/*
			 * Note pages allocated by racing tasks while
			 * scan for a free cluster is in progress, so
			 * that its final discard can exclude them.
			 */
			if (offset < si->lowest_alloc)
				si->lowest_alloc = offset;
			if (offset > si->highest_alloc)
				si->highest_alloc = offset;
		}
	}
371
	return offset;
372

373
scan:
374
	spin_unlock(&swap_lock);
375
	while (++offset <= si->highest_bit) {
376
		if (!si->swap_map[offset]) {
377
			spin_lock(&swap_lock);
378 379
			goto checks;
		}
380 381 382 383
		if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
			spin_lock(&swap_lock);
			goto checks;
		}
384 385 386 387
		if (unlikely(--latency_ration < 0)) {
			cond_resched();
			latency_ration = LATENCY_LIMIT;
		}
388
	}
389 390 391 392 393 394
	offset = si->lowest_bit;
	while (++offset < scan_base) {
		if (!si->swap_map[offset]) {
			spin_lock(&swap_lock);
			goto checks;
		}
395 396 397 398
		if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
			spin_lock(&swap_lock);
			goto checks;
		}
399 400 401 402 403
		if (unlikely(--latency_ration < 0)) {
			cond_resched();
			latency_ration = LATENCY_LIMIT;
		}
	}
404
	spin_lock(&swap_lock);
405 406

no_page:
407
	si->flags -= SWP_SCANNING;
L
Linus Torvalds 已提交
408 409 410 411 412
	return 0;
}

swp_entry_t get_swap_page(void)
{
413 414 415 416
	struct swap_info_struct *si;
	pgoff_t offset;
	int type, next;
	int wrapped = 0;
L
Linus Torvalds 已提交
417

418
	spin_lock(&swap_lock);
L
Linus Torvalds 已提交
419
	if (nr_swap_pages <= 0)
420 421 422 423
		goto noswap;
	nr_swap_pages--;

	for (type = swap_list.next; type >= 0 && wrapped < 2; type = next) {
424
		si = swap_info[type];
425 426
		next = si->next;
		if (next < 0 ||
427
		    (!wrapped && si->prio != swap_info[next]->prio)) {
428 429
			next = swap_list.head;
			wrapped++;
L
Linus Torvalds 已提交
430
		}
431 432 433 434 435 436 437

		if (!si->highest_bit)
			continue;
		if (!(si->flags & SWP_WRITEOK))
			continue;

		swap_list.next = next;
438
		/* This is called for allocating swap entry for cache */
H
Hugh Dickins 已提交
439
		offset = scan_swap_map(si, SWAP_HAS_CACHE);
440 441
		if (offset) {
			spin_unlock(&swap_lock);
442
			return swp_entry(type, offset);
443
		}
444
		next = swap_list.next;
L
Linus Torvalds 已提交
445
	}
446 447 448

	nr_swap_pages++;
noswap:
449
	spin_unlock(&swap_lock);
450
	return (swp_entry_t) {0};
L
Linus Torvalds 已提交
451 452
}

453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
/* The only caller of this function is now susupend routine */
swp_entry_t get_swap_page_of_type(int type)
{
	struct swap_info_struct *si;
	pgoff_t offset;

	spin_lock(&swap_lock);
	si = swap_info[type];
	if (si && (si->flags & SWP_WRITEOK)) {
		nr_swap_pages--;
		/* This is called for allocating swap entry, not cache */
		offset = scan_swap_map(si, 1);
		if (offset) {
			spin_unlock(&swap_lock);
			return swp_entry(type, offset);
		}
		nr_swap_pages++;
	}
	spin_unlock(&swap_lock);
	return (swp_entry_t) {0};
}

475
static struct swap_info_struct *swap_info_get(swp_entry_t entry)
L
Linus Torvalds 已提交
476
{
477
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
478 479 480 481 482 483 484
	unsigned long offset, type;

	if (!entry.val)
		goto out;
	type = swp_type(entry);
	if (type >= nr_swapfiles)
		goto bad_nofile;
485
	p = swap_info[type];
L
Linus Torvalds 已提交
486 487 488 489 490 491 492
	if (!(p->flags & SWP_USED))
		goto bad_device;
	offset = swp_offset(entry);
	if (offset >= p->max)
		goto bad_offset;
	if (!p->swap_map[offset])
		goto bad_free;
493
	spin_lock(&swap_lock);
L
Linus Torvalds 已提交
494 495 496 497 498 499 500 501 502 503 504 505 506 507 508
	return p;

bad_free:
	printk(KERN_ERR "swap_free: %s%08lx\n", Unused_offset, entry.val);
	goto out;
bad_offset:
	printk(KERN_ERR "swap_free: %s%08lx\n", Bad_offset, entry.val);
	goto out;
bad_device:
	printk(KERN_ERR "swap_free: %s%08lx\n", Unused_file, entry.val);
	goto out;
bad_nofile:
	printk(KERN_ERR "swap_free: %s%08lx\n", Bad_file, entry.val);
out:
	return NULL;
509
}
L
Linus Torvalds 已提交
510

511 512
static unsigned char swap_entry_free(struct swap_info_struct *p,
				     swp_entry_t entry, unsigned char usage)
L
Linus Torvalds 已提交
513
{
H
Hugh Dickins 已提交
514
	unsigned long offset = swp_offset(entry);
515 516
	unsigned char count;
	unsigned char has_cache;
517

H
Hugh Dickins 已提交
518 519 520
	count = p->swap_map[offset];
	has_cache = count & SWAP_HAS_CACHE;
	count &= ~SWAP_HAS_CACHE;
521

H
Hugh Dickins 已提交
522
	if (usage == SWAP_HAS_CACHE) {
523
		VM_BUG_ON(!has_cache);
H
Hugh Dickins 已提交
524
		has_cache = 0;
H
Hugh Dickins 已提交
525 526 527 528 529 530
	} else if (count == SWAP_MAP_SHMEM) {
		/*
		 * Or we could insist on shmem.c using a special
		 * swap_shmem_free() and free_shmem_swap_and_cache()...
		 */
		count = 0;
H
Hugh Dickins 已提交
531 532 533 534 535 536 537 538 539
	} else if ((count & ~COUNT_CONTINUED) <= SWAP_MAP_MAX) {
		if (count == COUNT_CONTINUED) {
			if (swap_count_continued(p, offset, count))
				count = SWAP_MAP_MAX | COUNT_CONTINUED;
			else
				count = SWAP_MAP_MAX;
		} else
			count--;
	}
H
Hugh Dickins 已提交
540 541 542 543 544 545

	if (!count)
		mem_cgroup_uncharge_swap(entry);

	usage = count | has_cache;
	p->swap_map[offset] = usage;
546 547

	/* free if no reference */
H
Hugh Dickins 已提交
548
	if (!usage) {
549
		struct gendisk *disk = p->bdev->bd_disk;
550 551 552 553
		if (offset < p->lowest_bit)
			p->lowest_bit = offset;
		if (offset > p->highest_bit)
			p->highest_bit = offset;
554 555 556
		if (swap_list.next >= 0 &&
		    p->prio > swap_info[swap_list.next]->prio)
			swap_list.next = p->type;
557 558
		nr_swap_pages++;
		p->inuse_pages--;
559 560 561
		if ((p->flags & SWP_BLKDEV) &&
				disk->fops->swap_slot_free_notify)
			disk->fops->swap_slot_free_notify(p->bdev, offset);
L
Linus Torvalds 已提交
562
	}
H
Hugh Dickins 已提交
563 564

	return usage;
L
Linus Torvalds 已提交
565 566 567 568 569 570 571 572
}

/*
 * Caller has made sure that the swapdevice corresponding to entry
 * is still around or has not been recycled.
 */
void swap_free(swp_entry_t entry)
{
573
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
574 575 576

	p = swap_info_get(entry);
	if (p) {
H
Hugh Dickins 已提交
577
		swap_entry_free(p, entry, 1);
578
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
579 580 581
	}
}

582 583 584 585 586
/*
 * Called after dropping swapcache to decrease refcnt to swap entries.
 */
void swapcache_free(swp_entry_t entry, struct page *page)
{
587
	struct swap_info_struct *p;
588
	unsigned char count;
589 590 591

	p = swap_info_get(entry);
	if (p) {
H
Hugh Dickins 已提交
592 593 594
		count = swap_entry_free(p, entry, SWAP_HAS_CACHE);
		if (page)
			mem_cgroup_uncharge_swapcache(page, entry, count != 0);
595 596
		spin_unlock(&swap_lock);
	}
597 598
}

L
Linus Torvalds 已提交
599
/*
600
 * How many references to page are currently swapped out?
H
Hugh Dickins 已提交
601 602
 * This does not give an exact answer when swap count is continued,
 * but does include the high COUNT_CONTINUED flag to allow for that.
L
Linus Torvalds 已提交
603
 */
604
static inline int page_swapcount(struct page *page)
L
Linus Torvalds 已提交
605
{
606 607
	int count = 0;
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
608 609
	swp_entry_t entry;

H
Hugh Dickins 已提交
610
	entry.val = page_private(page);
L
Linus Torvalds 已提交
611 612
	p = swap_info_get(entry);
	if (p) {
613
		count = swap_count(p->swap_map[swp_offset(entry)]);
614
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
615
	}
616
	return count;
L
Linus Torvalds 已提交
617 618 619
}

/*
620 621 622 623
 * We can write to an anon page without COW if there are no other references
 * to it.  And as a side-effect, free up its swap: because the old content
 * on disk will never be read, and seeking back there to write new content
 * later would only waste time away from clustering.
L
Linus Torvalds 已提交
624
 */
625
int reuse_swap_page(struct page *page)
L
Linus Torvalds 已提交
626
{
627 628
	int count;

629
	VM_BUG_ON(!PageLocked(page));
H
Hugh Dickins 已提交
630 631
	if (unlikely(PageKsm(page)))
		return 0;
632
	count = page_mapcount(page);
633
	if (count <= 1 && PageSwapCache(page)) {
634
		count += page_swapcount(page);
635 636 637 638 639
		if (count == 1 && !PageWriteback(page)) {
			delete_from_swap_cache(page);
			SetPageDirty(page);
		}
	}
H
Hugh Dickins 已提交
640
	return count <= 1;
L
Linus Torvalds 已提交
641 642 643
}

/*
644 645
 * If swap is getting full, or if there are no more mappings of this page,
 * then try_to_free_swap is called to free its swap space.
L
Linus Torvalds 已提交
646
 */
647
int try_to_free_swap(struct page *page)
L
Linus Torvalds 已提交
648
{
649
	VM_BUG_ON(!PageLocked(page));
L
Linus Torvalds 已提交
650 651 652 653 654

	if (!PageSwapCache(page))
		return 0;
	if (PageWriteback(page))
		return 0;
655
	if (page_swapcount(page))
L
Linus Torvalds 已提交
656 657
		return 0;

658 659 660 661 662 663 664 665 666 667 668 669
	/*
	 * Once hibernation has begun to create its image of memory,
	 * there's a danger that one of the calls to try_to_free_swap()
	 * - most probably a call from __try_to_reclaim_swap() while
	 * hibernation is allocating its own swap pages for the image,
	 * but conceivably even a call from memory reclaim - will free
	 * the swap from a page which has already been recorded in the
	 * image as a clean swapcache page, and then reuse its swap for
	 * another page of the image.  On waking from hibernation, the
	 * original page might be freed under memory pressure, then
	 * later read back in from swap, now with the wrong data.
	 *
670 671
	 * Hibration suspends storage while it is writing the image
	 * to disk so check that here.
672
	 */
673
	if (pm_suspended_storage())
674 675
		return 0;

676 677 678
	delete_from_swap_cache(page);
	SetPageDirty(page);
	return 1;
679 680
}

L
Linus Torvalds 已提交
681 682 683 684
/*
 * Free the swap entry like above, but also try to
 * free the page cache entry if it is the last user.
 */
685
int free_swap_and_cache(swp_entry_t entry)
L
Linus Torvalds 已提交
686
{
687
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
688 689
	struct page *page = NULL;

690
	if (non_swap_entry(entry))
691
		return 1;
692

L
Linus Torvalds 已提交
693 694
	p = swap_info_get(entry);
	if (p) {
H
Hugh Dickins 已提交
695
		if (swap_entry_free(p, entry, 1) == SWAP_HAS_CACHE) {
696
			page = find_get_page(&swapper_space, entry.val);
N
Nick Piggin 已提交
697
			if (page && !trylock_page(page)) {
698 699 700 701
				page_cache_release(page);
				page = NULL;
			}
		}
702
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
703 704
	}
	if (page) {
705 706 707 708
		/*
		 * Not mapped elsewhere, or swap space full? Free it!
		 * Also recheck PageSwapCache now page is locked (above).
		 */
709
		if (PageSwapCache(page) && !PageWriteback(page) &&
710
				(!page_mapped(page) || vm_swap_full())) {
L
Linus Torvalds 已提交
711 712 713 714 715 716
			delete_from_swap_cache(page);
			SetPageDirty(page);
		}
		unlock_page(page);
		page_cache_release(page);
	}
717
	return p != NULL;
L
Linus Torvalds 已提交
718 719
}

720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750
#ifdef CONFIG_CGROUP_MEM_RES_CTLR
/**
 * mem_cgroup_count_swap_user - count the user of a swap entry
 * @ent: the swap entry to be checked
 * @pagep: the pointer for the swap cache page of the entry to be stored
 *
 * Returns the number of the user of the swap entry. The number is valid only
 * for swaps of anonymous pages.
 * If the entry is found on swap cache, the page is stored to pagep with
 * refcount of it being incremented.
 */
int mem_cgroup_count_swap_user(swp_entry_t ent, struct page **pagep)
{
	struct page *page;
	struct swap_info_struct *p;
	int count = 0;

	page = find_get_page(&swapper_space, ent.val);
	if (page)
		count += page_mapcount(page);
	p = swap_info_get(ent);
	if (p) {
		count += swap_count(p->swap_map[swp_offset(ent)]);
		spin_unlock(&swap_lock);
	}

	*pagep = page;
	return count;
}
#endif

751
#ifdef CONFIG_HIBERNATION
752
/*
753
 * Find the swap type that corresponds to given device (if any).
754
 *
755 756 757 758
 * @offset - number of the PAGE_SIZE-sized block of the device, starting
 * from 0, in which the swap header is expected to be located.
 *
 * This is needed for the suspend to disk (aka swsusp).
759
 */
760
int swap_type_of(dev_t device, sector_t offset, struct block_device **bdev_p)
761
{
762
	struct block_device *bdev = NULL;
763
	int type;
764

765 766 767
	if (device)
		bdev = bdget(device);

768
	spin_lock(&swap_lock);
769 770
	for (type = 0; type < nr_swapfiles; type++) {
		struct swap_info_struct *sis = swap_info[type];
771

772
		if (!(sis->flags & SWP_WRITEOK))
773
			continue;
774

775
		if (!bdev) {
776
			if (bdev_p)
777
				*bdev_p = bdgrab(sis->bdev);
778

779
			spin_unlock(&swap_lock);
780
			return type;
781
		}
782
		if (bdev == sis->bdev) {
783
			struct swap_extent *se = &sis->first_swap_extent;
784 785

			if (se->start_block == offset) {
786
				if (bdev_p)
787
					*bdev_p = bdgrab(sis->bdev);
788

789 790
				spin_unlock(&swap_lock);
				bdput(bdev);
791
				return type;
792
			}
793 794 795
		}
	}
	spin_unlock(&swap_lock);
796 797 798
	if (bdev)
		bdput(bdev);

799 800 801
	return -ENODEV;
}

802 803 804 805 806 807 808 809 810 811 812 813
/*
 * Get the (PAGE_SIZE) block corresponding to given offset on the swapdev
 * corresponding to given index in swap_info (swap type).
 */
sector_t swapdev_block(int type, pgoff_t offset)
{
	struct block_device *bdev;

	if ((unsigned int)type >= nr_swapfiles)
		return 0;
	if (!(swap_info[type]->flags & SWP_WRITEOK))
		return 0;
814
	return map_swap_entry(swp_entry(type, offset), &bdev);
815 816
}

817 818 819 820 821 822 823 824 825 826
/*
 * Return either the total number of swap pages of given type, or the number
 * of free pages of that type (depending on @free)
 *
 * This is needed for software suspend
 */
unsigned int count_swap_pages(int type, int free)
{
	unsigned int n = 0;

827 828 829 830 831 832
	spin_lock(&swap_lock);
	if ((unsigned int)type < nr_swapfiles) {
		struct swap_info_struct *sis = swap_info[type];

		if (sis->flags & SWP_WRITEOK) {
			n = sis->pages;
833
			if (free)
834
				n -= sis->inuse_pages;
835 836
		}
	}
837
	spin_unlock(&swap_lock);
838 839
	return n;
}
840
#endif /* CONFIG_HIBERNATION */
841

L
Linus Torvalds 已提交
842
/*
843 844 845
 * No need to decide whether this PTE shares the swap entry with others,
 * just let do_wp_page work it out if a write is requested later - to
 * force COW, vm_page_prot omits write permission from any private vma.
L
Linus Torvalds 已提交
846
 */
H
Hugh Dickins 已提交
847
static int unuse_pte(struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
848 849
		unsigned long addr, swp_entry_t entry, struct page *page)
{
850
	struct mem_cgroup *memcg;
H
Hugh Dickins 已提交
851 852 853 854
	spinlock_t *ptl;
	pte_t *pte;
	int ret = 1;

855 856
	if (mem_cgroup_try_charge_swapin(vma->vm_mm, page,
					 GFP_KERNEL, &memcg)) {
H
Hugh Dickins 已提交
857
		ret = -ENOMEM;
858 859
		goto out_nolock;
	}
H
Hugh Dickins 已提交
860 861 862 863

	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	if (unlikely(!pte_same(*pte, swp_entry_to_pte(entry)))) {
		if (ret > 0)
864
			mem_cgroup_cancel_charge_swapin(memcg);
H
Hugh Dickins 已提交
865 866 867
		ret = 0;
		goto out;
	}
868

K
KAMEZAWA Hiroyuki 已提交
869
	dec_mm_counter(vma->vm_mm, MM_SWAPENTS);
K
KAMEZAWA Hiroyuki 已提交
870
	inc_mm_counter(vma->vm_mm, MM_ANONPAGES);
L
Linus Torvalds 已提交
871 872 873 874
	get_page(page);
	set_pte_at(vma->vm_mm, addr, pte,
		   pte_mkold(mk_pte(page, vma->vm_page_prot)));
	page_add_anon_rmap(page, vma, addr);
875
	mem_cgroup_commit_charge_swapin(page, memcg);
L
Linus Torvalds 已提交
876 877 878 879 880 881
	swap_free(entry);
	/*
	 * Move the page to the active list so it is not
	 * immediately swapped out again after swapon.
	 */
	activate_page(page);
H
Hugh Dickins 已提交
882 883
out:
	pte_unmap_unlock(pte, ptl);
884
out_nolock:
H
Hugh Dickins 已提交
885
	return ret;
L
Linus Torvalds 已提交
886 887 888 889 890 891 892
}

static int unuse_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
				unsigned long addr, unsigned long end,
				swp_entry_t entry, struct page *page)
{
	pte_t swp_pte = swp_entry_to_pte(entry);
893
	pte_t *pte;
894
	int ret = 0;
L
Linus Torvalds 已提交
895

H
Hugh Dickins 已提交
896 897 898 899 900 901 902 903 904 905
	/*
	 * We don't actually need pte lock while scanning for swp_pte: since
	 * we hold page lock and mmap_sem, swp_pte cannot be inserted into the
	 * page table while we're scanning; though it could get zapped, and on
	 * some architectures (e.g. x86_32 with PAE) we might catch a glimpse
	 * of unmatched parts which look like swp_pte, so unuse_pte must
	 * recheck under pte lock.  Scanning without pte lock lets it be
	 * preemptible whenever CONFIG_PREEMPT but not CONFIG_HIGHPTE.
	 */
	pte = pte_offset_map(pmd, addr);
L
Linus Torvalds 已提交
906 907 908 909 910 911
	do {
		/*
		 * swapoff spends a _lot_ of time in this loop!
		 * Test inline before going to call unuse_pte.
		 */
		if (unlikely(pte_same(*pte, swp_pte))) {
H
Hugh Dickins 已提交
912 913 914 915 916
			pte_unmap(pte);
			ret = unuse_pte(vma, pmd, addr, entry, page);
			if (ret)
				goto out;
			pte = pte_offset_map(pmd, addr);
L
Linus Torvalds 已提交
917 918
		}
	} while (pte++, addr += PAGE_SIZE, addr != end);
H
Hugh Dickins 已提交
919 920
	pte_unmap(pte - 1);
out:
921
	return ret;
L
Linus Torvalds 已提交
922 923 924 925 926 927 928 929
}

static inline int unuse_pmd_range(struct vm_area_struct *vma, pud_t *pud,
				unsigned long addr, unsigned long end,
				swp_entry_t entry, struct page *page)
{
	pmd_t *pmd;
	unsigned long next;
930
	int ret;
L
Linus Torvalds 已提交
931 932 933 934

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
A
Andrea Arcangeli 已提交
935 936
		if (unlikely(pmd_trans_huge(*pmd)))
			continue;
L
Linus Torvalds 已提交
937 938
		if (pmd_none_or_clear_bad(pmd))
			continue;
939 940 941
		ret = unuse_pte_range(vma, pmd, addr, next, entry, page);
		if (ret)
			return ret;
L
Linus Torvalds 已提交
942 943 944 945 946 947 948 949 950 951
	} while (pmd++, addr = next, addr != end);
	return 0;
}

static inline int unuse_pud_range(struct vm_area_struct *vma, pgd_t *pgd,
				unsigned long addr, unsigned long end,
				swp_entry_t entry, struct page *page)
{
	pud_t *pud;
	unsigned long next;
952
	int ret;
L
Linus Torvalds 已提交
953 954 955 956 957 958

	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
959 960 961
		ret = unuse_pmd_range(vma, pud, addr, next, entry, page);
		if (ret)
			return ret;
L
Linus Torvalds 已提交
962 963 964 965 966 967 968 969 970
	} while (pud++, addr = next, addr != end);
	return 0;
}

static int unuse_vma(struct vm_area_struct *vma,
				swp_entry_t entry, struct page *page)
{
	pgd_t *pgd;
	unsigned long addr, end, next;
971
	int ret;
L
Linus Torvalds 已提交
972

H
Hugh Dickins 已提交
973
	if (page_anon_vma(page)) {
L
Linus Torvalds 已提交
974 975 976 977 978 979 980 981 982 983 984 985 986 987 988
		addr = page_address_in_vma(page, vma);
		if (addr == -EFAULT)
			return 0;
		else
			end = addr + PAGE_SIZE;
	} else {
		addr = vma->vm_start;
		end = vma->vm_end;
	}

	pgd = pgd_offset(vma->vm_mm, addr);
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
989 990 991
		ret = unuse_pud_range(vma, pgd, addr, next, entry, page);
		if (ret)
			return ret;
L
Linus Torvalds 已提交
992 993 994 995 996 997 998 999
	} while (pgd++, addr = next, addr != end);
	return 0;
}

static int unuse_mm(struct mm_struct *mm,
				swp_entry_t entry, struct page *page)
{
	struct vm_area_struct *vma;
1000
	int ret = 0;
L
Linus Torvalds 已提交
1001 1002 1003

	if (!down_read_trylock(&mm->mmap_sem)) {
		/*
1004 1005
		 * Activate page so shrink_inactive_list is unlikely to unmap
		 * its ptes while lock is dropped, so swapoff can make progress.
L
Linus Torvalds 已提交
1006
		 */
1007
		activate_page(page);
L
Linus Torvalds 已提交
1008 1009 1010 1011 1012
		unlock_page(page);
		down_read(&mm->mmap_sem);
		lock_page(page);
	}
	for (vma = mm->mmap; vma; vma = vma->vm_next) {
1013
		if (vma->anon_vma && (ret = unuse_vma(vma, entry, page)))
L
Linus Torvalds 已提交
1014 1015 1016
			break;
	}
	up_read(&mm->mmap_sem);
1017
	return (ret < 0)? ret: 0;
L
Linus Torvalds 已提交
1018 1019 1020 1021 1022 1023
}

/*
 * Scan swap_map from current position to next entry still in use.
 * Recycle to start on reaching the end, returning 0 when empty.
 */
1024 1025
static unsigned int find_next_to_unuse(struct swap_info_struct *si,
					unsigned int prev)
L
Linus Torvalds 已提交
1026
{
1027 1028
	unsigned int max = si->max;
	unsigned int i = prev;
1029
	unsigned char count;
L
Linus Torvalds 已提交
1030 1031

	/*
1032
	 * No need for swap_lock here: we're just looking
L
Linus Torvalds 已提交
1033 1034
	 * for whether an entry is in use, not modifying it; false
	 * hits are okay, and sys_swapoff() has already prevented new
1035
	 * allocations from this area (while holding swap_lock).
L
Linus Torvalds 已提交
1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
	 */
	for (;;) {
		if (++i >= max) {
			if (!prev) {
				i = 0;
				break;
			}
			/*
			 * No entries in use at top of swap_map,
			 * loop back to start and recheck there.
			 */
			max = prev + 1;
			prev = 0;
			i = 1;
		}
		count = si->swap_map[i];
1052
		if (count && swap_count(count) != SWAP_MAP_BAD)
L
Linus Torvalds 已提交
1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064
			break;
	}
	return i;
}

/*
 * We completely avoid races by reading each swap page in advance,
 * and then search for the process using it.  All the necessary
 * page table adjustments can then be made atomically.
 */
static int try_to_unuse(unsigned int type)
{
1065
	struct swap_info_struct *si = swap_info[type];
L
Linus Torvalds 已提交
1066
	struct mm_struct *start_mm;
1067 1068
	unsigned char *swap_map;
	unsigned char swcount;
L
Linus Torvalds 已提交
1069 1070
	struct page *page;
	swp_entry_t entry;
1071
	unsigned int i = 0;
L
Linus Torvalds 已提交
1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
	int retval = 0;

	/*
	 * When searching mms for an entry, a good strategy is to
	 * start at the first mm we freed the previous entry from
	 * (though actually we don't notice whether we or coincidence
	 * freed the entry).  Initialize this start_mm with a hold.
	 *
	 * A simpler strategy would be to start at the last mm we
	 * freed the previous entry from; but that would take less
	 * advantage of mmlist ordering, which clusters forked mms
	 * together, child after parent.  If we race with dup_mmap(), we
	 * prefer to resolve parent before child, lest we miss entries
	 * duplicated after we scanned child: using last mm would invert
H
Hugh Dickins 已提交
1086
	 * that.
L
Linus Torvalds 已提交
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101
	 */
	start_mm = &init_mm;
	atomic_inc(&init_mm.mm_users);

	/*
	 * Keep on scanning until all entries have gone.  Usually,
	 * one pass through swap_map is enough, but not necessarily:
	 * there are races when an instance of an entry might be missed.
	 */
	while ((i = find_next_to_unuse(si, i)) != 0) {
		if (signal_pending(current)) {
			retval = -EINTR;
			break;
		}

1102
		/*
L
Linus Torvalds 已提交
1103 1104
		 * Get a page for the entry, using the existing swap
		 * cache page if there is one.  Otherwise, get a clean
1105
		 * page and read the swap into it.
L
Linus Torvalds 已提交
1106 1107 1108
		 */
		swap_map = &si->swap_map[i];
		entry = swp_entry(type, i);
1109 1110
		page = read_swap_cache_async(entry,
					GFP_HIGHUSER_MOVABLE, NULL, 0);
L
Linus Torvalds 已提交
1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149
		if (!page) {
			/*
			 * Either swap_duplicate() failed because entry
			 * has been freed independently, and will not be
			 * reused since sys_swapoff() already disabled
			 * allocation from here, or alloc_page() failed.
			 */
			if (!*swap_map)
				continue;
			retval = -ENOMEM;
			break;
		}

		/*
		 * Don't hold on to start_mm if it looks like exiting.
		 */
		if (atomic_read(&start_mm->mm_users) == 1) {
			mmput(start_mm);
			start_mm = &init_mm;
			atomic_inc(&init_mm.mm_users);
		}

		/*
		 * Wait for and lock page.  When do_swap_page races with
		 * try_to_unuse, do_swap_page can handle the fault much
		 * faster than try_to_unuse can locate the entry.  This
		 * apparently redundant "wait_on_page_locked" lets try_to_unuse
		 * defer to do_swap_page in such a case - in some tests,
		 * do_swap_page and try_to_unuse repeatedly compete.
		 */
		wait_on_page_locked(page);
		wait_on_page_writeback(page);
		lock_page(page);
		wait_on_page_writeback(page);

		/*
		 * Remove all references to entry.
		 */
		swcount = *swap_map;
H
Hugh Dickins 已提交
1150 1151 1152 1153 1154 1155
		if (swap_count(swcount) == SWAP_MAP_SHMEM) {
			retval = shmem_unuse(entry, page);
			/* page has already been unlocked and released */
			if (retval < 0)
				break;
			continue;
L
Linus Torvalds 已提交
1156
		}
H
Hugh Dickins 已提交
1157 1158 1159
		if (swap_count(swcount) && start_mm != &init_mm)
			retval = unuse_mm(start_mm, entry, page);

1160
		if (swap_count(*swap_map)) {
L
Linus Torvalds 已提交
1161 1162 1163 1164 1165 1166 1167 1168 1169
			int set_start_mm = (*swap_map >= swcount);
			struct list_head *p = &start_mm->mmlist;
			struct mm_struct *new_start_mm = start_mm;
			struct mm_struct *prev_mm = start_mm;
			struct mm_struct *mm;

			atomic_inc(&new_start_mm->mm_users);
			atomic_inc(&prev_mm->mm_users);
			spin_lock(&mmlist_lock);
H
Hugh Dickins 已提交
1170
			while (swap_count(*swap_map) && !retval &&
L
Linus Torvalds 已提交
1171 1172
					(p = p->next) != &start_mm->mmlist) {
				mm = list_entry(p, struct mm_struct, mmlist);
1173
				if (!atomic_inc_not_zero(&mm->mm_users))
L
Linus Torvalds 已提交
1174 1175 1176 1177 1178 1179 1180 1181
					continue;
				spin_unlock(&mmlist_lock);
				mmput(prev_mm);
				prev_mm = mm;

				cond_resched();

				swcount = *swap_map;
1182
				if (!swap_count(swcount)) /* any usage ? */
L
Linus Torvalds 已提交
1183
					;
H
Hugh Dickins 已提交
1184
				else if (mm == &init_mm)
L
Linus Torvalds 已提交
1185
					set_start_mm = 1;
H
Hugh Dickins 已提交
1186
				else
L
Linus Torvalds 已提交
1187
					retval = unuse_mm(mm, entry, page);
1188

1189
				if (set_start_mm && *swap_map < swcount) {
L
Linus Torvalds 已提交
1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219
					mmput(new_start_mm);
					atomic_inc(&mm->mm_users);
					new_start_mm = mm;
					set_start_mm = 0;
				}
				spin_lock(&mmlist_lock);
			}
			spin_unlock(&mmlist_lock);
			mmput(prev_mm);
			mmput(start_mm);
			start_mm = new_start_mm;
		}
		if (retval) {
			unlock_page(page);
			page_cache_release(page);
			break;
		}

		/*
		 * If a reference remains (rare), we would like to leave
		 * the page in the swap cache; but try_to_unmap could
		 * then re-duplicate the entry once we drop page lock,
		 * so we might loop indefinitely; also, that page could
		 * not be swapped out to other storage meanwhile.  So:
		 * delete from cache even if there's another reference,
		 * after ensuring that the data has been saved to disk -
		 * since if the reference remains (rarer), it will be
		 * read from disk into another page.  Splitting into two
		 * pages would be incorrect if swap supported "shared
		 * private" pages, but they are handled by tmpfs files.
H
Hugh Dickins 已提交
1220 1221 1222 1223 1224 1225
		 *
		 * Given how unuse_vma() targets one particular offset
		 * in an anon_vma, once the anon_vma has been determined,
		 * this splitting happens to be just what is needed to
		 * handle where KSM pages have been swapped out: re-reading
		 * is unnecessarily slow, but we can fix that later on.
L
Linus Torvalds 已提交
1226
		 */
1227 1228
		if (swap_count(*swap_map) &&
		     PageDirty(page) && PageSwapCache(page)) {
L
Linus Torvalds 已提交
1229 1230 1231 1232 1233 1234 1235 1236
			struct writeback_control wbc = {
				.sync_mode = WB_SYNC_NONE,
			};

			swap_writepage(page, &wbc);
			lock_page(page);
			wait_on_page_writeback(page);
		}
1237 1238 1239 1240 1241 1242 1243 1244 1245 1246

		/*
		 * It is conceivable that a racing task removed this page from
		 * swap cache just before we acquired the page lock at the top,
		 * or while we dropped it in unuse_mm().  The page might even
		 * be back in swap cache on another swap area: that we must not
		 * delete, since it may not have been written out to swap yet.
		 */
		if (PageSwapCache(page) &&
		    likely(page_private(page) == entry.val))
1247
			delete_from_swap_cache(page);
L
Linus Torvalds 已提交
1248 1249 1250 1251

		/*
		 * So we could skip searching mms once swap count went
		 * to 1, we did not mark any present ptes as dirty: must
1252
		 * mark page dirty so shrink_page_list will preserve it.
L
Linus Torvalds 已提交
1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269
		 */
		SetPageDirty(page);
		unlock_page(page);
		page_cache_release(page);

		/*
		 * Make sure that we aren't completely killing
		 * interactive performance.
		 */
		cond_resched();
	}

	mmput(start_mm);
	return retval;
}

/*
1270 1271 1272
 * After a successful try_to_unuse, if no swap is now in use, we know
 * we can empty the mmlist.  swap_lock must be held on entry and exit.
 * Note that mmlist_lock nests inside swap_lock, and an mm must be
L
Linus Torvalds 已提交
1273 1274 1275 1276 1277
 * added to the mmlist just after page_duplicate - before would be racy.
 */
static void drain_mmlist(void)
{
	struct list_head *p, *next;
1278
	unsigned int type;
L
Linus Torvalds 已提交
1279

1280 1281
	for (type = 0; type < nr_swapfiles; type++)
		if (swap_info[type]->inuse_pages)
L
Linus Torvalds 已提交
1282 1283 1284 1285 1286 1287 1288 1289 1290
			return;
	spin_lock(&mmlist_lock);
	list_for_each_safe(p, next, &init_mm.mmlist)
		list_del_init(p);
	spin_unlock(&mmlist_lock);
}

/*
 * Use this swapdev's extent info to locate the (PAGE_SIZE) block which
1291 1292 1293
 * corresponds to page offset for the specified swap entry.
 * Note that the type of this function is sector_t, but it returns page offset
 * into the bdev, not sector offset.
L
Linus Torvalds 已提交
1294
 */
1295
static sector_t map_swap_entry(swp_entry_t entry, struct block_device **bdev)
L
Linus Torvalds 已提交
1296
{
H
Hugh Dickins 已提交
1297 1298 1299 1300 1301
	struct swap_info_struct *sis;
	struct swap_extent *start_se;
	struct swap_extent *se;
	pgoff_t offset;

1302
	sis = swap_info[swp_type(entry)];
H
Hugh Dickins 已提交
1303 1304 1305 1306 1307
	*bdev = sis->bdev;

	offset = swp_offset(entry);
	start_se = sis->curr_swap_extent;
	se = start_se;
L
Linus Torvalds 已提交
1308 1309 1310 1311 1312 1313 1314 1315

	for ( ; ; ) {
		struct list_head *lh;

		if (se->start_page <= offset &&
				offset < (se->start_page + se->nr_pages)) {
			return se->start_block + (offset - se->start_page);
		}
1316
		lh = se->list.next;
L
Linus Torvalds 已提交
1317 1318 1319 1320 1321 1322
		se = list_entry(lh, struct swap_extent, list);
		sis->curr_swap_extent = se;
		BUG_ON(se == start_se);		/* It *must* be present */
	}
}

1323 1324 1325 1326 1327 1328 1329 1330 1331 1332
/*
 * Returns the page offset into bdev for the specified page's swap entry.
 */
sector_t map_swap_page(struct page *page, struct block_device **bdev)
{
	swp_entry_t entry;
	entry.val = page_private(page);
	return map_swap_entry(entry, bdev);
}

L
Linus Torvalds 已提交
1333 1334 1335 1336 1337
/*
 * Free all of a swapdev's extent information
 */
static void destroy_swap_extents(struct swap_info_struct *sis)
{
1338
	while (!list_empty(&sis->first_swap_extent.list)) {
L
Linus Torvalds 已提交
1339 1340
		struct swap_extent *se;

1341
		se = list_entry(sis->first_swap_extent.list.next,
L
Linus Torvalds 已提交
1342 1343 1344 1345 1346 1347 1348 1349
				struct swap_extent, list);
		list_del(&se->list);
		kfree(se);
	}
}

/*
 * Add a block range (and the corresponding page range) into this swapdev's
1350
 * extent list.  The extent list is kept sorted in page order.
L
Linus Torvalds 已提交
1351
 *
1352
 * This function rather assumes that it is called in ascending page order.
L
Linus Torvalds 已提交
1353 1354 1355 1356 1357 1358 1359 1360 1361
 */
static int
add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
		unsigned long nr_pages, sector_t start_block)
{
	struct swap_extent *se;
	struct swap_extent *new_se;
	struct list_head *lh;

1362 1363 1364 1365 1366 1367 1368 1369 1370
	if (start_page == 0) {
		se = &sis->first_swap_extent;
		sis->curr_swap_extent = se;
		se->start_page = 0;
		se->nr_pages = nr_pages;
		se->start_block = start_block;
		return 1;
	} else {
		lh = sis->first_swap_extent.list.prev;	/* Highest extent */
L
Linus Torvalds 已提交
1371
		se = list_entry(lh, struct swap_extent, list);
1372 1373
		BUG_ON(se->start_page + se->nr_pages != start_page);
		if (se->start_block + se->nr_pages == start_block) {
L
Linus Torvalds 已提交
1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389
			/* Merge it */
			se->nr_pages += nr_pages;
			return 0;
		}
	}

	/*
	 * No merge.  Insert a new extent, preserving ordering.
	 */
	new_se = kmalloc(sizeof(*se), GFP_KERNEL);
	if (new_se == NULL)
		return -ENOMEM;
	new_se->start_page = start_page;
	new_se->nr_pages = nr_pages;
	new_se->start_block = start_block;

1390
	list_add_tail(&new_se->list, &sis->first_swap_extent.list);
1391
	return 1;
L
Linus Torvalds 已提交
1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413
}

/*
 * A `swap extent' is a simple thing which maps a contiguous range of pages
 * onto a contiguous range of disk blocks.  An ordered list of swap extents
 * is built at swapon time and is then used at swap_writepage/swap_readpage
 * time for locating where on disk a page belongs.
 *
 * If the swapfile is an S_ISBLK block device, a single extent is installed.
 * This is done so that the main operating code can treat S_ISBLK and S_ISREG
 * swap files identically.
 *
 * Whether the swapdev is an S_ISREG file or an S_ISBLK blockdev, the swap
 * extent list operates in PAGE_SIZE disk blocks.  Both S_ISREG and S_ISBLK
 * swapfiles are handled *identically* after swapon time.
 *
 * For S_ISREG swapfiles, setup_swap_extents() will walk all the file's blocks
 * and will parse them into an ordered extent list, in PAGE_SIZE chunks.  If
 * some stray blocks are found which do not fall within the PAGE_SIZE alignment
 * requirements, they are simply tossed out - we will never use those blocks
 * for swapping.
 *
1414
 * For S_ISREG swapfiles we set S_SWAPFILE across the life of the swapon.  This
L
Linus Torvalds 已提交
1415 1416 1417 1418 1419 1420 1421 1422 1423 1424
 * prevents root from shooting her foot off by ftruncating an in-use swapfile,
 * which will scribble on the fs.
 *
 * The amount of disk space which a single swap extent represents varies.
 * Typically it is in the 1-4 megabyte range.  So we can have hundreds of
 * extents in the list.  To avoid much list walking, we cache the previous
 * search location in `curr_swap_extent', and start new searches from there.
 * This is extremely effective.  The average number of iterations in
 * map_swap_page() has been measured at about 0.3 per page.  - akpm.
 */
1425
static int setup_swap_extents(struct swap_info_struct *sis, sector_t *span)
L
Linus Torvalds 已提交
1426 1427 1428 1429 1430 1431 1432
{
	struct inode *inode;
	unsigned blocks_per_page;
	unsigned long page_no;
	unsigned blkbits;
	sector_t probe_block;
	sector_t last_block;
1433 1434 1435
	sector_t lowest_block = -1;
	sector_t highest_block = 0;
	int nr_extents = 0;
L
Linus Torvalds 已提交
1436 1437 1438 1439 1440
	int ret;

	inode = sis->swap_file->f_mapping->host;
	if (S_ISBLK(inode->i_mode)) {
		ret = add_swap_extent(sis, 0, sis->max, 0);
1441
		*span = sis->pages;
1442
		goto out;
L
Linus Torvalds 已提交
1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485
	}

	blkbits = inode->i_blkbits;
	blocks_per_page = PAGE_SIZE >> blkbits;

	/*
	 * Map all the blocks into the extent list.  This code doesn't try
	 * to be very smart.
	 */
	probe_block = 0;
	page_no = 0;
	last_block = i_size_read(inode) >> blkbits;
	while ((probe_block + blocks_per_page) <= last_block &&
			page_no < sis->max) {
		unsigned block_in_page;
		sector_t first_block;

		first_block = bmap(inode, probe_block);
		if (first_block == 0)
			goto bad_bmap;

		/*
		 * It must be PAGE_SIZE aligned on-disk
		 */
		if (first_block & (blocks_per_page - 1)) {
			probe_block++;
			goto reprobe;
		}

		for (block_in_page = 1; block_in_page < blocks_per_page;
					block_in_page++) {
			sector_t block;

			block = bmap(inode, probe_block + block_in_page);
			if (block == 0)
				goto bad_bmap;
			if (block != first_block + block_in_page) {
				/* Discontiguity */
				probe_block++;
				goto reprobe;
			}
		}

1486 1487 1488 1489 1490 1491 1492 1493
		first_block >>= (PAGE_SHIFT - blkbits);
		if (page_no) {	/* exclude the header page */
			if (first_block < lowest_block)
				lowest_block = first_block;
			if (first_block > highest_block)
				highest_block = first_block;
		}

L
Linus Torvalds 已提交
1494 1495 1496
		/*
		 * We found a PAGE_SIZE-length, PAGE_SIZE-aligned run of blocks
		 */
1497 1498
		ret = add_swap_extent(sis, page_no, 1, first_block);
		if (ret < 0)
L
Linus Torvalds 已提交
1499
			goto out;
1500
		nr_extents += ret;
L
Linus Torvalds 已提交
1501 1502 1503 1504 1505
		page_no++;
		probe_block += blocks_per_page;
reprobe:
		continue;
	}
1506 1507
	ret = nr_extents;
	*span = 1 + highest_block - lowest_block;
L
Linus Torvalds 已提交
1508
	if (page_no == 0)
1509
		page_no = 1;	/* force Empty message */
L
Linus Torvalds 已提交
1510
	sis->max = page_no;
1511
	sis->pages = page_no - 1;
L
Linus Torvalds 已提交
1512
	sis->highest_bit = page_no - 1;
1513 1514
out:
	return ret;
L
Linus Torvalds 已提交
1515 1516 1517
bad_bmap:
	printk(KERN_ERR "swapon: swapfile has holes\n");
	ret = -EINVAL;
1518
	goto out;
L
Linus Torvalds 已提交
1519 1520
}

1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550
static void enable_swap_info(struct swap_info_struct *p, int prio,
				unsigned char *swap_map)
{
	int i, prev;

	spin_lock(&swap_lock);
	if (prio >= 0)
		p->prio = prio;
	else
		p->prio = --least_priority;
	p->swap_map = swap_map;
	p->flags |= SWP_WRITEOK;
	nr_swap_pages += p->pages;
	total_swap_pages += p->pages;

	/* insert swap space into swap_list: */
	prev = -1;
	for (i = swap_list.head; i >= 0; i = swap_info[i]->next) {
		if (p->prio >= swap_info[i]->prio)
			break;
		prev = i;
	}
	p->next = i;
	if (prev < 0)
		swap_list.head = swap_list.next = p->type;
	else
		swap_info[prev]->next = p->type;
	spin_unlock(&swap_lock);
}

1551
SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
L
Linus Torvalds 已提交
1552
{
1553
	struct swap_info_struct *p = NULL;
1554
	unsigned char *swap_map;
L
Linus Torvalds 已提交
1555 1556 1557
	struct file *swap_file, *victim;
	struct address_space *mapping;
	struct inode *inode;
1558
	char *pathname;
1559
	int oom_score_adj;
L
Linus Torvalds 已提交
1560 1561
	int i, type, prev;
	int err;
1562

L
Linus Torvalds 已提交
1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;

	pathname = getname(specialfile);
	err = PTR_ERR(pathname);
	if (IS_ERR(pathname))
		goto out;

	victim = filp_open(pathname, O_RDWR|O_LARGEFILE, 0);
	putname(pathname);
	err = PTR_ERR(victim);
	if (IS_ERR(victim))
		goto out;

	mapping = victim->f_mapping;
	prev = -1;
1579
	spin_lock(&swap_lock);
1580 1581
	for (type = swap_list.head; type >= 0; type = swap_info[type]->next) {
		p = swap_info[type];
H
Hugh Dickins 已提交
1582
		if (p->flags & SWP_WRITEOK) {
L
Linus Torvalds 已提交
1583 1584 1585 1586 1587 1588 1589
			if (p->swap_file->f_mapping == mapping)
				break;
		}
		prev = type;
	}
	if (type < 0) {
		err = -EINVAL;
1590
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
1591 1592 1593 1594 1595 1596
		goto out_dput;
	}
	if (!security_vm_enough_memory(p->pages))
		vm_unacct_memory(p->pages);
	else {
		err = -ENOMEM;
1597
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
1598 1599
		goto out_dput;
	}
1600
	if (prev < 0)
L
Linus Torvalds 已提交
1601
		swap_list.head = p->next;
1602 1603
	else
		swap_info[prev]->next = p->next;
L
Linus Torvalds 已提交
1604 1605 1606 1607
	if (type == swap_list.next) {
		/* just pick something that's safe... */
		swap_list.next = swap_list.head;
	}
1608
	if (p->prio < 0) {
1609 1610
		for (i = p->next; i >= 0; i = swap_info[i]->next)
			swap_info[i]->prio = p->prio--;
1611 1612
		least_priority++;
	}
L
Linus Torvalds 已提交
1613 1614 1615
	nr_swap_pages -= p->pages;
	total_swap_pages -= p->pages;
	p->flags &= ~SWP_WRITEOK;
1616
	spin_unlock(&swap_lock);
1617

1618
	oom_score_adj = test_set_oom_score_adj(OOM_SCORE_ADJ_MAX);
L
Linus Torvalds 已提交
1619
	err = try_to_unuse(type);
1620
	compare_swap_oom_score_adj(OOM_SCORE_ADJ_MAX, oom_score_adj);
L
Linus Torvalds 已提交
1621 1622

	if (err) {
1623 1624 1625 1626 1627 1628
		/*
		 * reading p->prio and p->swap_map outside the lock is
		 * safe here because only sys_swapon and sys_swapoff
		 * change them, and there can be no other sys_swapon or
		 * sys_swapoff for this swap_info_struct at this point.
		 */
L
Linus Torvalds 已提交
1629
		/* re-insert swap space back into swap_list */
1630
		enable_swap_info(p, p->prio, p->swap_map);
L
Linus Torvalds 已提交
1631 1632
		goto out_dput;
	}
1633

1634
	destroy_swap_extents(p);
H
Hugh Dickins 已提交
1635 1636 1637
	if (p->flags & SWP_CONTINUED)
		free_swap_count_continuations(p);

I
Ingo Molnar 已提交
1638
	mutex_lock(&swapon_mutex);
1639 1640 1641
	spin_lock(&swap_lock);
	drain_mmlist();

1642 1643 1644
	/* wait for anyone still in scan_swap_map */
	p->highest_bit = 0;		/* cuts scans short */
	while (p->flags >= SWP_SCANNING) {
1645
		spin_unlock(&swap_lock);
1646
		schedule_timeout_uninterruptible(1);
1647
		spin_lock(&swap_lock);
1648 1649
	}

L
Linus Torvalds 已提交
1650 1651 1652 1653 1654 1655
	swap_file = p->swap_file;
	p->swap_file = NULL;
	p->max = 0;
	swap_map = p->swap_map;
	p->swap_map = NULL;
	p->flags = 0;
1656
	spin_unlock(&swap_lock);
I
Ingo Molnar 已提交
1657
	mutex_unlock(&swapon_mutex);
L
Linus Torvalds 已提交
1658
	vfree(swap_map);
1659 1660 1661
	/* Destroy swap account informatin */
	swap_cgroup_swapoff(type);

L
Linus Torvalds 已提交
1662 1663 1664 1665
	inode = mapping->host;
	if (S_ISBLK(inode->i_mode)) {
		struct block_device *bdev = I_BDEV(inode);
		set_blocksize(bdev, p->old_block_size);
1666
		blkdev_put(bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL);
L
Linus Torvalds 已提交
1667
	} else {
1668
		mutex_lock(&inode->i_mutex);
L
Linus Torvalds 已提交
1669
		inode->i_flags &= ~S_SWAPFILE;
1670
		mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
1671 1672 1673
	}
	filp_close(swap_file, NULL);
	err = 0;
K
Kay Sievers 已提交
1674 1675
	atomic_inc(&proc_poll_event);
	wake_up_interruptible(&proc_poll_wait);
L
Linus Torvalds 已提交
1676 1677 1678 1679 1680 1681 1682 1683

out_dput:
	filp_close(victim, NULL);
out:
	return err;
}

#ifdef CONFIG_PROC_FS
K
Kay Sievers 已提交
1684 1685
static unsigned swaps_poll(struct file *file, poll_table *wait)
{
1686
	struct seq_file *seq = file->private_data;
K
Kay Sievers 已提交
1687 1688 1689

	poll_wait(file, &proc_poll_wait, wait);

1690 1691
	if (seq->poll_event != atomic_read(&proc_poll_event)) {
		seq->poll_event = atomic_read(&proc_poll_event);
K
Kay Sievers 已提交
1692 1693 1694 1695 1696 1697
		return POLLIN | POLLRDNORM | POLLERR | POLLPRI;
	}

	return POLLIN | POLLRDNORM;
}

L
Linus Torvalds 已提交
1698 1699 1700
/* iterator */
static void *swap_start(struct seq_file *swap, loff_t *pos)
{
1701 1702
	struct swap_info_struct *si;
	int type;
L
Linus Torvalds 已提交
1703 1704
	loff_t l = *pos;

I
Ingo Molnar 已提交
1705
	mutex_lock(&swapon_mutex);
L
Linus Torvalds 已提交
1706

1707 1708 1709
	if (!l)
		return SEQ_START_TOKEN;

1710 1711 1712 1713
	for (type = 0; type < nr_swapfiles; type++) {
		smp_rmb();	/* read nr_swapfiles before swap_info[type] */
		si = swap_info[type];
		if (!(si->flags & SWP_USED) || !si->swap_map)
L
Linus Torvalds 已提交
1714
			continue;
1715
		if (!--l)
1716
			return si;
L
Linus Torvalds 已提交
1717 1718 1719 1720 1721 1722 1723
	}

	return NULL;
}

static void *swap_next(struct seq_file *swap, void *v, loff_t *pos)
{
1724 1725
	struct swap_info_struct *si = v;
	int type;
L
Linus Torvalds 已提交
1726

1727
	if (v == SEQ_START_TOKEN)
1728 1729 1730
		type = 0;
	else
		type = si->type + 1;
1731

1732 1733 1734 1735
	for (; type < nr_swapfiles; type++) {
		smp_rmb();	/* read nr_swapfiles before swap_info[type] */
		si = swap_info[type];
		if (!(si->flags & SWP_USED) || !si->swap_map)
L
Linus Torvalds 已提交
1736 1737
			continue;
		++*pos;
1738
		return si;
L
Linus Torvalds 已提交
1739 1740 1741 1742 1743 1744 1745
	}

	return NULL;
}

static void swap_stop(struct seq_file *swap, void *v)
{
I
Ingo Molnar 已提交
1746
	mutex_unlock(&swapon_mutex);
L
Linus Torvalds 已提交
1747 1748 1749 1750
}

static int swap_show(struct seq_file *swap, void *v)
{
1751
	struct swap_info_struct *si = v;
L
Linus Torvalds 已提交
1752 1753 1754
	struct file *file;
	int len;

1755
	if (si == SEQ_START_TOKEN) {
1756 1757 1758
		seq_puts(swap,"Filename\t\t\t\tType\t\tSize\tUsed\tPriority\n");
		return 0;
	}
L
Linus Torvalds 已提交
1759

1760
	file = si->swap_file;
1761
	len = seq_path(swap, &file->f_path, " \t\n\\");
1762
	seq_printf(swap, "%*s%s\t%u\t%u\t%d\n",
1763 1764
			len < 40 ? 40 - len : 1, " ",
			S_ISBLK(file->f_path.dentry->d_inode->i_mode) ?
L
Linus Torvalds 已提交
1765
				"partition" : "file\t",
1766 1767 1768
			si->pages << (PAGE_SHIFT - 10),
			si->inuse_pages << (PAGE_SHIFT - 10),
			si->prio);
L
Linus Torvalds 已提交
1769 1770 1771
	return 0;
}

1772
static const struct seq_operations swaps_op = {
L
Linus Torvalds 已提交
1773 1774 1775 1776 1777 1778 1779 1780
	.start =	swap_start,
	.next =		swap_next,
	.stop =		swap_stop,
	.show =		swap_show
};

static int swaps_open(struct inode *inode, struct file *file)
{
1781
	struct seq_file *seq;
K
Kay Sievers 已提交
1782 1783 1784
	int ret;

	ret = seq_open(file, &swaps_op);
1785
	if (ret)
K
Kay Sievers 已提交
1786 1787
		return ret;

1788 1789 1790
	seq = file->private_data;
	seq->poll_event = atomic_read(&proc_poll_event);
	return 0;
L
Linus Torvalds 已提交
1791 1792
}

1793
static const struct file_operations proc_swaps_operations = {
L
Linus Torvalds 已提交
1794 1795 1796 1797
	.open		= swaps_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= seq_release,
K
Kay Sievers 已提交
1798
	.poll		= swaps_poll,
L
Linus Torvalds 已提交
1799 1800 1801 1802
};

static int __init procswaps_init(void)
{
1803
	proc_create("swaps", 0, NULL, &proc_swaps_operations);
L
Linus Torvalds 已提交
1804 1805 1806 1807 1808
	return 0;
}
__initcall(procswaps_init);
#endif /* CONFIG_PROC_FS */

J
Jan Beulich 已提交
1809 1810 1811 1812 1813 1814 1815 1816 1817
#ifdef MAX_SWAPFILES_CHECK
static int __init max_swapfiles_check(void)
{
	MAX_SWAPFILES_CHECK();
	return 0;
}
late_initcall(max_swapfiles_check);
#endif

1818
static struct swap_info_struct *alloc_swap_info(void)
L
Linus Torvalds 已提交
1819
{
1820
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
1821
	unsigned int type;
1822 1823 1824

	p = kzalloc(sizeof(*p), GFP_KERNEL);
	if (!p)
1825
		return ERR_PTR(-ENOMEM);
1826

1827
	spin_lock(&swap_lock);
1828 1829
	for (type = 0; type < nr_swapfiles; type++) {
		if (!(swap_info[type]->flags & SWP_USED))
L
Linus Torvalds 已提交
1830
			break;
1831
	}
1832
	if (type >= MAX_SWAPFILES) {
1833
		spin_unlock(&swap_lock);
1834
		kfree(p);
1835
		return ERR_PTR(-EPERM);
L
Linus Torvalds 已提交
1836
	}
1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854
	if (type >= nr_swapfiles) {
		p->type = type;
		swap_info[type] = p;
		/*
		 * Write swap_info[type] before nr_swapfiles, in case a
		 * racing procfs swap_start() or swap_next() is reading them.
		 * (We never shrink nr_swapfiles, we never free this entry.)
		 */
		smp_wmb();
		nr_swapfiles++;
	} else {
		kfree(p);
		p = swap_info[type];
		/*
		 * Do not memset this entry: a racing procfs swap_next()
		 * would be relying on p->type to remain valid.
		 */
	}
1855
	INIT_LIST_HEAD(&p->first_swap_extent.list);
L
Linus Torvalds 已提交
1856 1857
	p->flags = SWP_USED;
	p->next = -1;
1858
	spin_unlock(&swap_lock);
1859

1860 1861 1862
	return p;
}

1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873
static int claim_swapfile(struct swap_info_struct *p, struct inode *inode)
{
	int error;

	if (S_ISBLK(inode->i_mode)) {
		p->bdev = bdgrab(I_BDEV(inode));
		error = blkdev_get(p->bdev,
				   FMODE_READ | FMODE_WRITE | FMODE_EXCL,
				   sys_swapon);
		if (error < 0) {
			p->bdev = NULL;
1874
			return -EINVAL;
1875 1876 1877 1878
		}
		p->old_block_size = block_size(p->bdev);
		error = set_blocksize(p->bdev, PAGE_SIZE);
		if (error < 0)
1879
			return error;
1880 1881 1882 1883
		p->flags |= SWP_BLKDEV;
	} else if (S_ISREG(inode->i_mode)) {
		p->bdev = inode->i_sb->s_bdev;
		mutex_lock(&inode->i_mutex);
1884 1885 1886 1887
		if (IS_SWAPFILE(inode))
			return -EBUSY;
	} else
		return -EINVAL;
1888 1889 1890 1891

	return 0;
}

1892 1893 1894 1895 1896 1897 1898 1899 1900 1901
static unsigned long read_swap_header(struct swap_info_struct *p,
					union swap_header *swap_header,
					struct inode *inode)
{
	int i;
	unsigned long maxpages;
	unsigned long swapfilepages;

	if (memcmp("SWAPSPACE2", swap_header->magic.magic, 10)) {
		printk(KERN_ERR "Unable to find swap-space signature\n");
1902
		return 0;
1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917
	}

	/* swap partition endianess hack... */
	if (swab32(swap_header->info.version) == 1) {
		swab32s(&swap_header->info.version);
		swab32s(&swap_header->info.last_page);
		swab32s(&swap_header->info.nr_badpages);
		for (i = 0; i < swap_header->info.nr_badpages; i++)
			swab32s(&swap_header->info.badpages[i]);
	}
	/* Check the swap header's sub-version */
	if (swap_header->info.version != 1) {
		printk(KERN_WARNING
		       "Unable to handle swap header version %d\n",
		       swap_header->info.version);
1918
		return 0;
1919 1920 1921 1922 1923 1924 1925 1926
	}

	p->lowest_bit  = 1;
	p->cluster_next = 1;
	p->cluster_nr = 0;

	/*
	 * Find out how many pages are allowed for a single swap
1927 1928 1929 1930 1931 1932
	 * device. There are three limiting factors: 1) the number
	 * of bits for the swap offset in the swp_entry_t type, and
	 * 2) the number of bits in the swap pte as defined by the
	 * the different architectures, and 3) the number of free bits
	 * in an exceptional radix_tree entry. In order to find the
	 * largest possible bit mask, a swap entry with swap type 0
1933
	 * and swap offset ~0UL is created, encoded to a swap pte,
1934
	 * decoded to a swp_entry_t again, and finally the swap
1935 1936 1937
	 * offset is extracted. This will mask all the bits from
	 * the initial ~0UL mask that can't be encoded in either
	 * the swp_entry_t or the architecture definition of a
1938
	 * swap pte.  Then the same is done for a radix_tree entry.
1939 1940
	 */
	maxpages = swp_offset(pte_to_swp_entry(
1941 1942 1943 1944
			swp_entry_to_pte(swp_entry(0, ~0UL))));
	maxpages = swp_offset(radix_to_swp_entry(
			swp_to_radix_entry(swp_entry(0, maxpages)))) + 1;

1945 1946 1947 1948 1949 1950 1951 1952 1953
	if (maxpages > swap_header->info.last_page) {
		maxpages = swap_header->info.last_page + 1;
		/* p->max is an unsigned int: don't overflow it */
		if ((unsigned int)maxpages == 0)
			maxpages = UINT_MAX;
	}
	p->highest_bit = maxpages - 1;

	if (!maxpages)
1954
		return 0;
1955 1956 1957 1958
	swapfilepages = i_size_read(inode) >> PAGE_SHIFT;
	if (swapfilepages && maxpages > swapfilepages) {
		printk(KERN_WARNING
		       "Swap area shorter than signature indicates\n");
1959
		return 0;
1960 1961
	}
	if (swap_header->info.nr_badpages && S_ISREG(inode->i_mode))
1962
		return 0;
1963
	if (swap_header->info.nr_badpages > MAX_SWAP_BADPAGES)
1964
		return 0;
1965 1966 1967 1968

	return maxpages;
}

1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982
static int setup_swap_map_and_extents(struct swap_info_struct *p,
					union swap_header *swap_header,
					unsigned char *swap_map,
					unsigned long maxpages,
					sector_t *span)
{
	int i;
	unsigned int nr_good_pages;
	int nr_extents;

	nr_good_pages = maxpages - 1;	/* omit header page */

	for (i = 0; i < swap_header->info.nr_badpages; i++) {
		unsigned int page_nr = swap_header->info.badpages[i];
1983 1984
		if (page_nr == 0 || page_nr > swap_header->info.last_page)
			return -EINVAL;
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995
		if (page_nr < maxpages) {
			swap_map[page_nr] = SWAP_MAP_BAD;
			nr_good_pages--;
		}
	}

	if (nr_good_pages) {
		swap_map[0] = SWAP_MAP_BAD;
		p->max = maxpages;
		p->pages = nr_good_pages;
		nr_extents = setup_swap_extents(p, span);
1996 1997
		if (nr_extents < 0)
			return nr_extents;
1998 1999 2000 2001
		nr_good_pages = p->pages;
	}
	if (!nr_good_pages) {
		printk(KERN_WARNING "Empty swap-file\n");
2002
		return -EINVAL;
2003 2004 2005 2006 2007
	}

	return nr_extents;
}

2008 2009 2010
SYSCALL_DEFINE2(swapon, const char __user *, specialfile, int, swap_flags)
{
	struct swap_info_struct *p;
2011
	char *name;
2012 2013
	struct file *swap_file = NULL;
	struct address_space *mapping;
2014 2015
	int i;
	int prio;
2016 2017
	int error;
	union swap_header *swap_header;
2018
	int nr_extents;
2019 2020 2021 2022 2023 2024 2025 2026 2027 2028
	sector_t span;
	unsigned long maxpages;
	unsigned char *swap_map = NULL;
	struct page *page = NULL;
	struct inode *inode = NULL;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;

	p = alloc_swap_info();
2029 2030
	if (IS_ERR(p))
		return PTR_ERR(p);
2031

L
Linus Torvalds 已提交
2032 2033
	name = getname(specialfile);
	if (IS_ERR(name)) {
2034
		error = PTR_ERR(name);
L
Linus Torvalds 已提交
2035
		name = NULL;
2036
		goto bad_swap;
L
Linus Torvalds 已提交
2037 2038 2039
	}
	swap_file = filp_open(name, O_RDWR|O_LARGEFILE, 0);
	if (IS_ERR(swap_file)) {
2040
		error = PTR_ERR(swap_file);
L
Linus Torvalds 已提交
2041
		swap_file = NULL;
2042
		goto bad_swap;
L
Linus Torvalds 已提交
2043 2044 2045 2046 2047 2048
	}

	p->swap_file = swap_file;
	mapping = swap_file->f_mapping;

	for (i = 0; i < nr_swapfiles; i++) {
2049
		struct swap_info_struct *q = swap_info[i];
L
Linus Torvalds 已提交
2050

2051
		if (q == p || !q->swap_file)
L
Linus Torvalds 已提交
2052
			continue;
2053 2054
		if (mapping == q->swap_file->f_mapping) {
			error = -EBUSY;
L
Linus Torvalds 已提交
2055
			goto bad_swap;
2056
		}
L
Linus Torvalds 已提交
2057 2058
	}

2059 2060
	inode = mapping->host;
	/* If S_ISREG(inode->i_mode) will do mutex_lock(&inode->i_mutex); */
2061 2062
	error = claim_swapfile(p, inode);
	if (unlikely(error))
L
Linus Torvalds 已提交
2063 2064 2065 2066 2067 2068 2069 2070 2071
		goto bad_swap;

	/*
	 * Read the swap header.
	 */
	if (!mapping->a_ops->readpage) {
		error = -EINVAL;
		goto bad_swap;
	}
2072
	page = read_mapping_page(mapping, 0, swap_file);
L
Linus Torvalds 已提交
2073 2074 2075 2076
	if (IS_ERR(page)) {
		error = PTR_ERR(page);
		goto bad_swap;
	}
2077
	swap_header = kmap(page);
L
Linus Torvalds 已提交
2078

2079 2080
	maxpages = read_swap_header(p, swap_header, inode);
	if (unlikely(!maxpages)) {
L
Linus Torvalds 已提交
2081 2082 2083
		error = -EINVAL;
		goto bad_swap;
	}
2084

2085
	/* OK, set up the swap map and apply the bad block list */
2086
	swap_map = vzalloc(maxpages);
2087 2088 2089 2090
	if (!swap_map) {
		error = -ENOMEM;
		goto bad_swap;
	}
L
Linus Torvalds 已提交
2091

2092 2093 2094 2095
	error = swap_cgroup_swapon(p->type, maxpages);
	if (error)
		goto bad_swap;

2096 2097 2098 2099
	nr_extents = setup_swap_map_and_extents(p, swap_header, swap_map,
		maxpages, &span);
	if (unlikely(nr_extents < 0)) {
		error = nr_extents;
L
Linus Torvalds 已提交
2100 2101 2102
		goto bad_swap;
	}

2103 2104 2105 2106 2107
	if (p->bdev) {
		if (blk_queue_nonrot(bdev_get_queue(p->bdev))) {
			p->flags |= SWP_SOLIDSTATE;
			p->cluster_next = 1 + (random32() % p->highest_bit);
		}
2108
		if (discard_swap(p) == 0 && (swap_flags & SWAP_FLAG_DISCARD))
2109
			p->flags |= SWP_DISCARDABLE;
2110
	}
2111

I
Ingo Molnar 已提交
2112
	mutex_lock(&swapon_mutex);
2113
	prio = -1;
2114
	if (swap_flags & SWAP_FLAG_PREFER)
2115
		prio =
2116
		  (swap_flags & SWAP_FLAG_PRIO_MASK) >> SWAP_FLAG_PRIO_SHIFT;
2117
	enable_swap_info(p, prio, swap_map);
2118 2119 2120 2121 2122 2123 2124 2125

	printk(KERN_INFO "Adding %uk swap on %s.  "
			"Priority:%d extents:%d across:%lluk %s%s\n",
		p->pages<<(PAGE_SHIFT-10), name, p->prio,
		nr_extents, (unsigned long long)span<<(PAGE_SHIFT-10),
		(p->flags & SWP_SOLIDSTATE) ? "SS" : "",
		(p->flags & SWP_DISCARDABLE) ? "D" : "");

I
Ingo Molnar 已提交
2126
	mutex_unlock(&swapon_mutex);
K
Kay Sievers 已提交
2127 2128 2129
	atomic_inc(&proc_poll_event);
	wake_up_interruptible(&proc_poll_wait);

2130 2131
	if (S_ISREG(inode->i_mode))
		inode->i_flags |= S_SWAPFILE;
L
Linus Torvalds 已提交
2132 2133 2134
	error = 0;
	goto out;
bad_swap:
2135
	if (inode && S_ISBLK(inode->i_mode) && p->bdev) {
2136 2137
		set_blocksize(p->bdev, p->old_block_size);
		blkdev_put(p->bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL);
L
Linus Torvalds 已提交
2138
	}
2139
	destroy_swap_extents(p);
2140
	swap_cgroup_swapoff(p->type);
2141
	spin_lock(&swap_lock);
L
Linus Torvalds 已提交
2142 2143
	p->swap_file = NULL;
	p->flags = 0;
2144
	spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
2145
	vfree(swap_map);
2146
	if (swap_file) {
2147
		if (inode && S_ISREG(inode->i_mode)) {
2148
			mutex_unlock(&inode->i_mutex);
2149 2150
			inode = NULL;
		}
L
Linus Torvalds 已提交
2151
		filp_close(swap_file, NULL);
2152
	}
L
Linus Torvalds 已提交
2153 2154 2155 2156 2157 2158 2159
out:
	if (page && !IS_ERR(page)) {
		kunmap(page);
		page_cache_release(page);
	}
	if (name)
		putname(name);
2160
	if (inode && S_ISREG(inode->i_mode))
2161
		mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
2162 2163 2164 2165 2166
	return error;
}

void si_swapinfo(struct sysinfo *val)
{
2167
	unsigned int type;
L
Linus Torvalds 已提交
2168 2169
	unsigned long nr_to_be_unused = 0;

2170
	spin_lock(&swap_lock);
2171 2172 2173 2174 2175
	for (type = 0; type < nr_swapfiles; type++) {
		struct swap_info_struct *si = swap_info[type];

		if ((si->flags & SWP_USED) && !(si->flags & SWP_WRITEOK))
			nr_to_be_unused += si->inuse_pages;
L
Linus Torvalds 已提交
2176 2177 2178
	}
	val->freeswap = nr_swap_pages + nr_to_be_unused;
	val->totalswap = total_swap_pages + nr_to_be_unused;
2179
	spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
2180 2181 2182 2183 2184
}

/*
 * Verify that a swap entry is valid and increment its swap map count.
 *
2185 2186 2187 2188 2189 2190
 * Returns error code in following case.
 * - success -> 0
 * - swp_entry is invalid -> EINVAL
 * - swp_entry is migration entry -> EINVAL
 * - swap-cache reference is requested but there is already one. -> EEXIST
 * - swap-cache reference is requested but the entry is not used. -> ENOENT
H
Hugh Dickins 已提交
2191
 * - swap-mapped reference requested but needs continued swap count. -> ENOMEM
L
Linus Torvalds 已提交
2192
 */
2193
static int __swap_duplicate(swp_entry_t entry, unsigned char usage)
L
Linus Torvalds 已提交
2194
{
2195
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
2196
	unsigned long offset, type;
2197 2198
	unsigned char count;
	unsigned char has_cache;
H
Hugh Dickins 已提交
2199
	int err = -EINVAL;
L
Linus Torvalds 已提交
2200

2201
	if (non_swap_entry(entry))
H
Hugh Dickins 已提交
2202
		goto out;
2203

L
Linus Torvalds 已提交
2204 2205 2206
	type = swp_type(entry);
	if (type >= nr_swapfiles)
		goto bad_file;
2207
	p = swap_info[type];
L
Linus Torvalds 已提交
2208 2209
	offset = swp_offset(entry);

2210
	spin_lock(&swap_lock);
2211 2212 2213
	if (unlikely(offset >= p->max))
		goto unlock_out;

H
Hugh Dickins 已提交
2214 2215 2216 2217
	count = p->swap_map[offset];
	has_cache = count & SWAP_HAS_CACHE;
	count &= ~SWAP_HAS_CACHE;
	err = 0;
2218

H
Hugh Dickins 已提交
2219
	if (usage == SWAP_HAS_CACHE) {
2220 2221

		/* set SWAP_HAS_CACHE if there is no cache and entry is used */
H
Hugh Dickins 已提交
2222 2223 2224 2225 2226 2227
		if (!has_cache && count)
			has_cache = SWAP_HAS_CACHE;
		else if (has_cache)		/* someone else added cache */
			err = -EEXIST;
		else				/* no users remaining */
			err = -ENOENT;
2228 2229

	} else if (count || has_cache) {
H
Hugh Dickins 已提交
2230

H
Hugh Dickins 已提交
2231 2232 2233
		if ((count & ~COUNT_CONTINUED) < SWAP_MAP_MAX)
			count += usage;
		else if ((count & ~COUNT_CONTINUED) > SWAP_MAP_MAX)
H
Hugh Dickins 已提交
2234
			err = -EINVAL;
H
Hugh Dickins 已提交
2235 2236 2237 2238
		else if (swap_count_continued(p, offset, count))
			count = COUNT_CONTINUED;
		else
			err = -ENOMEM;
2239
	} else
H
Hugh Dickins 已提交
2240 2241 2242 2243
		err = -ENOENT;			/* unused swap entry */

	p->swap_map[offset] = count | has_cache;

2244
unlock_out:
2245
	spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
2246
out:
H
Hugh Dickins 已提交
2247
	return err;
L
Linus Torvalds 已提交
2248 2249 2250 2251 2252

bad_file:
	printk(KERN_ERR "swap_dup: %s%08lx\n", Bad_file, entry.val);
	goto out;
}
H
Hugh Dickins 已提交
2253

H
Hugh Dickins 已提交
2254 2255 2256 2257 2258 2259 2260 2261 2262
/*
 * Help swapoff by noting that swap entry belongs to shmem/tmpfs
 * (in which case its reference count is never incremented).
 */
void swap_shmem_alloc(swp_entry_t entry)
{
	__swap_duplicate(entry, SWAP_MAP_SHMEM);
}

2263
/*
2264 2265 2266 2267 2268
 * Increase reference count of swap entry by 1.
 * Returns 0 for success, or -ENOMEM if a swap_count_continuation is required
 * but could not be atomically allocated.  Returns 0, just as if it succeeded,
 * if __swap_duplicate() fails for another reason (-EINVAL or -ENOENT), which
 * might occur if a page table entry has got corrupted.
2269
 */
H
Hugh Dickins 已提交
2270
int swap_duplicate(swp_entry_t entry)
2271
{
H
Hugh Dickins 已提交
2272 2273 2274 2275 2276
	int err = 0;

	while (!err && __swap_duplicate(entry, 1) == -ENOMEM)
		err = add_swap_count_continuation(entry, GFP_ATOMIC);
	return err;
2277
}
L
Linus Torvalds 已提交
2278

2279
/*
2280 2281
 * @entry: swap entry for which we allocate swap cache.
 *
2282
 * Called when allocating swap cache for existing swap entry,
2283 2284 2285
 * This can return error codes. Returns 0 at success.
 * -EBUSY means there is a swap cache.
 * Note: return code is different from swap_duplicate().
2286 2287 2288
 */
int swapcache_prepare(swp_entry_t entry)
{
H
Hugh Dickins 已提交
2289
	return __swap_duplicate(entry, SWAP_HAS_CACHE);
2290 2291
}

L
Linus Torvalds 已提交
2292
/*
2293
 * swap_lock prevents swap_map being freed. Don't grab an extra
L
Linus Torvalds 已提交
2294 2295 2296 2297
 * reference on the swaphandle, it doesn't matter if it becomes unused.
 */
int valid_swaphandles(swp_entry_t entry, unsigned long *offset)
{
2298
	struct swap_info_struct *si;
H
Hugh Dickins 已提交
2299
	int our_page_cluster = page_cluster;
2300 2301 2302
	pgoff_t target, toff;
	pgoff_t base, end;
	int nr_pages = 0;
L
Linus Torvalds 已提交
2303

H
Hugh Dickins 已提交
2304
	if (!our_page_cluster)	/* no readahead */
L
Linus Torvalds 已提交
2305
		return 0;
2306

2307
	si = swap_info[swp_type(entry)];
2308 2309 2310 2311 2312
	target = swp_offset(entry);
	base = (target >> our_page_cluster) << our_page_cluster;
	end = base + (1 << our_page_cluster);
	if (!base)		/* first page is swap header */
		base++;
L
Linus Torvalds 已提交
2313

2314
	spin_lock(&swap_lock);
2315 2316 2317 2318 2319 2320 2321 2322
	if (end > si->max)	/* don't go beyond end of map */
		end = si->max;

	/* Count contiguous allocated slots above our target */
	for (toff = target; ++toff < end; nr_pages++) {
		/* Don't read in free or bad pages */
		if (!si->swap_map[toff])
			break;
2323
		if (swap_count(si->swap_map[toff]) == SWAP_MAP_BAD)
L
Linus Torvalds 已提交
2324
			break;
2325 2326 2327
	}
	/* Count contiguous allocated slots below our target */
	for (toff = target; --toff >= base; nr_pages++) {
L
Linus Torvalds 已提交
2328
		/* Don't read in free or bad pages */
2329
		if (!si->swap_map[toff])
L
Linus Torvalds 已提交
2330
			break;
2331
		if (swap_count(si->swap_map[toff]) == SWAP_MAP_BAD)
L
Linus Torvalds 已提交
2332
			break;
2333
	}
2334
	spin_unlock(&swap_lock);
2335 2336 2337 2338 2339 2340 2341

	/*
	 * Indicate starting offset, and return number of pages to get:
	 * if only 1, say 0, since there's then no readahead to be done.
	 */
	*offset = ++toff;
	return nr_pages? ++nr_pages: 0;
L
Linus Torvalds 已提交
2342
}
H
Hugh Dickins 已提交
2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558

/*
 * add_swap_count_continuation - called when a swap count is duplicated
 * beyond SWAP_MAP_MAX, it allocates a new page and links that to the entry's
 * page of the original vmalloc'ed swap_map, to hold the continuation count
 * (for that entry and for its neighbouring PAGE_SIZE swap entries).  Called
 * again when count is duplicated beyond SWAP_MAP_MAX * SWAP_CONT_MAX, etc.
 *
 * These continuation pages are seldom referenced: the common paths all work
 * on the original swap_map, only referring to a continuation page when the
 * low "digit" of a count is incremented or decremented through SWAP_MAP_MAX.
 *
 * add_swap_count_continuation(, GFP_ATOMIC) can be called while holding
 * page table locks; if it fails, add_swap_count_continuation(, GFP_KERNEL)
 * can be called after dropping locks.
 */
int add_swap_count_continuation(swp_entry_t entry, gfp_t gfp_mask)
{
	struct swap_info_struct *si;
	struct page *head;
	struct page *page;
	struct page *list_page;
	pgoff_t offset;
	unsigned char count;

	/*
	 * When debugging, it's easier to use __GFP_ZERO here; but it's better
	 * for latency not to zero a page while GFP_ATOMIC and holding locks.
	 */
	page = alloc_page(gfp_mask | __GFP_HIGHMEM);

	si = swap_info_get(entry);
	if (!si) {
		/*
		 * An acceptable race has occurred since the failing
		 * __swap_duplicate(): the swap entry has been freed,
		 * perhaps even the whole swap_map cleared for swapoff.
		 */
		goto outer;
	}

	offset = swp_offset(entry);
	count = si->swap_map[offset] & ~SWAP_HAS_CACHE;

	if ((count & ~COUNT_CONTINUED) != SWAP_MAP_MAX) {
		/*
		 * The higher the swap count, the more likely it is that tasks
		 * will race to add swap count continuation: we need to avoid
		 * over-provisioning.
		 */
		goto out;
	}

	if (!page) {
		spin_unlock(&swap_lock);
		return -ENOMEM;
	}

	/*
	 * We are fortunate that although vmalloc_to_page uses pte_offset_map,
	 * no architecture is using highmem pages for kernel pagetables: so it
	 * will not corrupt the GFP_ATOMIC caller's atomic pagetable kmaps.
	 */
	head = vmalloc_to_page(si->swap_map + offset);
	offset &= ~PAGE_MASK;

	/*
	 * Page allocation does not initialize the page's lru field,
	 * but it does always reset its private field.
	 */
	if (!page_private(head)) {
		BUG_ON(count & COUNT_CONTINUED);
		INIT_LIST_HEAD(&head->lru);
		set_page_private(head, SWP_CONTINUED);
		si->flags |= SWP_CONTINUED;
	}

	list_for_each_entry(list_page, &head->lru, lru) {
		unsigned char *map;

		/*
		 * If the previous map said no continuation, but we've found
		 * a continuation page, free our allocation and use this one.
		 */
		if (!(count & COUNT_CONTINUED))
			goto out;

		map = kmap_atomic(list_page, KM_USER0) + offset;
		count = *map;
		kunmap_atomic(map, KM_USER0);

		/*
		 * If this continuation count now has some space in it,
		 * free our allocation and use this one.
		 */
		if ((count & ~COUNT_CONTINUED) != SWAP_CONT_MAX)
			goto out;
	}

	list_add_tail(&page->lru, &head->lru);
	page = NULL;			/* now it's attached, don't free it */
out:
	spin_unlock(&swap_lock);
outer:
	if (page)
		__free_page(page);
	return 0;
}

/*
 * swap_count_continued - when the original swap_map count is incremented
 * from SWAP_MAP_MAX, check if there is already a continuation page to carry
 * into, carry if so, or else fail until a new continuation page is allocated;
 * when the original swap_map count is decremented from 0 with continuation,
 * borrow from the continuation and report whether it still holds more.
 * Called while __swap_duplicate() or swap_entry_free() holds swap_lock.
 */
static bool swap_count_continued(struct swap_info_struct *si,
				 pgoff_t offset, unsigned char count)
{
	struct page *head;
	struct page *page;
	unsigned char *map;

	head = vmalloc_to_page(si->swap_map + offset);
	if (page_private(head) != SWP_CONTINUED) {
		BUG_ON(count & COUNT_CONTINUED);
		return false;		/* need to add count continuation */
	}

	offset &= ~PAGE_MASK;
	page = list_entry(head->lru.next, struct page, lru);
	map = kmap_atomic(page, KM_USER0) + offset;

	if (count == SWAP_MAP_MAX)	/* initial increment from swap_map */
		goto init_map;		/* jump over SWAP_CONT_MAX checks */

	if (count == (SWAP_MAP_MAX | COUNT_CONTINUED)) { /* incrementing */
		/*
		 * Think of how you add 1 to 999
		 */
		while (*map == (SWAP_CONT_MAX | COUNT_CONTINUED)) {
			kunmap_atomic(map, KM_USER0);
			page = list_entry(page->lru.next, struct page, lru);
			BUG_ON(page == head);
			map = kmap_atomic(page, KM_USER0) + offset;
		}
		if (*map == SWAP_CONT_MAX) {
			kunmap_atomic(map, KM_USER0);
			page = list_entry(page->lru.next, struct page, lru);
			if (page == head)
				return false;	/* add count continuation */
			map = kmap_atomic(page, KM_USER0) + offset;
init_map:		*map = 0;		/* we didn't zero the page */
		}
		*map += 1;
		kunmap_atomic(map, KM_USER0);
		page = list_entry(page->lru.prev, struct page, lru);
		while (page != head) {
			map = kmap_atomic(page, KM_USER0) + offset;
			*map = COUNT_CONTINUED;
			kunmap_atomic(map, KM_USER0);
			page = list_entry(page->lru.prev, struct page, lru);
		}
		return true;			/* incremented */

	} else {				/* decrementing */
		/*
		 * Think of how you subtract 1 from 1000
		 */
		BUG_ON(count != COUNT_CONTINUED);
		while (*map == COUNT_CONTINUED) {
			kunmap_atomic(map, KM_USER0);
			page = list_entry(page->lru.next, struct page, lru);
			BUG_ON(page == head);
			map = kmap_atomic(page, KM_USER0) + offset;
		}
		BUG_ON(*map == 0);
		*map -= 1;
		if (*map == 0)
			count = 0;
		kunmap_atomic(map, KM_USER0);
		page = list_entry(page->lru.prev, struct page, lru);
		while (page != head) {
			map = kmap_atomic(page, KM_USER0) + offset;
			*map = SWAP_CONT_MAX | count;
			count = COUNT_CONTINUED;
			kunmap_atomic(map, KM_USER0);
			page = list_entry(page->lru.prev, struct page, lru);
		}
		return count == COUNT_CONTINUED;
	}
}

/*
 * free_swap_count_continuations - swapoff free all the continuation pages
 * appended to the swap_map, after swap_map is quiesced, before vfree'ing it.
 */
static void free_swap_count_continuations(struct swap_info_struct *si)
{
	pgoff_t offset;

	for (offset = 0; offset < si->max; offset += PAGE_SIZE) {
		struct page *head;
		head = vmalloc_to_page(si->swap_map + offset);
		if (page_private(head)) {
			struct list_head *this, *next;
			list_for_each_safe(this, next, &head->lru) {
				struct page *page;
				page = list_entry(this, struct page, lru);
				list_del(this);
				__free_page(page);
			}
		}
	}
}