shmem.c 78.6 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 86 87 88 89 90 91 92 93
/*
 * 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 */
};

94
/* Flag allocation requirements to shmem_getpage */
Linus Torvalds's avatar
Linus Torvalds committed
95 96 97
enum sgp_type {
	SGP_READ,	/* don't exceed i_size, don't allocate page */
	SGP_CACHE,	/* don't exceed i_size, may allocate page */
98
	SGP_DIRTY,	/* like SGP_CACHE, but set new page dirty */
99 100
	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
101 102
};

Andrew Morton's avatar
Andrew Morton committed
103
#ifdef CONFIG_TMPFS
104 105 106 107 108 109 110 111 112
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
113
#endif
114

115 116 117
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);
118 119 120 121 122 123 124 125 126
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
127 128 129 130 131 132 133 134 135 136 137 138 139 140

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)
{
141
	return (flags & VM_NORESERVE) ?
142
		0 : security_vm_enough_memory_mm(current->mm, VM_ACCT(size));
Linus Torvalds's avatar
Linus Torvalds committed
143 144 145 146
}

static inline void shmem_unacct_size(unsigned long flags, loff_t size)
{
147
	if (!(flags & VM_NORESERVE))
Linus Torvalds's avatar
Linus Torvalds committed
148 149 150 151 152 153 154 155 156 157 158
		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)
{
159
	return (flags & VM_NORESERVE) ?
160
		security_vm_enough_memory_mm(current->mm, VM_ACCT(PAGE_CACHE_SIZE)) : 0;
Linus Torvalds's avatar
Linus Torvalds committed
161 162 163 164
}

static inline void shmem_unacct_blocks(unsigned long flags, long pages)
{
165
	if (flags & VM_NORESERVE)
Linus Torvalds's avatar
Linus Torvalds committed
166 167 168
		vm_unacct_memory(pages * VM_ACCT(PAGE_CACHE_SIZE));
}

169
static const struct super_operations shmem_ops;
170
static const struct address_space_operations shmem_aops;
171
static const struct file_operations shmem_file_operations;
172 173 174
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;
175
static const struct vm_operations_struct shmem_vm_ops;
Linus Torvalds's avatar
Linus Torvalds committed
176

177
static struct backing_dev_info shmem_backing_dev_info  __read_mostly = {
Linus Torvalds's avatar
Linus Torvalds committed
178
	.ra_pages	= 0,	/* No readahead */
179
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK | BDI_CAP_SWAP_BACKED,
Linus Torvalds's avatar
Linus Torvalds committed
180 181 182
};

static LIST_HEAD(shmem_swaplist);
183
static DEFINE_MUTEX(shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
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
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);
	}
}

210
/**
211
 * shmem_recalc_inode - recalculate the block usage of an inode
Linus Torvalds's avatar
Linus Torvalds committed
212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228
 * @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) {
229 230 231
		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
232
		info->alloced -= freed;
233
		inode->i_blocks -= freed * BLOCKS_PER_PAGE;
Linus Torvalds's avatar
Linus Torvalds committed
234 235 236 237
		shmem_unacct_blocks(info->flags, freed);
	}
}

238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
/*
 * 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;
}

261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
/*
 * 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);
}

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

288 289
	VM_BUG_ON_PAGE(!PageLocked(page), page);
	VM_BUG_ON_PAGE(!PageSwapBacked(page), page);
290

291 292 293 294 295
	page_cache_get(page);
	page->mapping = mapping;
	page->index = index;

	spin_lock_irq(&mapping->tree_lock);
296
	if (!expected)
297 298 299 300
		error = radix_tree_insert(&mapping->page_tree, index, page);
	else
		error = shmem_radix_tree_replace(mapping, index, expected,
								 page);
301
	if (!error) {
302 303 304 305 306 307 308 309
		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);
310 311 312 313
	}
	return error;
}

314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332
/*
 * 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);
}

333 334 335 336 337 338 339
/*
 * 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)
{
340 341 342 343 344 345
	void **slot;
	unsigned int ret = 0;
	struct radix_tree_iter iter;

	if (!nr_pages)
		return 0;
346 347 348

	rcu_read_lock();
restart:
349
	radix_tree_for_each_slot(slot, &mapping->page_tree, &iter, start) {
350 351
		struct page *page;
repeat:
352
		page = radix_tree_deref_slot(slot);
353 354 355
		if (unlikely(!page))
			continue;
		if (radix_tree_exception(page)) {
356 357 358 359 360 361 362 363
			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;
364 365 366 367 368
		}
		if (!page_cache_get_speculative(page))
			goto repeat;

		/* Has the page moved? */
369
		if (unlikely(page != *slot)) {
370 371 372 373
			page_cache_release(page);
			goto repeat;
		}
export:
374
		indices[ret] = iter.index;
375
		pages[ret] = page;
376 377
		if (++ret == nr_pages)
			break;
378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401
	}
	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.
 */
402
static void shmem_deswap_pagevec(struct pagevec *pvec)
403 404 405 406 407 408 409 410 411
{
	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;
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
}

/*
 * 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();
	}
442 443 444 445
}

/*
 * Remove range of pages and swap entries from radix tree, and free them.
446
 * If !unfalloc, truncate or punch hole; if unfalloc, undo failed fallocate.
447
 */
448 449
static void shmem_undo_range(struct inode *inode, loff_t lstart, loff_t lend,
								 bool unfalloc)
Linus Torvalds's avatar
Linus Torvalds committed
450
{
451
	struct address_space *mapping = inode->i_mapping;
Linus Torvalds's avatar
Linus Torvalds committed
452
	struct shmem_inode_info *info = SHMEM_I(inode);
453
	pgoff_t start = (lstart + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
454 455 456
	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);
457
	struct pagevec pvec;
458 459
	pgoff_t indices[PAGEVEC_SIZE];
	long nr_swaps_freed = 0;
460
	pgoff_t index;
461 462
	int i;

463 464
	if (lend == -1)
		end = -1;	/* unsigned, so actually very big */
465 466 467

	pagevec_init(&pvec, 0);
	index = start;
468
	while (index < end) {
469
		pvec.nr = shmem_find_get_pages_and_swap(mapping, index,
470
				min(end - index, (pgoff_t)PAGEVEC_SIZE),
471 472 473
							pvec.pages, indices);
		if (!pvec.nr)
			break;
474 475 476 477
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

478
			index = indices[i];
479
			if (index >= end)
480 481
				break;

482
			if (radix_tree_exceptional_entry(page)) {
483 484
				if (unfalloc)
					continue;
485 486
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
487
				continue;
488 489 490
			}

			if (!trylock_page(page))
491
				continue;
492 493
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
494
					VM_BUG_ON_PAGE(PageWriteback(page), page);
495 496
					truncate_inode_page(mapping, page);
				}
497 498 499
			}
			unlock_page(page);
		}
500 501
		shmem_deswap_pagevec(&pvec);
		pagevec_release(&pvec);
502 503 504 505
		mem_cgroup_uncharge_end();
		cond_resched();
		index++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
506

507
	if (partial_start) {
508 509 510
		struct page *page = NULL;
		shmem_getpage(inode, start - 1, &page, SGP_READ, NULL);
		if (page) {
511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526
			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);
527 528 529 530 531
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}
532 533
	if (start >= end)
		return;
534 535 536 537

	index = start;
	for ( ; ; ) {
		cond_resched();
538
		pvec.nr = shmem_find_get_pages_and_swap(mapping, index,
539
				min(end - index, (pgoff_t)PAGEVEC_SIZE),
540 541
							pvec.pages, indices);
		if (!pvec.nr) {
542
			if (index == start || unfalloc)
543 544 545 546
				break;
			index = start;
			continue;
		}
547
		if ((index == start || unfalloc) && indices[0] >= end) {
548 549
			shmem_deswap_pagevec(&pvec);
			pagevec_release(&pvec);
550 551 552 553 554 555
			break;
		}
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

556
			index = indices[i];
557
			if (index >= end)
558 559
				break;

560
			if (radix_tree_exceptional_entry(page)) {
561 562
				if (unfalloc)
					continue;
563 564 565 566 567
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
				continue;
			}

568
			lock_page(page);
569 570
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
571
					VM_BUG_ON_PAGE(PageWriteback(page), page);
572 573
					truncate_inode_page(mapping, page);
				}
574
			}
575 576
			unlock_page(page);
		}
577 578
		shmem_deswap_pagevec(&pvec);
		pagevec_release(&pvec);
579 580 581
		mem_cgroup_uncharge_end();
		index++;
	}
582

Linus Torvalds's avatar
Linus Torvalds committed
583
	spin_lock(&info->lock);
584
	info->swapped -= nr_swaps_freed;
Linus Torvalds's avatar
Linus Torvalds committed
585 586
	shmem_recalc_inode(inode);
	spin_unlock(&info->lock);
587
}
Linus Torvalds's avatar
Linus Torvalds committed
588

589 590 591
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
{
	shmem_undo_range(inode, lstart, lend, false);
592
	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
Linus Torvalds's avatar
Linus Torvalds committed
593
}
594
EXPORT_SYMBOL_GPL(shmem_truncate_range);
Linus Torvalds's avatar
Linus Torvalds committed
595

596
static int shmem_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds's avatar
Linus Torvalds committed
597 598 599 600
{
	struct inode *inode = dentry->d_inode;
	int error;

601 602 603 604
	error = inode_change_ok(inode, attr);
	if (error)
		return error;

605 606 607
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
		loff_t oldsize = inode->i_size;
		loff_t newsize = attr->ia_size;
608

609 610 611 612 613 614 615 616 617 618 619
		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
620 621
	}

622 623
	setattr_copy(inode, attr);
	if (attr->ia_valid & ATTR_MODE)
Christoph Hellwig's avatar
Christoph Hellwig committed
624
		error = posix_acl_chmod(inode, inode->i_mode);
Linus Torvalds's avatar
Linus Torvalds committed
625 626 627
	return error;
}

Al Viro's avatar
Al Viro committed
628
static void shmem_evict_inode(struct inode *inode)
Linus Torvalds's avatar
Linus Torvalds committed
629 630 631
{
	struct shmem_inode_info *info = SHMEM_I(inode);

632
	if (inode->i_mapping->a_ops == &shmem_aops) {
Linus Torvalds's avatar
Linus Torvalds committed
633 634
		shmem_unacct_size(info->flags, inode->i_size);
		inode->i_size = 0;
635
		shmem_truncate_range(inode, 0, (loff_t)-1);
Linus Torvalds's avatar
Linus Torvalds committed
636
		if (!list_empty(&info->swaplist)) {
637
			mutex_lock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
638
			list_del_init(&info->swaplist);
639
			mutex_unlock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
640
		}
641 642
	} else
		kfree(info->symlink);
643

644
	simple_xattrs_free(&info->xattrs);
645
	WARN_ON(inode->i_blocks);
646
	shmem_free_inode(inode->i_sb);
647
	clear_inode(inode);
Linus Torvalds's avatar
Linus Torvalds committed
648 649
}

650 651 652
/*
 * If swap found in inode, free it and move page from swapcache to filecache.
 */
653
static int shmem_unuse_inode(struct shmem_inode_info *info,
654
			     swp_entry_t swap, struct page **pagep)
Linus Torvalds's avatar
Linus Torvalds committed
655
{
656
	struct address_space *mapping = info->vfs_inode.i_mapping;
657
	void *radswap;
658
	pgoff_t index;
659 660
	gfp_t gfp;
	int error = 0;
Linus Torvalds's avatar
Linus Torvalds committed
661

662
	radswap = swp_to_radix_entry(swap);
663
	index = radix_tree_locate_item(&mapping->page_tree, radswap);
664
	if (index == -1)
665
		return 0;
666

Hugh Dickins's avatar
Hugh Dickins committed
667 668
	/*
	 * Move _head_ to start search for next from here.
Al Viro's avatar
Al Viro committed
669
	 * But be careful: shmem_evict_inode checks list_empty without taking
Hugh Dickins's avatar
Hugh Dickins committed
670
	 * mutex, and there's an instant in list_move_tail when info->swaplist
671
	 * would appear empty, if it were the only one on shmem_swaplist.
Hugh Dickins's avatar
Hugh Dickins committed
672 673 674
	 */
	if (shmem_swaplist.next != &info->swaplist)
		list_move_tail(&shmem_swaplist, &info->swaplist);
675

676 677 678 679 680 681 682
	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
683 684 685 686 687 688 689 690 691 692 693 694 695 696 697
		 * 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).
698 699 700 701 702
		 */
		if (!page_swapcount(*pagep))
			error = -ENOENT;
	}

KAMEZAWA Hiroyuki's avatar
KAMEZAWA Hiroyuki committed
703
	/*
704 705 706
	 * 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
707
	 */
708 709
	if (!error)
		error = shmem_add_to_page_cache(*pagep, mapping, index,
710
						GFP_NOWAIT, radswap);
711
	if (error != -ENOMEM) {
712 713 714 715
		/*
		 * Truncation and eviction use free_swap_and_cache(), which
		 * only does trylock page: if we raced, best clean up here.
		 */
716 717
		delete_from_swap_cache(*pagep);
		set_page_dirty(*pagep);
718 719 720 721 722 723
		if (!error) {
			spin_lock(&info->lock);
			info->swapped--;
			spin_unlock(&info->lock);
			swap_free(swap);
		}
724
		error = 1;	/* not an error, but entry was found */
Linus Torvalds's avatar
Linus Torvalds committed
725
	}
726
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
727 728 729
}

/*
730
 * Search through swapped inodes to find and replace swap by page.
Linus Torvalds's avatar
Linus Torvalds committed
731
 */
732
int shmem_unuse(swp_entry_t swap, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
733
{
734
	struct list_head *this, *next;
Linus Torvalds's avatar
Linus Torvalds committed
735 736
	struct shmem_inode_info *info;
	int found = 0;
737 738 739 740
	int error = 0;

	/*
	 * There's a faint possibility that swap page was replaced before
741
	 * caller locked it: caller will come back later with the right page.
742
	 */
743
	if (unlikely(!PageSwapCache(page) || page_private(page) != swap.val))
744
		goto out;
745 746 747 748 749 750 751 752 753

	/*
	 * 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;
754
	/* No radix_tree_preload: swap entry keeps a place for page in tree */
Linus Torvalds's avatar
Linus Torvalds committed
755

756
	mutex_lock(&shmem_swaplist_mutex);
757 758
	list_for_each_safe(this, next, &shmem_swaplist) {
		info = list_entry(this, struct shmem_inode_info, swaplist);
759
		if (info->swapped)
760
			found = shmem_unuse_inode(info, swap, &page);
761 762
		else
			list_del_init(&info->swaplist);
763
		cond_resched();
764
		if (found)
765
			break;
Linus Torvalds's avatar
Linus Torvalds committed
766
	}
767
	mutex_unlock(&shmem_swaplist_mutex);
768 769 770 771

	if (found < 0)
		error = found;
out:
Hugh Dickins's avatar
Hugh Dickins committed
772 773
	unlock_page(page);
	page_cache_release(page);
774
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
775 776 777 778 779 780 781 782 783 784
}

/*
 * 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;
785 786
	swp_entry_t swap;
	pgoff_t index;
Linus Torvalds's avatar
Linus Torvalds committed
787 788 789 790 791 792 793 794

	BUG_ON(!PageLocked(page));
	mapping = page->mapping;
	index = page->index;
	inode = mapping->host;
	info = SHMEM_I(inode);
	if (info->flags & VM_LOCKED)
		goto redirty;
795
	if (!total_swap_pages)
Linus Torvalds's avatar
Linus Torvalds committed
796 797
		goto redirty;

798 799 800
	/*
	 * shmem_backing_dev_info's capabilities prevent regular writeback or
	 * sync from ever calling shmem_writepage; but a stacking filesystem
801
	 * might use ->writepage of its underlying filesystem, in which case
802
	 * tmpfs should write out to swap only in response to memory pressure,
803
	 * and not for the writeback threads or sync.
804
	 */
805 806 807 808
	if (!wbc->for_reclaim) {
		WARN_ON_ONCE(1);	/* Still happens? Tell us about it! */
		goto redirty;
	}
809 810 811 812 813

	/*
	 * 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.
814 815 816 817 818 819
	 *
	 * 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.
820 821
	 */
	if (!PageUptodate(page)) {
822 823 824 825 826 827 828 829 830 831 832 833 834 835
		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;
		}
836 837 838 839 840
		clear_highpage(page);
		flush_dcache_page(page);
		SetPageUptodate(page);
	}

841 842 843
	swap = get_swap_page();
	if (!swap.val)
		goto redirty;
844

845 846
	/*
	 * Add inode to shmem_unuse()'s list of swapped-out inodes,
847 848
	 * if it's not already there.  Do it now before the page is
	 * moved to swap cache, when its pagelock no longer protects
849
	 * the inode from eviction.  But don't unlock the mutex until
850 851
	 * we've incremented swapped, because shmem_unuse_inode() will
	 * prune a !swapped inode from the swaplist under this mutex.
852
	 */
853 854 855
	mutex_lock(&shmem_swaplist_mutex);
	if (list_empty(&info->swaplist))
		list_add_tail(&info->swaplist, &shmem_swaplist);
856

857
	if (add_to_swap_cache(page, swap, GFP_ATOMIC) == 0) {
Hugh Dickins's avatar
Hugh Dickins committed
858
		swap_shmem_alloc(swap);
859 860 861 862 863
		shmem_delete_from_page_cache(page, swp_to_radix_entry(swap));

		spin_lock(&info->lock);
		info->swapped++;
		shmem_recalc_inode(inode);
864
		spin_unlock(&info->lock);
865 866

		mutex_unlock(&shmem_swaplist_mutex);
867
		BUG_ON(page_mapped(page));
868
		swap_writepage(page, wbc);
Linus Torvalds's avatar
Linus Torvalds committed
869 870 871
		return 0;
	}

872
	mutex_unlock(&shmem_swaplist_mutex);
873
	swapcache_free(swap, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
874 875
redirty:
	set_page_dirty(page);
876 877 878 879
	if (wbc->for_reclaim)
		return AOP_WRITEPAGE_ACTIVATE;	/* Return with page locked */
	unlock_page(page);
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
880 881 882
}

#ifdef CONFIG_NUMA
883
#ifdef CONFIG_TMPFS
884
static void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
885
{
886
	char buffer[64];
887

888
	if (!mpol || mpol->mode == MPOL_DEFAULT)
889
		return;		/* show nothing */
890

891
	mpol_to_str(buffer, sizeof(buffer), mpol);
892 893

	seq_printf(seq, ",mpol=%s", buffer);
894
}
895 896 897 898 899 900 901 902 903 904 905 906

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;
}
907 908
#endif /* CONFIG_TMPFS */

909 910
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
911 912
{
	struct vm_area_struct pvma;
Mel Gorman's avatar
Mel Gorman committed
913
	struct page *page;
914

Linus Torvalds's avatar
Linus Torvalds committed
915
	/* Create a pseudo vma that just contains the policy */
916
	pvma.vm_start = 0;
917 918
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
919
	pvma.vm_ops = NULL;
Mel Gorman's avatar
Mel Gorman committed
920 921 922 923 924 925 926 927
	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
928 929
}

930
static struct page *shmem_alloc_page(gfp_t gfp,
931
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
932 933
{
	struct vm_area_struct pvma;
Mel Gorman's avatar
Mel Gorman committed
934
	struct page *page;
Linus Torvalds's avatar
Linus Torvalds committed
935

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

Mel Gorman's avatar
Mel Gorman committed
943 944 945 946 947 948
	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
949
}
950 951
#else /* !CONFIG_NUMA */
#ifdef CONFIG_TMPFS
952
static inline void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
953 954 955 956
{
}
#endif /* CONFIG_TMPFS */

957 958
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
959
{
960
	return swapin_readahead(swap, gfp, NULL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
961 962
}

963
static inline struct page *shmem_alloc_page(gfp_t gfp,
964
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
965
{
966
	return alloc_page(gfp);
Linus Torvalds's avatar
Linus Torvalds committed
967
}
968
#endif /* CONFIG_NUMA */
Linus Torvalds's avatar
Linus Torvalds committed
969

970 971 972 973 974 975 976
#if !defined(CONFIG_NUMA) || !defined(CONFIG_TMPFS)
static inline struct mempolicy *shmem_get_sbmpol(struct shmem_sb_info *sbinfo)
{
	return NULL;
}
#endif

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 1012 1013 1014 1015 1016
/*
 * 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);
1017
	flush_dcache_page(newpage);
1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031

	__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);
1032 1033 1034 1035
	if (!error) {
		__inc_zone_page_state(newpage, NR_FILE_PAGES);
		__dec_zone_page_state(oldpage, NR_FILE_PAGES);
	}
1036 1037
	spin_unlock_irq(&swap_mapping->tree_lock);

Hugh Dickins's avatar