shmem.c 77.3 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4 5 6 7 8
/*
 * Resizable virtual memory filesystem for Linux.
 *
 * Copyright (C) 2000 Linus Torvalds.
 *		 2000 Transmeta Corp.
 *		 2000-2001 Christoph Rohland
 *		 2000-2001 SAP AG
 *		 2002 Red Hat Inc.
9 10
 * Copyright (C) 2002-2011 Hugh Dickins.
 * Copyright (C) 2011 Google Inc.
11
 * Copyright (C) 2002-2005 VERITAS Software Corporation.
Linus Torvalds's avatar
Linus Torvalds committed
12 13 14 15 16 17
 * Copyright (C) 2004 Andi Kleen, SuSE Labs
 *
 * Extended attribute support for tmpfs:
 * Copyright (c) 2004, Luke Kenneth Casson Leighton <lkcl@lkcl.net>
 * Copyright (c) 2004 Red Hat, Inc., James Morris <jmorris@redhat.com>
 *
18 19 20
 * tiny-shmem:
 * Copyright (c) 2004, 2008 Matt Mackall <mpm@selenic.com>
 *
Linus Torvalds's avatar
Linus Torvalds committed
21 22 23
 * This file is released under the GPL.
 */

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

static struct vfsmount *shm_mnt;

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

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

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

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

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

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

82 83 84 85 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
/*
 * 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;
245
	void *item;
246 247

	VM_BUG_ON(!expected);
248
	VM_BUG_ON(!replacement);
249
	pslot = radix_tree_lookup_slot(&mapping->page_tree, index);
250 251 252
	if (!pslot)
		return -ENOENT;
	item = radix_tree_deref_slot_protected(pslot, &mapping->tree_lock);
253 254
	if (item != expected)
		return -ENOENT;
255
	radix_tree_replace_slot(pslot, replacement);
256 257 258
	return 0;
}

259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276
/*
 * Sometimes, before we decide whether to proceed or to fail, we must check
 * that an entry was not already brought back from swap by a racing thread.
 *
 * Checking page is not enough: by the time a SwapCache page is locked, it
 * might be reused, and again be SwapCache, using the same swap as before.
 */
static bool shmem_confirm_swap(struct address_space *mapping,
			       pgoff_t index, swp_entry_t swap)
{
	void *item;

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

277 278 279 280 281 282 283
/*
 * Like add_to_page_cache_locked, but error if expected item has gone.
 */
static int shmem_add_to_page_cache(struct page *page,
				   struct address_space *mapping,
				   pgoff_t index, gfp_t gfp, void *expected)
{
284
	int error;
285

286 287
	VM_BUG_ON_PAGE(!PageLocked(page), page);
	VM_BUG_ON_PAGE(!PageSwapBacked(page), page);
288

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

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

312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330
/*
 * Like delete_from_page_cache, but substitutes swap for page.
 */
static void shmem_delete_from_page_cache(struct page *page, void *radswap)
{
	struct address_space *mapping = page->mapping;
	int error;

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

331 332 333 334 335 336
/*
 * 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)
{
337
	void *old;
338 339

	spin_lock_irq(&mapping->tree_lock);
340
	old = radix_tree_delete_item(&mapping->page_tree, index, radswap);
341
	spin_unlock_irq(&mapping->tree_lock);
342 343 344 345
	if (old != radswap)
		return -ENOENT;
	free_swap_and_cache(radix_to_swp_entry(radswap));
	return 0;
346 347
}

348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365
/*
 * 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.
		 */
366 367
		pvec.nr = find_get_entries(mapping, index,
					   PAGEVEC_SIZE, pvec.pages, indices);
368 369 370
		if (!pvec.nr)
			break;
		index = indices[pvec.nr - 1] + 1;
371
		pagevec_remove_exceptionals(&pvec);
372 373 374 375
		check_move_unevictable_pages(pvec.pages, pvec.nr);
		pagevec_release(&pvec);
		cond_resched();
	}
376 377 378 379
}

/*
 * Remove range of pages and swap entries from radix tree, and free them.
380
 * If !unfalloc, truncate or punch hole; if unfalloc, undo failed fallocate.
381
 */
382 383
static void shmem_undo_range(struct inode *inode, loff_t lstart, loff_t lend,
								 bool unfalloc)
Linus Torvalds's avatar
Linus Torvalds committed
384
{
385
	struct address_space *mapping = inode->i_mapping;
Linus Torvalds's avatar
Linus Torvalds committed
386
	struct shmem_inode_info *info = SHMEM_I(inode);
387
	pgoff_t start = (lstart + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
388 389 390
	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);
391
	struct pagevec pvec;
392 393
	pgoff_t indices[PAGEVEC_SIZE];
	long nr_swaps_freed = 0;
394
	pgoff_t index;
395 396
	int i;

397 398
	if (lend == -1)
		end = -1;	/* unsigned, so actually very big */
399 400 401

	pagevec_init(&pvec, 0);
	index = start;
402
	while (index < end) {
403 404 405
		pvec.nr = find_get_entries(mapping, index,
			min(end - index, (pgoff_t)PAGEVEC_SIZE),
			pvec.pages, indices);
406 407
		if (!pvec.nr)
			break;
408 409 410 411
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

412
			index = indices[i];
413
			if (index >= end)
414 415
				break;

416
			if (radix_tree_exceptional_entry(page)) {
417 418
				if (unfalloc)
					continue;
419 420
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
421
				continue;
422 423 424
			}

			if (!trylock_page(page))
425
				continue;
426 427
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
428
					VM_BUG_ON_PAGE(PageWriteback(page), page);
429 430
					truncate_inode_page(mapping, page);
				}
431 432 433
			}
			unlock_page(page);
		}
434
		pagevec_remove_exceptionals(&pvec);
435
		pagevec_release(&pvec);
436 437 438 439
		mem_cgroup_uncharge_end();
		cond_resched();
		index++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
440

441
	if (partial_start) {
442 443 444
		struct page *page = NULL;
		shmem_getpage(inode, start - 1, &page, SGP_READ, NULL);
		if (page) {
445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460
			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);
461 462 463 464 465
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}
466 467
	if (start >= end)
		return;
468 469 470 471

	index = start;
	for ( ; ; ) {
		cond_resched();
472 473

		pvec.nr = find_get_entries(mapping, index,
474
				min(end - index, (pgoff_t)PAGEVEC_SIZE),
475
				pvec.pages, indices);
476
		if (!pvec.nr) {
477
			if (index == start || unfalloc)
478 479 480 481
				break;
			index = start;
			continue;
		}
482
		if ((index == start || unfalloc) && indices[0] >= end) {
483
			pagevec_remove_exceptionals(&pvec);
484
			pagevec_release(&pvec);
485 486 487 488 489 490
			break;
		}
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

491
			index = indices[i];
492
			if (index >= end)
493 494
				break;

495
			if (radix_tree_exceptional_entry(page)) {
496 497
				if (unfalloc)
					continue;
498 499 500 501 502
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
				continue;
			}

503
			lock_page(page);
504 505
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
506
					VM_BUG_ON_PAGE(PageWriteback(page), page);
507 508
					truncate_inode_page(mapping, page);
				}
509
			}
510 511
			unlock_page(page);
		}
512
		pagevec_remove_exceptionals(&pvec);
513
		pagevec_release(&pvec);
514 515 516
		mem_cgroup_uncharge_end();
		index++;
	}
517

Linus Torvalds's avatar
Linus Torvalds committed
518
	spin_lock(&info->lock);
519
	info->swapped -= nr_swaps_freed;
Linus Torvalds's avatar
Linus Torvalds committed
520 521
	shmem_recalc_inode(inode);
	spin_unlock(&info->lock);
522
}
Linus Torvalds's avatar
Linus Torvalds committed
523

524 525 526
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
{
	shmem_undo_range(inode, lstart, lend, false);
527
	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
Linus Torvalds's avatar
Linus Torvalds committed
528
}
529
EXPORT_SYMBOL_GPL(shmem_truncate_range);
Linus Torvalds's avatar
Linus Torvalds committed
530

531
static int shmem_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds's avatar
Linus Torvalds committed
532 533 534 535
{
	struct inode *inode = dentry->d_inode;
	int error;

536 537 538 539
	error = inode_change_ok(inode, attr);
	if (error)
		return error;

540 541 542
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
		loff_t oldsize = inode->i_size;
		loff_t newsize = attr->ia_size;
543

544 545 546 547 548 549 550 551 552 553 554
		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
555 556
	}

557 558
	setattr_copy(inode, attr);
	if (attr->ia_valid & ATTR_MODE)
Christoph Hellwig's avatar
Christoph Hellwig committed
559
		error = posix_acl_chmod(inode, inode->i_mode);
Linus Torvalds's avatar
Linus Torvalds committed
560 561 562
	return error;
}

Al Viro's avatar
Al Viro committed
563
static void shmem_evict_inode(struct inode *inode)
Linus Torvalds's avatar
Linus Torvalds committed
564 565 566
{
	struct shmem_inode_info *info = SHMEM_I(inode);

567
	if (inode->i_mapping->a_ops == &shmem_aops) {
Linus Torvalds's avatar
Linus Torvalds committed
568 569
		shmem_unacct_size(info->flags, inode->i_size);
		inode->i_size = 0;
570
		shmem_truncate_range(inode, 0, (loff_t)-1);
Linus Torvalds's avatar
Linus Torvalds committed
571
		if (!list_empty(&info->swaplist)) {
572
			mutex_lock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
573
			list_del_init(&info->swaplist);
574
			mutex_unlock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
575
		}
576 577
	} else
		kfree(info->symlink);
578

579
	simple_xattrs_free(&info->xattrs);
580
	WARN_ON(inode->i_blocks);
581
	shmem_free_inode(inode->i_sb);
582
	clear_inode(inode);
Linus Torvalds's avatar
Linus Torvalds committed
583 584
}

585 586 587
/*
 * If swap found in inode, free it and move page from swapcache to filecache.
 */
588
static int shmem_unuse_inode(struct shmem_inode_info *info,
589
			     swp_entry_t swap, struct page **pagep)
Linus Torvalds's avatar
Linus Torvalds committed
590
{
591
	struct address_space *mapping = info->vfs_inode.i_mapping;
592
	void *radswap;
593
	pgoff_t index;
594 595
	gfp_t gfp;
	int error = 0;
Linus Torvalds's avatar
Linus Torvalds committed
596

597
	radswap = swp_to_radix_entry(swap);
598
	index = radix_tree_locate_item(&mapping->page_tree, radswap);
599
	if (index == -1)
600
		return 0;
601

Hugh Dickins's avatar
Hugh Dickins committed
602 603
	/*
	 * Move _head_ to start search for next from here.
Al Viro's avatar
Al Viro committed
604
	 * But be careful: shmem_evict_inode checks list_empty without taking
Hugh Dickins's avatar
Hugh Dickins committed
605
	 * mutex, and there's an instant in list_move_tail when info->swaplist
606
	 * would appear empty, if it were the only one on shmem_swaplist.
Hugh Dickins's avatar
Hugh Dickins committed
607 608 609
	 */
	if (shmem_swaplist.next != &info->swaplist)
		list_move_tail(&shmem_swaplist, &info->swaplist);
610

611 612 613 614 615 616 617
	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
618 619 620 621 622 623 624 625 626 627 628 629 630 631 632
		 * 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).
633 634 635 636 637
		 */
		if (!page_swapcount(*pagep))
			error = -ENOENT;
	}

KAMEZAWA Hiroyuki's avatar
KAMEZAWA Hiroyuki committed
638
	/*
639 640 641
	 * 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
642
	 */
643 644
	if (!error)
		error = shmem_add_to_page_cache(*pagep, mapping, index,
645
						GFP_NOWAIT, radswap);
646
	if (error != -ENOMEM) {
647 648 649 650
		/*
		 * Truncation and eviction use free_swap_and_cache(), which
		 * only does trylock page: if we raced, best clean up here.
		 */
651 652
		delete_from_swap_cache(*pagep);
		set_page_dirty(*pagep);
653 654 655 656 657 658
		if (!error) {
			spin_lock(&info->lock);
			info->swapped--;
			spin_unlock(&info->lock);
			swap_free(swap);
		}
659
		error = 1;	/* not an error, but entry was found */
Linus Torvalds's avatar
Linus Torvalds committed
660
	}
661
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
662 663 664
}

/*
665
 * Search through swapped inodes to find and replace swap by page.
Linus Torvalds's avatar
Linus Torvalds committed
666
 */
667
int shmem_unuse(swp_entry_t swap, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
668
{
669
	struct list_head *this, *next;
Linus Torvalds's avatar
Linus Torvalds committed
670 671
	struct shmem_inode_info *info;
	int found = 0;
672 673 674 675
	int error = 0;

	/*
	 * There's a faint possibility that swap page was replaced before
676
	 * caller locked it: caller will come back later with the right page.
677
	 */
678
	if (unlikely(!PageSwapCache(page) || page_private(page) != swap.val))
679
		goto out;
680 681 682 683 684 685 686 687 688

	/*
	 * 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;
689
	/* No radix_tree_preload: swap entry keeps a place for page in tree */
Linus Torvalds's avatar
Linus Torvalds committed
690

691
	mutex_lock(&shmem_swaplist_mutex);
692 693
	list_for_each_safe(this, next, &shmem_swaplist) {
		info = list_entry(this, struct shmem_inode_info, swaplist);
694
		if (info->swapped)
695
			found = shmem_unuse_inode(info, swap, &page);
696 697
		else
			list_del_init(&info->swaplist);
698
		cond_resched();
699
		if (found)
700
			break;
Linus Torvalds's avatar
Linus Torvalds committed
701
	}
702
	mutex_unlock(&shmem_swaplist_mutex);
703 704 705 706

	if (found < 0)
		error = found;
out:
Hugh Dickins's avatar
Hugh Dickins committed
707 708
	unlock_page(page);
	page_cache_release(page);
709
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
710 711 712 713 714 715 716 717 718 719
}

/*
 * 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;
720 721
	swp_entry_t swap;
	pgoff_t index;
Linus Torvalds's avatar
Linus Torvalds committed
722 723 724 725 726 727 728 729

	BUG_ON(!PageLocked(page));
	mapping = page->mapping;
	index = page->index;
	inode = mapping->host;
	info = SHMEM_I(inode);
	if (info->flags & VM_LOCKED)
		goto redirty;
730
	if (!total_swap_pages)
Linus Torvalds's avatar
Linus Torvalds committed
731 732
		goto redirty;

733 734 735
	/*
	 * shmem_backing_dev_info's capabilities prevent regular writeback or
	 * sync from ever calling shmem_writepage; but a stacking filesystem
736
	 * might use ->writepage of its underlying filesystem, in which case
737
	 * tmpfs should write out to swap only in response to memory pressure,
738
	 * and not for the writeback threads or sync.
739
	 */
740 741 742 743
	if (!wbc->for_reclaim) {
		WARN_ON_ONCE(1);	/* Still happens? Tell us about it! */
		goto redirty;
	}
744 745 746 747 748

	/*
	 * 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.
749 750 751 752 753 754
	 *
	 * 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.
755 756
	 */
	if (!PageUptodate(page)) {
757 758 759 760 761 762 763 764 765 766 767 768 769 770
		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;
		}
771 772 773 774 775
		clear_highpage(page);
		flush_dcache_page(page);
		SetPageUptodate(page);
	}

776 777 778
	swap = get_swap_page();
	if (!swap.val)
		goto redirty;
779

780 781
	/*
	 * Add inode to shmem_unuse()'s list of swapped-out inodes,
782 783
	 * if it's not already there.  Do it now before the page is
	 * moved to swap cache, when its pagelock no longer protects
784
	 * the inode from eviction.  But don't unlock the mutex until
785 786
	 * we've incremented swapped, because shmem_unuse_inode() will
	 * prune a !swapped inode from the swaplist under this mutex.
787
	 */
788 789 790
	mutex_lock(&shmem_swaplist_mutex);
	if (list_empty(&info->swaplist))
		list_add_tail(&info->swaplist, &shmem_swaplist);
791

792
	if (add_to_swap_cache(page, swap, GFP_ATOMIC) == 0) {
Hugh Dickins's avatar
Hugh Dickins committed
793
		swap_shmem_alloc(swap);
794 795 796 797 798
		shmem_delete_from_page_cache(page, swp_to_radix_entry(swap));

		spin_lock(&info->lock);
		info->swapped++;
		shmem_recalc_inode(inode);
799
		spin_unlock(&info->lock);
800 801

		mutex_unlock(&shmem_swaplist_mutex);
802
		BUG_ON(page_mapped(page));
803
		swap_writepage(page, wbc);
Linus Torvalds's avatar
Linus Torvalds committed
804 805 806
		return 0;
	}

807
	mutex_unlock(&shmem_swaplist_mutex);
808
	swapcache_free(swap, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
809 810
redirty:
	set_page_dirty(page);
811 812 813 814
	if (wbc->for_reclaim)
		return AOP_WRITEPAGE_ACTIVATE;	/* Return with page locked */
	unlock_page(page);
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
815 816 817
}

#ifdef CONFIG_NUMA
818
#ifdef CONFIG_TMPFS
819
static void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
820
{
821
	char buffer[64];
822

823
	if (!mpol || mpol->mode == MPOL_DEFAULT)
824
		return;		/* show nothing */
825

826
	mpol_to_str(buffer, sizeof(buffer), mpol);
827 828

	seq_printf(seq, ",mpol=%s", buffer);
829
}
830 831 832 833 834 835 836 837 838 839 840 841

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;
}
842 843
#endif /* CONFIG_TMPFS */

844 845
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
846 847
{
	struct vm_area_struct pvma;
Mel Gorman's avatar
Mel Gorman committed
848
	struct page *page;
849

Linus Torvalds's avatar
Linus Torvalds committed
850
	/* Create a pseudo vma that just contains the policy */
851
	pvma.vm_start = 0;
852 853
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
854
	pvma.vm_ops = NULL;
Mel Gorman's avatar
Mel Gorman committed
855 856 857 858 859 860 861 862
	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
863 864
}

865
static struct page *shmem_alloc_page(gfp_t gfp,
866
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
867 868
{
	struct vm_area_struct pvma;
Mel Gorman's avatar
Mel Gorman committed
869
	struct page *page;
Linus Torvalds's avatar
Linus Torvalds committed
870

871 872
	/* Create a pseudo vma that just contains the policy */
	pvma.vm_start = 0;
873 874
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
875
	pvma.vm_ops = NULL;
876
	pvma.vm_policy = mpol_shared_policy_lookup(&info->policy, index);
877

Mel Gorman's avatar
Mel Gorman committed
878 879 880 881 882 883
	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
884
}
885 886
#else /* !CONFIG_NUMA */
#ifdef CONFIG_TMPFS
887
static inline void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
888 889 890 891
{
}
#endif /* CONFIG_TMPFS */

892 893
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
894
{
895
	return swapin_readahead(swap, gfp, NULL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
896 897
}

898
static inline struct page *shmem_alloc_page(gfp_t gfp,
899
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
900
{
901
	return alloc_page(gfp);
Linus Torvalds's avatar
Linus Torvalds committed
902
}
903
#endif /* CONFIG_NUMA */
Linus Torvalds's avatar
Linus Torvalds committed
904

905 906 907 908 909 910 911
#if !defined(CONFIG_NUMA) || !defined(CONFIG_TMPFS)
static inline struct mempolicy *shmem_get_sbmpol(struct shmem_sb_info *sbinfo)
{
	return NULL;
}
#endif

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
/*
 * 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);
952
	flush_dcache_page(newpage);
953 954 955 956 957 958 959 960 961 962 963 964 965 966

	__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);
967 968 969 970
	if (!error) {
		__inc_zone_page_state(newpage, NR_FILE_PAGES);
		__dec_zone_page_state(oldpage, NR_FILE_PAGES);
	}
971 972
	spin_unlock_irq(&swap_mapping->tree_lock);

973 974 975 976 977 978 979 980 981 982 983 984
	if (unlikely(error)) {
		/*
		 * Is this possible?  I think not, now that our callers check
		 * both PageSwapCache and page_private after getting page lock;
		 * but be defensive.  Reverse old to newpage for clear and free.
		 */
		oldpage = newpage;
	} else {
		mem_cgroup_replace_page_cache(oldpage, newpage);
		lru_cache_add_anon(newpage);
		*pagep = newpage;
	}
985 986 987 988 989 990 991

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

	unlock_page(oldpage);
	page_cache_release(oldpage);
	page_cache_release(oldpage);
992
	return error;
993 994
}

Linus Torvalds's avatar
Linus Torvalds committed
995
/*
996
 * shmem_getpage_gfp - find page in cache, or get from swap, or allocate
Linus Torvalds's avatar