shmem.c 79.3 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4 5 6 7 8
/*
 * Resizable virtual memory filesystem for Linux.
 *
 * Copyright (C) 2000 Linus Torvalds.
 *		 2000 Transmeta Corp.
 *		 2000-2001 Christoph Rohland
 *		 2000-2001 SAP AG
 *		 2002 Red Hat Inc.
9 10
 * Copyright (C) 2002-2011 Hugh Dickins.
 * Copyright (C) 2011 Google Inc.
11
 * Copyright (C) 2002-2005 VERITAS Software Corporation.
Linus Torvalds's avatar
Linus Torvalds committed
12 13 14 15 16 17
 * Copyright (C) 2004 Andi Kleen, SuSE Labs
 *
 * Extended attribute support for tmpfs:
 * Copyright (c) 2004, Luke Kenneth Casson Leighton <lkcl@lkcl.net>
 * Copyright (c) 2004 Red Hat, Inc., James Morris <jmorris@redhat.com>
 *
18 19 20
 * tiny-shmem:
 * Copyright (c) 2004, 2008 Matt Mackall <mpm@selenic.com>
 *
Linus Torvalds's avatar
Linus Torvalds committed
21 22 23
 * This file is released under the GPL.
 */

24 25 26 27
#include <linux/fs.h>
#include <linux/init.h>
#include <linux/vfs.h>
#include <linux/mount.h>
Andrew Morton's avatar
Andrew Morton committed
28
#include <linux/ramfs.h>
Hugh Dickins's avatar
Hugh Dickins committed
29
#include <linux/pagemap.h>
30 31
#include <linux/file.h>
#include <linux/mm.h>
32
#include <linux/export.h>
33
#include <linux/swap.h>
34
#include <linux/aio.h>
35 36 37 38

static struct vfsmount *shm_mnt;

#ifdef CONFIG_SHMEM
Linus Torvalds's avatar
Linus Torvalds committed
39 40 41 42 43 44
/*
 * This virtual memory filesystem is heavily based on the ramfs. It
 * extends ramfs by the ability to use swap and honor resource limits
 * which makes it a completely usable filesystem.
 */

45
#include <linux/xattr.h>
46
#include <linux/exportfs.h>
47
#include <linux/posix_acl.h>
Christoph Hellwig's avatar
Christoph Hellwig committed
48
#include <linux/posix_acl_xattr.h>
Linus Torvalds's avatar
Linus Torvalds committed
49 50 51 52 53 54 55
#include <linux/mman.h>
#include <linux/string.h>
#include <linux/slab.h>
#include <linux/backing-dev.h>
#include <linux/shmem_fs.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>
56
#include <linux/pagevec.h>
57
#include <linux/percpu_counter.h>
58
#include <linux/falloc.h>
59
#include <linux/splice.h>
Linus Torvalds's avatar
Linus Torvalds committed
60 61 62 63
#include <linux/security.h>
#include <linux/swapops.h>
#include <linux/mempolicy.h>
#include <linux/namei.h>
64
#include <linux/ctype.h>
65
#include <linux/migrate.h>
66
#include <linux/highmem.h>
67
#include <linux/seq_file.h>
Mimi Zohar's avatar
Mimi Zohar committed
68
#include <linux/magic.h>
69

Linus Torvalds's avatar
Linus Torvalds committed
70 71 72
#include <asm/uaccess.h>
#include <asm/pgtable.h>

Hugh Dickins's avatar
Hugh Dickins committed
73
#define BLOCKS_PER_PAGE  (PAGE_CACHE_SIZE/512)
Linus Torvalds's avatar
Linus Torvalds committed
74 75 76 77 78
#define VM_ACCT(size)    (PAGE_CACHE_ALIGN(size) >> PAGE_SHIFT)

/* Pretend that each entry is of this size in directory's i_size */
#define BOGO_DIRENT_SIZE 20

79 80 81
/* Symlink up to this size is kmalloc'ed instead of using a swappable page */
#define SHORT_SYMLINK_LEN 128

82
/*
83 84 85
 * shmem_fallocate communicates with shmem_fault or shmem_writepage via
 * inode->i_private (with i_mutex making sure that it has only one user at
 * a time): we would prefer not to enlarge the shmem inode just for that.
86 87
 */
struct shmem_falloc {
88
	wait_queue_head_t *waitq; /* faults into hole wait for punch to end */
89 90 91 92 93 94
	pgoff_t start;		/* start of range currently being fallocated */
	pgoff_t next;		/* the next page offset to be fallocated */
	pgoff_t nr_falloced;	/* how many new pages have been fallocated */
	pgoff_t nr_unswapped;	/* how often writepage refused to swap out */
};

95
/* Flag allocation requirements to shmem_getpage */
Linus Torvalds's avatar
Linus Torvalds committed
96 97 98
enum sgp_type {
	SGP_READ,	/* don't exceed i_size, don't allocate page */
	SGP_CACHE,	/* don't exceed i_size, may allocate page */
99
	SGP_DIRTY,	/* like SGP_CACHE, but set new page dirty */
100 101
	SGP_WRITE,	/* may exceed i_size, may allocate !Uptodate page */
	SGP_FALLOC,	/* like SGP_WRITE, but make existing page Uptodate */
Linus Torvalds's avatar
Linus Torvalds committed
102 103
};

Andrew Morton's avatar
Andrew Morton committed
104
#ifdef CONFIG_TMPFS
105 106 107 108 109 110 111 112 113
static unsigned long shmem_default_max_blocks(void)
{
	return totalram_pages / 2;
}

static unsigned long shmem_default_max_inodes(void)
{
	return min(totalram_pages - totalhigh_pages, totalram_pages / 2);
}
Andrew Morton's avatar
Andrew Morton committed
114
#endif
115

116 117 118
static bool shmem_should_replace_page(struct page *page, gfp_t gfp);
static int shmem_replace_page(struct page **pagep, gfp_t gfp,
				struct shmem_inode_info *info, pgoff_t index);
119 120 121 122 123 124 125 126 127
static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
	struct page **pagep, enum sgp_type sgp, gfp_t gfp, int *fault_type);

static inline int shmem_getpage(struct inode *inode, pgoff_t index,
	struct page **pagep, enum sgp_type sgp, int *fault_type)
{
	return shmem_getpage_gfp(inode, index, pagep, sgp,
			mapping_gfp_mask(inode->i_mapping), fault_type);
}
Linus Torvalds's avatar
Linus Torvalds committed
128 129 130 131 132 133 134 135 136 137 138 139 140 141

static inline struct shmem_sb_info *SHMEM_SB(struct super_block *sb)
{
	return sb->s_fs_info;
}

/*
 * shmem_file_setup pre-accounts the whole fixed size of a VM object,
 * for shared memory and for shared anonymous (/dev/zero) mappings
 * (unless MAP_NORESERVE and sysctl_overcommit_memory <= 1),
 * consistent with the pre-accounting of private mappings ...
 */
static inline int shmem_acct_size(unsigned long flags, loff_t size)
{
142
	return (flags & VM_NORESERVE) ?
143
		0 : security_vm_enough_memory_mm(current->mm, VM_ACCT(size));
Linus Torvalds's avatar
Linus Torvalds committed
144 145 146 147
}

static inline void shmem_unacct_size(unsigned long flags, loff_t size)
{
148
	if (!(flags & VM_NORESERVE))
Linus Torvalds's avatar
Linus Torvalds committed
149 150 151 152 153 154 155 156 157 158 159
		vm_unacct_memory(VM_ACCT(size));
}

/*
 * ... whereas tmpfs objects are accounted incrementally as
 * pages are allocated, in order to allow huge sparse files.
 * shmem_getpage reports shmem_acct_block failure as -ENOSPC not -ENOMEM,
 * so that a failure on a sparse tmpfs mapping will give SIGBUS not OOM.
 */
static inline int shmem_acct_block(unsigned long flags)
{
160
	return (flags & VM_NORESERVE) ?
161
		security_vm_enough_memory_mm(current->mm, VM_ACCT(PAGE_CACHE_SIZE)) : 0;
Linus Torvalds's avatar
Linus Torvalds committed
162 163 164 165
}

static inline void shmem_unacct_blocks(unsigned long flags, long pages)
{
166
	if (flags & VM_NORESERVE)
Linus Torvalds's avatar
Linus Torvalds committed
167 168 169
		vm_unacct_memory(pages * VM_ACCT(PAGE_CACHE_SIZE));
}

170
static const struct super_operations shmem_ops;
171
static const struct address_space_operations shmem_aops;
172
static const struct file_operations shmem_file_operations;
173 174 175
static const struct inode_operations shmem_inode_operations;
static const struct inode_operations shmem_dir_inode_operations;
static const struct inode_operations shmem_special_inode_operations;
176
static const struct vm_operations_struct shmem_vm_ops;
Linus Torvalds's avatar
Linus Torvalds committed
177

178
static struct backing_dev_info shmem_backing_dev_info  __read_mostly = {
Linus Torvalds's avatar
Linus Torvalds committed
179
	.ra_pages	= 0,	/* No readahead */
180
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK | BDI_CAP_SWAP_BACKED,
Linus Torvalds's avatar
Linus Torvalds committed
181 182 183
};

static LIST_HEAD(shmem_swaplist);
184
static DEFINE_MUTEX(shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
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
static int shmem_reserve_inode(struct super_block *sb)
{
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);
	if (sbinfo->max_inodes) {
		spin_lock(&sbinfo->stat_lock);
		if (!sbinfo->free_inodes) {
			spin_unlock(&sbinfo->stat_lock);
			return -ENOSPC;
		}
		sbinfo->free_inodes--;
		spin_unlock(&sbinfo->stat_lock);
	}
	return 0;
}

static void shmem_free_inode(struct super_block *sb)
{
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);
	if (sbinfo->max_inodes) {
		spin_lock(&sbinfo->stat_lock);
		sbinfo->free_inodes++;
		spin_unlock(&sbinfo->stat_lock);
	}
}

211
/**
212
 * shmem_recalc_inode - recalculate the block usage of an inode
Linus Torvalds's avatar
Linus Torvalds committed
213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229
 * @inode: inode to recalc
 *
 * We have to calculate the free blocks since the mm can drop
 * undirtied hole pages behind our back.
 *
 * But normally   info->alloced == inode->i_mapping->nrpages + info->swapped
 * So mm freed is info->alloced - (inode->i_mapping->nrpages + info->swapped)
 *
 * It has to be called with the spinlock held.
 */
static void shmem_recalc_inode(struct inode *inode)
{
	struct shmem_inode_info *info = SHMEM_I(inode);
	long freed;

	freed = info->alloced - info->swapped - inode->i_mapping->nrpages;
	if (freed > 0) {
230 231 232
		struct shmem_sb_info *sbinfo = SHMEM_SB(inode->i_sb);
		if (sbinfo->max_blocks)
			percpu_counter_add(&sbinfo->used_blocks, -freed);
Linus Torvalds's avatar
Linus Torvalds committed
233
		info->alloced -= freed;
234
		inode->i_blocks -= freed * BLOCKS_PER_PAGE;
Linus Torvalds's avatar
Linus Torvalds committed
235 236 237 238
		shmem_unacct_blocks(info->flags, freed);
	}
}

239 240 241 242 243 244 245
/*
 * Replace item expected in radix tree by a new item, while holding tree lock.
 */
static int shmem_radix_tree_replace(struct address_space *mapping,
			pgoff_t index, void *expected, void *replacement)
{
	void **pslot;
246
	void *item;
247 248

	VM_BUG_ON(!expected);
249
	VM_BUG_ON(!replacement);
250
	pslot = radix_tree_lookup_slot(&mapping->page_tree, index);
251 252 253
	if (!pslot)
		return -ENOENT;
	item = radix_tree_deref_slot_protected(pslot, &mapping->tree_lock);
254 255
	if (item != expected)
		return -ENOENT;
256
	radix_tree_replace_slot(pslot, replacement);
257 258 259
	return 0;
}

260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277
/*
 * Sometimes, before we decide whether to proceed or to fail, we must check
 * that an entry was not already brought back from swap by a racing thread.
 *
 * Checking page is not enough: by the time a SwapCache page is locked, it
 * might be reused, and again be SwapCache, using the same swap as before.
 */
static bool shmem_confirm_swap(struct address_space *mapping,
			       pgoff_t index, swp_entry_t swap)
{
	void *item;

	rcu_read_lock();
	item = radix_tree_lookup(&mapping->page_tree, index);
	rcu_read_unlock();
	return item == swp_to_radix_entry(swap);
}

278 279 280 281 282 283 284
/*
 * Like add_to_page_cache_locked, but error if expected item has gone.
 */
static int shmem_add_to_page_cache(struct page *page,
				   struct address_space *mapping,
				   pgoff_t index, gfp_t gfp, void *expected)
{
285
	int error;
286

287 288
	VM_BUG_ON_PAGE(!PageLocked(page), page);
	VM_BUG_ON_PAGE(!PageSwapBacked(page), page);
289

290 291 292 293 294
	page_cache_get(page);
	page->mapping = mapping;
	page->index = index;

	spin_lock_irq(&mapping->tree_lock);
295
	if (!expected)
296 297 298 299
		error = radix_tree_insert(&mapping->page_tree, index, page);
	else
		error = shmem_radix_tree_replace(mapping, index, expected,
								 page);
300
	if (!error) {
301 302 303 304 305 306 307 308
		mapping->nrpages++;
		__inc_zone_page_state(page, NR_FILE_PAGES);
		__inc_zone_page_state(page, NR_SHMEM);
		spin_unlock_irq(&mapping->tree_lock);
	} else {
		page->mapping = NULL;
		spin_unlock_irq(&mapping->tree_lock);
		page_cache_release(page);
309 310 311 312
	}
	return error;
}

313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331
/*
 * Like delete_from_page_cache, but substitutes swap for page.
 */
static void shmem_delete_from_page_cache(struct page *page, void *radswap)
{
	struct address_space *mapping = page->mapping;
	int error;

	spin_lock_irq(&mapping->tree_lock);
	error = shmem_radix_tree_replace(mapping, page->index, page, radswap);
	page->mapping = NULL;
	mapping->nrpages--;
	__dec_zone_page_state(page, NR_FILE_PAGES);
	__dec_zone_page_state(page, NR_SHMEM);
	spin_unlock_irq(&mapping->tree_lock);
	page_cache_release(page);
	BUG_ON(error);
}

332 333 334 335 336 337
/*
 * Remove swap entry from radix tree, free the swap and its page cache.
 */
static int shmem_free_swap(struct address_space *mapping,
			   pgoff_t index, void *radswap)
{
338
	void *old;
339 340

	spin_lock_irq(&mapping->tree_lock);
341
	old = radix_tree_delete_item(&mapping->page_tree, index, radswap);
342
	spin_unlock_irq(&mapping->tree_lock);
343 344 345 346
	if (old != radswap)
		return -ENOENT;
	free_swap_and_cache(radix_to_swp_entry(radswap));
	return 0;
347 348
}

349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366
/*
 * SysV IPC SHM_UNLOCK restore Unevictable pages to their evictable lists.
 */
void shmem_unlock_mapping(struct address_space *mapping)
{
	struct pagevec pvec;
	pgoff_t indices[PAGEVEC_SIZE];
	pgoff_t index = 0;

	pagevec_init(&pvec, 0);
	/*
	 * Minor point, but we might as well stop if someone else SHM_LOCKs it.
	 */
	while (!mapping_unevictable(mapping)) {
		/*
		 * Avoid pagevec_lookup(): find_get_pages() returns 0 as if it
		 * has finished, if it hits a row of PAGEVEC_SIZE swap entries.
		 */
367 368
		pvec.nr = find_get_entries(mapping, index,
					   PAGEVEC_SIZE, pvec.pages, indices);
369 370 371
		if (!pvec.nr)
			break;
		index = indices[pvec.nr - 1] + 1;
372
		pagevec_remove_exceptionals(&pvec);
373 374 375 376
		check_move_unevictable_pages(pvec.pages, pvec.nr);
		pagevec_release(&pvec);
		cond_resched();
	}
377 378 379 380
}

/*
 * Remove range of pages and swap entries from radix tree, and free them.
381
 * If !unfalloc, truncate or punch hole; if unfalloc, undo failed fallocate.
382
 */
383 384
static void shmem_undo_range(struct inode *inode, loff_t lstart, loff_t lend,
								 bool unfalloc)
Linus Torvalds's avatar
Linus Torvalds committed
385
{
386
	struct address_space *mapping = inode->i_mapping;
Linus Torvalds's avatar
Linus Torvalds committed
387
	struct shmem_inode_info *info = SHMEM_I(inode);
388
	pgoff_t start = (lstart + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
389 390 391
	pgoff_t end = (lend + 1) >> PAGE_CACHE_SHIFT;
	unsigned int partial_start = lstart & (PAGE_CACHE_SIZE - 1);
	unsigned int partial_end = (lend + 1) & (PAGE_CACHE_SIZE - 1);
392
	struct pagevec pvec;
393 394
	pgoff_t indices[PAGEVEC_SIZE];
	long nr_swaps_freed = 0;
395
	pgoff_t index;
396 397
	int i;

398 399
	if (lend == -1)
		end = -1;	/* unsigned, so actually very big */
400 401 402

	pagevec_init(&pvec, 0);
	index = start;
403
	while (index < end) {
404 405 406
		pvec.nr = find_get_entries(mapping, index,
			min(end - index, (pgoff_t)PAGEVEC_SIZE),
			pvec.pages, indices);
407 408
		if (!pvec.nr)
			break;
409 410 411 412
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

413
			index = indices[i];
414
			if (index >= end)
415 416
				break;

417
			if (radix_tree_exceptional_entry(page)) {
418 419
				if (unfalloc)
					continue;
420 421
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
422
				continue;
423 424 425
			}

			if (!trylock_page(page))
426
				continue;
427 428
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
429
					VM_BUG_ON_PAGE(PageWriteback(page), page);
430 431
					truncate_inode_page(mapping, page);
				}
432 433 434
			}
			unlock_page(page);
		}
435
		pagevec_remove_exceptionals(&pvec);
436
		pagevec_release(&pvec);
437 438 439 440
		mem_cgroup_uncharge_end();
		cond_resched();
		index++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
441

442
	if (partial_start) {
443 444 445
		struct page *page = NULL;
		shmem_getpage(inode, start - 1, &page, SGP_READ, NULL);
		if (page) {
446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461
			unsigned int top = PAGE_CACHE_SIZE;
			if (start > end) {
				top = partial_end;
				partial_end = 0;
			}
			zero_user_segment(page, partial_start, top);
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}
	if (partial_end) {
		struct page *page = NULL;
		shmem_getpage(inode, end, &page, SGP_READ, NULL);
		if (page) {
			zero_user_segment(page, 0, partial_end);
462 463 464 465 466
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}
467 468
	if (start >= end)
		return;
469 470 471 472

	index = start;
	for ( ; ; ) {
		cond_resched();
473 474

		pvec.nr = find_get_entries(mapping, index,
475
				min(end - index, (pgoff_t)PAGEVEC_SIZE),
476
				pvec.pages, indices);
477
		if (!pvec.nr) {
478
			if (index == start || unfalloc)
479 480 481 482
				break;
			index = start;
			continue;
		}
483
		if ((index == start || unfalloc) && indices[0] >= end) {
484
			pagevec_remove_exceptionals(&pvec);
485
			pagevec_release(&pvec);
486 487 488 489 490 491
			break;
		}
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

492
			index = indices[i];
493
			if (index >= end)
494 495
				break;

496
			if (radix_tree_exceptional_entry(page)) {
497 498
				if (unfalloc)
					continue;
499 500 501 502 503
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
				continue;
			}

504
			lock_page(page);
505 506
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
507
					VM_BUG_ON_PAGE(PageWriteback(page), page);
508 509
					truncate_inode_page(mapping, page);
				}
510
			}
511 512
			unlock_page(page);
		}
513
		pagevec_remove_exceptionals(&pvec);
514
		pagevec_release(&pvec);
515 516 517
		mem_cgroup_uncharge_end();
		index++;
	}
518

Linus Torvalds's avatar
Linus Torvalds committed
519
	spin_lock(&info->lock);
520
	info->swapped -= nr_swaps_freed;
Linus Torvalds's avatar
Linus Torvalds committed
521 522
	shmem_recalc_inode(inode);
	spin_unlock(&info->lock);
523
}
Linus Torvalds's avatar
Linus Torvalds committed
524

525 526 527
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
{
	shmem_undo_range(inode, lstart, lend, false);
528
	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
Linus Torvalds's avatar
Linus Torvalds committed
529
}
530
EXPORT_SYMBOL_GPL(shmem_truncate_range);
Linus Torvalds's avatar
Linus Torvalds committed
531

532
static int shmem_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds's avatar
Linus Torvalds committed
533 534 535 536
{
	struct inode *inode = dentry->d_inode;
	int error;

537 538 539 540
	error = inode_change_ok(inode, attr);
	if (error)
		return error;

541 542 543
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
		loff_t oldsize = inode->i_size;
		loff_t newsize = attr->ia_size;
544

545 546 547 548 549 550 551 552 553 554 555
		if (newsize != oldsize) {
			i_size_write(inode, newsize);
			inode->i_ctime = inode->i_mtime = CURRENT_TIME;
		}
		if (newsize < oldsize) {
			loff_t holebegin = round_up(newsize, PAGE_SIZE);
			unmap_mapping_range(inode->i_mapping, holebegin, 0, 1);
			shmem_truncate_range(inode, newsize, (loff_t)-1);
			/* unmap again to remove racily COWed private pages */
			unmap_mapping_range(inode->i_mapping, holebegin, 0, 1);
		}
Linus Torvalds's avatar
Linus Torvalds committed
556 557
	}

558 559
	setattr_copy(inode, attr);
	if (attr->ia_valid & ATTR_MODE)
Christoph Hellwig's avatar
Christoph Hellwig committed
560
		error = posix_acl_chmod(inode, inode->i_mode);
Linus Torvalds's avatar
Linus Torvalds committed
561 562 563
	return error;
}

Al Viro's avatar
Al Viro committed
564
static void shmem_evict_inode(struct inode *inode)
Linus Torvalds's avatar
Linus Torvalds committed
565 566 567
{
	struct shmem_inode_info *info = SHMEM_I(inode);

568
	if (inode->i_mapping->a_ops == &shmem_aops) {
Linus Torvalds's avatar
Linus Torvalds committed
569 570
		shmem_unacct_size(info->flags, inode->i_size);
		inode->i_size = 0;
571
		shmem_truncate_range(inode, 0, (loff_t)-1);
Linus Torvalds's avatar
Linus Torvalds committed
572
		if (!list_empty(&info->swaplist)) {
573
			mutex_lock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
574
			list_del_init(&info->swaplist);
575
			mutex_unlock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
576
		}
577 578
	} else
		kfree(info->symlink);
579

580
	simple_xattrs_free(&info->xattrs);
581
	WARN_ON(inode->i_blocks);
582
	shmem_free_inode(inode->i_sb);
583
	clear_inode(inode);
Linus Torvalds's avatar
Linus Torvalds committed
584 585
}

586 587 588
/*
 * If swap found in inode, free it and move page from swapcache to filecache.
 */
589
static int shmem_unuse_inode(struct shmem_inode_info *info,
590
			     swp_entry_t swap, struct page **pagep)
Linus Torvalds's avatar
Linus Torvalds committed
591
{
592
	struct address_space *mapping = info->vfs_inode.i_mapping;
593
	void *radswap;
594
	pgoff_t index;
595 596
	gfp_t gfp;
	int error = 0;
Linus Torvalds's avatar
Linus Torvalds committed
597

598
	radswap = swp_to_radix_entry(swap);
599
	index = radix_tree_locate_item(&mapping->page_tree, radswap);
600
	if (index == -1)
601
		return 0;
602

Hugh Dickins's avatar
Hugh Dickins committed
603 604
	/*
	 * Move _head_ to start search for next from here.
Al Viro's avatar
Al Viro committed
605
	 * But be careful: shmem_evict_inode checks list_empty without taking
Hugh Dickins's avatar
Hugh Dickins committed
606
	 * mutex, and there's an instant in list_move_tail when info->swaplist
607
	 * would appear empty, if it were the only one on shmem_swaplist.
Hugh Dickins's avatar
Hugh Dickins committed
608 609 610
	 */
	if (shmem_swaplist.next != &info->swaplist)
		list_move_tail(&shmem_swaplist, &info->swaplist);
611

612 613 614 615 616 617 618
	gfp = mapping_gfp_mask(mapping);
	if (shmem_should_replace_page(*pagep, gfp)) {
		mutex_unlock(&shmem_swaplist_mutex);
		error = shmem_replace_page(pagep, gfp, info, index);
		mutex_lock(&shmem_swaplist_mutex);
		/*
		 * We needed to drop mutex to make that restrictive page
619 620 621 622 623 624 625 626 627 628 629 630 631 632 633
		 * allocation, but the inode might have been freed while we
		 * dropped it: although a racing shmem_evict_inode() cannot
		 * complete without emptying the radix_tree, our page lock
		 * on this swapcache page is not enough to prevent that -
		 * free_swap_and_cache() of our swap entry will only
		 * trylock_page(), removing swap from radix_tree whatever.
		 *
		 * We must not proceed to shmem_add_to_page_cache() if the
		 * inode has been freed, but of course we cannot rely on
		 * inode or mapping or info to check that.  However, we can
		 * safely check if our swap entry is still in use (and here
		 * it can't have got reused for another page): if it's still
		 * in use, then the inode cannot have been freed yet, and we
		 * can safely proceed (if it's no longer in use, that tells
		 * nothing about the inode, but we don't need to unuse swap).
634 635 636 637 638
		 */
		if (!page_swapcount(*pagep))
			error = -ENOENT;
	}

KAMEZAWA Hiroyuki's avatar
KAMEZAWA Hiroyuki committed
639
	/*
640 641 642
	 * We rely on shmem_swaplist_mutex, not only to protect the swaplist,
	 * but also to hold up shmem_evict_inode(): so inode cannot be freed
	 * beneath us (pagelock doesn't help until the page is in pagecache).
KAMEZAWA Hiroyuki's avatar
KAMEZAWA Hiroyuki committed
643
	 */
644 645
	if (!error)
		error = shmem_add_to_page_cache(*pagep, mapping, index,
646
						GFP_NOWAIT, radswap);
647
	if (error != -ENOMEM) {
648 649 650 651
		/*
		 * Truncation and eviction use free_swap_and_cache(), which
		 * only does trylock page: if we raced, best clean up here.
		 */
652 653
		delete_from_swap_cache(*pagep);
		set_page_dirty(*pagep);
654 655 656 657 658 659
		if (!error) {
			spin_lock(&info->lock);
			info->swapped--;
			spin_unlock(&info->lock);
			swap_free(swap);
		}
660
		error = 1;	/* not an error, but entry was found */
Linus Torvalds's avatar
Linus Torvalds committed
661
	}
662
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
663 664 665
}

/*
666
 * Search through swapped inodes to find and replace swap by page.
Linus Torvalds's avatar
Linus Torvalds committed
667
 */
668
int shmem_unuse(swp_entry_t swap, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
669
{
670
	struct list_head *this, *next;
Linus Torvalds's avatar
Linus Torvalds committed
671 672
	struct shmem_inode_info *info;
	int found = 0;
673 674 675 676
	int error = 0;

	/*
	 * There's a faint possibility that swap page was replaced before
677
	 * caller locked it: caller will come back later with the right page.
678
	 */
679
	if (unlikely(!PageSwapCache(page) || page_private(page) != swap.val))
680
		goto out;
681 682 683 684 685 686

	/*
	 * Charge page using GFP_KERNEL while we can wait, before taking
	 * the shmem_swaplist_mutex which might hold up shmem_writepage().
	 * Charged back to the user (not to caller) when swap account is used.
	 */
687
	error = mem_cgroup_charge_file(page, current->mm, GFP_KERNEL);
688 689
	if (error)
		goto out;
690
	/* No radix_tree_preload: swap entry keeps a place for page in tree */
Linus Torvalds's avatar
Linus Torvalds committed
691

692
	mutex_lock(&shmem_swaplist_mutex);
693 694
	list_for_each_safe(this, next, &shmem_swaplist) {
		info = list_entry(this, struct shmem_inode_info, swaplist);
695
		if (info->swapped)
696
			found = shmem_unuse_inode(info, swap, &page);
697 698
		else
			list_del_init(&info->swaplist);
699
		cond_resched();
700
		if (found)
701
			break;
Linus Torvalds's avatar
Linus Torvalds committed
702
	}
703
	mutex_unlock(&shmem_swaplist_mutex);
704 705 706 707

	if (found < 0)
		error = found;
out:
Hugh Dickins's avatar
Hugh Dickins committed
708 709
	unlock_page(page);
	page_cache_release(page);
710
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
711 712 713 714 715 716 717 718 719 720
}

/*
 * Move the page from the page cache to the swap cache.
 */
static int shmem_writepage(struct page *page, struct writeback_control *wbc)
{
	struct shmem_inode_info *info;
	struct address_space *mapping;
	struct inode *inode;
721 722
	swp_entry_t swap;
	pgoff_t index;
Linus Torvalds's avatar
Linus Torvalds committed
723 724 725 726 727 728 729 730

	BUG_ON(!PageLocked(page));
	mapping = page->mapping;
	index = page->index;
	inode = mapping->host;
	info = SHMEM_I(inode);
	if (info->flags & VM_LOCKED)
		goto redirty;
731
	if (!total_swap_pages)
Linus Torvalds's avatar
Linus Torvalds committed
732 733
		goto redirty;

734 735 736
	/*
	 * shmem_backing_dev_info's capabilities prevent regular writeback or
	 * sync from ever calling shmem_writepage; but a stacking filesystem
737
	 * might use ->writepage of its underlying filesystem, in which case
738
	 * tmpfs should write out to swap only in response to memory pressure,
739
	 * and not for the writeback threads or sync.
740
	 */
741 742 743 744
	if (!wbc->for_reclaim) {
		WARN_ON_ONCE(1);	/* Still happens? Tell us about it! */
		goto redirty;
	}
745 746 747 748 749

	/*
	 * This is somewhat ridiculous, but without plumbing a SWAP_MAP_FALLOC
	 * value into swapfile.c, the only way we can correctly account for a
	 * fallocated page arriving here is now to initialize it and write it.
750 751 752 753 754 755
	 *
	 * That's okay for a page already fallocated earlier, but if we have
	 * not yet completed the fallocation, then (a) we want to keep track
	 * of this page in case we have to undo it, and (b) it may not be a
	 * good idea to continue anyway, once we're pushing into swap.  So
	 * reactivate the page, and let shmem_fallocate() quit when too many.
756 757
	 */
	if (!PageUptodate(page)) {
758 759 760 761 762
		if (inode->i_private) {
			struct shmem_falloc *shmem_falloc;
			spin_lock(&inode->i_lock);
			shmem_falloc = inode->i_private;
			if (shmem_falloc &&
763
			    !shmem_falloc->waitq &&
764 765 766 767 768 769 770 771 772
			    index >= shmem_falloc->start &&
			    index < shmem_falloc->next)
				shmem_falloc->nr_unswapped++;
			else
				shmem_falloc = NULL;
			spin_unlock(&inode->i_lock);
			if (shmem_falloc)
				goto redirty;
		}
773 774 775 776 777
		clear_highpage(page);
		flush_dcache_page(page);
		SetPageUptodate(page);
	}

778 779 780
	swap = get_swap_page();
	if (!swap.val)
		goto redirty;
781

782 783
	/*
	 * Add inode to shmem_unuse()'s list of swapped-out inodes,
784 785
	 * if it's not already there.  Do it now before the page is
	 * moved to swap cache, when its pagelock no longer protects
786
	 * the inode from eviction.  But don't unlock the mutex until
787 788
	 * we've incremented swapped, because shmem_unuse_inode() will
	 * prune a !swapped inode from the swaplist under this mutex.
789
	 */
790 791 792
	mutex_lock(&shmem_swaplist_mutex);
	if (list_empty(&info->swaplist))
		list_add_tail(&info->swaplist, &shmem_swaplist);
793

794
	if (add_to_swap_cache(page, swap, GFP_ATOMIC) == 0) {
Hugh Dickins's avatar
Hugh Dickins committed
795
		swap_shmem_alloc(swap);
796 797 798 799 800
		shmem_delete_from_page_cache(page, swp_to_radix_entry(swap));

		spin_lock(&info->lock);
		info->swapped++;
		shmem_recalc_inode(inode);
801
		spin_unlock(&info->lock);
802 803

		mutex_unlock(&shmem_swaplist_mutex);
804
		BUG_ON(page_mapped(page));
805
		swap_writepage(page, wbc);
Linus Torvalds's avatar
Linus Torvalds committed
806 807 808
		return 0;
	}

809
	mutex_unlock(&shmem_swaplist_mutex);
810
	swapcache_free(swap, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
811 812
redirty:
	set_page_dirty(page);
813 814 815 816
	if (wbc->for_reclaim)
		return AOP_WRITEPAGE_ACTIVATE;	/* Return with page locked */
	unlock_page(page);
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
817 818 819
}

#ifdef CONFIG_NUMA
820
#ifdef CONFIG_TMPFS
821
static void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
822
{
823
	char buffer[64];
824

825
	if (!mpol || mpol->mode == MPOL_DEFAULT)
826
		return;		/* show nothing */
827

828
	mpol_to_str(buffer, sizeof(buffer), mpol);
829 830

	seq_printf(seq, ",mpol=%s", buffer);
831
}
832 833 834 835 836 837 838 839 840 841 842 843

static struct mempolicy *shmem_get_sbmpol(struct shmem_sb_info *sbinfo)
{
	struct mempolicy *mpol = NULL;
	if (sbinfo->mpol) {
		spin_lock(&sbinfo->stat_lock);	/* prevent replace/use races */
		mpol = sbinfo->mpol;
		mpol_get(mpol);
		spin_unlock(&sbinfo->stat_lock);
	}
	return mpol;
}
844 845
#endif /* CONFIG_TMPFS */

846 847
static struct page *shmem_swapin(swp_entry_t swap, gfp_t gfp,
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
848 849
{
	struct vm_area_struct pvma;
Mel Gorman's avatar
Mel Gorman committed
850
	struct page *page;
851

Linus Torvalds's avatar
Linus Torvalds committed
852
	/* Create a pseudo vma that just contains the policy */
853
	pvma.vm_start = 0;
854 855
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
856
	pvma.vm_ops = NULL;
Mel Gorman's avatar
Mel Gorman committed
857 858 859 860 861 862 863 864
	pvma.vm_policy = mpol_shared_policy_lookup(&info->policy, index);

	page = swapin_readahead(swap, gfp, &pvma, 0);

	/* Drop reference taken by mpol_shared_policy_lookup() */
	mpol_cond_put(pvma.vm_policy);

	return page;
Linus Torvalds's avatar
Linus Torvalds committed
865 866
}

867
static struct page *shmem_alloc_page(gfp_t gfp,
868
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
869 870
{
	struct vm_area_struct pvma;
Mel Gorman's avatar
Mel Gorman committed
871
	struct page *page;
Linus Torvalds's avatar
Linus Torvalds committed
872

873 874
	/* Create a pseudo vma that just contains the policy */
	pvma.vm_start = 0;
875 876
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
877
	pvma.vm_ops = NULL;
878
	pvma.vm_policy = mpol_shared_policy_lookup(&info->policy, index);
879

Mel Gorman's avatar
Mel Gorman committed
880 881 882 883 884 885
	page = alloc_page_vma(gfp, &pvma, 0);

	/* Drop reference taken by mpol_shared_policy_lookup() */
	mpol_cond_put(pvma.vm_policy);

	return page;
Linus Torvalds's avatar
Linus Torvalds committed
886
}
887 888
#else /* !CONFIG_NUMA */
#ifdef CONFIG_TMPFS
889
static inline void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
890 891 892 893
{
}
#endif /* CONFIG_TMPFS */

894 895
static inline struct page *shmem_swapin(swp_entry_t swap, gfp_t gfp,
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
896
{
897
	return swapin_readahead(swap, gfp, NULL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
898 899
}

900
static inline struct page *shmem_alloc_page(gfp_t gfp,
901
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
902
{
903
	return alloc_page(gfp);
Linus Torvalds's avatar
Linus Torvalds committed
904
}
905
#endif /* CONFIG_NUMA */
Linus Torvalds's avatar
Linus Torvalds committed
906

907 908 909 910 911 912 913
#if !defined(CONFIG_NUMA) || !defined(CONFIG_TMPFS)
static inline struct mempolicy *shmem_get_sbmpol(struct shmem_sb_info *sbinfo)
{
	return NULL;
}
#endif

914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953
/*
 * When a page is moved from swapcache to shmem filecache (either by the
 * usual swapin of shmem_getpage_gfp(), or by the less common swapoff of
 * shmem_unuse_inode()), it may have been read in earlier from swap, in
 * ignorance of the mapping it belongs to.  If that mapping has special
 * constraints (like the gma500 GEM driver, which requires RAM below 4GB),
 * we may need to copy to a suitable page before moving to filecache.
 *
 * In a future release, this may well be extended to respect cpuset and
 * NUMA mempolicy, and applied also to anonymous pages in do_swap_page();
 * but for now it is a simple matter of zone.
 */
static bool shmem_should_replace_page(struct page *page, gfp_t gfp)
{
	return page_zonenum(page) > gfp_zone(gfp);
}

static int shmem_replace_page(struct page **pagep, gfp_t gfp,
				struct shmem_inode_info *info, pgoff_t index)
{
	struct page *oldpage, *newpage;
	struct address_space *swap_mapping;
	pgoff_t swap_index;
	int error;

	oldpage = *pagep;
	swap_index = page_private(oldpage);
	swap_mapping = page_mapping(oldpage);

	/*
	 * We have arrived here because our zones are constrained, so don't
	 * limit chance of success by further cpuset and node constraints.
	 */
	gfp &= ~GFP_CONSTRAINT_MASK;
	newpage = shmem_alloc_page(gfp, info, index);
	if (!newpage)
		return -ENOMEM;

	page_cache_get(newpage);
	copy_highpage(newpage, oldpage);
954
	flush_dcache_page(newpage);
955 956 957 958 959 960 961 962 963 964 965 966 967 968

	__set_page_locked(newpage);
	SetPageUptodate(newpage);
	SetPageSwapBacked(newpage);
	set_page_private(newpage, swap_index);
	SetPageSwapCache(newpage);

	/*
	 * Our caller will very soon move newpage out of swapcache, but it's
	 * a nice clean interface for us to replace oldpage by newpage there.
	 */
	spin_lock_irq(&swap_mapping->tree_lock);
	error = shmem_radix_tree_replace(swap_mapping, swap_index, oldpage,
								   newpage);
969 970 971 972
	if (!error) {
		__inc_zone_page_state(newpage, NR_FILE_PAGES);
		__dec_zone_page_state(oldpage, NR_FILE_PAGES);
	}
973 974
	spin_unlock_irq(&swap_mapping->tree_lock);

975 976 977 978 979 980 981 982 983 984 985 986
	if (unlikely(error)) {
		/*
		 * Is this possible?  I think not, now that our callers check
		 * both PageSwapCache and page_private after getting page lock;
		 * but be defensive.  Reverse old to newpage for clear and free.
		 */
		oldpage = newpage;