shmem.c 74.5 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>
Hugh Dickins's avatar
Hugh Dickins committed
28
#include <linux/pagemap.h>
29 30
#include <linux/file.h>
#include <linux/mm.h>
31
#include <linux/export.h>
32 33 34 35 36
#include <linux/swap.h>

static struct vfsmount *shm_mnt;

#ifdef CONFIG_SHMEM
Linus Torvalds's avatar
Linus Torvalds committed
37 38 39 40 41 42
/*
 * 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.
 */

43
#include <linux/xattr.h>
44
#include <linux/exportfs.h>
45
#include <linux/posix_acl.h>
46
#include <linux/generic_acl.h>
Linus Torvalds's avatar
Linus Torvalds committed
47 48 49 50 51 52 53
#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>
54
#include <linux/pagevec.h>
55
#include <linux/percpu_counter.h>
56
#include <linux/falloc.h>
57
#include <linux/splice.h>
Linus Torvalds's avatar
Linus Torvalds committed
58 59 60 61
#include <linux/security.h>
#include <linux/swapops.h>
#include <linux/mempolicy.h>
#include <linux/namei.h>
62
#include <linux/ctype.h>
63
#include <linux/migrate.h>
64
#include <linux/highmem.h>
65
#include <linux/seq_file.h>
Mimi Zohar's avatar
Mimi Zohar committed
66
#include <linux/magic.h>
67

Linus Torvalds's avatar
Linus Torvalds committed
68 69 70
#include <asm/uaccess.h>
#include <asm/pgtable.h>

Hugh Dickins's avatar
Hugh Dickins committed
71
#define BLOCKS_PER_PAGE  (PAGE_CACHE_SIZE/512)
Linus Torvalds's avatar
Linus Torvalds committed
72 73 74 75 76
#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

77 78 79
/* Symlink up to this size is kmalloc'ed instead of using a swappable page */
#define SHORT_SYMLINK_LEN 128

80 81 82 83 84 85 86 87 88 89 90 91
/*
 * shmem_fallocate and shmem_writepage communicate 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.
 */
struct shmem_falloc {
	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 */
};

92
/* Flag allocation requirements to shmem_getpage */
Linus Torvalds's avatar
Linus Torvalds committed
93 94 95
enum sgp_type {
	SGP_READ,	/* don't exceed i_size, don't allocate page */
	SGP_CACHE,	/* don't exceed i_size, may allocate page */
96
	SGP_DIRTY,	/* like SGP_CACHE, but set new page dirty */
97 98
	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
99 100
};

Andrew Morton's avatar
Andrew Morton committed
101
#ifdef CONFIG_TMPFS
102 103 104 105 106 107 108 109 110
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
111
#endif
112

113 114 115
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);
116 117 118 119 120 121 122 123 124
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
125 126 127 128 129 130 131 132 133 134 135 136 137 138

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)
{
139
	return (flags & VM_NORESERVE) ?
140
		0 : security_vm_enough_memory_mm(current->mm, VM_ACCT(size));
Linus Torvalds's avatar
Linus Torvalds committed
141 142 143 144
}

static inline void shmem_unacct_size(unsigned long flags, loff_t size)
{
145
	if (!(flags & VM_NORESERVE))
Linus Torvalds's avatar
Linus Torvalds committed
146 147 148 149 150 151 152 153 154 155 156
		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)
{
157
	return (flags & VM_NORESERVE) ?
158
		security_vm_enough_memory_mm(current->mm, VM_ACCT(PAGE_CACHE_SIZE)) : 0;
Linus Torvalds's avatar
Linus Torvalds committed
159 160 161 162
}

static inline void shmem_unacct_blocks(unsigned long flags, long pages)
{
163
	if (flags & VM_NORESERVE)
Linus Torvalds's avatar
Linus Torvalds committed
164 165 166
		vm_unacct_memory(pages * VM_ACCT(PAGE_CACHE_SIZE));
}

167
static const struct super_operations shmem_ops;
168
static const struct address_space_operations shmem_aops;
169
static const struct file_operations shmem_file_operations;
170 171 172
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;
173
static const struct vm_operations_struct shmem_vm_ops;
Linus Torvalds's avatar
Linus Torvalds committed
174

175
static struct backing_dev_info shmem_backing_dev_info  __read_mostly = {
Linus Torvalds's avatar
Linus Torvalds committed
176
	.ra_pages	= 0,	/* No readahead */
177
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK | BDI_CAP_SWAP_BACKED,
Linus Torvalds's avatar
Linus Torvalds committed
178 179 180
};

static LIST_HEAD(shmem_swaplist);
181
static DEFINE_MUTEX(shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
182

183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207
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);
	}
}

208
/**
209
 * shmem_recalc_inode - recalculate the block usage of an inode
Linus Torvalds's avatar
Linus Torvalds committed
210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226
 * @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) {
227 228 229
		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
230
		info->alloced -= freed;
231
		inode->i_blocks -= freed * BLOCKS_PER_PAGE;
Linus Torvalds's avatar
Linus Torvalds committed
232 233 234 235
		shmem_unacct_blocks(info->flags, freed);
	}
}

236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258
/*
 * 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;
	void *item = NULL;

	VM_BUG_ON(!expected);
	pslot = radix_tree_lookup_slot(&mapping->page_tree, index);
	if (pslot)
		item = radix_tree_deref_slot_protected(pslot,
							&mapping->tree_lock);
	if (item != expected)
		return -ENOENT;
	if (replacement)
		radix_tree_replace_slot(pslot, replacement);
	else
		radix_tree_delete(&mapping->page_tree, index);
	return 0;
}

259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276
/*
 * 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);
}

277 278 279 280 281 282 283
/*
 * 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)
{
284
	int error;
285 286 287 288

	VM_BUG_ON(!PageLocked(page));
	VM_BUG_ON(!PageSwapBacked(page));

289 290 291 292 293
	page_cache_get(page);
	page->mapping = mapping;
	page->index = index;

	spin_lock_irq(&mapping->tree_lock);
294
	if (!expected)
295 296 297 298
		error = radix_tree_insert(&mapping->page_tree, index, page);
	else
		error = shmem_radix_tree_replace(mapping, index, expected,
								 page);
299
	if (!error) {
300 301 302 303 304 305 306 307
		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);
308 309 310 311
	}
	return error;
}

312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330
/*
 * 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);
}

331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353
/*
 * Like find_get_pages, but collecting swap entries as well as pages.
 */
static unsigned shmem_find_get_pages_and_swap(struct address_space *mapping,
					pgoff_t start, unsigned int nr_pages,
					struct page **pages, pgoff_t *indices)
{
	unsigned int i;
	unsigned int ret;
	unsigned int nr_found;

	rcu_read_lock();
restart:
	nr_found = radix_tree_gang_lookup_slot(&mapping->page_tree,
				(void ***)pages, indices, start, nr_pages);
	ret = 0;
	for (i = 0; i < nr_found; i++) {
		struct page *page;
repeat:
		page = radix_tree_deref_slot((void **)pages[i]);
		if (unlikely(!page))
			continue;
		if (radix_tree_exception(page)) {
354 355 356 357 358 359 360 361
			if (radix_tree_deref_retry(page))
				goto restart;
			/*
			 * Otherwise, we must be storing a swap entry
			 * here as an exceptional entry: so return it
			 * without attempting to raise page count.
			 */
			goto export;
362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400
		}
		if (!page_cache_get_speculative(page))
			goto repeat;

		/* Has the page moved? */
		if (unlikely(page != *((void **)pages[i]))) {
			page_cache_release(page);
			goto repeat;
		}
export:
		indices[ret] = indices[i];
		pages[ret] = page;
		ret++;
	}
	if (unlikely(!ret && nr_found))
		goto restart;
	rcu_read_unlock();
	return ret;
}

/*
 * 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)
{
	int error;

	spin_lock_irq(&mapping->tree_lock);
	error = shmem_radix_tree_replace(mapping, index, radswap, NULL);
	spin_unlock_irq(&mapping->tree_lock);
	if (!error)
		free_swap_and_cache(radix_to_swp_entry(radswap));
	return error;
}

/*
 * Pagevec may contain swap entries, so shuffle up pages before releasing.
 */
401
static void shmem_deswap_pagevec(struct pagevec *pvec)
402 403 404 405 406 407 408 409 410
{
	int i, j;

	for (i = 0, j = 0; i < pagevec_count(pvec); i++) {
		struct page *page = pvec->pages[i];
		if (!radix_tree_exceptional_entry(page))
			pvec->pages[j++] = page;
	}
	pvec->nr = j;
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
}

/*
 * 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.
		 */
		pvec.nr = shmem_find_get_pages_and_swap(mapping, index,
					PAGEVEC_SIZE, pvec.pages, indices);
		if (!pvec.nr)
			break;
		index = indices[pvec.nr - 1] + 1;
		shmem_deswap_pagevec(&pvec);
		check_move_unevictable_pages(pvec.pages, pvec.nr);
		pagevec_release(&pvec);
		cond_resched();
	}
441 442 443 444
}

/*
 * Remove range of pages and swap entries from radix tree, and free them.
445
 * If !unfalloc, truncate or punch hole; if unfalloc, undo failed fallocate.
446
 */
447 448
static void shmem_undo_range(struct inode *inode, loff_t lstart, loff_t lend,
								 bool unfalloc)
Linus Torvalds's avatar
Linus Torvalds committed
449
{
450
	struct address_space *mapping = inode->i_mapping;
Linus Torvalds's avatar
Linus Torvalds committed
451
	struct shmem_inode_info *info = SHMEM_I(inode);
452
	pgoff_t start = (lstart + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
453 454 455
	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);
456
	struct pagevec pvec;
457 458
	pgoff_t indices[PAGEVEC_SIZE];
	long nr_swaps_freed = 0;
459
	pgoff_t index;
460 461
	int i;

462 463
	if (lend == -1)
		end = -1;	/* unsigned, so actually very big */
464 465 466

	pagevec_init(&pvec, 0);
	index = start;
467
	while (index < end) {
468
		pvec.nr = shmem_find_get_pages_and_swap(mapping, index,
469
				min(end - index, (pgoff_t)PAGEVEC_SIZE),
470 471 472
							pvec.pages, indices);
		if (!pvec.nr)
			break;
473 474 475 476
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

477
			index = indices[i];
478
			if (index >= end)
479 480
				break;

481
			if (radix_tree_exceptional_entry(page)) {
482 483
				if (unfalloc)
					continue;
484 485
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
486
				continue;
487 488 489
			}

			if (!trylock_page(page))
490
				continue;
491 492 493 494 495
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
					VM_BUG_ON(PageWriteback(page));
					truncate_inode_page(mapping, page);
				}
496 497 498
			}
			unlock_page(page);
		}
499 500
		shmem_deswap_pagevec(&pvec);
		pagevec_release(&pvec);
501 502 503 504
		mem_cgroup_uncharge_end();
		cond_resched();
		index++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
505

506
	if (partial_start) {
507 508 509
		struct page *page = NULL;
		shmem_getpage(inode, start - 1, &page, SGP_READ, NULL);
		if (page) {
510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525
			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);
526 527 528 529 530
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}
531 532
	if (start >= end)
		return;
533 534 535 536

	index = start;
	for ( ; ; ) {
		cond_resched();
537
		pvec.nr = shmem_find_get_pages_and_swap(mapping, index,
538
				min(end - index, (pgoff_t)PAGEVEC_SIZE),
539 540
							pvec.pages, indices);
		if (!pvec.nr) {
541
			if (index == start || unfalloc)
542 543 544 545
				break;
			index = start;
			continue;
		}
546
		if ((index == start || unfalloc) && indices[0] >= end) {
547 548
			shmem_deswap_pagevec(&pvec);
			pagevec_release(&pvec);
549 550 551 552 553 554
			break;
		}
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

555
			index = indices[i];
556
			if (index >= end)
557 558
				break;

559
			if (radix_tree_exceptional_entry(page)) {
560 561
				if (unfalloc)
					continue;
562 563 564 565 566
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
				continue;
			}

567
			lock_page(page);
568 569 570 571 572
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
					VM_BUG_ON(PageWriteback(page));
					truncate_inode_page(mapping, page);
				}
573
			}
574 575
			unlock_page(page);
		}
576 577
		shmem_deswap_pagevec(&pvec);
		pagevec_release(&pvec);
578 579 580
		mem_cgroup_uncharge_end();
		index++;
	}
581

Linus Torvalds's avatar
Linus Torvalds committed
582
	spin_lock(&info->lock);
583
	info->swapped -= nr_swaps_freed;
Linus Torvalds's avatar
Linus Torvalds committed
584 585
	shmem_recalc_inode(inode);
	spin_unlock(&info->lock);
586
}
Linus Torvalds's avatar
Linus Torvalds committed
587

588 589 590
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
{
	shmem_undo_range(inode, lstart, lend, false);
591
	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
Linus Torvalds's avatar
Linus Torvalds committed
592
}
593
EXPORT_SYMBOL_GPL(shmem_truncate_range);
Linus Torvalds's avatar
Linus Torvalds committed
594

595
static int shmem_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds's avatar
Linus Torvalds committed
596 597 598 599
{
	struct inode *inode = dentry->d_inode;
	int error;

600 601 602 603
	error = inode_change_ok(inode, attr);
	if (error)
		return error;

604 605 606
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
		loff_t oldsize = inode->i_size;
		loff_t newsize = attr->ia_size;
607

608 609 610 611 612 613 614 615 616 617 618
		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
619 620
	}

621
	setattr_copy(inode, attr);
622
#ifdef CONFIG_TMPFS_POSIX_ACL
623
	if (attr->ia_valid & ATTR_MODE)
624
		error = generic_acl_chmod(inode);
625
#endif
Linus Torvalds's avatar
Linus Torvalds committed
626 627 628
	return error;
}

Al Viro's avatar
Al Viro committed
629
static void shmem_evict_inode(struct inode *inode)
Linus Torvalds's avatar
Linus Torvalds committed
630 631 632
{
	struct shmem_inode_info *info = SHMEM_I(inode);

633
	if (inode->i_mapping->a_ops == &shmem_aops) {
Linus Torvalds's avatar
Linus Torvalds committed
634 635
		shmem_unacct_size(info->flags, inode->i_size);
		inode->i_size = 0;
636
		shmem_truncate_range(inode, 0, (loff_t)-1);
Linus Torvalds's avatar
Linus Torvalds committed
637
		if (!list_empty(&info->swaplist)) {
638
			mutex_lock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
639
			list_del_init(&info->swaplist);
640
			mutex_unlock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
641
		}
642 643
	} else
		kfree(info->symlink);
644

645
	simple_xattrs_free(&info->xattrs);
646
	BUG_ON(inode->i_blocks);
647
	shmem_free_inode(inode->i_sb);
648
	clear_inode(inode);
Linus Torvalds's avatar
Linus Torvalds committed
649 650
}

651 652 653
/*
 * If swap found in inode, free it and move page from swapcache to filecache.
 */
654
static int shmem_unuse_inode(struct shmem_inode_info *info,
655
			     swp_entry_t swap, struct page **pagep)
Linus Torvalds's avatar
Linus Torvalds committed
656
{
657
	struct address_space *mapping = info->vfs_inode.i_mapping;
658
	void *radswap;
659
	pgoff_t index;
660 661
	gfp_t gfp;
	int error = 0;
Linus Torvalds's avatar
Linus Torvalds committed
662

663
	radswap = swp_to_radix_entry(swap);
664
	index = radix_tree_locate_item(&mapping->page_tree, radswap);
665
	if (index == -1)
666
		return 0;
667

Hugh Dickins's avatar
Hugh Dickins committed
668 669
	/*
	 * Move _head_ to start search for next from here.
Al Viro's avatar
Al Viro committed
670
	 * But be careful: shmem_evict_inode checks list_empty without taking
Hugh Dickins's avatar
Hugh Dickins committed
671
	 * mutex, and there's an instant in list_move_tail when info->swaplist
672
	 * would appear empty, if it were the only one on shmem_swaplist.
Hugh Dickins's avatar
Hugh Dickins committed
673 674 675
	 */
	if (shmem_swaplist.next != &info->swaplist)
		list_move_tail(&shmem_swaplist, &info->swaplist);
676

677 678 679 680 681 682 683
	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
684 685 686 687 688 689 690 691 692 693 694 695 696 697 698
		 * 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).
699 700 701 702 703
		 */
		if (!page_swapcount(*pagep))
			error = -ENOENT;
	}

KAMEZAWA Hiroyuki's avatar
KAMEZAWA Hiroyuki committed
704
	/*
705 706 707
	 * 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
708
	 */
709 710
	if (!error)
		error = shmem_add_to_page_cache(*pagep, mapping, index,
711
						GFP_NOWAIT, radswap);
712
	if (error != -ENOMEM) {
713 714 715 716
		/*
		 * Truncation and eviction use free_swap_and_cache(), which
		 * only does trylock page: if we raced, best clean up here.
		 */
717 718
		delete_from_swap_cache(*pagep);
		set_page_dirty(*pagep);
719 720 721 722 723 724
		if (!error) {
			spin_lock(&info->lock);
			info->swapped--;
			spin_unlock(&info->lock);
			swap_free(swap);
		}
725
		error = 1;	/* not an error, but entry was found */
Linus Torvalds's avatar
Linus Torvalds committed
726
	}
727
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
728 729 730
}

/*
731
 * Search through swapped inodes to find and replace swap by page.
Linus Torvalds's avatar
Linus Torvalds committed
732
 */
733
int shmem_unuse(swp_entry_t swap, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
734
{
735
	struct list_head *this, *next;
Linus Torvalds's avatar
Linus Torvalds committed
736 737
	struct shmem_inode_info *info;
	int found = 0;
738 739 740 741
	int error = 0;

	/*
	 * There's a faint possibility that swap page was replaced before
742
	 * caller locked it: caller will come back later with the right page.
743
	 */
744
	if (unlikely(!PageSwapCache(page) || page_private(page) != swap.val))
745
		goto out;
746 747 748 749 750 751 752 753 754

	/*
	 * 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.
	 */
	error = mem_cgroup_cache_charge(page, current->mm, GFP_KERNEL);
	if (error)
		goto out;
755
	/* No radix_tree_preload: swap entry keeps a place for page in tree */
Linus Torvalds's avatar
Linus Torvalds committed
756

757
	mutex_lock(&shmem_swaplist_mutex);
758 759
	list_for_each_safe(this, next, &shmem_swaplist) {
		info = list_entry(this, struct shmem_inode_info, swaplist);
760
		if (info->swapped)
761
			found = shmem_unuse_inode(info, swap, &page);
762 763
		else
			list_del_init(&info->swaplist);
764
		cond_resched();
765
		if (found)
766
			break;
Linus Torvalds's avatar
Linus Torvalds committed
767
	}
768
	mutex_unlock(&shmem_swaplist_mutex);
769 770 771 772

	if (found < 0)
		error = found;
out:
Hugh Dickins's avatar
Hugh Dickins committed
773 774
	unlock_page(page);
	page_cache_release(page);
775
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
776 777 778 779 780 781 782 783 784 785
}

/*
 * 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;
786 787
	swp_entry_t swap;
	pgoff_t index;
Linus Torvalds's avatar
Linus Torvalds committed
788 789 790 791 792 793 794 795

	BUG_ON(!PageLocked(page));
	mapping = page->mapping;
	index = page->index;
	inode = mapping->host;
	info = SHMEM_I(inode);
	if (info->flags & VM_LOCKED)
		goto redirty;
796
	if (!total_swap_pages)
Linus Torvalds's avatar
Linus Torvalds committed
797 798
		goto redirty;

799 800 801
	/*
	 * shmem_backing_dev_info's capabilities prevent regular writeback or
	 * sync from ever calling shmem_writepage; but a stacking filesystem
802
	 * might use ->writepage of its underlying filesystem, in which case
803
	 * tmpfs should write out to swap only in response to memory pressure,
804
	 * and not for the writeback threads or sync.
805
	 */
806 807 808 809
	if (!wbc->for_reclaim) {
		WARN_ON_ONCE(1);	/* Still happens? Tell us about it! */
		goto redirty;
	}
810 811 812 813 814

	/*
	 * 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.
815 816 817 818 819 820
	 *
	 * 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.
821 822
	 */
	if (!PageUptodate(page)) {
823 824 825 826 827 828 829 830 831 832 833 834 835 836
		if (inode->i_private) {
			struct shmem_falloc *shmem_falloc;
			spin_lock(&inode->i_lock);
			shmem_falloc = inode->i_private;
			if (shmem_falloc &&
			    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;
		}
837 838 839 840 841
		clear_highpage(page);
		flush_dcache_page(page);
		SetPageUptodate(page);
	}

842 843 844
	swap = get_swap_page();
	if (!swap.val)
		goto redirty;
845

846 847
	/*
	 * Add inode to shmem_unuse()'s list of swapped-out inodes,
848 849
	 * if it's not already there.  Do it now before the page is
	 * moved to swap cache, when its pagelock no longer protects
850
	 * the inode from eviction.  But don't unlock the mutex until
851 852
	 * we've incremented swapped, because shmem_unuse_inode() will
	 * prune a !swapped inode from the swaplist under this mutex.
853
	 */
854 855 856
	mutex_lock(&shmem_swaplist_mutex);
	if (list_empty(&info->swaplist))
		list_add_tail(&info->swaplist, &shmem_swaplist);
857

858
	if (add_to_swap_cache(page, swap, GFP_ATOMIC) == 0) {
Hugh Dickins's avatar
Hugh Dickins committed
859
		swap_shmem_alloc(swap);
860 861 862 863 864
		shmem_delete_from_page_cache(page, swp_to_radix_entry(swap));

		spin_lock(&info->lock);
		info->swapped++;
		shmem_recalc_inode(inode);
865
		spin_unlock(&info->lock);
866 867

		mutex_unlock(&shmem_swaplist_mutex);
868
		BUG_ON(page_mapped(page));
869
		swap_writepage(page, wbc);
Linus Torvalds's avatar
Linus Torvalds committed
870 871 872
		return 0;
	}

873
	mutex_unlock(&shmem_swaplist_mutex);
874
	swapcache_free(swap, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
875 876
redirty:
	set_page_dirty(page);
877 878 879 880
	if (wbc->for_reclaim)
		return AOP_WRITEPAGE_ACTIVATE;	/* Return with page locked */
	unlock_page(page);
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
881 882 883
}

#ifdef CONFIG_NUMA
884
#ifdef CONFIG_TMPFS
885
static void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
886
{
887
	char buffer[64];
888

889
	if (!mpol || mpol->mode == MPOL_DEFAULT)
890
		return;		/* show nothing */
891

892
	mpol_to_str(buffer, sizeof(buffer), mpol, 1);
893 894

	seq_printf(seq, ",mpol=%s", buffer);
895
}
896 897 898 899 900 901 902 903 904 905 906 907

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;
}
908 909
#endif /* CONFIG_TMPFS */

910 911
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
912
{
913
	struct mempolicy mpol, *spol;
Linus Torvalds's avatar
Linus Torvalds committed
914 915
	struct vm_area_struct pvma;

916
	spol = mpol_cond_copy(&mpol,
917
			mpol_shared_policy_lookup(&info->policy, index));
918

Linus Torvalds's avatar
Linus Torvalds committed
919
	/* Create a pseudo vma that just contains the policy */
920
	pvma.vm_start = 0;
921 922
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
923
	pvma.vm_ops = NULL;
924
	pvma.vm_policy = spol;
925
	return swapin_readahead(swap, gfp, &pvma, 0);
Linus Torvalds's avatar
Linus Torvalds committed
926 927
}

928
static struct page *shmem_alloc_page(gfp_t gfp,
929
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
930 931 932
{
	struct vm_area_struct pvma;

933 934
	/* Create a pseudo vma that just contains the policy */
	pvma.vm_start = 0;
935 936
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
937
	pvma.vm_ops = NULL;
938
	pvma.vm_policy = mpol_shared_policy_lookup(&info->policy, index);
939 940 941 942 943

	/*
	 * alloc_page_vma() will drop the shared policy reference
	 */
	return alloc_page_vma(gfp, &pvma, 0);
Linus Torvalds's avatar
Linus Torvalds committed
944
}
945 946
#else /* !CONFIG_NUMA */
#ifdef CONFIG_TMPFS
947
static inline void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
948 949 950 951
{
}
#endif /* CONFIG_TMPFS */

952 953
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
954
{
955
	return swapin_readahead(swap, gfp, NULL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
956 957
}

958
static inline struct page *shmem_alloc_page(gfp_t gfp,
959
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
960
{
961
	return alloc_page(gfp);
Linus Torvalds's avatar
Linus Torvalds committed
962
}
963
#endif /* CONFIG_NUMA */
Linus Torvalds's avatar
Linus Torvalds committed
964

965 966 967 968 969 970 971
#if !defined(CONFIG_NUMA) || !defined(CONFIG_TMPFS)
static inline struct mempolicy *shmem_get_sbmpol(struct shmem_sb_info *sbinfo)
{
	return NULL;
}
#endif

972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
/*
 * 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);
1012
	flush_dcache_page(newpage);
1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026

	__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);
1027 1028 1029 1030
	if (!error) {
		__inc_zone_page_state(newpage, NR_FILE_PAGES);
		__dec_zone_page_state(oldpage, NR_FILE_PAGES);
	}
1031 1032
	spin_unlock_irq(&swap_mapping->tree_lock);

1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044
	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;
	} else {
		mem_cgroup_replace_page_cache(oldpage, newpage);
		lru_cache_add_anon(newpage);
		*pagep = newpage;
	}
1045 1046 1047 1048 1049 1050 1051

	ClearPageSwapCache(oldpage);
	set_page_private(oldpage, 0);

	unlock_page(oldpage);
	page_cache_release(oldpage);
	page_cache_release(oldpage);
1052
	return error;
1053 1054
}

Linus Torvalds's avatar
Linus Torvalds committed
1055
/*
1056
 * shmem_getpage_gfp - find page in cache, or get from swap, or allocate
Linus Torvalds's avatar
Linus Torvalds committed
1057 1058 1059 1060 1061
 *
 * If we allocate a new one we do not mark it dirty. That's up to the
 * vm. If we swap it in we mark it dirty since we also free the swap
 * entry since a page cannot live in both the swap and page cache
 */
1062
static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
1063
	struct page **pagep, enum sgp_type sgp, gfp_t gfp, int *fault_type)
Linus Torvalds's avatar
Linus Torvalds committed