swapfile.c 76.1 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>
34 35
#include <linux/frontswap.h>
#include <linux/swapfile.h>
36
#include <linux/export.h>
L
Linus Torvalds 已提交
37 38 39 40

#include <asm/pgtable.h>
#include <asm/tlbflush.h>
#include <linux/swapops.h>
41
#include <linux/page_cgroup.h>
L
Linus Torvalds 已提交
42

H
Hugh Dickins 已提交
43 44 45
static bool swap_count_continued(struct swap_info_struct *, pgoff_t,
				 unsigned char);
static void free_swap_count_continuations(struct swap_info_struct *);
46
static sector_t map_swap_entry(swp_entry_t, struct block_device**);
H
Hugh Dickins 已提交
47

48
DEFINE_SPINLOCK(swap_lock);
A
Adrian Bunk 已提交
49
static unsigned int nr_swapfiles;
50 51
atomic_long_t nr_swap_pages;
/* protected with swap_lock. reading in vm_swap_full() doesn't need lock */
L
Linus Torvalds 已提交
52
long total_swap_pages;
53
static int least_priority;
L
Linus Torvalds 已提交
54 55 56 57 58 59

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 ";

60 61 62 63 64
/*
 * all active swap_info_structs
 * protected with swap_lock, and ordered by priority.
 */
LIST_HEAD(swap_list_head);
L
Linus Torvalds 已提交
65

66
struct swap_info_struct *swap_info[MAX_SWAPFILES];
L
Linus Torvalds 已提交
67

I
Ingo Molnar 已提交
68
static DEFINE_MUTEX(swapon_mutex);
L
Linus Torvalds 已提交
69

K
Kay Sievers 已提交
70 71 72 73
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);

74
static inline unsigned char swap_count(unsigned char ent)
75
{
H
Hugh Dickins 已提交
76
	return ent & ~SWAP_HAS_CACHE;	/* may include SWAP_HAS_CONT flag */
77 78
}

79
/* returns 1 if swap entry is freed */
80 81 82
static int
__try_to_reclaim_swap(struct swap_info_struct *si, unsigned long offset)
{
83
	swp_entry_t entry = swp_entry(si->type, offset);
84 85 86
	struct page *page;
	int ret = 0;

87
	page = find_get_page(swap_address_space(entry), entry.val);
88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103
	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;
}
104

105 106 107 108 109 110 111
/*
 * 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;
112 113
	sector_t start_block;
	sector_t nr_blocks;
114 115
	int err = 0;

116 117 118 119 120 121
	/* 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,
122
				nr_blocks, GFP_KERNEL, 0);
123 124 125 126
		if (err)
			return err;
		cond_resched();
	}
127

128 129 130
	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);
131 132

		err = blkdev_issue_discard(si->bdev, start_block,
133
				nr_blocks, GFP_KERNEL, 0);
134 135 136 137 138 139 140 141
		if (err)
			break;

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

142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158
/*
 * 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;
159
			sector_t nr_blocks = se->nr_pages - offset;
160 161 162 163 164 165 166 167 168 169 170 171

			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,
172
				    nr_blocks, GFP_NOIO, 0))
173 174 175 176 177 178 179 180
				break;
		}

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

181 182 183
#define SWAPFILE_CLUSTER	256
#define LATENCY_LIMIT		256

184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241
static inline void cluster_set_flag(struct swap_cluster_info *info,
	unsigned int flag)
{
	info->flags = flag;
}

static inline unsigned int cluster_count(struct swap_cluster_info *info)
{
	return info->data;
}

static inline void cluster_set_count(struct swap_cluster_info *info,
				     unsigned int c)
{
	info->data = c;
}

static inline void cluster_set_count_flag(struct swap_cluster_info *info,
					 unsigned int c, unsigned int f)
{
	info->flags = f;
	info->data = c;
}

static inline unsigned int cluster_next(struct swap_cluster_info *info)
{
	return info->data;
}

static inline void cluster_set_next(struct swap_cluster_info *info,
				    unsigned int n)
{
	info->data = n;
}

static inline void cluster_set_next_flag(struct swap_cluster_info *info,
					 unsigned int n, unsigned int f)
{
	info->flags = f;
	info->data = n;
}

static inline bool cluster_is_free(struct swap_cluster_info *info)
{
	return info->flags & CLUSTER_FLAG_FREE;
}

static inline bool cluster_is_null(struct swap_cluster_info *info)
{
	return info->flags & CLUSTER_FLAG_NEXT_NULL;
}

static inline void cluster_set_null(struct swap_cluster_info *info)
{
	info->flags = CLUSTER_FLAG_NEXT_NULL;
	info->data = 0;
}

S
Shaohua Li 已提交
242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
/* Add a cluster to discard list and schedule it to do discard */
static void swap_cluster_schedule_discard(struct swap_info_struct *si,
		unsigned int idx)
{
	/*
	 * If scan_swap_map() can't find a free cluster, it will check
	 * si->swap_map directly. To make sure the discarding cluster isn't
	 * taken by scan_swap_map(), mark the swap entries bad (occupied). It
	 * will be cleared after discard
	 */
	memset(si->swap_map + idx * SWAPFILE_CLUSTER,
			SWAP_MAP_BAD, SWAPFILE_CLUSTER);

	if (cluster_is_null(&si->discard_cluster_head)) {
		cluster_set_next_flag(&si->discard_cluster_head,
						idx, 0);
		cluster_set_next_flag(&si->discard_cluster_tail,
						idx, 0);
	} else {
		unsigned int tail = cluster_next(&si->discard_cluster_tail);
		cluster_set_next(&si->cluster_info[tail], idx);
		cluster_set_next_flag(&si->discard_cluster_tail,
						idx, 0);
	}

	schedule_work(&si->discard_work);
}

/*
 * Doing discard actually. After a cluster discard is finished, the cluster
 * will be added to free cluster list. caller should hold si->lock.
*/
static void swap_do_scheduled_discard(struct swap_info_struct *si)
{
	struct swap_cluster_info *info;
	unsigned int idx;

	info = si->cluster_info;

	while (!cluster_is_null(&si->discard_cluster_head)) {
		idx = cluster_next(&si->discard_cluster_head);

		cluster_set_next_flag(&si->discard_cluster_head,
						cluster_next(&info[idx]), 0);
		if (cluster_next(&si->discard_cluster_tail) == idx) {
			cluster_set_null(&si->discard_cluster_head);
			cluster_set_null(&si->discard_cluster_tail);
		}
		spin_unlock(&si->lock);

		discard_swap_cluster(si, idx * SWAPFILE_CLUSTER,
				SWAPFILE_CLUSTER);

		spin_lock(&si->lock);
		cluster_set_flag(&info[idx], CLUSTER_FLAG_FREE);
		if (cluster_is_null(&si->free_cluster_head)) {
			cluster_set_next_flag(&si->free_cluster_head,
						idx, 0);
			cluster_set_next_flag(&si->free_cluster_tail,
						idx, 0);
		} else {
			unsigned int tail;

			tail = cluster_next(&si->free_cluster_tail);
			cluster_set_next(&info[tail], idx);
			cluster_set_next_flag(&si->free_cluster_tail,
						idx, 0);
		}
		memset(si->swap_map + idx * SWAPFILE_CLUSTER,
				0, SWAPFILE_CLUSTER);
	}
}

static void swap_discard_work(struct work_struct *work)
{
	struct swap_info_struct *si;

	si = container_of(work, struct swap_info_struct, discard_work);

	spin_lock(&si->lock);
	swap_do_scheduled_discard(si);
	spin_unlock(&si->lock);
}

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
/*
 * The cluster corresponding to page_nr will be used. The cluster will be
 * removed from free cluster list and its usage counter will be increased.
 */
static void inc_cluster_info_page(struct swap_info_struct *p,
	struct swap_cluster_info *cluster_info, unsigned long page_nr)
{
	unsigned long idx = page_nr / SWAPFILE_CLUSTER;

	if (!cluster_info)
		return;
	if (cluster_is_free(&cluster_info[idx])) {
		VM_BUG_ON(cluster_next(&p->free_cluster_head) != idx);
		cluster_set_next_flag(&p->free_cluster_head,
			cluster_next(&cluster_info[idx]), 0);
		if (cluster_next(&p->free_cluster_tail) == idx) {
			cluster_set_null(&p->free_cluster_tail);
			cluster_set_null(&p->free_cluster_head);
		}
		cluster_set_count_flag(&cluster_info[idx], 0, 0);
	}

	VM_BUG_ON(cluster_count(&cluster_info[idx]) >= SWAPFILE_CLUSTER);
	cluster_set_count(&cluster_info[idx],
		cluster_count(&cluster_info[idx]) + 1);
}

/*
 * The cluster corresponding to page_nr decreases one usage. If the usage
 * counter becomes 0, which means no page in the cluster is in using, we can
 * optionally discard the cluster and add it to free cluster list.
 */
static void dec_cluster_info_page(struct swap_info_struct *p,
	struct swap_cluster_info *cluster_info, unsigned long page_nr)
{
	unsigned long idx = page_nr / SWAPFILE_CLUSTER;

	if (!cluster_info)
		return;

	VM_BUG_ON(cluster_count(&cluster_info[idx]) == 0);
	cluster_set_count(&cluster_info[idx],
		cluster_count(&cluster_info[idx]) - 1);

	if (cluster_count(&cluster_info[idx]) == 0) {
S
Shaohua Li 已提交
371 372 373 374 375
		/*
		 * If the swap is discardable, prepare discard the cluster
		 * instead of free it immediately. The cluster will be freed
		 * after discard.
		 */
376 377
		if ((p->flags & (SWP_WRITEOK | SWP_PAGE_DISCARD)) ==
				 (SWP_WRITEOK | SWP_PAGE_DISCARD)) {
S
Shaohua Li 已提交
378 379 380 381
			swap_cluster_schedule_discard(p, idx);
			return;
		}

382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397
		cluster_set_flag(&cluster_info[idx], CLUSTER_FLAG_FREE);
		if (cluster_is_null(&p->free_cluster_head)) {
			cluster_set_next_flag(&p->free_cluster_head, idx, 0);
			cluster_set_next_flag(&p->free_cluster_tail, idx, 0);
		} else {
			unsigned int tail = cluster_next(&p->free_cluster_tail);
			cluster_set_next(&cluster_info[tail], idx);
			cluster_set_next_flag(&p->free_cluster_tail, idx, 0);
		}
	}
}

/*
 * It's possible scan_swap_map() uses a free cluster in the middle of free
 * cluster list. Avoiding such abuse to avoid list corruption.
 */
398 399
static bool
scan_swap_map_ssd_cluster_conflict(struct swap_info_struct *si,
400 401
	unsigned long offset)
{
402 403 404
	struct percpu_cluster *percpu_cluster;
	bool conflict;

405
	offset /= SWAPFILE_CLUSTER;
406
	conflict = !cluster_is_null(&si->free_cluster_head) &&
407 408
		offset != cluster_next(&si->free_cluster_head) &&
		cluster_is_free(&si->cluster_info[offset]);
409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469

	if (!conflict)
		return false;

	percpu_cluster = this_cpu_ptr(si->percpu_cluster);
	cluster_set_null(&percpu_cluster->index);
	return true;
}

/*
 * Try to get a swap entry from current cpu's swap entry pool (a cluster). This
 * might involve allocating a new cluster for current CPU too.
 */
static void scan_swap_map_try_ssd_cluster(struct swap_info_struct *si,
	unsigned long *offset, unsigned long *scan_base)
{
	struct percpu_cluster *cluster;
	bool found_free;
	unsigned long tmp;

new_cluster:
	cluster = this_cpu_ptr(si->percpu_cluster);
	if (cluster_is_null(&cluster->index)) {
		if (!cluster_is_null(&si->free_cluster_head)) {
			cluster->index = si->free_cluster_head;
			cluster->next = cluster_next(&cluster->index) *
					SWAPFILE_CLUSTER;
		} else if (!cluster_is_null(&si->discard_cluster_head)) {
			/*
			 * we don't have free cluster but have some clusters in
			 * discarding, do discard now and reclaim them
			 */
			swap_do_scheduled_discard(si);
			*scan_base = *offset = si->cluster_next;
			goto new_cluster;
		} else
			return;
	}

	found_free = false;

	/*
	 * Other CPUs can use our cluster if they can't find a free cluster,
	 * check if there is still free entry in the cluster
	 */
	tmp = cluster->next;
	while (tmp < si->max && tmp < (cluster_next(&cluster->index) + 1) *
	       SWAPFILE_CLUSTER) {
		if (!si->swap_map[tmp]) {
			found_free = true;
			break;
		}
		tmp++;
	}
	if (!found_free) {
		cluster_set_null(&cluster->index);
		goto new_cluster;
	}
	cluster->next = tmp + 1;
	*offset = tmp;
	*scan_base = tmp;
470 471
}

472 473
static unsigned long scan_swap_map(struct swap_info_struct *si,
				   unsigned char usage)
L
Linus Torvalds 已提交
474
{
475
	unsigned long offset;
476
	unsigned long scan_base;
477
	unsigned long last_in_cluster = 0;
478
	int latency_ration = LATENCY_LIMIT;
479

480
	/*
481 482 483 484 485 486 487
	 * 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
488
	 * And we let swap pages go all over an SSD partition.  Hugh
489 490
	 */

491
	si->flags += SWP_SCANNING;
492
	scan_base = offset = si->cluster_next;
493

494 495 496 497 498 499
	/* SSD algorithm */
	if (si->cluster_info) {
		scan_swap_map_try_ssd_cluster(si, &offset, &scan_base);
		goto checks;
	}

500 501 502 503 504
	if (unlikely(!si->cluster_nr--)) {
		if (si->pages - si->inuse_pages < SWAPFILE_CLUSTER) {
			si->cluster_nr = SWAPFILE_CLUSTER - 1;
			goto checks;
		}
505

506
		spin_unlock(&si->lock);
507

508 509 510 511 512 513 514 515 516 517
		/*
		 * 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;
518 519 520 521
		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
522
			if (si->swap_map[offset])
523 524
				last_in_cluster = offset + SWAPFILE_CLUSTER;
			else if (offset == last_in_cluster) {
525
				spin_lock(&si->lock);
526 527 528 529
				offset -= SWAPFILE_CLUSTER - 1;
				si->cluster_next = offset;
				si->cluster_nr = SWAPFILE_CLUSTER - 1;
				goto checks;
L
Linus Torvalds 已提交
530
			}
531 532 533 534
			if (unlikely(--latency_ration < 0)) {
				cond_resched();
				latency_ration = LATENCY_LIMIT;
			}
535
		}
536 537

		offset = si->lowest_bit;
538 539 540 541 542 543 544
		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) {
545
				spin_lock(&si->lock);
546 547 548 549 550 551 552 553 554 555 556 557
				offset -= SWAPFILE_CLUSTER - 1;
				si->cluster_next = offset;
				si->cluster_nr = SWAPFILE_CLUSTER - 1;
				goto checks;
			}
			if (unlikely(--latency_ration < 0)) {
				cond_resched();
				latency_ration = LATENCY_LIMIT;
			}
		}

		offset = scan_base;
558
		spin_lock(&si->lock);
559
		si->cluster_nr = SWAPFILE_CLUSTER - 1;
L
Linus Torvalds 已提交
560
	}
561

562
checks:
563 564 565 566
	if (si->cluster_info) {
		while (scan_swap_map_ssd_cluster_conflict(si, offset))
			scan_swap_map_try_ssd_cluster(si, &offset, &scan_base);
	}
567
	if (!(si->flags & SWP_WRITEOK))
568
		goto no_page;
569 570
	if (!si->highest_bit)
		goto no_page;
571
	if (offset > si->highest_bit)
572
		scan_base = offset = si->lowest_bit;
573

574 575
	/* reuse swap entry of cache-only swap if not busy. */
	if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
576
		int swap_was_freed;
577
		spin_unlock(&si->lock);
578
		swap_was_freed = __try_to_reclaim_swap(si, offset);
579
		spin_lock(&si->lock);
580 581 582 583 584 585
		/* entry was freed successfully, try to use this again */
		if (swap_was_freed)
			goto checks;
		goto scan; /* check next one */
	}

586 587 588 589 590 591 592 593 594 595 596
	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
597
	}
H
Hugh Dickins 已提交
598
	si->swap_map[offset] = usage;
599
	inc_cluster_info_page(si, si->cluster_info, offset);
600 601
	si->cluster_next = offset + 1;
	si->flags -= SWP_SCANNING;
602

603
	return offset;
604

605
scan:
606
	spin_unlock(&si->lock);
607
	while (++offset <= si->highest_bit) {
608
		if (!si->swap_map[offset]) {
609
			spin_lock(&si->lock);
610 611
			goto checks;
		}
612
		if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
613
			spin_lock(&si->lock);
614 615
			goto checks;
		}
616 617 618 619
		if (unlikely(--latency_ration < 0)) {
			cond_resched();
			latency_ration = LATENCY_LIMIT;
		}
620
	}
621
	offset = si->lowest_bit;
622
	while (offset < scan_base) {
623
		if (!si->swap_map[offset]) {
624
			spin_lock(&si->lock);
625 626
			goto checks;
		}
627
		if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
628
			spin_lock(&si->lock);
629 630
			goto checks;
		}
631 632 633 634
		if (unlikely(--latency_ration < 0)) {
			cond_resched();
			latency_ration = LATENCY_LIMIT;
		}
635
		offset++;
636
	}
637
	spin_lock(&si->lock);
638 639

no_page:
640
	si->flags -= SWP_SCANNING;
L
Linus Torvalds 已提交
641 642 643 644 645
	return 0;
}

swp_entry_t get_swap_page(void)
{
646
	struct swap_info_struct *si, *next;
647
	pgoff_t offset;
648
	struct list_head *tmp;
L
Linus Torvalds 已提交
649

650
	spin_lock(&swap_lock);
651
	if (atomic_long_read(&nr_swap_pages) <= 0)
652
		goto noswap;
653
	atomic_long_dec(&nr_swap_pages);
654

655 656
	list_for_each(tmp, &swap_list_head) {
		si = list_entry(tmp, typeof(*si), list);
657
		spin_lock(&si->lock);
658
		if (!si->highest_bit || !(si->flags & SWP_WRITEOK)) {
659
			spin_unlock(&si->lock);
660
			continue;
661
		}
662

663 664 665 666 667 668 669 670 671 672 673 674
		/*
		 * rotate the current swap_info that we're going to use
		 * to after any other swap_info that have the same prio,
		 * so that all equal-priority swap_info get used equally
		 */
		next = si;
		list_for_each_entry_continue(next, &swap_list_head, list) {
			if (si->prio != next->prio)
				break;
			list_rotate_left(&si->list);
			next = si;
		}
675 676

		spin_unlock(&swap_lock);
677
		/* This is called for allocating swap entry for cache */
H
Hugh Dickins 已提交
678
		offset = scan_swap_map(si, SWAP_HAS_CACHE);
679 680
		spin_unlock(&si->lock);
		if (offset)
681
			return swp_entry(si->type, offset);
682
		spin_lock(&swap_lock);
683 684 685 686 687 688 689 690 691 692 693
		/*
		 * if we got here, it's likely that si was almost full before,
		 * and since scan_swap_map() can drop the si->lock, multiple
		 * callers probably all tried to get a page from the same si
		 * and it filled up before we could get one.  So we need to
		 * try again.  Since we dropped the swap_lock, there may now
		 * be non-full higher priority swap_infos, and this si may have
		 * even been removed from the list (although very unlikely).
		 * Let's start over.
		 */
		tmp = &swap_list_head;
L
Linus Torvalds 已提交
694
	}
695

696
	atomic_long_inc(&nr_swap_pages);
697
noswap:
698
	spin_unlock(&swap_lock);
699
	return (swp_entry_t) {0};
L
Linus Torvalds 已提交
700 701
}

S
Seth Jennings 已提交
702
/* The only caller of this function is now suspend routine */
703 704 705 706 707 708
swp_entry_t get_swap_page_of_type(int type)
{
	struct swap_info_struct *si;
	pgoff_t offset;

	si = swap_info[type];
709
	spin_lock(&si->lock);
710
	if (si && (si->flags & SWP_WRITEOK)) {
711
		atomic_long_dec(&nr_swap_pages);
712 713 714
		/* This is called for allocating swap entry, not cache */
		offset = scan_swap_map(si, 1);
		if (offset) {
715
			spin_unlock(&si->lock);
716 717
			return swp_entry(type, offset);
		}
718
		atomic_long_inc(&nr_swap_pages);
719
	}
720
	spin_unlock(&si->lock);
721 722 723
	return (swp_entry_t) {0};
}

724
static struct swap_info_struct *swap_info_get(swp_entry_t entry)
L
Linus Torvalds 已提交
725
{
726
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
727 728 729 730 731 732 733
	unsigned long offset, type;

	if (!entry.val)
		goto out;
	type = swp_type(entry);
	if (type >= nr_swapfiles)
		goto bad_nofile;
734
	p = swap_info[type];
L
Linus Torvalds 已提交
735 736 737 738 739 740 741
	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;
742
	spin_lock(&p->lock);
L
Linus Torvalds 已提交
743 744 745
	return p;

bad_free:
746
	pr_err("swap_free: %s%08lx\n", Unused_offset, entry.val);
L
Linus Torvalds 已提交
747 748
	goto out;
bad_offset:
749
	pr_err("swap_free: %s%08lx\n", Bad_offset, entry.val);
L
Linus Torvalds 已提交
750 751
	goto out;
bad_device:
752
	pr_err("swap_free: %s%08lx\n", Unused_file, entry.val);
L
Linus Torvalds 已提交
753 754
	goto out;
bad_nofile:
755
	pr_err("swap_free: %s%08lx\n", Bad_file, entry.val);
L
Linus Torvalds 已提交
756 757
out:
	return NULL;
758
}
L
Linus Torvalds 已提交
759

760 761
static unsigned char swap_entry_free(struct swap_info_struct *p,
				     swp_entry_t entry, unsigned char usage)
L
Linus Torvalds 已提交
762
{
H
Hugh Dickins 已提交
763
	unsigned long offset = swp_offset(entry);
764 765
	unsigned char count;
	unsigned char has_cache;
766

H
Hugh Dickins 已提交
767 768 769
	count = p->swap_map[offset];
	has_cache = count & SWAP_HAS_CACHE;
	count &= ~SWAP_HAS_CACHE;
770

H
Hugh Dickins 已提交
771
	if (usage == SWAP_HAS_CACHE) {
772
		VM_BUG_ON(!has_cache);
H
Hugh Dickins 已提交
773
		has_cache = 0;
H
Hugh Dickins 已提交
774 775 776 777 778 779
	} 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
780 781 782 783 784 785 786 787 788
	} 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
789 790 791 792 793 794

	if (!count)
		mem_cgroup_uncharge_swap(entry);

	usage = count | has_cache;
	p->swap_map[offset] = usage;
795 796

	/* free if no reference */
H
Hugh Dickins 已提交
797
	if (!usage) {
798
		dec_cluster_info_page(p, p->cluster_info, offset);
799 800 801 802
		if (offset < p->lowest_bit)
			p->lowest_bit = offset;
		if (offset > p->highest_bit)
			p->highest_bit = offset;
803
		atomic_long_inc(&nr_swap_pages);
804
		p->inuse_pages--;
805
		frontswap_invalidate_page(p->type, offset);
806 807 808 809 810 811
		if (p->flags & SWP_BLKDEV) {
			struct gendisk *disk = p->bdev->bd_disk;
			if (disk->fops->swap_slot_free_notify)
				disk->fops->swap_slot_free_notify(p->bdev,
								  offset);
		}
L
Linus Torvalds 已提交
812
	}
H
Hugh Dickins 已提交
813 814

	return usage;
L
Linus Torvalds 已提交
815 816 817
}

/*
S
Seth Jennings 已提交
818
 * Caller has made sure that the swap device corresponding to entry
L
Linus Torvalds 已提交
819 820 821 822
 * is still around or has not been recycled.
 */
void swap_free(swp_entry_t entry)
{
823
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
824 825 826

	p = swap_info_get(entry);
	if (p) {
H
Hugh Dickins 已提交
827
		swap_entry_free(p, entry, 1);
828
		spin_unlock(&p->lock);
L
Linus Torvalds 已提交
829 830 831
	}
}

832 833 834 835 836
/*
 * Called after dropping swapcache to decrease refcnt to swap entries.
 */
void swapcache_free(swp_entry_t entry, struct page *page)
{
837
	struct swap_info_struct *p;
838
	unsigned char count;
839 840 841

	p = swap_info_get(entry);
	if (p) {
H
Hugh Dickins 已提交
842 843 844
		count = swap_entry_free(p, entry, SWAP_HAS_CACHE);
		if (page)
			mem_cgroup_uncharge_swapcache(page, entry, count != 0);
845
		spin_unlock(&p->lock);
846
	}
847 848
}

L
Linus Torvalds 已提交
849
/*
850
 * How many references to page are currently swapped out?
H
Hugh Dickins 已提交
851 852
 * 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
853
 */
854
int page_swapcount(struct page *page)
L
Linus Torvalds 已提交
855
{
856 857
	int count = 0;
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
858 859
	swp_entry_t entry;

H
Hugh Dickins 已提交
860
	entry.val = page_private(page);
L
Linus Torvalds 已提交
861 862
	p = swap_info_get(entry);
	if (p) {
863
		count = swap_count(p->swap_map[swp_offset(entry)]);
864
		spin_unlock(&p->lock);
L
Linus Torvalds 已提交
865
	}
866
	return count;
L
Linus Torvalds 已提交
867 868 869
}

/*
870 871 872 873
 * 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
874
 */
875
int reuse_swap_page(struct page *page)
L
Linus Torvalds 已提交
876
{
877 878
	int count;

879
	VM_BUG_ON_PAGE(!PageLocked(page), page);
H
Hugh Dickins 已提交
880 881
	if (unlikely(PageKsm(page)))
		return 0;
882
	count = page_mapcount(page);
883
	if (count <= 1 && PageSwapCache(page)) {
884
		count += page_swapcount(page);
885 886 887 888 889
		if (count == 1 && !PageWriteback(page)) {
			delete_from_swap_cache(page);
			SetPageDirty(page);
		}
	}
H
Hugh Dickins 已提交
890
	return count <= 1;
L
Linus Torvalds 已提交
891 892 893
}

/*
894 895
 * 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
896
 */
897
int try_to_free_swap(struct page *page)
L
Linus Torvalds 已提交
898
{
899
	VM_BUG_ON_PAGE(!PageLocked(page), page);
L
Linus Torvalds 已提交
900 901 902 903 904

	if (!PageSwapCache(page))
		return 0;
	if (PageWriteback(page))
		return 0;
905
	if (page_swapcount(page))
L
Linus Torvalds 已提交
906 907
		return 0;

908 909 910 911 912 913 914 915 916 917 918 919
	/*
	 * 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.
	 *
S
Seth Jennings 已提交
920
	 * Hibernation suspends storage while it is writing the image
921
	 * to disk so check that here.
922
	 */
923
	if (pm_suspended_storage())
924 925
		return 0;

926 927 928
	delete_from_swap_cache(page);
	SetPageDirty(page);
	return 1;
929 930
}

L
Linus Torvalds 已提交
931 932 933 934
/*
 * Free the swap entry like above, but also try to
 * free the page cache entry if it is the last user.
 */
935
int free_swap_and_cache(swp_entry_t entry)
L
Linus Torvalds 已提交
936
{
937
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
938 939
	struct page *page = NULL;

940
	if (non_swap_entry(entry))
941
		return 1;
942

L
Linus Torvalds 已提交
943 944
	p = swap_info_get(entry);
	if (p) {
H
Hugh Dickins 已提交
945
		if (swap_entry_free(p, entry, 1) == SWAP_HAS_CACHE) {
946 947
			page = find_get_page(swap_address_space(entry),
						entry.val);
N
Nick Piggin 已提交
948
			if (page && !trylock_page(page)) {
949 950 951 952
				page_cache_release(page);
				page = NULL;
			}
		}
953
		spin_unlock(&p->lock);
L
Linus Torvalds 已提交
954 955
	}
	if (page) {
956 957 958 959
		/*
		 * Not mapped elsewhere, or swap space full? Free it!
		 * Also recheck PageSwapCache now page is locked (above).
		 */
960
		if (PageSwapCache(page) && !PageWriteback(page) &&
961
				(!page_mapped(page) || vm_swap_full())) {
L
Linus Torvalds 已提交
962 963 964 965 966 967
			delete_from_swap_cache(page);
			SetPageDirty(page);
		}
		unlock_page(page);
		page_cache_release(page);
	}
968
	return p != NULL;
L
Linus Torvalds 已提交
969 970
}

971
#ifdef CONFIG_HIBERNATION
972
/*
973
 * Find the swap type that corresponds to given device (if any).
974
 *
975 976 977 978
 * @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).
979
 */
980
int swap_type_of(dev_t device, sector_t offset, struct block_device **bdev_p)
981
{
982
	struct block_device *bdev = NULL;
983
	int type;
984

985 986 987
	if (device)
		bdev = bdget(device);

988
	spin_lock(&swap_lock);
989 990
	for (type = 0; type < nr_swapfiles; type++) {
		struct swap_info_struct *sis = swap_info[type];
991

992
		if (!(sis->flags & SWP_WRITEOK))
993
			continue;
994

995
		if (!bdev) {
996
			if (bdev_p)
997
				*bdev_p = bdgrab(sis->bdev);
998

999
			spin_unlock(&swap_lock);
1000
			return type;
1001
		}
1002
		if (bdev == sis->bdev) {
1003
			struct swap_extent *se = &sis->first_swap_extent;
1004 1005

			if (se->start_block == offset) {
1006
				if (bdev_p)
1007
					*bdev_p = bdgrab(sis->bdev);
1008

1009 1010
				spin_unlock(&swap_lock);
				bdput(bdev);
1011
				return type;
1012
			}
1013 1014 1015
		}
	}
	spin_unlock(&swap_lock);
1016 1017 1018
	if (bdev)
		bdput(bdev);

1019 1020 1021
	return -ENODEV;
}

1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
/*
 * 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;
1034
	return map_swap_entry(swp_entry(type, offset), &bdev);
1035 1036
}

1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
/*
 * 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;

1047 1048 1049 1050
	spin_lock(&swap_lock);
	if ((unsigned int)type < nr_swapfiles) {
		struct swap_info_struct *sis = swap_info[type];

1051
		spin_lock(&sis->lock);
1052 1053
		if (sis->flags & SWP_WRITEOK) {
			n = sis->pages;
1054
			if (free)
1055
				n -= sis->inuse_pages;
1056
		}
1057
		spin_unlock(&sis->lock);
1058
	}
1059
	spin_unlock(&swap_lock);
1060 1061
	return n;
}
1062
#endif /* CONFIG_HIBERNATION */
1063

1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
static inline int maybe_same_pte(pte_t pte, pte_t swp_pte)
{
#ifdef CONFIG_MEM_SOFT_DIRTY
	/*
	 * When pte keeps soft dirty bit the pte generated
	 * from swap entry does not has it, still it's same
	 * pte from logical point of view.
	 */
	pte_t swp_pte_dirty = pte_swp_mksoft_dirty(swp_pte);
	return pte_same(pte, swp_pte) || pte_same(pte, swp_pte_dirty);
#else
	return pte_same(pte, swp_pte);
#endif
}

L
Linus Torvalds 已提交
1079
/*
1080 1081 1082
 * 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
1083
 */
H
Hugh Dickins 已提交
1084
static int unuse_pte(struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
1085 1086
		unsigned long addr, swp_entry_t entry, struct page *page)
{
1087
	struct page *swapcache;
1088
	struct mem_cgroup *memcg;
H
Hugh Dickins 已提交
1089 1090 1091 1092
	spinlock_t *ptl;
	pte_t *pte;
	int ret = 1;

1093 1094 1095 1096 1097
	swapcache = page;
	page = ksm_might_need_to_copy(page, vma, addr);
	if (unlikely(!page))
		return -ENOMEM;

1098 1099
	if (mem_cgroup_try_charge_swapin(vma->vm_mm, page,
					 GFP_KERNEL, &memcg)) {
H
Hugh Dickins 已提交
1100
		ret = -ENOMEM;
1101 1102
		goto out_nolock;
	}
H
Hugh Dickins 已提交
1103 1104

	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
1105
	if (unlikely(!maybe_same_pte(*pte, swp_entry_to_pte(entry)))) {
1106
		mem_cgroup_cancel_charge_swapin(memcg);
H
Hugh Dickins 已提交
1107 1108 1109
		ret = 0;
		goto out;
	}
1110

K
KAMEZAWA Hiroyuki 已提交
1111
	dec_mm_counter(vma->vm_mm, MM_SWAPENTS);
K
KAMEZAWA Hiroyuki 已提交
1112
	inc_mm_counter(vma->vm_mm, MM_ANONPAGES);
L
Linus Torvalds 已提交
1113 1114 1115
	get_page(page);
	set_pte_at(vma->vm_mm, addr, pte,
		   pte_mkold(mk_pte(page, vma->vm_page_prot)));
1116 1117 1118 1119
	if (page == swapcache)
		page_add_anon_rmap(page, vma, addr);
	else /* ksm created a completely new copy */
		page_add_new_anon_rmap(page, vma, addr);
1120
	mem_cgroup_commit_charge_swapin(page, memcg);
L
Linus Torvalds 已提交
1121 1122 1123 1124 1125 1126
	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
1127 1128
out:
	pte_unmap_unlock(pte, ptl);
1129
out_nolock:
1130 1131 1132 1133
	if (page != swapcache) {
		unlock_page(page);
		put_page(page);
	}
H
Hugh Dickins 已提交
1134
	return ret;
L
Linus Torvalds 已提交
1135 1136 1137 1138 1139 1140 1141
}

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);
1142
	pte_t *pte;
1143
	int ret = 0;
L
Linus Torvalds 已提交
1144

H
Hugh Dickins 已提交
1145 1146 1147 1148 1149 1150 1151
	/*
	 * 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
S
Seth Jennings 已提交
1152
	 * preemptable whenever CONFIG_PREEMPT but not CONFIG_HIGHPTE.
H
Hugh Dickins 已提交
1153 1154
	 */
	pte = pte_offset_map(pmd, addr);
L
Linus Torvalds 已提交
1155 1156 1157 1158 1159
	do {
		/*
		 * swapoff spends a _lot_ of time in this loop!
		 * Test inline before going to call unuse_pte.
		 */
1160
		if (unlikely(maybe_same_pte(*pte, swp_pte))) {
H
Hugh Dickins 已提交
1161 1162 1163 1164 1165
			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
1166 1167
		}
	} while (pte++, addr += PAGE_SIZE, addr != end);
H
Hugh Dickins 已提交
1168 1169
	pte_unmap(pte - 1);
out:
1170
	return ret;
L
Linus Torvalds 已提交
1171 1172 1173 1174 1175 1176 1177 1178
}

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;
1179
	int ret;
L
Linus Torvalds 已提交
1180 1181 1182 1183

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1184
		if (pmd_none_or_trans_huge_or_clear_bad(pmd))
L
Linus Torvalds 已提交
1185
			continue;
1186 1187 1188
		ret = unuse_pte_range(vma, pmd, addr, next, entry, page);
		if (ret)
			return ret;
L
Linus Torvalds 已提交
1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
	} 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;
1199
	int ret;
L
Linus Torvalds 已提交
1200 1201 1202 1203 1204 1205

	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
1206 1207 1208
		ret = unuse_pmd_range(vma, pud, addr, next, entry, page);
		if (ret)
			return ret;
L
Linus Torvalds 已提交
1209 1210 1211 1212 1213 1214 1215 1216 1217
	} 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;
1218
	int ret;
L
Linus Torvalds 已提交
1219

H
Hugh Dickins 已提交
1220
	if (page_anon_vma(page)) {
L
Linus Torvalds 已提交
1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235
		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;
1236 1237 1238
		ret = unuse_pud_range(vma, pgd, addr, next, entry, page);
		if (ret)
			return ret;
L
Linus Torvalds 已提交
1239 1240 1241 1242 1243 1244 1245 1246
	} 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;
1247
	int ret = 0;
L
Linus Torvalds 已提交
1248 1249 1250

	if (!down_read_trylock(&mm->mmap_sem)) {
		/*
1251 1252
		 * 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
1253
		 */
1254
		activate_page(page);
L
Linus Torvalds 已提交
1255 1256 1257 1258 1259
		unlock_page(page);
		down_read(&mm->mmap_sem);
		lock_page(page);
	}
	for (vma = mm->mmap; vma; vma = vma->vm_next) {
1260
		if (vma->anon_vma && (ret = unuse_vma(vma, entry, page)))
L
Linus Torvalds 已提交
1261 1262 1263
			break;
	}
	up_read(&mm->mmap_sem);
1264
	return (ret < 0)? ret: 0;
L
Linus Torvalds 已提交
1265 1266 1267
}

/*
1268 1269
 * Scan swap_map (or frontswap_map if frontswap parameter is true)
 * from current position to next entry still in use.
L
Linus Torvalds 已提交
1270 1271
 * Recycle to start on reaching the end, returning 0 when empty.
 */
1272
static unsigned int find_next_to_unuse(struct swap_info_struct *si,
1273
					unsigned int prev, bool frontswap)
L
Linus Torvalds 已提交
1274
{
1275 1276
	unsigned int max = si->max;
	unsigned int i = prev;
1277
	unsigned char count;
L
Linus Torvalds 已提交
1278 1279

	/*
1280
	 * No need for swap_lock here: we're just looking
L
Linus Torvalds 已提交
1281 1282
	 * for whether an entry is in use, not modifying it; false
	 * hits are okay, and sys_swapoff() has already prevented new
1283
	 * allocations from this area (while holding swap_lock).
L
Linus Torvalds 已提交
1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298
	 */
	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;
		}
1299 1300 1301 1302 1303 1304
		if (frontswap) {
			if (frontswap_test(si, i))
				break;
			else
				continue;
		}
1305
		count = ACCESS_ONCE(si->swap_map[i]);
1306
		if (count && swap_count(count) != SWAP_MAP_BAD)
L
Linus Torvalds 已提交
1307 1308 1309 1310 1311 1312 1313 1314 1315
			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.
1316 1317 1318
 *
 * if the boolean frontswap is true, only unuse pages_to_unuse pages;
 * pages_to_unuse==0 means all pages; ignored if frontswap is false
L
Linus Torvalds 已提交
1319
 */
1320 1321
int try_to_unuse(unsigned int type, bool frontswap,
		 unsigned long pages_to_unuse)
L
Linus Torvalds 已提交
1322
{
1323
	struct swap_info_struct *si = swap_info[type];
L
Linus Torvalds 已提交
1324
	struct mm_struct *start_mm;
1325 1326 1327 1328 1329
	volatile unsigned char *swap_map; /* swap_map is accessed without
					   * locking. Mark it as volatile
					   * to prevent compiler doing
					   * something odd.
					   */
1330
	unsigned char swcount;
L
Linus Torvalds 已提交
1331 1332
	struct page *page;
	swp_entry_t entry;
1333
	unsigned int i = 0;
L
Linus Torvalds 已提交
1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347
	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
1348
	 * that.
L
Linus Torvalds 已提交
1349 1350 1351 1352 1353 1354 1355 1356 1357
	 */
	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.
	 */
1358
	while ((i = find_next_to_unuse(si, i, frontswap)) != 0) {
L
Linus Torvalds 已提交
1359 1360 1361 1362 1363
		if (signal_pending(current)) {
			retval = -EINTR;
			break;
		}

1364
		/*
L
Linus Torvalds 已提交
1365 1366
		 * Get a page for the entry, using the existing swap
		 * cache page if there is one.  Otherwise, get a clean
1367
		 * page and read the swap into it.
L
Linus Torvalds 已提交
1368 1369 1370
		 */
		swap_map = &si->swap_map[i];
		entry = swp_entry(type, i);
1371 1372
		page = read_swap_cache_async(entry,
					GFP_HIGHUSER_MOVABLE, NULL, 0);
L
Linus Torvalds 已提交
1373 1374 1375 1376 1377 1378 1379
		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.
			 */
1380 1381 1382 1383 1384 1385 1386 1387 1388
			swcount = *swap_map;
			/*
			 * We don't hold lock here, so the swap entry could be
			 * SWAP_MAP_BAD (when the cluster is discarding).
			 * Instead of fail out, We can just skip the swap
			 * entry because swapoff will wait for discarding
			 * finish anyway.
			 */
			if (!swcount || swcount == SWAP_MAP_BAD)
L
Linus Torvalds 已提交
1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419
				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
1420 1421 1422 1423 1424 1425
		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
1426
		}
H
Hugh Dickins 已提交
1427 1428 1429
		if (swap_count(swcount) && start_mm != &init_mm)
			retval = unuse_mm(start_mm, entry, page);

1430
		if (swap_count(*swap_map)) {
L
Linus Torvalds 已提交
1431 1432 1433 1434 1435 1436 1437 1438 1439
			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
1440
			while (swap_count(*swap_map) && !retval &&
L
Linus Torvalds 已提交
1441 1442
					(p = p->next) != &start_mm->mmlist) {
				mm = list_entry(p, struct mm_struct, mmlist);
1443
				if (!atomic_inc_not_zero(&mm->mm_users))
L
Linus Torvalds 已提交
1444 1445 1446 1447 1448 1449 1450 1451
					continue;
				spin_unlock(&mmlist_lock);
				mmput(prev_mm);
				prev_mm = mm;

				cond_resched();

				swcount = *swap_map;
1452
				if (!swap_count(swcount)) /* any usage ? */
L
Linus Torvalds 已提交
1453
					;
H
Hugh Dickins 已提交
1454
				else if (mm == &init_mm)
L
Linus Torvalds 已提交
1455
					set_start_mm = 1;
H
Hugh Dickins 已提交
1456
				else
L
Linus Torvalds 已提交
1457
					retval = unuse_mm(mm, entry, page);
1458

1459
				if (set_start_mm && *swap_map < swcount) {
L
Linus Torvalds 已提交
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 1486 1487 1488 1489
					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
1490 1491 1492 1493 1494 1495
		 *
		 * 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
1496
		 */
1497 1498
		if (swap_count(*swap_map) &&
		     PageDirty(page) && PageSwapCache(page)) {
L
Linus Torvalds 已提交
1499 1500 1501 1502 1503 1504 1505 1506
			struct writeback_control wbc = {
				.sync_mode = WB_SYNC_NONE,
			};

			swap_writepage(page, &wbc);
			lock_page(page);
			wait_on_page_writeback(page);
		}
1507 1508 1509 1510 1511 1512 1513 1514 1515 1516

		/*
		 * 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))
1517
			delete_from_swap_cache(page);
L
Linus Torvalds 已提交
1518 1519 1520 1521

		/*
		 * So we could skip searching mms once swap count went
		 * to 1, we did not mark any present ptes as dirty: must
1522
		 * mark page dirty so shrink_page_list will preserve it.
L
Linus Torvalds 已提交
1523 1524 1525 1526 1527 1528 1529 1530 1531 1532
		 */
		SetPageDirty(page);
		unlock_page(page);
		page_cache_release(page);

		/*
		 * Make sure that we aren't completely killing
		 * interactive performance.
		 */
		cond_resched();
1533 1534 1535 1536
		if (frontswap && pages_to_unuse > 0) {
			if (!--pages_to_unuse)
				break;
		}
L
Linus Torvalds 已提交
1537 1538 1539 1540 1541 1542 1543
	}

	mmput(start_mm);
	return retval;
}

/*
1544 1545 1546
 * 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
1547 1548 1549 1550 1551
 * added to the mmlist just after page_duplicate - before would be racy.
 */
static void drain_mmlist(void)
{
	struct list_head *p, *next;
1552
	unsigned int type;
L
Linus Torvalds 已提交
1553

1554 1555
	for (type = 0; type < nr_swapfiles; type++)
		if (swap_info[type]->inuse_pages)
L
Linus Torvalds 已提交
1556 1557 1558 1559 1560 1561 1562 1563 1564
			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
1565 1566 1567
 * 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
1568
 */
1569
static sector_t map_swap_entry(swp_entry_t entry, struct block_device **bdev)
L
Linus Torvalds 已提交
1570
{
H
Hugh Dickins 已提交
1571 1572 1573 1574 1575
	struct swap_info_struct *sis;
	struct swap_extent *start_se;
	struct swap_extent *se;
	pgoff_t offset;

1576
	sis = swap_info[swp_type(entry)];
H
Hugh Dickins 已提交
1577 1578 1579 1580 1581
	*bdev = sis->bdev;

	offset = swp_offset(entry);
	start_se = sis->curr_swap_extent;
	se = start_se;
L
Linus Torvalds 已提交
1582 1583 1584 1585 1586 1587 1588 1589

	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);
		}
1590
		lh = se->list.next;
L
Linus Torvalds 已提交
1591 1592 1593 1594 1595 1596
		se = list_entry(lh, struct swap_extent, list);
		sis->curr_swap_extent = se;
		BUG_ON(se == start_se);		/* It *must* be present */
	}
}

1597 1598 1599 1600 1601 1602 1603 1604 1605 1606
/*
 * 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
1607 1608 1609 1610 1611
/*
 * Free all of a swapdev's extent information
 */
static void destroy_swap_extents(struct swap_info_struct *sis)
{
1612
	while (!list_empty(&sis->first_swap_extent.list)) {
L
Linus Torvalds 已提交
1613 1614
		struct swap_extent *se;

1615
		se = list_entry(sis->first_swap_extent.list.next,
L
Linus Torvalds 已提交
1616 1617 1618 1619
				struct swap_extent, list);
		list_del(&se->list);
		kfree(se);
	}
1620 1621 1622 1623 1624 1625 1626 1627

	if (sis->flags & SWP_FILE) {
		struct file *swap_file = sis->swap_file;
		struct address_space *mapping = swap_file->f_mapping;

		sis->flags &= ~SWP_FILE;
		mapping->a_ops->swap_deactivate(swap_file);
	}
L
Linus Torvalds 已提交
1628 1629 1630 1631
}

/*
 * Add a block range (and the corresponding page range) into this swapdev's
1632
 * extent list.  The extent list is kept sorted in page order.
L
Linus Torvalds 已提交
1633
 *
1634
 * This function rather assumes that it is called in ascending page order.
L
Linus Torvalds 已提交
1635
 */
1636
int
L
Linus Torvalds 已提交
1637 1638 1639 1640 1641 1642 1643
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;

1644 1645 1646 1647 1648 1649 1650 1651 1652
	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
1653
		se = list_entry(lh, struct swap_extent, list);
1654 1655
		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
1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671
			/* 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;

1672
	list_add_tail(&new_se->list, &sis->first_swap_extent.list);
1673
	return 1;
L
Linus Torvalds 已提交
1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695
}

/*
 * 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.
 *
1696
 * For S_ISREG swapfiles we set S_SWAPFILE across the life of the swapon.  This
L
Linus Torvalds 已提交
1697 1698 1699 1700 1701 1702 1703 1704 1705 1706
 * 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.
 */
1707
static int setup_swap_extents(struct swap_info_struct *sis, sector_t *span)
L
Linus Torvalds 已提交
1708
{
1709 1710 1711
	struct file *swap_file = sis->swap_file;
	struct address_space *mapping = swap_file->f_mapping;
	struct inode *inode = mapping->host;
L
Linus Torvalds 已提交
1712 1713 1714 1715
	int ret;

	if (S_ISBLK(inode->i_mode)) {
		ret = add_swap_extent(sis, 0, sis->max, 0);
1716
		*span = sis->pages;
1717
		return ret;
L
Linus Torvalds 已提交
1718 1719
	}

1720
	if (mapping->a_ops->swap_activate) {
1721
		ret = mapping->a_ops->swap_activate(sis, swap_file, span);
1722 1723 1724 1725 1726
		if (!ret) {
			sis->flags |= SWP_FILE;
			ret = add_swap_extent(sis, 0, sis->max, 0);
			*span = sis->pages;
		}
1727
		return ret;
1728 1729
	}

1730
	return generic_swapfile_activate(sis, swap_file, span);
L
Linus Torvalds 已提交
1731 1732
}

1733
static void _enable_swap_info(struct swap_info_struct *p, int prio,
1734 1735
				unsigned char *swap_map,
				struct swap_cluster_info *cluster_info)
1736
{
1737
	struct swap_info_struct *si;
1738 1739 1740 1741 1742 1743

	if (prio >= 0)
		p->prio = prio;
	else
		p->prio = --least_priority;
	p->swap_map = swap_map;
1744
	p->cluster_info = cluster_info;
1745
	p->flags |= SWP_WRITEOK;
1746
	atomic_long_add(p->pages, &nr_swap_pages);
1747 1748
	total_swap_pages += p->pages;

1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763
	assert_spin_locked(&swap_lock);
	BUG_ON(!list_empty(&p->list));
	/*
	 * insert into swap list; the list is in priority order,
	 * so that get_swap_page() can get a page from the highest
	 * priority swap_info_struct with available page(s), and
	 * swapoff can adjust the auto-assigned (i.e. negative) prio
	 * values for any lower-priority swap_info_structs when
	 * removing a negative-prio swap_info_struct
	 */
	list_for_each_entry(si, &swap_list_head, list) {
		if (p->prio >= si->prio) {
			list_add_tail(&p->list, &si->list);
			return;
		}
1764
	}
1765 1766 1767 1768 1769 1770
	/*
	 * this covers two cases:
	 * 1) p->prio is less than all existing prio
	 * 2) the swap list is empty
	 */
	list_add_tail(&p->list, &swap_list_head);
1771 1772 1773 1774
}

static void enable_swap_info(struct swap_info_struct *p, int prio,
				unsigned char *swap_map,
1775
				struct swap_cluster_info *cluster_info,
1776 1777
				unsigned long *frontswap_map)
{
1778
	frontswap_init(p->type, frontswap_map);
1779
	spin_lock(&swap_lock);
1780
	spin_lock(&p->lock);
1781
	 _enable_swap_info(p, prio, swap_map, cluster_info);
1782
	spin_unlock(&p->lock);
1783 1784 1785 1786 1787 1788
	spin_unlock(&swap_lock);
}

static void reinsert_swap_info(struct swap_info_struct *p)
{
	spin_lock(&swap_lock);
1789
	spin_lock(&p->lock);
1790
	_enable_swap_info(p, p->prio, p->swap_map, p->cluster_info);
1791
	spin_unlock(&p->lock);
1792 1793 1794
	spin_unlock(&swap_lock);
}

1795
SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
L
Linus Torvalds 已提交
1796
{
1797
	struct swap_info_struct *p = NULL;
1798
	unsigned char *swap_map;
1799
	struct swap_cluster_info *cluster_info;
1800
	unsigned long *frontswap_map;
L
Linus Torvalds 已提交
1801 1802 1803
	struct file *swap_file, *victim;
	struct address_space *mapping;
	struct inode *inode;
1804
	struct filename *pathname;
1805
	int err, found = 0;
1806
	unsigned int old_block_size;
1807

L
Linus Torvalds 已提交
1808 1809 1810
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;

1811 1812
	BUG_ON(!current->mm);

L
Linus Torvalds 已提交
1813 1814
	pathname = getname(specialfile);
	if (IS_ERR(pathname))
X
Xiaotian Feng 已提交
1815
		return PTR_ERR(pathname);
L
Linus Torvalds 已提交
1816

1817
	victim = file_open_name(pathname, O_RDWR|O_LARGEFILE, 0);
L
Linus Torvalds 已提交
1818 1819 1820 1821 1822
	err = PTR_ERR(victim);
	if (IS_ERR(victim))
		goto out;

	mapping = victim->f_mapping;
1823
	spin_lock(&swap_lock);
1824
	list_for_each_entry(p, &swap_list_head, list) {
H
Hugh Dickins 已提交
1825
		if (p->flags & SWP_WRITEOK) {
1826 1827
			if (p->swap_file->f_mapping == mapping) {
				found = 1;
L
Linus Torvalds 已提交
1828
				break;
1829
			}
L
Linus Torvalds 已提交
1830 1831
		}
	}
1832
	if (!found) {
L
Linus Torvalds 已提交
1833
		err = -EINVAL;
1834
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
1835 1836
		goto out_dput;
	}
1837
	if (!security_vm_enough_memory_mm(current->mm, p->pages))
L
Linus Torvalds 已提交
1838 1839 1840
		vm_unacct_memory(p->pages);
	else {
		err = -ENOMEM;
1841
		spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
1842 1843
		goto out_dput;
	}
1844
	spin_lock(&p->lock);
1845
	if (p->prio < 0) {
1846 1847 1848 1849 1850
		struct swap_info_struct *si = p;

		list_for_each_entry_continue(si, &swap_list_head, list) {
			si->prio++;
		}
1851 1852
		least_priority++;
	}
1853
	list_del_init(&p->list);
1854
	atomic_long_sub(p->pages, &nr_swap_pages);
L
Linus Torvalds 已提交
1855 1856
	total_swap_pages -= p->pages;
	p->flags &= ~SWP_WRITEOK;
1857
	spin_unlock(&p->lock);
1858
	spin_unlock(&swap_lock);
1859

1860
	set_current_oom_origin();
1861
	err = try_to_unuse(p->type, false, 0); /* force unuse all pages */
1862
	clear_current_oom_origin();
L
Linus Torvalds 已提交
1863 1864 1865

	if (err) {
		/* re-insert swap space back into swap_list */
1866
		reinsert_swap_info(p);
L
Linus Torvalds 已提交
1867 1868
		goto out_dput;
	}
1869

S
Shaohua Li 已提交
1870 1871
	flush_work(&p->discard_work);

1872
	destroy_swap_extents(p);
H
Hugh Dickins 已提交
1873 1874 1875
	if (p->flags & SWP_CONTINUED)
		free_swap_count_continuations(p);

I
Ingo Molnar 已提交
1876
	mutex_lock(&swapon_mutex);
1877
	spin_lock(&swap_lock);
1878
	spin_lock(&p->lock);
1879 1880
	drain_mmlist();

1881 1882 1883
	/* wait for anyone still in scan_swap_map */
	p->highest_bit = 0;		/* cuts scans short */
	while (p->flags >= SWP_SCANNING) {
1884
		spin_unlock(&p->lock);
1885
		spin_unlock(&swap_lock);
1886
		schedule_timeout_uninterruptible(1);
1887
		spin_lock(&swap_lock);
1888
		spin_lock(&p->lock);
1889 1890
	}

L
Linus Torvalds 已提交
1891
	swap_file = p->swap_file;
1892
	old_block_size = p->old_block_size;
L
Linus Torvalds 已提交
1893 1894 1895 1896
	p->swap_file = NULL;
	p->max = 0;
	swap_map = p->swap_map;
	p->swap_map = NULL;
1897 1898
	cluster_info = p->cluster_info;
	p->cluster_info = NULL;
1899
	frontswap_map = frontswap_map_get(p);
1900
	spin_unlock(&p->lock);
1901
	spin_unlock(&swap_lock);
1902
	frontswap_invalidate_area(p->type);
1903
	frontswap_map_set(p, NULL);
I
Ingo Molnar 已提交
1904
	mutex_unlock(&swapon_mutex);
1905 1906
	free_percpu(p->percpu_cluster);
	p->percpu_cluster = NULL;
L
Linus Torvalds 已提交
1907
	vfree(swap_map);
1908
	vfree(cluster_info);
1909
	vfree(frontswap_map);
S
Seth Jennings 已提交
1910
	/* Destroy swap account information */
1911
	swap_cgroup_swapoff(p->type);
1912

L
Linus Torvalds 已提交
1913 1914 1915
	inode = mapping->host;
	if (S_ISBLK(inode->i_mode)) {
		struct block_device *bdev = I_BDEV(inode);
1916
		set_blocksize(bdev, old_block_size);
1917
		blkdev_put(bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL);
L
Linus Torvalds 已提交
1918
	} else {
1919
		mutex_lock(&inode->i_mutex);
L
Linus Torvalds 已提交
1920
		inode->i_flags &= ~S_SWAPFILE;
1921
		mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
1922 1923
	}
	filp_close(swap_file, NULL);
1924 1925 1926 1927 1928 1929 1930 1931 1932 1933

	/*
	 * Clear the SWP_USED flag after all resources are freed so that swapon
	 * can reuse this swap_info in alloc_swap_info() safely.  It is ok to
	 * not hold p->lock after we cleared its SWP_WRITEOK.
	 */
	spin_lock(&swap_lock);
	p->flags = 0;
	spin_unlock(&swap_lock);

L
Linus Torvalds 已提交
1934
	err = 0;
K
Kay Sievers 已提交
1935 1936
	atomic_inc(&proc_poll_event);
	wake_up_interruptible(&proc_poll_wait);
L
Linus Torvalds 已提交
1937 1938 1939 1940

out_dput:
	filp_close(victim, NULL);
out:
X
Xiaotian Feng 已提交
1941
	putname(pathname);
L
Linus Torvalds 已提交
1942 1943 1944 1945
	return err;
}

#ifdef CONFIG_PROC_FS
K
Kay Sievers 已提交
1946 1947
static unsigned swaps_poll(struct file *file, poll_table *wait)
{
1948
	struct seq_file *seq = file->private_data;
K
Kay Sievers 已提交
1949 1950 1951

	poll_wait(file, &proc_poll_wait, wait);

1952 1953
	if (seq->poll_event != atomic_read(&proc_poll_event)) {
		seq->poll_event = atomic_read(&proc_poll_event);
K
Kay Sievers 已提交
1954 1955 1956 1957 1958 1959
		return POLLIN | POLLRDNORM | POLLERR | POLLPRI;
	}

	return POLLIN | POLLRDNORM;
}

L
Linus Torvalds 已提交
1960 1961 1962
/* iterator */
static void *swap_start(struct seq_file *swap, loff_t *pos)
{
1963 1964
	struct swap_info_struct *si;
	int type;
L
Linus Torvalds 已提交
1965 1966
	loff_t l = *pos;

I
Ingo Molnar 已提交
1967
	mutex_lock(&swapon_mutex);
L
Linus Torvalds 已提交
1968

1969 1970 1971
	if (!l)
		return SEQ_START_TOKEN;

1972 1973 1974 1975
	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
1976
			continue;
1977
		if (!--l)
1978
			return si;
L
Linus Torvalds 已提交
1979 1980 1981 1982 1983 1984 1985
	}

	return NULL;
}

static void *swap_next(struct seq_file *swap, void *v, loff_t *pos)
{
1986 1987
	struct swap_info_struct *si = v;
	int type;
L
Linus Torvalds 已提交
1988

1989
	if (v == SEQ_START_TOKEN)
1990 1991 1992
		type = 0;
	else
		type = si->type + 1;
1993

1994 1995 1996 1997
	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
1998 1999
			continue;
		++*pos;
2000
		return si;
L
Linus Torvalds 已提交
2001 2002 2003 2004 2005 2006 2007
	}

	return NULL;
}

static void swap_stop(struct seq_file *swap, void *v)
{
I
Ingo Molnar 已提交
2008
	mutex_unlock(&swapon_mutex);
L
Linus Torvalds 已提交
2009 2010 2011 2012
}

static int swap_show(struct seq_file *swap, void *v)
{
2013
	struct swap_info_struct *si = v;
L
Linus Torvalds 已提交
2014 2015 2016
	struct file *file;
	int len;

2017
	if (si == SEQ_START_TOKEN) {
2018 2019 2020
		seq_puts(swap,"Filename\t\t\t\tType\t\tSize\tUsed\tPriority\n");
		return 0;
	}
L
Linus Torvalds 已提交
2021

2022
	file = si->swap_file;
2023
	len = seq_path(swap, &file->f_path, " \t\n\\");
2024
	seq_printf(swap, "%*s%s\t%u\t%u\t%d\n",
2025
			len < 40 ? 40 - len : 1, " ",
A
Al Viro 已提交
2026
			S_ISBLK(file_inode(file)->i_mode) ?
L
Linus Torvalds 已提交
2027
				"partition" : "file\t",
2028 2029 2030
			si->pages << (PAGE_SHIFT - 10),
			si->inuse_pages << (PAGE_SHIFT - 10),
			si->prio);
L
Linus Torvalds 已提交
2031 2032 2033
	return 0;
}

2034
static const struct seq_operations swaps_op = {
L
Linus Torvalds 已提交
2035 2036 2037 2038 2039 2040 2041 2042
	.start =	swap_start,
	.next =		swap_next,
	.stop =		swap_stop,
	.show =		swap_show
};

static int swaps_open(struct inode *inode, struct file *file)
{
2043
	struct seq_file *seq;
K
Kay Sievers 已提交
2044 2045 2046
	int ret;

	ret = seq_open(file, &swaps_op);
2047
	if (ret)
K
Kay Sievers 已提交
2048 2049
		return ret;

2050 2051 2052
	seq = file->private_data;
	seq->poll_event = atomic_read(&proc_poll_event);
	return 0;
L
Linus Torvalds 已提交
2053 2054
}

2055
static const struct file_operations proc_swaps_operations = {
L
Linus Torvalds 已提交
2056 2057 2058 2059
	.open		= swaps_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= seq_release,
K
Kay Sievers 已提交
2060
	.poll		= swaps_poll,
L
Linus Torvalds 已提交
2061 2062 2063 2064
};

static int __init procswaps_init(void)
{
2065
	proc_create("swaps", 0, NULL, &proc_swaps_operations);
L
Linus Torvalds 已提交
2066 2067 2068 2069 2070
	return 0;
}
__initcall(procswaps_init);
#endif /* CONFIG_PROC_FS */

J
Jan Beulich 已提交
2071 2072 2073 2074 2075 2076 2077 2078 2079
#ifdef MAX_SWAPFILES_CHECK
static int __init max_swapfiles_check(void)
{
	MAX_SWAPFILES_CHECK();
	return 0;
}
late_initcall(max_swapfiles_check);
#endif

2080
static struct swap_info_struct *alloc_swap_info(void)
L
Linus Torvalds 已提交
2081
{
2082
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
2083
	unsigned int type;
2084 2085 2086

	p = kzalloc(sizeof(*p), GFP_KERNEL);
	if (!p)
2087
		return ERR_PTR(-ENOMEM);
2088

2089
	spin_lock(&swap_lock);
2090 2091
	for (type = 0; type < nr_swapfiles; type++) {
		if (!(swap_info[type]->flags & SWP_USED))
L
Linus Torvalds 已提交
2092
			break;
2093
	}
2094
	if (type >= MAX_SWAPFILES) {
2095
		spin_unlock(&swap_lock);
2096
		kfree(p);
2097
		return ERR_PTR(-EPERM);
L
Linus Torvalds 已提交
2098
	}
2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116
	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.
		 */
	}
2117
	INIT_LIST_HEAD(&p->first_swap_extent.list);
2118
	INIT_LIST_HEAD(&p->list);
L
Linus Torvalds 已提交
2119
	p->flags = SWP_USED;
2120
	spin_unlock(&swap_lock);
2121
	spin_lock_init(&p->lock);
2122

2123 2124 2125
	return p;
}

2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136
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;
2137
			return -EINVAL;
2138 2139 2140 2141
		}
		p->old_block_size = block_size(p->bdev);
		error = set_blocksize(p->bdev, PAGE_SIZE);
		if (error < 0)
2142
			return error;
2143 2144 2145 2146
		p->flags |= SWP_BLKDEV;
	} else if (S_ISREG(inode->i_mode)) {
		p->bdev = inode->i_sb->s_bdev;
		mutex_lock(&inode->i_mutex);
2147 2148 2149 2150
		if (IS_SWAPFILE(inode))
			return -EBUSY;
	} else
		return -EINVAL;
2151 2152 2153 2154

	return 0;
}

2155 2156 2157 2158 2159 2160 2161
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;
2162
	unsigned long last_page;
2163 2164

	if (memcmp("SWAPSPACE2", swap_header->magic.magic, 10)) {
2165
		pr_err("Unable to find swap-space signature\n");
2166
		return 0;
2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178
	}

	/* 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) {
2179 2180
		pr_warn("Unable to handle swap header version %d\n",
			swap_header->info.version);
2181
		return 0;
2182 2183 2184 2185 2186 2187 2188 2189
	}

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

	/*
	 * Find out how many pages are allowed for a single swap
2190
	 * device. There are two limiting factors: 1) the number
2191 2192
	 * 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
2193
	 * different architectures. In order to find the
2194
	 * largest possible bit mask, a swap entry with swap type 0
2195
	 * and swap offset ~0UL is created, encoded to a swap pte,
2196
	 * decoded to a swp_entry_t again, and finally the swap
2197 2198 2199
	 * 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
2200
	 * swap pte.
2201 2202
	 */
	maxpages = swp_offset(pte_to_swp_entry(
2203
			swp_entry_to_pte(swp_entry(0, ~0UL)))) + 1;
2204 2205
	last_page = swap_header->info.last_page;
	if (last_page > maxpages) {
2206
		pr_warn("Truncating oversized swap area, only using %luk out of %luk\n",
2207 2208 2209 2210 2211
			maxpages << (PAGE_SHIFT - 10),
			last_page << (PAGE_SHIFT - 10));
	}
	if (maxpages > last_page) {
		maxpages = last_page + 1;
2212 2213 2214 2215 2216 2217 2218
		/* 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)
2219
		return 0;
2220 2221
	swapfilepages = i_size_read(inode) >> PAGE_SHIFT;
	if (swapfilepages && maxpages > swapfilepages) {
2222
		pr_warn("Swap area shorter than signature indicates\n");
2223
		return 0;
2224 2225
	}
	if (swap_header->info.nr_badpages && S_ISREG(inode->i_mode))
2226
		return 0;
2227
	if (swap_header->info.nr_badpages > MAX_SWAP_BADPAGES)
2228
		return 0;
2229 2230 2231 2232

	return maxpages;
}

2233 2234 2235
static int setup_swap_map_and_extents(struct swap_info_struct *p,
					union swap_header *swap_header,
					unsigned char *swap_map,
2236
					struct swap_cluster_info *cluster_info,
2237 2238 2239 2240 2241 2242
					unsigned long maxpages,
					sector_t *span)
{
	int i;
	unsigned int nr_good_pages;
	int nr_extents;
2243 2244
	unsigned long nr_clusters = DIV_ROUND_UP(maxpages, SWAPFILE_CLUSTER);
	unsigned long idx = p->cluster_next / SWAPFILE_CLUSTER;
2245 2246 2247

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

2248 2249
	cluster_set_null(&p->free_cluster_head);
	cluster_set_null(&p->free_cluster_tail);
S
Shaohua Li 已提交
2250 2251
	cluster_set_null(&p->discard_cluster_head);
	cluster_set_null(&p->discard_cluster_tail);
2252

2253 2254
	for (i = 0; i < swap_header->info.nr_badpages; i++) {
		unsigned int page_nr = swap_header->info.badpages[i];
2255 2256
		if (page_nr == 0 || page_nr > swap_header->info.last_page)
			return -EINVAL;
2257 2258 2259
		if (page_nr < maxpages) {
			swap_map[page_nr] = SWAP_MAP_BAD;
			nr_good_pages--;
2260 2261 2262 2263 2264
			/*
			 * Haven't marked the cluster free yet, no list
			 * operation involved
			 */
			inc_cluster_info_page(p, cluster_info, page_nr);
2265 2266 2267
		}
	}

2268 2269 2270 2271
	/* Haven't marked the cluster free yet, no list operation involved */
	for (i = maxpages; i < round_up(maxpages, SWAPFILE_CLUSTER); i++)
		inc_cluster_info_page(p, cluster_info, i);

2272 2273
	if (nr_good_pages) {
		swap_map[0] = SWAP_MAP_BAD;
2274 2275 2276 2277 2278
		/*
		 * Not mark the cluster free yet, no list
		 * operation involved
		 */
		inc_cluster_info_page(p, cluster_info, 0);
2279 2280 2281
		p->max = maxpages;
		p->pages = nr_good_pages;
		nr_extents = setup_swap_extents(p, span);
2282 2283
		if (nr_extents < 0)
			return nr_extents;
2284 2285 2286
		nr_good_pages = p->pages;
	}
	if (!nr_good_pages) {
2287
		pr_warn("Empty swap-file\n");
2288
		return -EINVAL;
2289 2290
	}

2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314
	if (!cluster_info)
		return nr_extents;

	for (i = 0; i < nr_clusters; i++) {
		if (!cluster_count(&cluster_info[idx])) {
			cluster_set_flag(&cluster_info[idx], CLUSTER_FLAG_FREE);
			if (cluster_is_null(&p->free_cluster_head)) {
				cluster_set_next_flag(&p->free_cluster_head,
								idx, 0);
				cluster_set_next_flag(&p->free_cluster_tail,
								idx, 0);
			} else {
				unsigned int tail;

				tail = cluster_next(&p->free_cluster_tail);
				cluster_set_next(&cluster_info[tail], idx);
				cluster_set_next_flag(&p->free_cluster_tail,
								idx, 0);
			}
		}
		idx++;
		if (idx == nr_clusters)
			idx = 0;
	}
2315 2316 2317
	return nr_extents;
}

2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331
/*
 * Helper to sys_swapon determining if a given swap
 * backing device queue supports DISCARD operations.
 */
static bool swap_discardable(struct swap_info_struct *si)
{
	struct request_queue *q = bdev_get_queue(si->bdev);

	if (!q || !blk_queue_discard(q))
		return false;

	return true;
}

2332 2333 2334
SYSCALL_DEFINE2(swapon, const char __user *, specialfile, int, swap_flags)
{
	struct swap_info_struct *p;
2335
	struct filename *name;
2336 2337
	struct file *swap_file = NULL;
	struct address_space *mapping;
2338 2339
	int i;
	int prio;
2340 2341
	int error;
	union swap_header *swap_header;
2342
	int nr_extents;
2343 2344 2345
	sector_t span;
	unsigned long maxpages;
	unsigned char *swap_map = NULL;
2346
	struct swap_cluster_info *cluster_info = NULL;
2347
	unsigned long *frontswap_map = NULL;
2348 2349 2350
	struct page *page = NULL;
	struct inode *inode = NULL;

2351 2352 2353
	if (swap_flags & ~SWAP_FLAGS_VALID)
		return -EINVAL;

2354 2355 2356 2357
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;

	p = alloc_swap_info();
2358 2359
	if (IS_ERR(p))
		return PTR_ERR(p);
2360

S
Shaohua Li 已提交
2361 2362
	INIT_WORK(&p->discard_work, swap_discard_work);

L
Linus Torvalds 已提交
2363 2364
	name = getname(specialfile);
	if (IS_ERR(name)) {
2365
		error = PTR_ERR(name);
L
Linus Torvalds 已提交
2366
		name = NULL;
2367
		goto bad_swap;
L
Linus Torvalds 已提交
2368
	}
2369
	swap_file = file_open_name(name, O_RDWR|O_LARGEFILE, 0);
L
Linus Torvalds 已提交
2370
	if (IS_ERR(swap_file)) {
2371
		error = PTR_ERR(swap_file);
L
Linus Torvalds 已提交
2372
		swap_file = NULL;
2373
		goto bad_swap;
L
Linus Torvalds 已提交
2374 2375 2376 2377 2378 2379
	}

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

	for (i = 0; i < nr_swapfiles; i++) {
2380
		struct swap_info_struct *q = swap_info[i];
L
Linus Torvalds 已提交
2381

2382
		if (q == p || !q->swap_file)
L
Linus Torvalds 已提交
2383
			continue;
2384 2385
		if (mapping == q->swap_file->f_mapping) {
			error = -EBUSY;
L
Linus Torvalds 已提交
2386
			goto bad_swap;
2387
		}
L
Linus Torvalds 已提交
2388 2389
	}

2390 2391
	inode = mapping->host;
	/* If S_ISREG(inode->i_mode) will do mutex_lock(&inode->i_mutex); */
2392 2393
	error = claim_swapfile(p, inode);
	if (unlikely(error))
L
Linus Torvalds 已提交
2394 2395 2396 2397 2398 2399 2400 2401 2402
		goto bad_swap;

	/*
	 * Read the swap header.
	 */
	if (!mapping->a_ops->readpage) {
		error = -EINVAL;
		goto bad_swap;
	}
2403
	page = read_mapping_page(mapping, 0, swap_file);
L
Linus Torvalds 已提交
2404 2405 2406 2407
	if (IS_ERR(page)) {
		error = PTR_ERR(page);
		goto bad_swap;
	}
2408
	swap_header = kmap(page);
L
Linus Torvalds 已提交
2409

2410 2411
	maxpages = read_swap_header(p, swap_header, inode);
	if (unlikely(!maxpages)) {
L
Linus Torvalds 已提交
2412 2413 2414
		error = -EINVAL;
		goto bad_swap;
	}
2415

2416
	/* OK, set up the swap map and apply the bad block list */
2417
	swap_map = vzalloc(maxpages);
2418 2419 2420 2421
	if (!swap_map) {
		error = -ENOMEM;
		goto bad_swap;
	}
2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435
	if (p->bdev && blk_queue_nonrot(bdev_get_queue(p->bdev))) {
		p->flags |= SWP_SOLIDSTATE;
		/*
		 * select a random position to start with to help wear leveling
		 * SSD
		 */
		p->cluster_next = 1 + (prandom_u32() % p->highest_bit);

		cluster_info = vzalloc(DIV_ROUND_UP(maxpages,
			SWAPFILE_CLUSTER) * sizeof(*cluster_info));
		if (!cluster_info) {
			error = -ENOMEM;
			goto bad_swap;
		}
2436 2437 2438 2439 2440 2441 2442 2443 2444 2445
		p->percpu_cluster = alloc_percpu(struct percpu_cluster);
		if (!p->percpu_cluster) {
			error = -ENOMEM;
			goto bad_swap;
		}
		for_each_possible_cpu(i) {
			struct percpu_cluster *cluster;
			cluster = per_cpu_ptr(p->percpu_cluster, i);
			cluster_set_null(&cluster->index);
		}
2446
	}
L
Linus Torvalds 已提交
2447

2448 2449 2450 2451
	error = swap_cgroup_swapon(p->type, maxpages);
	if (error)
		goto bad_swap;

2452
	nr_extents = setup_swap_map_and_extents(p, swap_header, swap_map,
2453
		cluster_info, maxpages, &span);
2454 2455
	if (unlikely(nr_extents < 0)) {
		error = nr_extents;
L
Linus Torvalds 已提交
2456 2457
		goto bad_swap;
	}
2458 2459
	/* frontswap enabled? set up bit-per-page map for frontswap */
	if (frontswap_enabled)
2460
		frontswap_map = vzalloc(BITS_TO_LONGS(maxpages) * sizeof(long));
L
Linus Torvalds 已提交
2461

2462 2463 2464 2465 2466 2467 2468 2469 2470
	if (p->bdev &&(swap_flags & SWAP_FLAG_DISCARD) && swap_discardable(p)) {
		/*
		 * When discard is enabled for swap with no particular
		 * policy flagged, we set all swap discard flags here in
		 * order to sustain backward compatibility with older
		 * swapon(8) releases.
		 */
		p->flags |= (SWP_DISCARDABLE | SWP_AREA_DISCARD |
			     SWP_PAGE_DISCARD);
2471

2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488
		/*
		 * By flagging sys_swapon, a sysadmin can tell us to
		 * either do single-time area discards only, or to just
		 * perform discards for released swap page-clusters.
		 * Now it's time to adjust the p->flags accordingly.
		 */
		if (swap_flags & SWAP_FLAG_DISCARD_ONCE)
			p->flags &= ~SWP_PAGE_DISCARD;
		else if (swap_flags & SWAP_FLAG_DISCARD_PAGES)
			p->flags &= ~SWP_AREA_DISCARD;

		/* issue a swapon-time discard if it's still required */
		if (p->flags & SWP_AREA_DISCARD) {
			int err = discard_swap(p);
			if (unlikely(err))
				pr_err("swapon: discard_swap(%p): %d\n",
					p, err);
2489
		}
2490
	}
2491

I
Ingo Molnar 已提交
2492
	mutex_lock(&swapon_mutex);
2493
	prio = -1;
2494
	if (swap_flags & SWAP_FLAG_PREFER)
2495
		prio =
2496
		  (swap_flags & SWAP_FLAG_PRIO_MASK) >> SWAP_FLAG_PRIO_SHIFT;
2497
	enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
2498

2499
	pr_info("Adding %uk swap on %s.  "
2500
			"Priority:%d extents:%d across:%lluk %s%s%s%s%s\n",
2501
		p->pages<<(PAGE_SHIFT-10), name->name, p->prio,
2502 2503
		nr_extents, (unsigned long long)span<<(PAGE_SHIFT-10),
		(p->flags & SWP_SOLIDSTATE) ? "SS" : "",
2504
		(p->flags & SWP_DISCARDABLE) ? "D" : "",
2505 2506
		(p->flags & SWP_AREA_DISCARD) ? "s" : "",
		(p->flags & SWP_PAGE_DISCARD) ? "c" : "",
2507
		(frontswap_map) ? "FS" : "");
2508

I
Ingo Molnar 已提交
2509
	mutex_unlock(&swapon_mutex);
K
Kay Sievers 已提交
2510 2511 2512
	atomic_inc(&proc_poll_event);
	wake_up_interruptible(&proc_poll_wait);

2513 2514
	if (S_ISREG(inode->i_mode))
		inode->i_flags |= S_SWAPFILE;
L
Linus Torvalds 已提交
2515 2516 2517
	error = 0;
	goto out;
bad_swap:
2518 2519
	free_percpu(p->percpu_cluster);
	p->percpu_cluster = NULL;
2520
	if (inode && S_ISBLK(inode->i_mode) && p->bdev) {
2521 2522
		set_blocksize(p->bdev, p->old_block_size);
		blkdev_put(p->bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL);
L
Linus Torvalds 已提交
2523
	}
2524
	destroy_swap_extents(p);
2525
	swap_cgroup_swapoff(p->type);
2526
	spin_lock(&swap_lock);
L
Linus Torvalds 已提交
2527 2528
	p->swap_file = NULL;
	p->flags = 0;
2529
	spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
2530
	vfree(swap_map);
2531
	vfree(cluster_info);
2532
	if (swap_file) {
2533
		if (inode && S_ISREG(inode->i_mode)) {
2534
			mutex_unlock(&inode->i_mutex);
2535 2536
			inode = NULL;
		}
L
Linus Torvalds 已提交
2537
		filp_close(swap_file, NULL);
2538
	}
L
Linus Torvalds 已提交
2539 2540 2541 2542 2543 2544 2545
out:
	if (page && !IS_ERR(page)) {
		kunmap(page);
		page_cache_release(page);
	}
	if (name)
		putname(name);
2546
	if (inode && S_ISREG(inode->i_mode))
2547
		mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
2548 2549 2550 2551 2552
	return error;
}

void si_swapinfo(struct sysinfo *val)
{
2553
	unsigned int type;
L
Linus Torvalds 已提交
2554 2555
	unsigned long nr_to_be_unused = 0;

2556
	spin_lock(&swap_lock);
2557 2558 2559 2560 2561
	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
2562
	}
2563
	val->freeswap = atomic_long_read(&nr_swap_pages) + nr_to_be_unused;
L
Linus Torvalds 已提交
2564
	val->totalswap = total_swap_pages + nr_to_be_unused;
2565
	spin_unlock(&swap_lock);
L
Linus Torvalds 已提交
2566 2567 2568 2569 2570
}

/*
 * Verify that a swap entry is valid and increment its swap map count.
 *
2571 2572 2573 2574 2575 2576
 * 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
2577
 * - swap-mapped reference requested but needs continued swap count. -> ENOMEM
L
Linus Torvalds 已提交
2578
 */
2579
static int __swap_duplicate(swp_entry_t entry, unsigned char usage)
L
Linus Torvalds 已提交
2580
{
2581
	struct swap_info_struct *p;
L
Linus Torvalds 已提交
2582
	unsigned long offset, type;
2583 2584
	unsigned char count;
	unsigned char has_cache;
H
Hugh Dickins 已提交
2585
	int err = -EINVAL;
L
Linus Torvalds 已提交
2586

2587
	if (non_swap_entry(entry))
H
Hugh Dickins 已提交
2588
		goto out;
2589

L
Linus Torvalds 已提交
2590 2591 2592
	type = swp_type(entry);
	if (type >= nr_swapfiles)
		goto bad_file;
2593
	p = swap_info[type];
L
Linus Torvalds 已提交
2594 2595
	offset = swp_offset(entry);

2596
	spin_lock(&p->lock);
2597 2598 2599
	if (unlikely(offset >= p->max))
		goto unlock_out;

H
Hugh Dickins 已提交
2600
	count = p->swap_map[offset];
2601 2602 2603 2604 2605 2606 2607 2608 2609 2610

	/*
	 * swapin_readahead() doesn't check if a swap entry is valid, so the
	 * swap entry could be SWAP_MAP_BAD. Check here with lock held.
	 */
	if (unlikely(swap_count(count) == SWAP_MAP_BAD)) {
		err = -ENOENT;
		goto unlock_out;
	}

H
Hugh Dickins 已提交
2611 2612 2613
	has_cache = count & SWAP_HAS_CACHE;
	count &= ~SWAP_HAS_CACHE;
	err = 0;
2614

H
Hugh Dickins 已提交
2615
	if (usage == SWAP_HAS_CACHE) {
2616 2617

		/* set SWAP_HAS_CACHE if there is no cache and entry is used */
H
Hugh Dickins 已提交
2618 2619 2620 2621 2622 2623
		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;
2624 2625

	} else if (count || has_cache) {
H
Hugh Dickins 已提交
2626

H
Hugh Dickins 已提交
2627 2628 2629
		if ((count & ~COUNT_CONTINUED) < SWAP_MAP_MAX)
			count += usage;
		else if ((count & ~COUNT_CONTINUED) > SWAP_MAP_MAX)
H
Hugh Dickins 已提交
2630
			err = -EINVAL;
H
Hugh Dickins 已提交
2631 2632 2633 2634
		else if (swap_count_continued(p, offset, count))
			count = COUNT_CONTINUED;
		else
			err = -ENOMEM;
2635
	} else
H
Hugh Dickins 已提交
2636 2637 2638 2639
		err = -ENOENT;			/* unused swap entry */

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

2640
unlock_out:
2641
	spin_unlock(&p->lock);
L
Linus Torvalds 已提交
2642
out:
H
Hugh Dickins 已提交
2643
	return err;
L
Linus Torvalds 已提交
2644 2645

bad_file:
2646
	pr_err("swap_dup: %s%08lx\n", Bad_file, entry.val);
L
Linus Torvalds 已提交
2647 2648
	goto out;
}
H
Hugh Dickins 已提交
2649

H
Hugh Dickins 已提交
2650 2651 2652 2653 2654 2655 2656 2657 2658
/*
 * 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);
}

2659
/*
2660 2661 2662 2663 2664
 * 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.
2665
 */
H
Hugh Dickins 已提交
2666
int swap_duplicate(swp_entry_t entry)
2667
{
H
Hugh Dickins 已提交
2668 2669 2670 2671 2672
	int err = 0;

	while (!err && __swap_duplicate(entry, 1) == -ENOMEM)
		err = add_swap_count_continuation(entry, GFP_ATOMIC);
	return err;
2673
}
L
Linus Torvalds 已提交
2674

2675
/*
2676 2677
 * @entry: swap entry for which we allocate swap cache.
 *
2678
 * Called when allocating swap cache for existing swap entry,
2679 2680 2681
 * This can return error codes. Returns 0 at success.
 * -EBUSY means there is a swap cache.
 * Note: return code is different from swap_duplicate().
2682 2683 2684
 */
int swapcache_prepare(swp_entry_t entry)
{
H
Hugh Dickins 已提交
2685
	return __swap_duplicate(entry, SWAP_HAS_CACHE);
2686 2687
}

2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699
struct swap_info_struct *page_swap_info(struct page *page)
{
	swp_entry_t swap = { .val = page_private(page) };
	BUG_ON(!PageSwapCache(page));
	return swap_info[swp_type(swap)];
}

/*
 * out-of-line __page_file_ methods to avoid include hell.
 */
struct address_space *__page_file_mapping(struct page *page)
{
2700
	VM_BUG_ON_PAGE(!PageSwapCache(page), page);
2701 2702 2703 2704 2705 2706 2707
	return page_swap_info(page)->swap_file->f_mapping;
}
EXPORT_SYMBOL_GPL(__page_file_mapping);

pgoff_t __page_file_index(struct page *page)
{
	swp_entry_t swap = { .val = page_private(page) };
2708
	VM_BUG_ON_PAGE(!PageSwapCache(page), page);
2709 2710 2711 2712
	return swp_offset(swap);
}
EXPORT_SYMBOL_GPL(__page_file_index);

H
Hugh Dickins 已提交
2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765
/*
 * 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) {
2766
		spin_unlock(&si->lock);
H
Hugh Dickins 已提交
2767 2768 2769 2770 2771
		return -ENOMEM;
	}

	/*
	 * We are fortunate that although vmalloc_to_page uses pte_offset_map,
S
Seth Jennings 已提交
2772 2773
	 * no architecture is using highmem pages for kernel page tables: so it
	 * will not corrupt the GFP_ATOMIC caller's atomic page table kmaps.
H
Hugh Dickins 已提交
2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798
	 */
	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;

2799
		map = kmap_atomic(list_page) + offset;
H
Hugh Dickins 已提交
2800
		count = *map;
2801
		kunmap_atomic(map);
H
Hugh Dickins 已提交
2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813

		/*
		 * 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:
2814
	spin_unlock(&si->lock);
H
Hugh Dickins 已提交
2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843
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);
2844
	map = kmap_atomic(page) + offset;
H
Hugh Dickins 已提交
2845 2846 2847 2848 2849 2850 2851 2852 2853

	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)) {
2854
			kunmap_atomic(map);
H
Hugh Dickins 已提交
2855 2856
			page = list_entry(page->lru.next, struct page, lru);
			BUG_ON(page == head);
2857
			map = kmap_atomic(page) + offset;
H
Hugh Dickins 已提交
2858 2859
		}
		if (*map == SWAP_CONT_MAX) {
2860
			kunmap_atomic(map);
H
Hugh Dickins 已提交
2861 2862 2863
			page = list_entry(page->lru.next, struct page, lru);
			if (page == head)
				return false;	/* add count continuation */
2864
			map = kmap_atomic(page) + offset;
H
Hugh Dickins 已提交
2865 2866 2867
init_map:		*map = 0;		/* we didn't zero the page */
		}
		*map += 1;
2868
		kunmap_atomic(map);
H
Hugh Dickins 已提交
2869 2870
		page = list_entry(page->lru.prev, struct page, lru);
		while (page != head) {
2871
			map = kmap_atomic(page) + offset;
H
Hugh Dickins 已提交
2872
			*map = COUNT_CONTINUED;
2873
			kunmap_atomic(map);
H
Hugh Dickins 已提交
2874 2875 2876 2877 2878 2879 2880 2881 2882 2883
			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) {
2884
			kunmap_atomic(map);
H
Hugh Dickins 已提交
2885 2886
			page = list_entry(page->lru.next, struct page, lru);
			BUG_ON(page == head);
2887
			map = kmap_atomic(page) + offset;
H
Hugh Dickins 已提交
2888 2889 2890 2891 2892
		}
		BUG_ON(*map == 0);
		*map -= 1;
		if (*map == 0)
			count = 0;
2893
		kunmap_atomic(map);
H
Hugh Dickins 已提交
2894 2895
		page = list_entry(page->lru.prev, struct page, lru);
		while (page != head) {
2896
			map = kmap_atomic(page) + offset;
H
Hugh Dickins 已提交
2897 2898
			*map = SWAP_CONT_MAX | count;
			count = COUNT_CONTINUED;
2899
			kunmap_atomic(map);
H
Hugh Dickins 已提交
2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927
			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);
			}
		}
	}
}