shmem.c 72.5 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
struct shmem_xattr {
	struct list_head list;	/* anchored by shmem_inode_info->xattr_list */
	char *name;		/* xattr name */
	size_t size;
	char value[0];
};

87
/* Flag allocation requirements to shmem_getpage */
Linus Torvalds's avatar
Linus Torvalds committed
88 89 90
enum sgp_type {
	SGP_READ,	/* don't exceed i_size, don't allocate page */
	SGP_CACHE,	/* don't exceed i_size, may allocate page */
91
	SGP_DIRTY,	/* like SGP_CACHE, but set new page dirty */
Linus Torvalds's avatar
Linus Torvalds committed
92 93 94
	SGP_WRITE,	/* may exceed i_size, may allocate page */
};

Andrew Morton's avatar
Andrew Morton committed
95
#ifdef CONFIG_TMPFS
96 97 98 99 100 101 102 103 104
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
105
#endif
106

107 108 109
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);
110 111 112 113 114 115 116 117 118
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
119 120 121 122 123 124 125 126 127 128 129 130 131 132

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)
{
133
	return (flags & VM_NORESERVE) ?
134
		0 : security_vm_enough_memory_mm(current->mm, VM_ACCT(size));
Linus Torvalds's avatar
Linus Torvalds committed
135 136 137 138
}

static inline void shmem_unacct_size(unsigned long flags, loff_t size)
{
139
	if (!(flags & VM_NORESERVE))
Linus Torvalds's avatar
Linus Torvalds committed
140 141 142 143 144 145 146 147 148 149 150
		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)
{
151
	return (flags & VM_NORESERVE) ?
152
		security_vm_enough_memory_mm(current->mm, VM_ACCT(PAGE_CACHE_SIZE)) : 0;
Linus Torvalds's avatar
Linus Torvalds committed
153 154 155 156
}

static inline void shmem_unacct_blocks(unsigned long flags, long pages)
{
157
	if (flags & VM_NORESERVE)
Linus Torvalds's avatar
Linus Torvalds committed
158 159 160
		vm_unacct_memory(pages * VM_ACCT(PAGE_CACHE_SIZE));
}

161
static const struct super_operations shmem_ops;
162
static const struct address_space_operations shmem_aops;
163
static const struct file_operations shmem_file_operations;
164 165 166
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;
167
static const struct vm_operations_struct shmem_vm_ops;
Linus Torvalds's avatar
Linus Torvalds committed
168

169
static struct backing_dev_info shmem_backing_dev_info  __read_mostly = {
Linus Torvalds's avatar
Linus Torvalds committed
170
	.ra_pages	= 0,	/* No readahead */
171
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK | BDI_CAP_SWAP_BACKED,
Linus Torvalds's avatar
Linus Torvalds committed
172 173 174
};

static LIST_HEAD(shmem_swaplist);
175
static DEFINE_MUTEX(shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
176

177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
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);
	}
}

202
/**
203
 * shmem_recalc_inode - recalculate the block usage of an inode
Linus Torvalds's avatar
Linus Torvalds committed
204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
 * @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) {
221 222 223
		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
224
		info->alloced -= freed;
225
		inode->i_blocks -= freed * BLOCKS_PER_PAGE;
Linus Torvalds's avatar
Linus Torvalds committed
226 227 228 229
		shmem_unacct_blocks(info->flags, freed);
	}
}

230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252
/*
 * 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;
}

253 254 255 256 257 258 259
/*
 * 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)
{
260
	int error = 0;
261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296

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

	if (!expected)
		error = radix_tree_preload(gfp & GFP_RECLAIM_MASK);
	if (!error) {
		page_cache_get(page);
		page->mapping = mapping;
		page->index = index;

		spin_lock_irq(&mapping->tree_lock);
		if (!expected)
			error = radix_tree_insert(&mapping->page_tree,
							index, page);
		else
			error = shmem_radix_tree_replace(mapping, index,
							expected, page);
		if (!error) {
			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);
		}
		if (!expected)
			radix_tree_preload_end();
	}
	if (error)
		mem_cgroup_uncharge_cache_page(page);
	return error;
}

297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315
/*
 * 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);
}

316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338
/*
 * 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)) {
339 340 341 342 343 344 345 346
			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;
347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385
		}
		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.
 */
386
static void shmem_deswap_pagevec(struct pagevec *pvec)
387 388 389 390 391 392 393 394 395
{
	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;
396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425
}

/*
 * 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();
	}
426 427 428 429 430
}

/*
 * Remove range of pages and swap entries from radix tree, and free them.
 */
431
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
Linus Torvalds's avatar
Linus Torvalds committed
432
{
433
	struct address_space *mapping = inode->i_mapping;
Linus Torvalds's avatar
Linus Torvalds committed
434
	struct shmem_inode_info *info = SHMEM_I(inode);
435
	pgoff_t start = (lstart + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
436 437 438
	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);
439
	struct pagevec pvec;
440 441
	pgoff_t indices[PAGEVEC_SIZE];
	long nr_swaps_freed = 0;
442
	pgoff_t index;
443 444
	int i;

445 446
	if (lend == -1)
		end = -1;	/* unsigned, so actually very big */
447 448 449

	pagevec_init(&pvec, 0);
	index = start;
450
	while (index < end) {
451
		pvec.nr = shmem_find_get_pages_and_swap(mapping, index,
452
				min(end - index, (pgoff_t)PAGEVEC_SIZE),
453 454 455
							pvec.pages, indices);
		if (!pvec.nr)
			break;
456 457 458 459
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

460
			index = indices[i];
461
			if (index >= end)
462 463
				break;

464 465 466
			if (radix_tree_exceptional_entry(page)) {
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
467
				continue;
468 469 470
			}

			if (!trylock_page(page))
471
				continue;
472 473 474
			if (page->mapping == mapping) {
				VM_BUG_ON(PageWriteback(page));
				truncate_inode_page(mapping, page);
475 476 477
			}
			unlock_page(page);
		}
478 479
		shmem_deswap_pagevec(&pvec);
		pagevec_release(&pvec);
480 481 482 483
		mem_cgroup_uncharge_end();
		cond_resched();
		index++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
484

485
	if (partial_start) {
486 487 488
		struct page *page = NULL;
		shmem_getpage(inode, start - 1, &page, SGP_READ, NULL);
		if (page) {
489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504
			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);
505 506 507 508 509
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}
510 511
	if (start >= end)
		return;
512 513 514 515

	index = start;
	for ( ; ; ) {
		cond_resched();
516
		pvec.nr = shmem_find_get_pages_and_swap(mapping, index,
517
				min(end - index, (pgoff_t)PAGEVEC_SIZE),
518 519
							pvec.pages, indices);
		if (!pvec.nr) {
520 521 522 523 524
			if (index == start)
				break;
			index = start;
			continue;
		}
525
		if (index == start && indices[0] >= end) {
526 527
			shmem_deswap_pagevec(&pvec);
			pagevec_release(&pvec);
528 529 530 531 532 533
			break;
		}
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

534
			index = indices[i];
535
			if (index >= end)
536 537
				break;

538 539 540 541 542 543
			if (radix_tree_exceptional_entry(page)) {
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
				continue;
			}

544
			lock_page(page);
545 546 547 548
			if (page->mapping == mapping) {
				VM_BUG_ON(PageWriteback(page));
				truncate_inode_page(mapping, page);
			}
549 550
			unlock_page(page);
		}
551 552
		shmem_deswap_pagevec(&pvec);
		pagevec_release(&pvec);
553 554 555
		mem_cgroup_uncharge_end();
		index++;
	}
556

Linus Torvalds's avatar
Linus Torvalds committed
557
	spin_lock(&info->lock);
558
	info->swapped -= nr_swaps_freed;
Linus Torvalds's avatar
Linus Torvalds committed
559 560 561
	shmem_recalc_inode(inode);
	spin_unlock(&info->lock);

562
	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
Linus Torvalds's avatar
Linus Torvalds committed
563
}
564
EXPORT_SYMBOL_GPL(shmem_truncate_range);
Linus Torvalds's avatar
Linus Torvalds committed
565

566
static int shmem_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds's avatar
Linus Torvalds committed
567 568 569 570
{
	struct inode *inode = dentry->d_inode;
	int error;

571 572 573 574
	error = inode_change_ok(inode, attr);
	if (error)
		return error;

575 576 577
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
		loff_t oldsize = inode->i_size;
		loff_t newsize = attr->ia_size;
578

579 580 581 582 583 584 585 586 587 588 589
		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
590 591
	}

592
	setattr_copy(inode, attr);
593
#ifdef CONFIG_TMPFS_POSIX_ACL
594
	if (attr->ia_valid & ATTR_MODE)
595
		error = generic_acl_chmod(inode);
596
#endif
Linus Torvalds's avatar
Linus Torvalds committed
597 598 599
	return error;
}

Al Viro's avatar
Al Viro committed
600
static void shmem_evict_inode(struct inode *inode)
Linus Torvalds's avatar
Linus Torvalds committed
601 602
{
	struct shmem_inode_info *info = SHMEM_I(inode);
603
	struct shmem_xattr *xattr, *nxattr;
Linus Torvalds's avatar
Linus Torvalds committed
604

605
	if (inode->i_mapping->a_ops == &shmem_aops) {
Linus Torvalds's avatar
Linus Torvalds committed
606 607
		shmem_unacct_size(info->flags, inode->i_size);
		inode->i_size = 0;
608
		shmem_truncate_range(inode, 0, (loff_t)-1);
Linus Torvalds's avatar
Linus Torvalds committed
609
		if (!list_empty(&info->swaplist)) {
610
			mutex_lock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
611
			list_del_init(&info->swaplist);
612
			mutex_unlock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
613
		}
614 615
	} else
		kfree(info->symlink);
616 617 618 619 620

	list_for_each_entry_safe(xattr, nxattr, &info->xattr_list, list) {
		kfree(xattr->name);
		kfree(xattr);
	}
621
	BUG_ON(inode->i_blocks);
622
	shmem_free_inode(inode->i_sb);
623
	clear_inode(inode);
Linus Torvalds's avatar
Linus Torvalds committed
624 625
}

626 627 628
/*
 * If swap found in inode, free it and move page from swapcache to filecache.
 */
629
static int shmem_unuse_inode(struct shmem_inode_info *info,
630
			     swp_entry_t swap, struct page **pagep)
Linus Torvalds's avatar
Linus Torvalds committed
631
{
632
	struct address_space *mapping = info->vfs_inode.i_mapping;
633
	void *radswap;
634
	pgoff_t index;
635 636
	gfp_t gfp;
	int error = 0;
Linus Torvalds's avatar
Linus Torvalds committed
637

638
	radswap = swp_to_radix_entry(swap);
639
	index = radix_tree_locate_item(&mapping->page_tree, radswap);
640
	if (index == -1)
641
		return 0;
642

Hugh Dickins's avatar
Hugh Dickins committed
643 644
	/*
	 * Move _head_ to start search for next from here.
Al Viro's avatar
Al Viro committed
645
	 * But be careful: shmem_evict_inode checks list_empty without taking
Hugh Dickins's avatar
Hugh Dickins committed
646
	 * mutex, and there's an instant in list_move_tail when info->swaplist
647
	 * would appear empty, if it were the only one on shmem_swaplist.
Hugh Dickins's avatar
Hugh Dickins committed
648 649 650
	 */
	if (shmem_swaplist.next != &info->swaplist)
		list_move_tail(&shmem_swaplist, &info->swaplist);
651

652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667
	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
		 * allocation; but the inode might already be freed by now,
		 * and we cannot refer to inode or mapping or info to check.
		 * However, we do hold page lock on the PageSwapCache page,
		 * so can check if that still has our reference remaining.
		 */
		if (!page_swapcount(*pagep))
			error = -ENOENT;
	}

KAMEZAWA Hiroyuki's avatar
KAMEZAWA Hiroyuki committed
668
	/*
669 670 671
	 * 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
672
	 */
673 674
	if (!error)
		error = shmem_add_to_page_cache(*pagep, mapping, index,
675
						GFP_NOWAIT, radswap);
676
	if (error != -ENOMEM) {
677 678 679 680
		/*
		 * Truncation and eviction use free_swap_and_cache(), which
		 * only does trylock page: if we raced, best clean up here.
		 */
681 682
		delete_from_swap_cache(*pagep);
		set_page_dirty(*pagep);
683 684 685 686 687 688
		if (!error) {
			spin_lock(&info->lock);
			info->swapped--;
			spin_unlock(&info->lock);
			swap_free(swap);
		}
689
		error = 1;	/* not an error, but entry was found */
Linus Torvalds's avatar
Linus Torvalds committed
690
	}
691
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
692 693 694
}

/*
695
 * Search through swapped inodes to find and replace swap by page.
Linus Torvalds's avatar
Linus Torvalds committed
696
 */
697
int shmem_unuse(swp_entry_t swap, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
698
{
699
	struct list_head *this, *next;
Linus Torvalds's avatar
Linus Torvalds committed
700 701
	struct shmem_inode_info *info;
	int found = 0;
702 703 704 705 706 707 708 709
	int error = 0;

	/*
	 * There's a faint possibility that swap page was replaced before
	 * caller locked it: it will come back later with the right page.
	 */
	if (unlikely(!PageSwapCache(page)))
		goto out;
710 711 712 713 714 715 716 717 718

	/*
	 * 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;
719
	/* No radix_tree_preload: swap entry keeps a place for page in tree */
Linus Torvalds's avatar
Linus Torvalds committed
720

721
	mutex_lock(&shmem_swaplist_mutex);
722 723
	list_for_each_safe(this, next, &shmem_swaplist) {
		info = list_entry(this, struct shmem_inode_info, swaplist);
724
		if (info->swapped)
725
			found = shmem_unuse_inode(info, swap, &page);
726 727
		else
			list_del_init(&info->swaplist);
728
		cond_resched();
729
		if (found)
730
			break;
Linus Torvalds's avatar
Linus Torvalds committed
731
	}
732
	mutex_unlock(&shmem_swaplist_mutex);
733 734 735 736

	if (found < 0)
		error = found;
out:
Hugh Dickins's avatar
Hugh Dickins committed
737 738
	unlock_page(page);
	page_cache_release(page);
739
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
740 741 742 743 744 745 746 747 748 749
}

/*
 * 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;
750 751
	swp_entry_t swap;
	pgoff_t index;
Linus Torvalds's avatar
Linus Torvalds committed
752 753 754 755 756 757 758 759

	BUG_ON(!PageLocked(page));
	mapping = page->mapping;
	index = page->index;
	inode = mapping->host;
	info = SHMEM_I(inode);
	if (info->flags & VM_LOCKED)
		goto redirty;
760
	if (!total_swap_pages)
Linus Torvalds's avatar
Linus Torvalds committed
761 762
		goto redirty;

763 764 765
	/*
	 * shmem_backing_dev_info's capabilities prevent regular writeback or
	 * sync from ever calling shmem_writepage; but a stacking filesystem
766
	 * might use ->writepage of its underlying filesystem, in which case
767
	 * tmpfs should write out to swap only in response to memory pressure,
768
	 * and not for the writeback threads or sync.
769
	 */
770 771 772 773 774 775 776
	if (!wbc->for_reclaim) {
		WARN_ON_ONCE(1);	/* Still happens? Tell us about it! */
		goto redirty;
	}
	swap = get_swap_page();
	if (!swap.val)
		goto redirty;
777

778 779
	/*
	 * Add inode to shmem_unuse()'s list of swapped-out inodes,
780 781
	 * if it's not already there.  Do it now before the page is
	 * moved to swap cache, when its pagelock no longer protects
782
	 * the inode from eviction.  But don't unlock the mutex until
783 784
	 * we've incremented swapped, because shmem_unuse_inode() will
	 * prune a !swapped inode from the swaplist under this mutex.
785
	 */
786 787 788
	mutex_lock(&shmem_swaplist_mutex);
	if (list_empty(&info->swaplist))
		list_add_tail(&info->swaplist, &shmem_swaplist);
789

790
	if (add_to_swap_cache(page, swap, GFP_ATOMIC) == 0) {
Hugh Dickins's avatar
Hugh Dickins committed
791
		swap_shmem_alloc(swap);
792 793 794 795 796
		shmem_delete_from_page_cache(page, swp_to_radix_entry(swap));

		spin_lock(&info->lock);
		info->swapped++;
		shmem_recalc_inode(inode);
797
		spin_unlock(&info->lock);
798 799

		mutex_unlock(&shmem_swaplist_mutex);
800
		BUG_ON(page_mapped(page));
801
		swap_writepage(page, wbc);
Linus Torvalds's avatar
Linus Torvalds committed
802 803 804
		return 0;
	}

805
	mutex_unlock(&shmem_swaplist_mutex);
806
	swapcache_free(swap, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
807 808
redirty:
	set_page_dirty(page);
809 810 811 812
	if (wbc->for_reclaim)
		return AOP_WRITEPAGE_ACTIVATE;	/* Return with page locked */
	unlock_page(page);
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
813 814 815
}

#ifdef CONFIG_NUMA
816
#ifdef CONFIG_TMPFS
817
static void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
818
{
819
	char buffer[64];
820

821
	if (!mpol || mpol->mode == MPOL_DEFAULT)
822
		return;		/* show nothing */
823

824
	mpol_to_str(buffer, sizeof(buffer), mpol, 1);
825 826

	seq_printf(seq, ",mpol=%s", buffer);
827
}
828 829 830 831 832 833 834 835 836 837 838 839

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;
}
840 841
#endif /* CONFIG_TMPFS */

842 843
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
844
{
845
	struct mempolicy mpol, *spol;
Linus Torvalds's avatar
Linus Torvalds committed
846 847
	struct vm_area_struct pvma;

848
	spol = mpol_cond_copy(&mpol,
849
			mpol_shared_policy_lookup(&info->policy, index));
850

Linus Torvalds's avatar
Linus Torvalds committed
851
	/* Create a pseudo vma that just contains the policy */
852
	pvma.vm_start = 0;
853
	pvma.vm_pgoff = index;
854
	pvma.vm_ops = NULL;
855
	pvma.vm_policy = spol;
856
	return swapin_readahead(swap, gfp, &pvma, 0);
Linus Torvalds's avatar
Linus Torvalds committed
857 858
}

859
static struct page *shmem_alloc_page(gfp_t gfp,
860
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
861 862 863
{
	struct vm_area_struct pvma;

864 865
	/* Create a pseudo vma that just contains the policy */
	pvma.vm_start = 0;
866
	pvma.vm_pgoff = index;
867
	pvma.vm_ops = NULL;
868
	pvma.vm_policy = mpol_shared_policy_lookup(&info->policy, index);
869 870 871 872 873

	/*
	 * alloc_page_vma() will drop the shared policy reference
	 */
	return alloc_page_vma(gfp, &pvma, 0);
Linus Torvalds's avatar
Linus Torvalds committed
874
}
875 876
#else /* !CONFIG_NUMA */
#ifdef CONFIG_TMPFS
877
static inline void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
878 879 880 881
{
}
#endif /* CONFIG_TMPFS */

882 883
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
884
{
885
	return swapin_readahead(swap, gfp, NULL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
886 887
}

888
static inline struct page *shmem_alloc_page(gfp_t gfp,
889
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
890
{
891
	return alloc_page(gfp);
Linus Torvalds's avatar
Linus Torvalds committed
892
}
893
#endif /* CONFIG_NUMA */
Linus Torvalds's avatar
Linus Torvalds committed
894

895 896 897 898 899 900 901
#if !defined(CONFIG_NUMA) || !defined(CONFIG_TMPFS)
static inline struct mempolicy *shmem_get_sbmpol(struct shmem_sb_info *sbinfo)
{
	return NULL;
}
#endif

902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979
/*
 * 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;
	VM_BUG_ON(shmem_should_replace_page(newpage, gfp));

	*pagep = newpage;
	page_cache_get(newpage);
	copy_highpage(newpage, oldpage);

	VM_BUG_ON(!PageLocked(oldpage));
	__set_page_locked(newpage);
	VM_BUG_ON(!PageUptodate(oldpage));
	SetPageUptodate(newpage);
	VM_BUG_ON(!PageSwapBacked(oldpage));
	SetPageSwapBacked(newpage);
	VM_BUG_ON(!swap_index);
	set_page_private(newpage, swap_index);
	VM_BUG_ON(!PageSwapCache(oldpage));
	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);
	__inc_zone_page_state(newpage, NR_FILE_PAGES);
	__dec_zone_page_state(oldpage, NR_FILE_PAGES);
	spin_unlock_irq(&swap_mapping->tree_lock);
	BUG_ON(error);

	mem_cgroup_replace_page_cache(oldpage, newpage);
	lru_cache_add_anon(newpage);

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

	unlock_page(oldpage);
	page_cache_release(oldpage);
	page_cache_release(oldpage);
	return 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
980
/*
981
 * shmem_getpage_gfp - find page in cache, or get from swap, or allocate
Linus Torvalds's avatar
Linus Torvalds committed
982 983 984 985 986
 *
 * 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
 */
987
static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
988
	struct page **pagep, enum sgp_type sgp, gfp_t gfp, int *fault_type)
Linus Torvalds's avatar
Linus Torvalds committed
989 990
{
	struct address_space *mapping = inode->i_mapping;
991
	struct shmem_inode_info *info;
Linus Torvalds's avatar
Linus Torvalds committed
992
	struct shmem_sb_info *sbinfo;
993
	struct page *page;
Linus Torvalds's avatar
Linus Torvalds committed
994 995
	swp_entry_t swap;
	int error;
996
	int once = 0;
Linus Torvalds's avatar
Linus Torvalds committed
997

998
	if (index > (MAX_LFS_FILESIZE >> PAGE_CACHE_SHIFT))
Linus Torvalds's avatar
Linus Torvalds committed
999 1000
		return -EFBIG;
repeat:
1001
	swap.val = 0;
1002
	page = find_lock_page(mapping, index);
1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014
	if (radix_tree_exceptional_entry(page)) {
		swap = radix_to_swp_entry(page);
		page = NULL;
	}

	if (sgp != SGP_WRITE &&
	    ((loff_t)index << PAGE_CACHE_SHIFT) >= i_size_read(inode)) {
		error = -EINVAL;
		goto failed;
	}

	if (page || (sgp == SGP_READ && !swap.val)) {
Hugh Dickins's avatar
Hugh Dickins committed
1015
		/*
1016 1017 1018
		 * Once we can get the page lock, it must be uptodate:
		 * if there were an error in reading back from swap,
		 * the page would not be inserted into the filecache.
Hugh Dickins's avatar
Hugh Dickins committed
1019
		 */
1020 1021 1022
		BUG_ON(page && !PageUptodate(page));
		*pagep = page;
		return 0;
1023 1024 1025
	}

	/*
1026 1027
	 * Fast cache lookup did not find it:
	 * bring it back from swap or allocate.
1028
	 */
1029 1030
	info = SHMEM_I(inode);
	sbinfo = SHMEM_SB(inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
1031 1032 1033

	if (swap.val) {
		/* Look it up and read it in.. */
1034 1035
		page = lookup_swap_cache(swap);
		if (!page) {
Linus Torvalds's avatar
Linus Torvalds committed
1036
			/* here we actually do the io */
1037 1038
			if (fault_type)
				*fault_type |= VM_FAULT_MAJOR;
1039
			page = shmem_swapin(swap, gfp, info, index);
1040
			if (!page) {
1041 1042
				error = -ENOMEM;
				goto failed;
Linus Torvalds's avatar
Linus Torvalds committed
1043 1044 1045 1046
			}
		}

		/* We have to do this with page locked to prevent races */
1047
		lock_page(page);
1048 1049 1050 1051
		if (!PageSwapCache(page) || page->mapping) {
			error = -EEXIST;	/* try again */
			goto failed;
		}
1052
		if (!PageUptodate(page)) {
Linus Torvalds's avatar
Linus Torvalds committed
1053
			error = -EIO;
1054
			goto failed;
Linus Torvalds's avatar
Linus Torvalds committed
1055
		}
1056 1057
		wait_on_page_writeback(page);

1058 1059 1060 1061
		if (shmem_should_replace_page(page, gfp)) {
			error = shmem_replace_page(&page, gfp, info, index);
			if (error)
				goto failed;
Linus Torvalds's avatar
Linus Torvalds committed
1062
		}
1063

1064 1065 1066 1067 1068
		error = mem_cgroup_cache_charge(page, current->mm,