shmem.c 63.5 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/splice.h>
Linus Torvalds's avatar
Linus Torvalds committed
57 58 59 60
#include <linux/security.h>
#include <linux/swapops.h>
#include <linux/mempolicy.h>
#include <linux/namei.h>
61
#include <linux/ctype.h>
62
#include <linux/migrate.h>
63
#include <linux/highmem.h>
64
#include <linux/seq_file.h>
Mimi Zohar's avatar
Mimi Zohar committed
65
#include <linux/magic.h>
66

Linus Torvalds's avatar
Linus Torvalds committed
67 68 69
#include <asm/uaccess.h>
#include <asm/pgtable.h>

Hugh Dickins's avatar
Hugh Dickins committed
70
#define BLOCKS_PER_PAGE  (PAGE_CACHE_SIZE/512)
Linus Torvalds's avatar
Linus Torvalds committed
71 72 73 74 75
#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

76 77 78
/* Symlink up to this size is kmalloc'ed instead of using a swappable page */
#define SHORT_SYMLINK_LEN 128

79 80 81 82 83 84 85
struct shmem_xattr {
	struct list_head list;	/* anchored by shmem_inode_info->xattr_list */
	char *name;		/* xattr name */
	size_t size;
	char value[0];
};

86
/* Flag allocation requirements to shmem_getpage */
Linus Torvalds's avatar
Linus Torvalds committed
87 88 89
enum sgp_type {
	SGP_READ,	/* don't exceed i_size, don't allocate page */
	SGP_CACHE,	/* don't exceed i_size, may allocate page */
90
	SGP_DIRTY,	/* like SGP_CACHE, but set new page dirty */
Linus Torvalds's avatar
Linus Torvalds committed
91 92 93
	SGP_WRITE,	/* may exceed i_size, may allocate page */
};

Andrew Morton's avatar
Andrew Morton committed
94
#ifdef CONFIG_TMPFS
95 96 97 98 99 100 101 102 103
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
104
#endif
105

106 107 108 109 110 111 112 113 114
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
115 116 117 118 119 120 121 122 123 124 125 126 127 128

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)
{
129 130
	return (flags & VM_NORESERVE) ?
		0 : security_vm_enough_memory_kern(VM_ACCT(size));
Linus Torvalds's avatar
Linus Torvalds committed
131 132 133 134
}

static inline void shmem_unacct_size(unsigned long flags, loff_t size)
{
135
	if (!(flags & VM_NORESERVE))
Linus Torvalds's avatar
Linus Torvalds committed
136 137 138 139 140 141 142 143 144 145 146
		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)
{
147 148
	return (flags & VM_NORESERVE) ?
		security_vm_enough_memory_kern(VM_ACCT(PAGE_CACHE_SIZE)) : 0;
Linus Torvalds's avatar
Linus Torvalds committed
149 150 151 152
}

static inline void shmem_unacct_blocks(unsigned long flags, long pages)
{
153
	if (flags & VM_NORESERVE)
Linus Torvalds's avatar
Linus Torvalds committed
154 155 156
		vm_unacct_memory(pages * VM_ACCT(PAGE_CACHE_SIZE));
}

157
static const struct super_operations shmem_ops;
158
static const struct address_space_operations shmem_aops;
159
static const struct file_operations shmem_file_operations;
160 161 162
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;
163
static const struct vm_operations_struct shmem_vm_ops;
Linus Torvalds's avatar
Linus Torvalds committed
164

165
static struct backing_dev_info shmem_backing_dev_info  __read_mostly = {
Linus Torvalds's avatar
Linus Torvalds committed
166
	.ra_pages	= 0,	/* No readahead */
167
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK | BDI_CAP_SWAP_BACKED,
Linus Torvalds's avatar
Linus Torvalds committed
168 169 170
};

static LIST_HEAD(shmem_swaplist);
171
static DEFINE_MUTEX(shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
172

173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197
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);
	}
}

198
/**
199
 * shmem_recalc_inode - recalculate the block usage of an inode
Linus Torvalds's avatar
Linus Torvalds committed
200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216
 * @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) {
217 218 219
		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
220
		info->alloced -= freed;
221
		inode->i_blocks -= freed * BLOCKS_PER_PAGE;
Linus Torvalds's avatar
Linus Torvalds committed
222 223 224 225
		shmem_unacct_blocks(info->flags, freed);
	}
}

226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248
/*
 * 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;
}

249 250 251 252 253 254 255
/*
 * 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)
{
256
	int error = 0;
257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292

	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;
}

293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311
/*
 * 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);
}

312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334
/*
 * 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)) {
335 336 337 338 339 340 341 342
			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;
343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397
		}
		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.
 */
static void shmem_pagevec_release(struct pagevec *pvec)
{
	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;
	pagevec_release(pvec);
}

/*
 * Remove range of pages and swap entries from radix tree, and free them.
 */
398
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
Linus Torvalds's avatar
Linus Torvalds committed
399
{
400
	struct address_space *mapping = inode->i_mapping;
Linus Torvalds's avatar
Linus Torvalds committed
401
	struct shmem_inode_info *info = SHMEM_I(inode);
402
	pgoff_t start = (lstart + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
403
	unsigned partial = lstart & (PAGE_CACHE_SIZE - 1);
404
	pgoff_t end = (lend >> PAGE_CACHE_SHIFT);
405
	struct pagevec pvec;
406 407
	pgoff_t indices[PAGEVEC_SIZE];
	long nr_swaps_freed = 0;
408
	pgoff_t index;
409 410 411 412 413 414
	int i;

	BUG_ON((lend & (PAGE_CACHE_SIZE - 1)) != (PAGE_CACHE_SIZE - 1));

	pagevec_init(&pvec, 0);
	index = start;
415 416 417 418 419 420
	while (index <= end) {
		pvec.nr = shmem_find_get_pages_and_swap(mapping, index,
			min(end - index, (pgoff_t)PAGEVEC_SIZE - 1) + 1,
							pvec.pages, indices);
		if (!pvec.nr)
			break;
421 422 423 424
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

425
			index = indices[i];
426 427 428
			if (index > end)
				break;

429 430 431
			if (radix_tree_exceptional_entry(page)) {
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
432
				continue;
433 434 435
			}

			if (!trylock_page(page))
436
				continue;
437 438 439
			if (page->mapping == mapping) {
				VM_BUG_ON(PageWriteback(page));
				truncate_inode_page(mapping, page);
440 441 442
			}
			unlock_page(page);
		}
443
		shmem_pagevec_release(&pvec);
444 445 446 447
		mem_cgroup_uncharge_end();
		cond_resched();
		index++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
448

449 450 451 452 453 454 455 456 457 458 459 460 461 462
	if (partial) {
		struct page *page = NULL;
		shmem_getpage(inode, start - 1, &page, SGP_READ, NULL);
		if (page) {
			zero_user_segment(page, partial, PAGE_CACHE_SIZE);
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}

	index = start;
	for ( ; ; ) {
		cond_resched();
463 464 465 466
		pvec.nr = shmem_find_get_pages_and_swap(mapping, index,
			min(end - index, (pgoff_t)PAGEVEC_SIZE - 1) + 1,
							pvec.pages, indices);
		if (!pvec.nr) {
467 468 469 470 471
			if (index == start)
				break;
			index = start;
			continue;
		}
472 473
		if (index == start && indices[0] > end) {
			shmem_pagevec_release(&pvec);
474 475 476 477 478 479
			break;
		}
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

480
			index = indices[i];
481 482 483
			if (index > end)
				break;

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

490
			lock_page(page);
491 492 493 494
			if (page->mapping == mapping) {
				VM_BUG_ON(PageWriteback(page));
				truncate_inode_page(mapping, page);
			}
495 496
			unlock_page(page);
		}
497
		shmem_pagevec_release(&pvec);
498 499 500
		mem_cgroup_uncharge_end();
		index++;
	}
501

Linus Torvalds's avatar
Linus Torvalds committed
502
	spin_lock(&info->lock);
503
	info->swapped -= nr_swaps_freed;
Linus Torvalds's avatar
Linus Torvalds committed
504 505 506
	shmem_recalc_inode(inode);
	spin_unlock(&info->lock);

507
	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
Linus Torvalds's avatar
Linus Torvalds committed
508
}
509
EXPORT_SYMBOL_GPL(shmem_truncate_range);
Linus Torvalds's avatar
Linus Torvalds committed
510

511
static int shmem_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds's avatar
Linus Torvalds committed
512 513 514 515
{
	struct inode *inode = dentry->d_inode;
	int error;

516 517 518 519
	error = inode_change_ok(inode, attr);
	if (error)
		return error;

520 521 522
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
		loff_t oldsize = inode->i_size;
		loff_t newsize = attr->ia_size;
523

524 525 526 527 528 529 530 531 532 533 534
		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
535 536
	}

537
	setattr_copy(inode, attr);
538
#ifdef CONFIG_TMPFS_POSIX_ACL
539
	if (attr->ia_valid & ATTR_MODE)
540
		error = generic_acl_chmod(inode);
541
#endif
Linus Torvalds's avatar
Linus Torvalds committed
542 543 544
	return error;
}

Al Viro's avatar
Al Viro committed
545
static void shmem_evict_inode(struct inode *inode)
Linus Torvalds's avatar
Linus Torvalds committed
546 547
{
	struct shmem_inode_info *info = SHMEM_I(inode);
548
	struct shmem_xattr *xattr, *nxattr;
Linus Torvalds's avatar
Linus Torvalds committed
549

550
	if (inode->i_mapping->a_ops == &shmem_aops) {
Linus Torvalds's avatar
Linus Torvalds committed
551 552
		shmem_unacct_size(info->flags, inode->i_size);
		inode->i_size = 0;
553
		shmem_truncate_range(inode, 0, (loff_t)-1);
Linus Torvalds's avatar
Linus Torvalds committed
554
		if (!list_empty(&info->swaplist)) {
555
			mutex_lock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
556
			list_del_init(&info->swaplist);
557
			mutex_unlock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
558
		}
559 560
	} else
		kfree(info->symlink);
561 562 563 564 565

	list_for_each_entry_safe(xattr, nxattr, &info->xattr_list, list) {
		kfree(xattr->name);
		kfree(xattr);
	}
566
	BUG_ON(inode->i_blocks);
567
	shmem_free_inode(inode->i_sb);
Al Viro's avatar
Al Viro committed
568
	end_writeback(inode);
Linus Torvalds's avatar
Linus Torvalds committed
569 570
}

571 572 573
/*
 * If swap found in inode, free it and move page from swapcache to filecache.
 */
574 575
static int shmem_unuse_inode(struct shmem_inode_info *info,
			     swp_entry_t swap, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
576
{
577
	struct address_space *mapping = info->vfs_inode.i_mapping;
578
	void *radswap;
579
	pgoff_t index;
580
	int error;
Linus Torvalds's avatar
Linus Torvalds committed
581

582
	radswap = swp_to_radix_entry(swap);
583
	index = radix_tree_locate_item(&mapping->page_tree, radswap);
584
	if (index == -1)
585
		return 0;
586

Hugh Dickins's avatar
Hugh Dickins committed
587 588
	/*
	 * Move _head_ to start search for next from here.
Al Viro's avatar
Al Viro committed
589
	 * But be careful: shmem_evict_inode checks list_empty without taking
Hugh Dickins's avatar
Hugh Dickins committed
590
	 * mutex, and there's an instant in list_move_tail when info->swaplist
591
	 * would appear empty, if it were the only one on shmem_swaplist.
Hugh Dickins's avatar
Hugh Dickins committed
592 593 594
	 */
	if (shmem_swaplist.next != &info->swaplist)
		list_move_tail(&shmem_swaplist, &info->swaplist);
595

KAMEZAWA Hiroyuki's avatar
KAMEZAWA Hiroyuki committed
596
	/*
597 598 599
	 * 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
600
	 */
601 602
	error = shmem_add_to_page_cache(page, mapping, index,
						GFP_NOWAIT, radswap);
603
	/* which does mem_cgroup_uncharge_cache_page on error */
604

605
	if (error != -ENOMEM) {
606 607 608 609
		/*
		 * Truncation and eviction use free_swap_and_cache(), which
		 * only does trylock page: if we raced, best clean up here.
		 */
610 611
		delete_from_swap_cache(page);
		set_page_dirty(page);
612 613 614 615 616 617
		if (!error) {
			spin_lock(&info->lock);
			info->swapped--;
			spin_unlock(&info->lock);
			swap_free(swap);
		}
618
		error = 1;	/* not an error, but entry was found */
Linus Torvalds's avatar
Linus Torvalds committed
619
	}
620
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
621 622 623
}

/*
624
 * Search through swapped inodes to find and replace swap by page.
Linus Torvalds's avatar
Linus Torvalds committed
625
 */
626
int shmem_unuse(swp_entry_t swap, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
627
{
628
	struct list_head *this, *next;
Linus Torvalds's avatar
Linus Torvalds committed
629 630
	struct shmem_inode_info *info;
	int found = 0;
631 632 633 634 635 636 637 638 639 640
	int error;

	/*
	 * 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;
641
	/* No radix_tree_preload: swap entry keeps a place for page in tree */
Linus Torvalds's avatar
Linus Torvalds committed
642

643
	mutex_lock(&shmem_swaplist_mutex);
644 645
	list_for_each_safe(this, next, &shmem_swaplist) {
		info = list_entry(this, struct shmem_inode_info, swaplist);
646
		if (info->swapped)
647
			found = shmem_unuse_inode(info, swap, page);
648 649
		else
			list_del_init(&info->swaplist);
650
		cond_resched();
651
		if (found)
652
			break;
Linus Torvalds's avatar
Linus Torvalds committed
653
	}
654
	mutex_unlock(&shmem_swaplist_mutex);
655 656 657 658 659 660

	if (!found)
		mem_cgroup_uncharge_cache_page(page);
	if (found < 0)
		error = found;
out:
Hugh Dickins's avatar
Hugh Dickins committed
661 662
	unlock_page(page);
	page_cache_release(page);
663
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
664 665 666 667 668 669 670 671 672 673
}

/*
 * 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;
674 675
	swp_entry_t swap;
	pgoff_t index;
Linus Torvalds's avatar
Linus Torvalds committed
676 677 678 679 680 681 682 683

	BUG_ON(!PageLocked(page));
	mapping = page->mapping;
	index = page->index;
	inode = mapping->host;
	info = SHMEM_I(inode);
	if (info->flags & VM_LOCKED)
		goto redirty;
684
	if (!total_swap_pages)
Linus Torvalds's avatar
Linus Torvalds committed
685 686
		goto redirty;

687 688 689
	/*
	 * shmem_backing_dev_info's capabilities prevent regular writeback or
	 * sync from ever calling shmem_writepage; but a stacking filesystem
690
	 * might use ->writepage of its underlying filesystem, in which case
691
	 * tmpfs should write out to swap only in response to memory pressure,
692
	 * and not for the writeback threads or sync.
693
	 */
694 695 696 697 698 699 700
	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;
701

702 703
	/*
	 * Add inode to shmem_unuse()'s list of swapped-out inodes,
704 705
	 * if it's not already there.  Do it now before the page is
	 * moved to swap cache, when its pagelock no longer protects
706
	 * the inode from eviction.  But don't unlock the mutex until
707 708
	 * we've incremented swapped, because shmem_unuse_inode() will
	 * prune a !swapped inode from the swaplist under this mutex.
709
	 */
710 711 712
	mutex_lock(&shmem_swaplist_mutex);
	if (list_empty(&info->swaplist))
		list_add_tail(&info->swaplist, &shmem_swaplist);
713

714
	if (add_to_swap_cache(page, swap, GFP_ATOMIC) == 0) {
Hugh Dickins's avatar
Hugh Dickins committed
715
		swap_shmem_alloc(swap);
716 717 718 719 720
		shmem_delete_from_page_cache(page, swp_to_radix_entry(swap));

		spin_lock(&info->lock);
		info->swapped++;
		shmem_recalc_inode(inode);
721
		spin_unlock(&info->lock);
722 723

		mutex_unlock(&shmem_swaplist_mutex);
724
		BUG_ON(page_mapped(page));
725
		swap_writepage(page, wbc);
Linus Torvalds's avatar
Linus Torvalds committed
726 727 728
		return 0;
	}

729
	mutex_unlock(&shmem_swaplist_mutex);
730
	swapcache_free(swap, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
731 732
redirty:
	set_page_dirty(page);
733 734 735 736
	if (wbc->for_reclaim)
		return AOP_WRITEPAGE_ACTIVATE;	/* Return with page locked */
	unlock_page(page);
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
737 738 739
}

#ifdef CONFIG_NUMA
740
#ifdef CONFIG_TMPFS
741
static void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
742
{
743
	char buffer[64];
744

745
	if (!mpol || mpol->mode == MPOL_DEFAULT)
746
		return;		/* show nothing */
747

748
	mpol_to_str(buffer, sizeof(buffer), mpol, 1);
749 750

	seq_printf(seq, ",mpol=%s", buffer);
751
}
752 753 754 755 756 757 758 759 760 761 762 763

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;
}
764 765
#endif /* CONFIG_TMPFS */

766 767
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
768
{
769
	struct mempolicy mpol, *spol;
Linus Torvalds's avatar
Linus Torvalds committed
770 771
	struct vm_area_struct pvma;

772
	spol = mpol_cond_copy(&mpol,
773
			mpol_shared_policy_lookup(&info->policy, index));
774

Linus Torvalds's avatar
Linus Torvalds committed
775
	/* Create a pseudo vma that just contains the policy */
776
	pvma.vm_start = 0;
777
	pvma.vm_pgoff = index;
778
	pvma.vm_ops = NULL;
779
	pvma.vm_policy = spol;
780
	return swapin_readahead(swap, gfp, &pvma, 0);
Linus Torvalds's avatar
Linus Torvalds committed
781 782
}

783
static struct page *shmem_alloc_page(gfp_t gfp,
784
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
785 786 787
{
	struct vm_area_struct pvma;

788 789
	/* Create a pseudo vma that just contains the policy */
	pvma.vm_start = 0;
790
	pvma.vm_pgoff = index;
791
	pvma.vm_ops = NULL;
792
	pvma.vm_policy = mpol_shared_policy_lookup(&info->policy, index);
793 794 795 796 797

	/*
	 * alloc_page_vma() will drop the shared policy reference
	 */
	return alloc_page_vma(gfp, &pvma, 0);
Linus Torvalds's avatar
Linus Torvalds committed
798
}
799 800
#else /* !CONFIG_NUMA */
#ifdef CONFIG_TMPFS
801
static inline void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
802 803 804 805
{
}
#endif /* CONFIG_TMPFS */

806 807
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
808
{
809
	return swapin_readahead(swap, gfp, NULL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
810 811
}

812
static inline struct page *shmem_alloc_page(gfp_t gfp,
813
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
814
{
815
	return alloc_page(gfp);
Linus Torvalds's avatar
Linus Torvalds committed
816
}
817
#endif /* CONFIG_NUMA */
Linus Torvalds's avatar
Linus Torvalds committed
818

819 820 821 822 823 824 825
#if !defined(CONFIG_NUMA) || !defined(CONFIG_TMPFS)
static inline struct mempolicy *shmem_get_sbmpol(struct shmem_sb_info *sbinfo)
{
	return NULL;
}
#endif

Linus Torvalds's avatar
Linus Torvalds committed
826
/*
827
 * shmem_getpage_gfp - find page in cache, or get from swap, or allocate
Linus Torvalds's avatar
Linus Torvalds committed
828 829 830 831 832
 *
 * 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
 */
833
static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
834
	struct page **pagep, enum sgp_type sgp, gfp_t gfp, int *fault_type)
Linus Torvalds's avatar
Linus Torvalds committed
835 836
{
	struct address_space *mapping = inode->i_mapping;
837
	struct shmem_inode_info *info;
Linus Torvalds's avatar
Linus Torvalds committed
838
	struct shmem_sb_info *sbinfo;
839
	struct page *page;
Linus Torvalds's avatar
Linus Torvalds committed
840 841
	swp_entry_t swap;
	int error;
842
	int once = 0;
Linus Torvalds's avatar
Linus Torvalds committed
843

844
	if (index > (MAX_LFS_FILESIZE >> PAGE_CACHE_SHIFT))
Linus Torvalds's avatar
Linus Torvalds committed
845 846
		return -EFBIG;
repeat:
847
	swap.val = 0;
848
	page = find_lock_page(mapping, index);
849 850 851 852 853 854 855 856 857 858 859 860
	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
861
		/*
862 863 864
		 * 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
865
		 */
866 867 868
		BUG_ON(page && !PageUptodate(page));
		*pagep = page;
		return 0;
869 870 871
	}

	/*
872 873
	 * Fast cache lookup did not find it:
	 * bring it back from swap or allocate.
874
	 */
875 876
	info = SHMEM_I(inode);
	sbinfo = SHMEM_SB(inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
877 878 879

	if (swap.val) {
		/* Look it up and read it in.. */
880 881
		page = lookup_swap_cache(swap);
		if (!page) {
Linus Torvalds's avatar
Linus Torvalds committed
882
			/* here we actually do the io */
883 884
			if (fault_type)
				*fault_type |= VM_FAULT_MAJOR;
885
			page = shmem_swapin(swap, gfp, info, index);
886
			if (!page) {
887 888
				error = -ENOMEM;
				goto failed;
Linus Torvalds's avatar
Linus Torvalds committed
889 890 891 892
			}
		}

		/* We have to do this with page locked to prevent races */
893
		lock_page(page);
894
		if (!PageUptodate(page)) {
Linus Torvalds's avatar
Linus Torvalds committed
895
			error = -EIO;
896
			goto failed;
Linus Torvalds's avatar
Linus Torvalds committed
897
		}
898 899 900 901 902 903 904 905 906
		wait_on_page_writeback(page);

		/* Someone may have already done it for us */
		if (page->mapping) {
			if (page->mapping == mapping &&
			    page->index == index)
				goto done;
			error = -EEXIST;
			goto failed;
Linus Torvalds's avatar
Linus Torvalds committed
907
		}
908

909 910 911 912 913
		error = mem_cgroup_cache_charge(page, current->mm,
						gfp & GFP_RECLAIM_MASK);
		if (!error)
			error = shmem_add_to_page_cache(page, mapping, index,
						gfp, swp_to_radix_entry(swap));
914 915 916 917
		if (error)
			goto failed;

		spin_lock(&info->lock);
918
		info->swapped--;
919
		shmem_recalc_inode(inode);
920
		spin_unlock(&info->lock);
921 922

		delete_from_swap_cache(page);
923 924 925
		set_page_dirty(page);
		swap_free(swap);

926 927 928 929
	} else {
		if (shmem_acct_block(info->flags)) {
			error = -ENOSPC;
			goto failed;
Linus Torvalds's avatar
Linus Torvalds committed
930
		}
931
		if (sbinfo->max_blocks) {
932
			if (percpu_counter_compare(&sbinfo->used_blocks,
933 934 935 936
						sbinfo->max_blocks) >= 0) {
				error = -ENOSPC;
				goto unacct;
			}
937
			percpu_counter_inc(&sbinfo->used_blocks);
938
		}
Linus Torvalds's avatar
Linus Torvalds committed
939

940 941 942 943
		page = shmem_alloc_page(gfp, info, index);
		if (!page) {
			error = -ENOMEM;
			goto decused;
Linus Torvalds's avatar
Linus Torvalds committed
944 945
		}

946 947
		SetPageSwapBacked(page);
		__set_page_locked(page);
948 949 950 951 952
		error = mem_cgroup_cache_charge(page, current->mm,
						gfp & GFP_RECLAIM_MASK);
		if (!error)
			error = shmem_add_to_page_cache(page, mapping, index,
						gfp, NULL);
953 954 955 956 957
		if (error)
			goto decused;
		lru_cache_add_anon(page);

		spin_lock(&info->lock);
Linus Torvalds's avatar
Linus Torvalds committed
958
		info->alloced++;
959 960
		inode->i_blocks += BLOCKS_PER_PAGE;
		shmem_recalc_inode(inode);
Linus Torvalds's avatar
Linus Torvalds committed
961
		spin_unlock(&info->lock);
962

963 964 965
		clear_highpage(page);
		flush_dcache_page(page);
		SetPageUptodate(page);
966
		if (sgp == SGP_DIRTY)
967
			set_page_dirty(page);
Linus Torvalds's avatar
Linus Torvalds committed
968 969
	}
done:
970 971 972 973 974
	/* Perhaps the file has been truncated since we checked */
	if (sgp != SGP_WRITE &&
	    ((loff_t)index << PAGE_CACHE_SHIFT) >= i_size_read(inode)) {
		error = -EINVAL;
		goto trunc;
Hugh Dickins's avatar
Hugh Dickins committed
975
	}
976 977
	*pagep = page;
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
978

979
	/*
980
	 * Error recovery.
981
	 */
982 983 984 985 986 987
trunc:
	ClearPageDirty(page);
	delete_from_page_cache(page);
	spin_lock(&info->lock);
	info->alloced--;
	inode->i_blocks -= BLOCKS_PER_PAGE;
988
	spin_unlock(&info->lock);
989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002
decused:
	if (sbinfo->max_blocks)
		percpu_counter_add(&sbinfo->used_blocks, -1);
unacct:
	shmem_unacct_blocks(info->flags, 1);
failed:
	if (swap.val && error != -EINVAL) {
		struct page *test = find_get_page(mapping, index);
		if (test && !radix_tree_exceptional_entry(test))
			page_cache_release(test);
		/* Have another try if the entry has changed */
		if (test != swp_to_radix_entry(swap))
			error = -EEXIST;
	}
1003
	if (page) {
1004
		unlock_page(page);
1005
		page_cache_release(page);
1006 1007 1008 1009 1010 1011
	}
	if (error == -ENOSPC && !once++) {
		info = SHMEM_I(inode);
		spin_lock(&info->lock);
		shmem_recalc_inode(inode);
		spin_unlock(&info->lock);
1012
		goto repeat;
1013
	}
1014 1015 1016
	if (error == -EEXIST)
		goto repeat;
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
1017 1018
}

Nick Piggin's avatar
Nick Piggin committed
1019
static int shmem_f