swapfile.c 65.6 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 17 18
/*
 *  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>
#include <linux/shm.h>
#include <linux/blkdev.h>
19
#include <linux/random.h>
L
Linus Torvalds 已提交
20 21 22 23 24
#include <linux/writeback.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/init.h>
#include <linux/module.h>
H
Hugh Dickins 已提交
25
#include <linux/ksm.h>
L
Linus Torvalds 已提交
26 27 28
#include <linux/rmap.h>
#include <linux/security.h>
#include <linux/backing-dev.h>
I
Ingo Molnar 已提交
29
#include <linux/mutex.h>
30
#include <linux/capability.h>
L
Linus Torvalds 已提交
31
#include <linux/syscalls.h>
32
#include <linux/memcontrol.h>
K
Kay Sievers 已提交
33
#include <linux/poll.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

L
Linus Torvalds 已提交
97 98
/*
 * We need this because the bdev->unplug_fn can sleep and we cannot
99
 * hold swap_lock while calling the unplug_fn. And swap_lock
I
Ingo Molnar 已提交
100
 * cannot be turned into a mutex.
L
Linus Torvalds 已提交
101 102 103 104 105 106 107 108
 */
static DECLARE_RWSEM(swap_unplug_sem);

void swap_unplug_io_fn(struct backing_dev_info *unused_bdi, struct page *page)
{
	swp_entry_t entry;

	down_read(&swap_unplug_sem);
H
Hugh Dickins 已提交
109
	entry.val = page_private(page);
L
Linus Torvalds 已提交
110
	if (PageSwapCache(page)) {
111
		struct block_device *bdev = swap_info[swp_type(entry)]->bdev;
L
Linus Torvalds 已提交
112 113 114 115 116
		struct backing_dev_info *bdi;

		/*
		 * If the page is removed from swapcache from under us (with a
		 * racy try_to_unuse/swapoff) we need an additional reference
H
Hugh Dickins 已提交
117 118
		 * count to avoid reading garbage from page_private(page) above.
		 * If the WARN_ON triggers during a swapoff it maybe the race
L
Linus Torvalds 已提交
119 120 121 122 123 124
		 * condition and it's harmless. However if it triggers without
		 * swapoff it signals a problem.
		 */
		WARN_ON(page_count(page) <= 1);

		bdi = bdev->bd_inode->i_mapping->backing_dev_info;
M
McMullan, Jason 已提交
125
		blk_run_backing_dev(bdi, page);
L
Linus Torvalds 已提交
126 127 128 129
	}
	up_read(&swap_unplug_sem);
}

130 131 132 133 134 135 136
/*
 * 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;
137 138
	sector_t start_block;
	sector_t nr_blocks;
139 140
	int err = 0;

141 142 143 144 145 146
	/* 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,
147
				nr_blocks, GFP_KERNEL, 0);
148 149 150 151
		if (err)
			return err;
		cond_resched();
	}
152

153 154 155
	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);
156 157

		err = blkdev_issue_discard(si->bdev, start_block,
158
				nr_blocks, GFP_KERNEL, 0);
159 160 161 162 163 164 165 166
		if (err)
			break;

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

167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
/*
 * 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;
184
			sector_t nr_blocks = se->nr_pages - offset;
185 186 187 188 189 190 191 192 193 194 195 196

			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,
197
				    nr_blocks, GFP_NOIO, 0))
198 199 200 201 202 203 204 205 206 207 208 209 210 211
				break;
		}

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

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

212 213 214
#define SWAPFILE_CLUSTER	256
#define LATENCY_LIMIT		256

215
static inline unsigned long scan_swap_map(struct swap_info_struct *si,
216
					  unsigned char usage)
L
Linus Torvalds 已提交
217
{
218
	unsigned long offset;
219
	unsigned long scan_base;
220
	unsigned long last_in_cluster = 0;
221
	int latency_ration = LATENCY_LIMIT;
222
	int found_free_cluster = 0;
223

224
	/*
225 226 227 228 229 230 231
	 * 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
232
	 * And we let swap pages go all over an SSD partition.  Hugh
233 234
	 */

235
	si->flags += SWP_SCANNING;
236
	scan_base = offset = si->cluster_next;
237 238 239 240 241 242

	if (unlikely(!si->cluster_nr--)) {
		if (si->pages - si->inuse_pages < SWAPFILE_CLUSTER) {
			si->cluster_nr = SWAPFILE_CLUSTER - 1;
			goto checks;
		}
243 244 245 246 247 248 249 250 251 252 253 254 255
		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;
		}
256
		spin_unlock(&swap_lock);
257

258 259 260 261 262 263 264 265 266 267
		/*
		 * 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;
268 269 270 271
		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
272
			if (si->swap_map[offset])
273 274
				last_in_cluster = offset + SWAPFILE_CLUSTER;
			else if (offset == last_in_cluster) {
275
				spin_lock(&swap_lock);
276 277 278
				offset -= SWAPFILE_CLUSTER - 1;
				si->cluster_next = offset;
				si->cluster_nr = SWAPFILE_CLUSTER - 1;
279
				found_free_cluster = 1;
280
				goto checks;
L
Linus Torvalds 已提交
281
			}
282 283 284 285
			if (unlikely(--latency_ration < 0)) {
				cond_resched();
				latency_ration = LATENCY_LIMIT;
			}
286
		}
287 288

		offset = si->lowest_bit;
289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309
		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;
310
		spin_lock(&swap_lock);
311
		si->cluster_nr = SWAPFILE_CLUSTER - 1;
312
		si->lowest_alloc = 0;
L
Linus Torvalds 已提交
313
	}
314

315 316
checks:
	if (!(si->flags & SWP_WRITEOK))
317
		goto no_page;
318 319
	if (!si->highest_bit)
		goto no_page;
320
	if (offset > si->highest_bit)
321
		scan_base = offset = si->lowest_bit;
322

323 324
	/* reuse swap entry of cache-only swap if not busy. */
	if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
325 326 327 328 329 330 331 332 333 334
		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 */
	}

335 336 337 338 339 340 341 342 343 344 345
	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
346
	}
H
Hugh Dickins 已提交
347
	si->swap_map[offset] = usage;
348 349
	si->cluster_next = offset + 1;
	si->flags -= SWP_SCANNING;
350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403

	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;
		}
	}
404
	return offset;
405

406
scan:
407
	spin_unlock(&swap_lock);
408
	while (++offset <= si->highest_bit) {
409
		if (!si->swap_map[offset]) {
410
			spin_lock(&swap_lock);
411 412
			goto checks;
		}
413 414 415 416
		if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
			spin_lock(&swap_lock);
			goto checks;
		}
417 418 419 420
		if (unlikely(--latency_ration < 0)) {
			cond_resched();
			latency_ration = LATENCY_LIMIT;
		}
421
	}
422 423 424 425 426 427
	offset = si->lowest_bit;
	while (++offset < scan_base) {
		if (!si->swap_map[offset]) {
			spin_lock(&swap_lock);
			goto checks;
		}
428 429 430 431
		if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
			spin_lock(&swap_lock);
			goto checks;
		}
432 433 434 435 436
		if (unlikely(--latency_ration < 0)) {
			cond_resched();
			latency_ration = LATENCY_LIMIT;
		}
	}
437
	spin_lock(&swap_lock);
438 439

no_page:
440
	si->flags -= SWP_SCANNING;
L
Linus Torvalds 已提交
441 442 443 444 445
	return 0;
}

swp_entry_t get_swap_page(void)
{
446 447 448 449
	struct swap_info_struct *si;
	pgoff_t offset;
	int type, next;
	int wrapped = 0;
L
Linus Torvalds 已提交
450

451
	spin_lock(&swap_lock);
L
Linus Torvalds 已提交
452
	if (nr_swap_pages <= 0)
453 454 455 456
		goto noswap;
	nr_swap_pages--;

	for (type = swap_list.next; type >= 0 && wrapped < 2; type = next) {
457
		si = swap_info[type];
458 459
		next = si->next;
		if (next < 0 ||
460
		    (!wrapped && si->prio != swap_info[next]->prio)) {
461 462
			next = swap_list.head;
			wrapped++;
L
Linus Torvalds 已提交
463
		}
464 465 466 467 468 469 470

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

		swap_list.next = next;
471
		/* This is called for allocating swap entry for cache */
H
Hugh Dickins 已提交
472
		offset = scan_swap_map(si, SWAP_HAS_CACHE);
473 474
		if (offset) {
			spin_unlock(&swap_lock);
475
			return swp_entry(type, offset);
476
		}
477
		next = swap_list.next;
L
Linus Torvalds 已提交
478
	}
479 480 481

	nr_swap_pages++;
noswap:
482
	spin_unlock(&swap_lock);
483
	return (swp_entry_t) {0};
L
Linus Torvalds 已提交
484 485
}

486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507
/* 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};
}

508
static struct swap_info_struct *swap_info_get(swp_entry_t entry)
L
Linus Torvalds 已提交
509
{
510
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
511 512 513 514 515 516 517
	unsigned long offset, type;

	if (!entry.val)
		goto out;
	type = swp_type(entry);
	if (type >= nr_swapfiles)
		goto bad_nofile;
518
	p = swap_info[type];
L
Linus Torvalds 已提交
519 520 521 522 523 524 525
	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;
526
	spin_lock(&swap_lock);
L
Linus Torvalds 已提交
527 528 529 530 531 532 533 534 535 536 537 538 539 540 541
	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;
542
}
L
Linus Torvalds 已提交
543

544 545
static unsigned char swap_entry_free(struct swap_info_struct *p,
				     swp_entry_t entry, unsigned char usage)
L
Linus Torvalds 已提交
546
{
H
Hugh Dickins 已提交
547
	unsigned long offset = swp_offset(entry);
548 549
	unsigned char count;
	unsigned char has_cache;
550

H
Hugh Dickins 已提交
551 552 553
	count = p->swap_map[offset];
	has_cache = count & SWAP_HAS_CACHE;
	count &= ~SWAP_HAS_CACHE;
554

H
Hugh Dickins 已提交
555
	if (usage == SWAP_HAS_CACHE) {
556
		VM_BUG_ON(!has_cache);
H
Hugh Dickins 已提交
557
		has_cache = 0;
H
Hugh Dickins 已提交
558 559 560 561 562 563
	} 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
564 565 566 567 568 569 570 571 572
	} 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
573 574 575 576 577 578

	if (!count)
		mem_cgroup_uncharge_swap(entry);

	usage = count | has_cache;
	p->swap_map[offset] = usage;
579 580

	/* free if no reference */
H
Hugh Dickins 已提交
581
	if (!usage) {
582
		struct gendisk *disk = p->bdev->bd_disk;
583 584 585 586
		if (offset < p->lowest_bit)
			p->lowest_bit = offset;
		if (offset > p->highest_bit)
			p->highest_bit = offset;
587 588 589
		if (swap_list.next >= 0 &&
		    p->prio > swap_info[swap_list.next]->prio)
			swap_list.next = p->type;
590 591
		nr_swap_pages++;
		p->inuse_pages--;
592 593 594
		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
595
	}
H
Hugh Dickins 已提交
596 597

	return usage;
L
Linus Torvalds 已提交
598 599 600 601 602 603 604 605
}

/*
 * 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)
{
606
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
607 608 609

	p = swap_info_get(entry);
	if (p) {
H
Hugh Dickins 已提交
610
		swap_entry_free(p, entry, 1);
611
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
612 613 614
	}
}

615 616 617 618 619
/*
 * Called after dropping swapcache to decrease refcnt to swap entries.
 */
void swapcache_free(swp_entry_t entry, struct page *page)
{
620
	struct swap_info_struct *p;
621
	unsigned char count;
622 623 624

	p = swap_info_get(entry);
	if (p) {
H
Hugh Dickins 已提交
625 626 627
		count = swap_entry_free(p, entry, SWAP_HAS_CACHE);
		if (page)
			mem_cgroup_uncharge_swapcache(page, entry, count != 0);
628 629
		spin_unlock(&swap_lock);
	}
630 631
}

L
Linus Torvalds 已提交
632
/*
633
 * How many references to page are currently swapped out?
H
Hugh Dickins 已提交
634 635
 * 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
636
 */
637
static inline int page_swapcount(struct page *page)
L
Linus Torvalds 已提交
638
{
639 640
	int count = 0;
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
641 642
	swp_entry_t entry;

H
Hugh Dickins 已提交
643
	entry.val = page_private(page);
L
Linus Torvalds 已提交
644 645
	p = swap_info_get(entry);
	if (p) {
646
		count = swap_count(p->swap_map[swp_offset(entry)]);
647
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
648
	}
649
	return count;
L
Linus Torvalds 已提交
650 651 652
}

/*
653 654 655 656
 * 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
657
 */
658
int reuse_swap_page(struct page *page)
L
Linus Torvalds 已提交
659
{
660 661
	int count;

662
	VM_BUG_ON(!PageLocked(page));
H
Hugh Dickins 已提交
663 664
	if (unlikely(PageKsm(page)))
		return 0;
665
	count = page_mapcount(page);
666
	if (count <= 1 && PageSwapCache(page)) {
667
		count += page_swapcount(page);
668 669 670 671 672
		if (count == 1 && !PageWriteback(page)) {
			delete_from_swap_cache(page);
			SetPageDirty(page);
		}
	}
H
Hugh Dickins 已提交
673
	return count <= 1;
L
Linus Torvalds 已提交
674 675 676
}

/*
677 678
 * 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
679
 */
680
int try_to_free_swap(struct page *page)
L
Linus Torvalds 已提交
681
{
682
	VM_BUG_ON(!PageLocked(page));
L
Linus Torvalds 已提交
683 684 685 686 687

	if (!PageSwapCache(page))
		return 0;
	if (PageWriteback(page))
		return 0;
688
	if (page_swapcount(page))
L
Linus Torvalds 已提交
689 690
		return 0;

691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708
	/*
	 * 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.
	 *
	 * Hibernation clears bits from gfp_allowed_mask to prevent
	 * memory reclaim from writing to disk, so check that here.
	 */
	if (!(gfp_allowed_mask & __GFP_IO))
		return 0;

709 710 711
	delete_from_swap_cache(page);
	SetPageDirty(page);
	return 1;
712 713
}

L
Linus Torvalds 已提交
714 715 716 717
/*
 * Free the swap entry like above, but also try to
 * free the page cache entry if it is the last user.
 */
718
int free_swap_and_cache(swp_entry_t entry)
L
Linus Torvalds 已提交
719
{
720
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
721 722
	struct page *page = NULL;

723
	if (non_swap_entry(entry))
724
		return 1;
725

L
Linus Torvalds 已提交
726 727
	p = swap_info_get(entry);
	if (p) {
H
Hugh Dickins 已提交
728
		if (swap_entry_free(p, entry, 1) == SWAP_HAS_CACHE) {
729
			page = find_get_page(&swapper_space, entry.val);
N
Nick Piggin 已提交
730
			if (page && !trylock_page(page)) {
731 732 733 734
				page_cache_release(page);
				page = NULL;
			}
		}
735
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
736 737
	}
	if (page) {
738 739 740 741
		/*
		 * Not mapped elsewhere, or swap space full? Free it!
		 * Also recheck PageSwapCache now page is locked (above).
		 */
742
		if (PageSwapCache(page) && !PageWriteback(page) &&
743
				(!page_mapped(page) || vm_swap_full())) {
L
Linus Torvalds 已提交
744 745 746 747 748 749
			delete_from_swap_cache(page);
			SetPageDirty(page);
		}
		unlock_page(page);
		page_cache_release(page);
	}
750
	return p != NULL;
L
Linus Torvalds 已提交
751 752
}

753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783
#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

784
#ifdef CONFIG_HIBERNATION
785
/*
786
 * Find the swap type that corresponds to given device (if any).
787
 *
788 789 790 791
 * @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).
792
 */
793
int swap_type_of(dev_t device, sector_t offset, struct block_device **bdev_p)
794
{
795
	struct block_device *bdev = NULL;
796
	int type;
797

798 799 800
	if (device)
		bdev = bdget(device);

801
	spin_lock(&swap_lock);
802 803
	for (type = 0; type < nr_swapfiles; type++) {
		struct swap_info_struct *sis = swap_info[type];
804

805
		if (!(sis->flags & SWP_WRITEOK))
806
			continue;
807

808
		if (!bdev) {
809
			if (bdev_p)
810
				*bdev_p = bdgrab(sis->bdev);
811

812
			spin_unlock(&swap_lock);
813
			return type;
814
		}
815
		if (bdev == sis->bdev) {
816
			struct swap_extent *se = &sis->first_swap_extent;
817 818

			if (se->start_block == offset) {
819
				if (bdev_p)
820
					*bdev_p = bdgrab(sis->bdev);
821

822 823
				spin_unlock(&swap_lock);
				bdput(bdev);
824
				return type;
825
			}
826 827 828
		}
	}
	spin_unlock(&swap_lock);
829 830 831
	if (bdev)
		bdput(bdev);

832 833 834
	return -ENODEV;
}

835 836 837 838 839 840 841 842 843 844 845 846
/*
 * 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;
847
	return map_swap_entry(swp_entry(type, offset), &bdev);
848 849
}

850 851 852 853 854 855 856 857 858 859
/*
 * 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;

860 861 862 863 864 865
	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;
866
			if (free)
867
				n -= sis->inuse_pages;
868 869
		}
	}
870
	spin_unlock(&swap_lock);
871 872
	return n;
}
873
#endif /* CONFIG_HIBERNATION */
874

L
Linus Torvalds 已提交
875
/*
876 877 878
 * 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
879
 */
H
Hugh Dickins 已提交
880
static int unuse_pte(struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
881 882
		unsigned long addr, swp_entry_t entry, struct page *page)
{
883
	struct mem_cgroup *ptr = NULL;
H
Hugh Dickins 已提交
884 885 886 887
	spinlock_t *ptl;
	pte_t *pte;
	int ret = 1;

888
	if (mem_cgroup_try_charge_swapin(vma->vm_mm, page, GFP_KERNEL, &ptr)) {
H
Hugh Dickins 已提交
889
		ret = -ENOMEM;
890 891
		goto out_nolock;
	}
H
Hugh Dickins 已提交
892 893 894 895

	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	if (unlikely(!pte_same(*pte, swp_entry_to_pte(entry)))) {
		if (ret > 0)
896
			mem_cgroup_cancel_charge_swapin(ptr);
H
Hugh Dickins 已提交
897 898 899
		ret = 0;
		goto out;
	}
900

K
KAMEZAWA Hiroyuki 已提交
901
	dec_mm_counter(vma->vm_mm, MM_SWAPENTS);
K
KAMEZAWA Hiroyuki 已提交
902
	inc_mm_counter(vma->vm_mm, MM_ANONPAGES);
L
Linus Torvalds 已提交
903 904 905 906
	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);
907
	mem_cgroup_commit_charge_swapin(page, ptr);
L
Linus Torvalds 已提交
908 909 910 911 912 913
	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
914 915
out:
	pte_unmap_unlock(pte, ptl);
916
out_nolock:
H
Hugh Dickins 已提交
917
	return ret;
L
Linus Torvalds 已提交
918 919 920 921 922 923 924
}

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);
925
	pte_t *pte;
926
	int ret = 0;
L
Linus Torvalds 已提交
927

H
Hugh Dickins 已提交
928 929 930 931 932 933 934 935 936 937
	/*
	 * 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
938 939 940 941 942 943
	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
944 945 946 947 948
			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
949 950
		}
	} while (pte++, addr += PAGE_SIZE, addr != end);
H
Hugh Dickins 已提交
951 952
	pte_unmap(pte - 1);
out:
953
	return ret;
L
Linus Torvalds 已提交
954 955 956 957 958 959 960 961
}

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;
962
	int ret;
L
Linus Torvalds 已提交
963 964 965 966

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
A
Andrea Arcangeli 已提交
967 968
		if (unlikely(pmd_trans_huge(*pmd)))
			continue;
L
Linus Torvalds 已提交
969 970
		if (pmd_none_or_clear_bad(pmd))
			continue;
971 972 973
		ret = unuse_pte_range(vma, pmd, addr, next, entry, page);
		if (ret)
			return ret;
L
Linus Torvalds 已提交
974 975 976 977 978 979 980 981 982 983
	} 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;
984
	int ret;
L
Linus Torvalds 已提交
985 986 987 988 989 990

	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
991 992 993
		ret = unuse_pmd_range(vma, pud, addr, next, entry, page);
		if (ret)
			return ret;
L
Linus Torvalds 已提交
994 995 996 997 998 999 1000 1001 1002
	} 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;
1003
	int ret;
L
Linus Torvalds 已提交
1004

H
Hugh Dickins 已提交
1005
	if (page_anon_vma(page)) {
L
Linus Torvalds 已提交
1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020
		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;
1021 1022 1023
		ret = unuse_pud_range(vma, pgd, addr, next, entry, page);
		if (ret)
			return ret;
L
Linus Torvalds 已提交
1024 1025 1026 1027 1028 1029 1030 1031
	} 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;
1032
	int ret = 0;
L
Linus Torvalds 已提交
1033 1034 1035

	if (!down_read_trylock(&mm->mmap_sem)) {
		/*
1036 1037
		 * 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
1038
		 */
1039
		activate_page(page);
L
Linus Torvalds 已提交
1040 1041 1042 1043 1044
		unlock_page(page);
		down_read(&mm->mmap_sem);
		lock_page(page);
	}
	for (vma = mm->mmap; vma; vma = vma->vm_next) {
1045
		if (vma->anon_vma && (ret = unuse_vma(vma, entry, page)))
L
Linus Torvalds 已提交
1046 1047 1048
			break;
	}
	up_read(&mm->mmap_sem);
1049
	return (ret < 0)? ret: 0;
L
Linus Torvalds 已提交
1050 1051 1052 1053 1054 1055
}

/*
 * Scan swap_map from current position to next entry still in use.
 * Recycle to start on reaching the end, returning 0 when empty.
 */
1056 1057
static unsigned int find_next_to_unuse(struct swap_info_struct *si,
					unsigned int prev)
L
Linus Torvalds 已提交
1058
{
1059 1060
	unsigned int max = si->max;
	unsigned int i = prev;
1061
	unsigned char count;
L
Linus Torvalds 已提交
1062 1063

	/*
1064
	 * No need for swap_lock here: we're just looking
L
Linus Torvalds 已提交
1065 1066
	 * for whether an entry is in use, not modifying it; false
	 * hits are okay, and sys_swapoff() has already prevented new
1067
	 * allocations from this area (while holding swap_lock).
L
Linus Torvalds 已提交
1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083
	 */
	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];
1084
		if (count && swap_count(count) != SWAP_MAP_BAD)
L
Linus Torvalds 已提交
1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096
			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)
{
1097
	struct swap_info_struct *si = swap_info[type];
L
Linus Torvalds 已提交
1098
	struct mm_struct *start_mm;
1099 1100
	unsigned char *swap_map;
	unsigned char swcount;
L
Linus Torvalds 已提交
1101 1102
	struct page *page;
	swp_entry_t entry;
1103
	unsigned int i = 0;
L
Linus Torvalds 已提交
1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117
	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
1118
	 * that.
L
Linus Torvalds 已提交
1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133
	 */
	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;
		}

1134
		/*
L
Linus Torvalds 已提交
1135 1136
		 * Get a page for the entry, using the existing swap
		 * cache page if there is one.  Otherwise, get a clean
1137
		 * page and read the swap into it.
L
Linus Torvalds 已提交
1138 1139 1140
		 */
		swap_map = &si->swap_map[i];
		entry = swp_entry(type, i);
1141 1142
		page = read_swap_cache_async(entry,
					GFP_HIGHUSER_MOVABLE, NULL, 0);
L
Linus Torvalds 已提交
1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181
		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
1182 1183 1184 1185 1186 1187
		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
1188
		}
H
Hugh Dickins 已提交
1189 1190 1191
		if (swap_count(swcount) && start_mm != &init_mm)
			retval = unuse_mm(start_mm, entry, page);

1192
		if (swap_count(*swap_map)) {
L
Linus Torvalds 已提交
1193 1194 1195 1196 1197 1198 1199 1200 1201
			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
1202
			while (swap_count(*swap_map) && !retval &&
L
Linus Torvalds 已提交
1203 1204
					(p = p->next) != &start_mm->mmlist) {
				mm = list_entry(p, struct mm_struct, mmlist);
1205
				if (!atomic_inc_not_zero(&mm->mm_users))
L
Linus Torvalds 已提交
1206 1207 1208 1209 1210 1211 1212 1213
					continue;
				spin_unlock(&mmlist_lock);
				mmput(prev_mm);
				prev_mm = mm;

				cond_resched();

				swcount = *swap_map;
1214
				if (!swap_count(swcount)) /* any usage ? */
L
Linus Torvalds 已提交
1215
					;
H
Hugh Dickins 已提交
1216
				else if (mm == &init_mm)
L
Linus Torvalds 已提交
1217
					set_start_mm = 1;
H
Hugh Dickins 已提交
1218
				else
L
Linus Torvalds 已提交
1219
					retval = unuse_mm(mm, entry, page);
1220

1221
				if (set_start_mm && *swap_map < swcount) {
L
Linus Torvalds 已提交
1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
					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
1252 1253 1254 1255 1256 1257
		 *
		 * 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
1258
		 */
1259 1260
		if (swap_count(*swap_map) &&
		     PageDirty(page) && PageSwapCache(page)) {
L
Linus Torvalds 已提交
1261 1262 1263 1264 1265 1266 1267 1268
			struct writeback_control wbc = {
				.sync_mode = WB_SYNC_NONE,
			};

			swap_writepage(page, &wbc);
			lock_page(page);
			wait_on_page_writeback(page);
		}
1269 1270 1271 1272 1273 1274 1275 1276 1277 1278

		/*
		 * 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))
1279
			delete_from_swap_cache(page);
L
Linus Torvalds 已提交
1280 1281 1282 1283

		/*
		 * So we could skip searching mms once swap count went
		 * to 1, we did not mark any present ptes as dirty: must
1284
		 * mark page dirty so shrink_page_list will preserve it.
L
Linus Torvalds 已提交
1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301
		 */
		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;
}

/*
1302 1303 1304
 * 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
1305 1306 1307 1308 1309
 * added to the mmlist just after page_duplicate - before would be racy.
 */
static void drain_mmlist(void)
{
	struct list_head *p, *next;
1310
	unsigned int type;
L
Linus Torvalds 已提交
1311

1312 1313
	for (type = 0; type < nr_swapfiles; type++)
		if (swap_info[type]->inuse_pages)
L
Linus Torvalds 已提交
1314 1315 1316 1317 1318 1319 1320 1321 1322
			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
1323 1324 1325
 * 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
1326
 */
1327
static sector_t map_swap_entry(swp_entry_t entry, struct block_device **bdev)
L
Linus Torvalds 已提交
1328
{
H
Hugh Dickins 已提交
1329 1330 1331 1332 1333
	struct swap_info_struct *sis;
	struct swap_extent *start_se;
	struct swap_extent *se;
	pgoff_t offset;

1334
	sis = swap_info[swp_type(entry)];
H
Hugh Dickins 已提交
1335 1336 1337 1338 1339
	*bdev = sis->bdev;

	offset = swp_offset(entry);
	start_se = sis->curr_swap_extent;
	se = start_se;
L
Linus Torvalds 已提交
1340 1341 1342 1343 1344 1345 1346 1347

	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);
		}
1348
		lh = se->list.next;
L
Linus Torvalds 已提交
1349 1350 1351 1352 1353 1354
		se = list_entry(lh, struct swap_extent, list);
		sis->curr_swap_extent = se;
		BUG_ON(se == start_se);		/* It *must* be present */
	}
}

1355 1356 1357 1358 1359 1360 1361 1362 1363 1364
/*
 * 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
1365 1366 1367 1368 1369
/*
 * Free all of a swapdev's extent information
 */
static void destroy_swap_extents(struct swap_info_struct *sis)
{
1370
	while (!list_empty(&sis->first_swap_extent.list)) {
L
Linus Torvalds 已提交
1371 1372
		struct swap_extent *se;

1373
		se = list_entry(sis->first_swap_extent.list.next,
L
Linus Torvalds 已提交
1374 1375 1376 1377 1378 1379 1380 1381
				struct swap_extent, list);
		list_del(&se->list);
		kfree(se);
	}
}

/*
 * Add a block range (and the corresponding page range) into this swapdev's
1382
 * extent list.  The extent list is kept sorted in page order.
L
Linus Torvalds 已提交
1383
 *
1384
 * This function rather assumes that it is called in ascending page order.
L
Linus Torvalds 已提交
1385 1386 1387 1388 1389 1390 1391 1392 1393
 */
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;

1394 1395 1396 1397 1398 1399 1400 1401 1402
	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
1403
		se = list_entry(lh, struct swap_extent, list);
1404 1405
		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
1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421
			/* 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;

1422
	list_add_tail(&new_se->list, &sis->first_swap_extent.list);
1423
	return 1;
L
Linus Torvalds 已提交
1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445
}

/*
 * 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.
 *
1446
 * For S_ISREG swapfiles we set S_SWAPFILE across the life of the swapon.  This
L
Linus Torvalds 已提交
1447 1448 1449 1450 1451 1452 1453 1454 1455 1456
 * 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.
 */
1457
static int setup_swap_extents(struct swap_info_struct *sis, sector_t *span)
L
Linus Torvalds 已提交
1458 1459 1460 1461 1462 1463 1464
{
	struct inode *inode;
	unsigned blocks_per_page;
	unsigned long page_no;
	unsigned blkbits;
	sector_t probe_block;
	sector_t last_block;
1465 1466 1467
	sector_t lowest_block = -1;
	sector_t highest_block = 0;
	int nr_extents = 0;
L
Linus Torvalds 已提交
1468 1469 1470 1471 1472
	int ret;

	inode = sis->swap_file->f_mapping->host;
	if (S_ISBLK(inode->i_mode)) {
		ret = add_swap_extent(sis, 0, sis->max, 0);
1473
		*span = sis->pages;
1474
		goto out;
L
Linus Torvalds 已提交
1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517
	}

	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;
			}
		}

1518 1519 1520 1521 1522 1523 1524 1525
		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
1526 1527 1528
		/*
		 * We found a PAGE_SIZE-length, PAGE_SIZE-aligned run of blocks
		 */
1529 1530
		ret = add_swap_extent(sis, page_no, 1, first_block);
		if (ret < 0)
L
Linus Torvalds 已提交
1531
			goto out;
1532
		nr_extents += ret;
L
Linus Torvalds 已提交
1533 1534 1535 1536 1537
		page_no++;
		probe_block += blocks_per_page;
reprobe:
		continue;
	}
1538 1539
	ret = nr_extents;
	*span = 1 + highest_block - lowest_block;
L
Linus Torvalds 已提交
1540
	if (page_no == 0)
1541
		page_no = 1;	/* force Empty message */
L
Linus Torvalds 已提交
1542
	sis->max = page_no;
1543
	sis->pages = page_no - 1;
L
Linus Torvalds 已提交
1544
	sis->highest_bit = page_no - 1;
1545 1546
out:
	return ret;
L
Linus Torvalds 已提交
1547 1548 1549
bad_bmap:
	printk(KERN_ERR "swapon: swapfile has holes\n");
	ret = -EINVAL;
1550
	goto out;
L
Linus Torvalds 已提交
1551 1552
}

1553
SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
L
Linus Torvalds 已提交
1554
{
1555
	struct swap_info_struct *p = NULL;
1556
	unsigned char *swap_map;
L
Linus Torvalds 已提交
1557 1558 1559
	struct file *swap_file, *victim;
	struct address_space *mapping;
	struct inode *inode;
1560
	char *pathname;
L
Linus Torvalds 已提交
1561 1562
	int i, type, prev;
	int err;
1563

L
Linus Torvalds 已提交
1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579
	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;
1580
	spin_lock(&swap_lock);
1581 1582
	for (type = swap_list.head; type >= 0; type = swap_info[type]->next) {
		p = swap_info[type];
H
Hugh Dickins 已提交
1583
		if (p->flags & SWP_WRITEOK) {
L
Linus Torvalds 已提交
1584 1585 1586 1587 1588 1589 1590
			if (p->swap_file->f_mapping == mapping)
				break;
		}
		prev = type;
	}
	if (type < 0) {
		err = -EINVAL;
1591
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
1592 1593 1594 1595 1596 1597
		goto out_dput;
	}
	if (!security_vm_enough_memory(p->pages))
		vm_unacct_memory(p->pages);
	else {
		err = -ENOMEM;
1598
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
1599 1600
		goto out_dput;
	}
1601
	if (prev < 0)
L
Linus Torvalds 已提交
1602
		swap_list.head = p->next;
1603 1604
	else
		swap_info[prev]->next = p->next;
L
Linus Torvalds 已提交
1605 1606 1607 1608
	if (type == swap_list.next) {
		/* just pick something that's safe... */
		swap_list.next = swap_list.head;
	}
1609
	if (p->prio < 0) {
1610 1611
		for (i = p->next; i >= 0; i = swap_info[i]->next)
			swap_info[i]->prio = p->prio--;
1612 1613
		least_priority++;
	}
L
Linus Torvalds 已提交
1614 1615 1616
	nr_swap_pages -= p->pages;
	total_swap_pages -= p->pages;
	p->flags &= ~SWP_WRITEOK;
1617
	spin_unlock(&swap_lock);
1618

H
Hugh Dickins 已提交
1619
	current->flags |= PF_OOM_ORIGIN;
L
Linus Torvalds 已提交
1620
	err = try_to_unuse(type);
H
Hugh Dickins 已提交
1621
	current->flags &= ~PF_OOM_ORIGIN;
L
Linus Torvalds 已提交
1622 1623 1624

	if (err) {
		/* re-insert swap space back into swap_list */
1625
		spin_lock(&swap_lock);
1626 1627 1628
		if (p->prio < 0)
			p->prio = --least_priority;
		prev = -1;
1629 1630
		for (i = swap_list.head; i >= 0; i = swap_info[i]->next) {
			if (p->prio >= swap_info[i]->prio)
L
Linus Torvalds 已提交
1631
				break;
1632 1633
			prev = i;
		}
L
Linus Torvalds 已提交
1634 1635
		p->next = i;
		if (prev < 0)
1636
			swap_list.head = swap_list.next = type;
L
Linus Torvalds 已提交
1637
		else
1638
			swap_info[prev]->next = type;
L
Linus Torvalds 已提交
1639 1640 1641
		nr_swap_pages += p->pages;
		total_swap_pages += p->pages;
		p->flags |= SWP_WRITEOK;
1642
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
1643 1644
		goto out_dput;
	}
1645 1646 1647 1648 1649

	/* wait for any unplug function to finish */
	down_write(&swap_unplug_sem);
	up_write(&swap_unplug_sem);

1650
	destroy_swap_extents(p);
H
Hugh Dickins 已提交
1651 1652 1653
	if (p->flags & SWP_CONTINUED)
		free_swap_count_continuations(p);

I
Ingo Molnar 已提交
1654
	mutex_lock(&swapon_mutex);
1655 1656 1657
	spin_lock(&swap_lock);
	drain_mmlist();

1658 1659 1660
	/* wait for anyone still in scan_swap_map */
	p->highest_bit = 0;		/* cuts scans short */
	while (p->flags >= SWP_SCANNING) {
1661
		spin_unlock(&swap_lock);
1662
		schedule_timeout_uninterruptible(1);
1663
		spin_lock(&swap_lock);
1664 1665
	}

L
Linus Torvalds 已提交
1666 1667 1668 1669 1670 1671
	swap_file = p->swap_file;
	p->swap_file = NULL;
	p->max = 0;
	swap_map = p->swap_map;
	p->swap_map = NULL;
	p->flags = 0;
1672
	spin_unlock(&swap_lock);
I
Ingo Molnar 已提交
1673
	mutex_unlock(&swapon_mutex);
L
Linus Torvalds 已提交
1674
	vfree(swap_map);
1675 1676 1677
	/* Destroy swap account informatin */
	swap_cgroup_swapoff(type);

L
Linus Torvalds 已提交
1678 1679 1680 1681
	inode = mapping->host;
	if (S_ISBLK(inode->i_mode)) {
		struct block_device *bdev = I_BDEV(inode);
		set_blocksize(bdev, p->old_block_size);
1682
		blkdev_put(bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL);
L
Linus Torvalds 已提交
1683
	} else {
1684
		mutex_lock(&inode->i_mutex);
L
Linus Torvalds 已提交
1685
		inode->i_flags &= ~S_SWAPFILE;
1686
		mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
1687 1688 1689
	}
	filp_close(swap_file, NULL);
	err = 0;
K
Kay Sievers 已提交
1690 1691
	atomic_inc(&proc_poll_event);
	wake_up_interruptible(&proc_poll_wait);
L
Linus Torvalds 已提交
1692 1693 1694 1695 1696 1697 1698 1699

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

#ifdef CONFIG_PROC_FS
K
Kay Sievers 已提交
1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718
struct proc_swaps {
	struct seq_file seq;
	int event;
};

static unsigned swaps_poll(struct file *file, poll_table *wait)
{
	struct proc_swaps *s = file->private_data;

	poll_wait(file, &proc_poll_wait, wait);

	if (s->event != atomic_read(&proc_poll_event)) {
		s->event = atomic_read(&proc_poll_event);
		return POLLIN | POLLRDNORM | POLLERR | POLLPRI;
	}

	return POLLIN | POLLRDNORM;
}

L
Linus Torvalds 已提交
1719 1720 1721
/* iterator */
static void *swap_start(struct seq_file *swap, loff_t *pos)
{
1722 1723
	struct swap_info_struct *si;
	int type;
L
Linus Torvalds 已提交
1724 1725
	loff_t l = *pos;

I
Ingo Molnar 已提交
1726
	mutex_lock(&swapon_mutex);
L
Linus Torvalds 已提交
1727

1728 1729 1730
	if (!l)
		return SEQ_START_TOKEN;

1731 1732 1733 1734
	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
1735
			continue;
1736
		if (!--l)
1737
			return si;
L
Linus Torvalds 已提交
1738 1739 1740 1741 1742 1743 1744
	}

	return NULL;
}

static void *swap_next(struct seq_file *swap, void *v, loff_t *pos)
{
1745 1746
	struct swap_info_struct *si = v;
	int type;
L
Linus Torvalds 已提交
1747

1748
	if (v == SEQ_START_TOKEN)
1749 1750 1751
		type = 0;
	else
		type = si->type + 1;
1752

1753 1754 1755 1756
	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
1757 1758
			continue;
		++*pos;
1759
		return si;
L
Linus Torvalds 已提交
1760 1761 1762 1763 1764 1765 1766
	}

	return NULL;
}

static void swap_stop(struct seq_file *swap, void *v)
{
I
Ingo Molnar 已提交
1767
	mutex_unlock(&swapon_mutex);
L
Linus Torvalds 已提交
1768 1769 1770 1771
}

static int swap_show(struct seq_file *swap, void *v)
{
1772
	struct swap_info_struct *si = v;
L
Linus Torvalds 已提交
1773 1774 1775
	struct file *file;
	int len;

1776
	if (si == SEQ_START_TOKEN) {
1777 1778 1779
		seq_puts(swap,"Filename\t\t\t\tType\t\tSize\tUsed\tPriority\n");
		return 0;
	}
L
Linus Torvalds 已提交
1780

1781
	file = si->swap_file;
1782
	len = seq_path(swap, &file->f_path, " \t\n\\");
1783
	seq_printf(swap, "%*s%s\t%u\t%u\t%d\n",
1784 1785
			len < 40 ? 40 - len : 1, " ",
			S_ISBLK(file->f_path.dentry->d_inode->i_mode) ?
L
Linus Torvalds 已提交
1786
				"partition" : "file\t",
1787 1788 1789
			si->pages << (PAGE_SHIFT - 10),
			si->inuse_pages << (PAGE_SHIFT - 10),
			si->prio);
L
Linus Torvalds 已提交
1790 1791 1792
	return 0;
}

1793
static const struct seq_operations swaps_op = {
L
Linus Torvalds 已提交
1794 1795 1796 1797 1798 1799 1800 1801
	.start =	swap_start,
	.next =		swap_next,
	.stop =		swap_stop,
	.show =		swap_show
};

static int swaps_open(struct inode *inode, struct file *file)
{
K
Kay Sievers 已提交
1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819
	struct proc_swaps *s;
	int ret;

	s = kmalloc(sizeof(struct proc_swaps), GFP_KERNEL);
	if (!s)
		return -ENOMEM;

	file->private_data = s;

	ret = seq_open(file, &swaps_op);
	if (ret) {
		kfree(s);
		return ret;
	}

	s->seq.private = s;
	s->event = atomic_read(&proc_poll_event);
	return ret;
L
Linus Torvalds 已提交
1820 1821
}

1822
static const struct file_operations proc_swaps_operations = {
L
Linus Torvalds 已提交
1823 1824 1825 1826
	.open		= swaps_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= seq_release,
K
Kay Sievers 已提交
1827
	.poll		= swaps_poll,
L
Linus Torvalds 已提交
1828 1829 1830 1831
};

static int __init procswaps_init(void)
{
1832
	proc_create("swaps", 0, NULL, &proc_swaps_operations);
L
Linus Torvalds 已提交
1833 1834 1835 1836 1837
	return 0;
}
__initcall(procswaps_init);
#endif /* CONFIG_PROC_FS */

J
Jan Beulich 已提交
1838 1839 1840 1841 1842 1843 1844 1845 1846
#ifdef MAX_SWAPFILES_CHECK
static int __init max_swapfiles_check(void)
{
	MAX_SWAPFILES_CHECK();
	return 0;
}
late_initcall(max_swapfiles_check);
#endif

1847
static struct swap_info_struct *alloc_swap_info(void)
L
Linus Torvalds 已提交
1848
{
1849
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
1850
	unsigned int type;
1851 1852 1853

	p = kzalloc(sizeof(*p), GFP_KERNEL);
	if (!p)
1854
		return ERR_PTR(-ENOMEM);
1855

1856
	spin_lock(&swap_lock);
1857 1858
	for (type = 0; type < nr_swapfiles; type++) {
		if (!(swap_info[type]->flags & SWP_USED))
L
Linus Torvalds 已提交
1859
			break;
1860
	}
1861
	if (type >= MAX_SWAPFILES) {
1862
		spin_unlock(&swap_lock);
1863
		kfree(p);
1864
		return ERR_PTR(-EPERM);
L
Linus Torvalds 已提交
1865
	}
1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883
	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.
		 */
	}
1884
	INIT_LIST_HEAD(&p->first_swap_extent.list);
L
Linus Torvalds 已提交
1885 1886
	p->flags = SWP_USED;
	p->next = -1;
1887
	spin_unlock(&swap_lock);
1888

1889 1890 1891 1892 1893 1894
	return p;
}

SYSCALL_DEFINE2(swapon, const char __user *, specialfile, int, swap_flags)
{
	struct swap_info_struct *p;
1895
	char *name;
1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914
	struct block_device *bdev = NULL;
	struct file *swap_file = NULL;
	struct address_space *mapping;
	int i, prev;
	int error;
	union swap_header *swap_header;
	unsigned int nr_good_pages;
	int nr_extents = 0;
	sector_t span;
	unsigned long maxpages;
	unsigned long swapfilepages;
	unsigned char *swap_map = NULL;
	struct page *page = NULL;
	struct inode *inode = NULL;

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

	p = alloc_swap_info();
1915 1916
	if (IS_ERR(p))
		return PTR_ERR(p);
1917

L
Linus Torvalds 已提交
1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936
	name = getname(specialfile);
	error = PTR_ERR(name);
	if (IS_ERR(name)) {
		name = NULL;
		goto bad_swap_2;
	}
	swap_file = filp_open(name, O_RDWR|O_LARGEFILE, 0);
	error = PTR_ERR(swap_file);
	if (IS_ERR(swap_file)) {
		swap_file = NULL;
		goto bad_swap_2;
	}

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

	error = -EBUSY;
	for (i = 0; i < nr_swapfiles; i++) {
1937
		struct swap_info_struct *q = swap_info[i];
L
Linus Torvalds 已提交
1938

1939
		if (q == p || !q->swap_file)
L
Linus Torvalds 已提交
1940 1941 1942 1943 1944 1945 1946
			continue;
		if (mapping == q->swap_file->f_mapping)
			goto bad_swap;
	}

	error = -EINVAL;
	if (S_ISBLK(inode->i_mode)) {
M
Miklos Szeredi 已提交
1947
		bdev = bdgrab(I_BDEV(inode));
1948 1949
		error = blkdev_get(bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL,
				   sys_swapon);
L
Linus Torvalds 已提交
1950 1951
		if (error < 0) {
			bdev = NULL;
R
Rob Landley 已提交
1952
			error = -EINVAL;
L
Linus Torvalds 已提交
1953 1954 1955 1956 1957 1958 1959
			goto bad_swap;
		}
		p->old_block_size = block_size(bdev);
		error = set_blocksize(bdev, PAGE_SIZE);
		if (error < 0)
			goto bad_swap;
		p->bdev = bdev;
1960
		p->flags |= SWP_BLKDEV;
L
Linus Torvalds 已提交
1961 1962
	} else if (S_ISREG(inode->i_mode)) {
		p->bdev = inode->i_sb->s_bdev;
1963
		mutex_lock(&inode->i_mutex);
L
Linus Torvalds 已提交
1964 1965 1966 1967 1968 1969 1970 1971
		if (IS_SWAPFILE(inode)) {
			error = -EBUSY;
			goto bad_swap;
		}
	} else {
		goto bad_swap;
	}

1972
	swapfilepages = i_size_read(inode) >> PAGE_SHIFT;
L
Linus Torvalds 已提交
1973 1974 1975 1976 1977 1978 1979 1980

	/*
	 * Read the swap header.
	 */
	if (!mapping->a_ops->readpage) {
		error = -EINVAL;
		goto bad_swap;
	}
1981
	page = read_mapping_page(mapping, 0, swap_file);
L
Linus Torvalds 已提交
1982 1983 1984 1985
	if (IS_ERR(page)) {
		error = PTR_ERR(page);
		goto bad_swap;
	}
1986
	swap_header = kmap(page);
L
Linus Torvalds 已提交
1987

1988
	if (memcmp("SWAPSPACE2", swap_header->magic.magic, 10)) {
1989
		printk(KERN_ERR "Unable to find swap-space signature\n");
L
Linus Torvalds 已提交
1990 1991 1992
		error = -EINVAL;
		goto bad_swap;
	}
1993

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
	/* 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);
L
Linus Torvalds 已提交
2007 2008
		error = -EINVAL;
		goto bad_swap;
2009
	}
L
Linus Torvalds 已提交
2010

2011 2012
	p->lowest_bit  = 1;
	p->cluster_next = 1;
2013
	p->cluster_nr = 0;
2014

2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029
	/*
	 * Find out how many pages are allowed for a single swap
	 * device. There are two 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 a swap pte as defined by
	 * the different architectures. In order to find the
	 * largest possible bit mask a swap entry with swap type 0
	 * and swap offset ~0UL is created, encoded to a swap pte,
	 * decoded to a swp_entry_t again and finally the swap
	 * 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
	 * swap pte.
	 */
	maxpages = swp_offset(pte_to_swp_entry(
2030 2031 2032 2033 2034 2035 2036
			swp_entry_to_pte(swp_entry(0, ~0UL)))) + 1;
	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;
	}
2037
	p->highest_bit = maxpages - 1;
L
Linus Torvalds 已提交
2038

2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050
	error = -EINVAL;
	if (!maxpages)
		goto bad_swap;
	if (swapfilepages && maxpages > swapfilepages) {
		printk(KERN_WARNING
		       "Swap area shorter than signature indicates\n");
		goto bad_swap;
	}
	if (swap_header->info.nr_badpages && S_ISREG(inode->i_mode))
		goto bad_swap;
	if (swap_header->info.nr_badpages > MAX_SWAP_BADPAGES)
		goto bad_swap;
2051

2052
	/* OK, set up the swap map and apply the bad block list */
2053
	swap_map = vzalloc(maxpages);
2054 2055 2056 2057
	if (!swap_map) {
		error = -ENOMEM;
		goto bad_swap;
	}
L
Linus Torvalds 已提交
2058

2059 2060
	nr_good_pages = maxpages - 1;	/* omit header page */

2061
	for (i = 0; i < swap_header->info.nr_badpages; i++) {
2062 2063
		unsigned int page_nr = swap_header->info.badpages[i];
		if (page_nr == 0 || page_nr > swap_header->info.last_page) {
2064
			error = -EINVAL;
L
Linus Torvalds 已提交
2065
			goto bad_swap;
2066
		}
2067 2068 2069 2070
		if (page_nr < maxpages) {
			swap_map[page_nr] = SWAP_MAP_BAD;
			nr_good_pages--;
		}
L
Linus Torvalds 已提交
2071
	}
2072

2073
	error = swap_cgroup_swapon(p->type, maxpages);
2074 2075 2076
	if (error)
		goto bad_swap;

2077
	if (nr_good_pages) {
2078
		swap_map[0] = SWAP_MAP_BAD;
2079 2080
		p->max = maxpages;
		p->pages = nr_good_pages;
2081 2082 2083
		nr_extents = setup_swap_extents(p, &span);
		if (nr_extents < 0) {
			error = nr_extents;
2084
			goto bad_swap;
2085
		}
2086 2087
		nr_good_pages = p->pages;
	}
L
Linus Torvalds 已提交
2088 2089 2090 2091 2092 2093
	if (!nr_good_pages) {
		printk(KERN_WARNING "Empty swap-file\n");
		error = -EINVAL;
		goto bad_swap;
	}

2094 2095 2096 2097 2098
	if (p->bdev) {
		if (blk_queue_nonrot(bdev_get_queue(p->bdev))) {
			p->flags |= SWP_SOLIDSTATE;
			p->cluster_next = 1 + (random32() % p->highest_bit);
		}
2099
		if (discard_swap(p) == 0 && (swap_flags & SWAP_FLAG_DISCARD))
2100
			p->flags |= SWP_DISCARDABLE;
2101
	}
2102

I
Ingo Molnar 已提交
2103
	mutex_lock(&swapon_mutex);
2104
	spin_lock(&swap_lock);
2105 2106 2107 2108 2109 2110
	if (swap_flags & SWAP_FLAG_PREFER)
		p->prio =
		  (swap_flags & SWAP_FLAG_PRIO_MASK) >> SWAP_FLAG_PRIO_SHIFT;
	else
		p->prio = --least_priority;
	p->swap_map = swap_map;
H
Hugh Dickins 已提交
2111
	p->flags |= SWP_WRITEOK;
L
Linus Torvalds 已提交
2112 2113
	nr_swap_pages += nr_good_pages;
	total_swap_pages += nr_good_pages;
2114

2115
	printk(KERN_INFO "Adding %uk swap on %s.  "
2116
			"Priority:%d extents:%d across:%lluk %s%s\n",
2117
		nr_good_pages<<(PAGE_SHIFT-10), name, p->prio,
2118
		nr_extents, (unsigned long long)span<<(PAGE_SHIFT-10),
2119 2120
		(p->flags & SWP_SOLIDSTATE) ? "SS" : "",
		(p->flags & SWP_DISCARDABLE) ? "D" : "");
L
Linus Torvalds 已提交
2121 2122 2123

	/* insert swap space into swap_list: */
	prev = -1;
2124 2125
	for (i = swap_list.head; i >= 0; i = swap_info[i]->next) {
		if (p->prio >= swap_info[i]->prio)
L
Linus Torvalds 已提交
2126 2127 2128 2129
			break;
		prev = i;
	}
	p->next = i;
2130
	if (prev < 0)
2131
		swap_list.head = swap_list.next = p->type;
2132
	else
2133
		swap_info[prev]->next = p->type;
2134
	spin_unlock(&swap_lock);
I
Ingo Molnar 已提交
2135
	mutex_unlock(&swapon_mutex);
K
Kay Sievers 已提交
2136 2137 2138
	atomic_inc(&proc_poll_event);
	wake_up_interruptible(&proc_poll_wait);

L
Linus Torvalds 已提交
2139 2140 2141 2142 2143
	error = 0;
	goto out;
bad_swap:
	if (bdev) {
		set_blocksize(bdev, p->old_block_size);
2144
		blkdev_put(bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL);
L
Linus Torvalds 已提交
2145
	}
2146
	destroy_swap_extents(p);
2147
	swap_cgroup_swapoff(p->type);
L
Linus Torvalds 已提交
2148
bad_swap_2:
2149
	spin_lock(&swap_lock);
L
Linus Torvalds 已提交
2150 2151
	p->swap_file = NULL;
	p->flags = 0;
2152
	spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
2153
	vfree(swap_map);
2154
	if (swap_file) {
2155
		if (inode && S_ISREG(inode->i_mode))
2156
			mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
2157
		filp_close(swap_file, NULL);
2158
	}
L
Linus Torvalds 已提交
2159 2160 2161 2162 2163 2164 2165
out:
	if (page && !IS_ERR(page)) {
		kunmap(page);
		page_cache_release(page);
	}
	if (name)
		putname(name);
2166
	if (inode && S_ISREG(inode->i_mode)) {
L
Linus Torvalds 已提交
2167 2168
		if (!error)
			inode->i_flags |= S_SWAPFILE;
2169
		mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
2170 2171 2172 2173 2174 2175
	}
	return error;
}

void si_swapinfo(struct sysinfo *val)
{
2176
	unsigned int type;
L
Linus Torvalds 已提交
2177 2178
	unsigned long nr_to_be_unused = 0;

2179
	spin_lock(&swap_lock);
2180 2181 2182 2183 2184
	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
2185 2186 2187
	}
	val->freeswap = nr_swap_pages + nr_to_be_unused;
	val->totalswap = total_swap_pages + nr_to_be_unused;
2188
	spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
2189 2190 2191 2192 2193
}

/*
 * Verify that a swap entry is valid and increment its swap map count.
 *
2194 2195 2196 2197 2198 2199
 * 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
2200
 * - swap-mapped reference requested but needs continued swap count. -> ENOMEM
L
Linus Torvalds 已提交
2201
 */
2202
static int __swap_duplicate(swp_entry_t entry, unsigned char usage)
L
Linus Torvalds 已提交
2203
{
2204
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
2205
	unsigned long offset, type;
2206 2207
	unsigned char count;
	unsigned char has_cache;
H
Hugh Dickins 已提交
2208
	int err = -EINVAL;
L
Linus Torvalds 已提交
2209

2210
	if (non_swap_entry(entry))
H
Hugh Dickins 已提交
2211
		goto out;
2212

L
Linus Torvalds 已提交
2213 2214 2215
	type = swp_type(entry);
	if (type >= nr_swapfiles)
		goto bad_file;
2216
	p = swap_info[type];
L
Linus Torvalds 已提交
2217 2218
	offset = swp_offset(entry);

2219
	spin_lock(&swap_lock);
2220 2221 2222
	if (unlikely(offset >= p->max))
		goto unlock_out;

H
Hugh Dickins 已提交
2223 2224 2225 2226
	count = p->swap_map[offset];
	has_cache = count & SWAP_HAS_CACHE;
	count &= ~SWAP_HAS_CACHE;
	err = 0;
2227

H
Hugh Dickins 已提交
2228
	if (usage == SWAP_HAS_CACHE) {
2229 2230

		/* set SWAP_HAS_CACHE if there is no cache and entry is used */
H
Hugh Dickins 已提交
2231 2232 2233 2234 2235 2236
		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;
2237 2238

	} else if (count || has_cache) {
H
Hugh Dickins 已提交
2239

H
Hugh Dickins 已提交
2240 2241 2242
		if ((count & ~COUNT_CONTINUED) < SWAP_MAP_MAX)
			count += usage;
		else if ((count & ~COUNT_CONTINUED) > SWAP_MAP_MAX)
H
Hugh Dickins 已提交
2243
			err = -EINVAL;
H
Hugh Dickins 已提交
2244 2245 2246 2247
		else if (swap_count_continued(p, offset, count))
			count = COUNT_CONTINUED;
		else
			err = -ENOMEM;
2248
	} else
H
Hugh Dickins 已提交
2249 2250 2251 2252
		err = -ENOENT;			/* unused swap entry */

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

2253
unlock_out:
2254
	spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
2255
out:
H
Hugh Dickins 已提交
2256
	return err;
L
Linus Torvalds 已提交
2257 2258 2259 2260 2261

bad_file:
	printk(KERN_ERR "swap_dup: %s%08lx\n", Bad_file, entry.val);
	goto out;
}
H
Hugh Dickins 已提交
2262

H
Hugh Dickins 已提交
2263 2264 2265 2266 2267 2268 2269 2270 2271
/*
 * 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);
}

2272
/*
2273 2274 2275 2276 2277
 * 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.
2278
 */
H
Hugh Dickins 已提交
2279
int swap_duplicate(swp_entry_t entry)
2280
{
H
Hugh Dickins 已提交
2281 2282 2283 2284 2285
	int err = 0;

	while (!err && __swap_duplicate(entry, 1) == -ENOMEM)
		err = add_swap_count_continuation(entry, GFP_ATOMIC);
	return err;
2286
}
L
Linus Torvalds 已提交
2287

2288
/*
2289 2290
 * @entry: swap entry for which we allocate swap cache.
 *
2291
 * Called when allocating swap cache for existing swap entry,
2292 2293 2294
 * This can return error codes. Returns 0 at success.
 * -EBUSY means there is a swap cache.
 * Note: return code is different from swap_duplicate().
2295 2296 2297
 */
int swapcache_prepare(swp_entry_t entry)
{
H
Hugh Dickins 已提交
2298
	return __swap_duplicate(entry, SWAP_HAS_CACHE);
2299 2300
}

L
Linus Torvalds 已提交
2301
/*
2302
 * swap_lock prevents swap_map being freed. Don't grab an extra
L
Linus Torvalds 已提交
2303 2304 2305 2306
 * reference on the swaphandle, it doesn't matter if it becomes unused.
 */
int valid_swaphandles(swp_entry_t entry, unsigned long *offset)
{
2307
	struct swap_info_struct *si;
H
Hugh Dickins 已提交
2308
	int our_page_cluster = page_cluster;
2309 2310 2311
	pgoff_t target, toff;
	pgoff_t base, end;
	int nr_pages = 0;
L
Linus Torvalds 已提交
2312

H
Hugh Dickins 已提交
2313
	if (!our_page_cluster)	/* no readahead */
L
Linus Torvalds 已提交
2314
		return 0;
2315

2316
	si = swap_info[swp_type(entry)];
2317 2318 2319 2320 2321
	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
2322

2323
	spin_lock(&swap_lock);
2324 2325 2326 2327 2328 2329 2330 2331
	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;
2332
		if (swap_count(si->swap_map[toff]) == SWAP_MAP_BAD)
L
Linus Torvalds 已提交
2333
			break;
2334 2335 2336
	}
	/* Count contiguous allocated slots below our target */
	for (toff = target; --toff >= base; nr_pages++) {
L
Linus Torvalds 已提交
2337
		/* Don't read in free or bad pages */
2338
		if (!si->swap_map[toff])
L
Linus Torvalds 已提交
2339
			break;
2340
		if (swap_count(si->swap_map[toff]) == SWAP_MAP_BAD)
L
Linus Torvalds 已提交
2341
			break;
2342
	}
2343
	spin_unlock(&swap_lock);
2344 2345 2346 2347 2348 2349 2350

	/*
	 * 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
2351
}
H
Hugh Dickins 已提交
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 2559 2560 2561 2562 2563 2564 2565 2566 2567

/*
 * 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);
			}
		}
	}
}