shmem.c 79.9 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
 * shmem_fallocate communicates with shmem_fault or shmem_writepage 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.
86 87
 */
struct shmem_falloc {
88
	wait_queue_head_t *waitq; /* faults into hole wait for punch to end */
89 90 91 92 93 94
	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 */
};

95
/* Flag allocation requirements to shmem_getpage */
Linus Torvalds's avatar
Linus Torvalds committed
96 97 98
enum sgp_type {
	SGP_READ,	/* don't exceed i_size, don't allocate page */
	SGP_CACHE,	/* don't exceed i_size, may allocate page */
99
	SGP_DIRTY,	/* like SGP_CACHE, but set new page dirty */
100 101
	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
102 103
};

Andrew Morton's avatar
Andrew Morton committed
104
#ifdef CONFIG_TMPFS
105 106 107 108 109 110 111 112 113
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
114
#endif
115

116 117 118
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);
119 120 121 122 123 124 125 126 127
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
128 129 130 131 132 133 134 135 136 137 138 139 140 141

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)
{
142
	return (flags & VM_NORESERVE) ?
143
		0 : security_vm_enough_memory_mm(current->mm, VM_ACCT(size));
Linus Torvalds's avatar
Linus Torvalds committed
144 145 146 147
}

static inline void shmem_unacct_size(unsigned long flags, loff_t size)
{
148
	if (!(flags & VM_NORESERVE))
Linus Torvalds's avatar
Linus Torvalds committed
149 150 151
		vm_unacct_memory(VM_ACCT(size));
}

152 153 154 155 156 157 158 159 160 161 162 163 164
static inline int shmem_reacct_size(unsigned long flags,
		loff_t oldsize, loff_t newsize)
{
	if (!(flags & VM_NORESERVE)) {
		if (VM_ACCT(newsize) > VM_ACCT(oldsize))
			return security_vm_enough_memory_mm(current->mm,
					VM_ACCT(newsize) - VM_ACCT(oldsize));
		else if (VM_ACCT(newsize) < VM_ACCT(oldsize))
			vm_unacct_memory(VM_ACCT(oldsize) - VM_ACCT(newsize));
	}
	return 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
165 166 167 168 169 170 171 172
/*
 * ... 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)
{
173
	return (flags & VM_NORESERVE) ?
174
		security_vm_enough_memory_mm(current->mm, VM_ACCT(PAGE_CACHE_SIZE)) : 0;
Linus Torvalds's avatar
Linus Torvalds committed
175 176 177 178
}

static inline void shmem_unacct_blocks(unsigned long flags, long pages)
{
179
	if (flags & VM_NORESERVE)
Linus Torvalds's avatar
Linus Torvalds committed
180 181 182
		vm_unacct_memory(pages * VM_ACCT(PAGE_CACHE_SIZE));
}

183
static const struct super_operations shmem_ops;
184
static const struct address_space_operations shmem_aops;
185
static const struct file_operations shmem_file_operations;
186 187 188
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;
189
static const struct vm_operations_struct shmem_vm_ops;
Linus Torvalds's avatar
Linus Torvalds committed
190

191
static struct backing_dev_info shmem_backing_dev_info  __read_mostly = {
Linus Torvalds's avatar
Linus Torvalds committed
192
	.ra_pages	= 0,	/* No readahead */
193
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK | BDI_CAP_SWAP_BACKED,
Linus Torvalds's avatar
Linus Torvalds committed
194 195 196
};

static LIST_HEAD(shmem_swaplist);
197
static DEFINE_MUTEX(shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
198

199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
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);
	}
}

224
/**
225
 * shmem_recalc_inode - recalculate the block usage of an inode
Linus Torvalds's avatar
Linus Torvalds committed
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242
 * @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) {
243 244 245
		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
246
		info->alloced -= freed;
247
		inode->i_blocks -= freed * BLOCKS_PER_PAGE;
Linus Torvalds's avatar
Linus Torvalds committed
248 249 250 251
		shmem_unacct_blocks(info->flags, freed);
	}
}

252 253 254 255 256 257 258
/*
 * Replace item expected in radix tree by a new item, while holding tree lock.
 */
static int shmem_radix_tree_replace(struct address_space *mapping,
			pgoff_t index, void *expected, void *replacement)
{
	void **pslot;
259
	void *item;
260 261

	VM_BUG_ON(!expected);
262
	VM_BUG_ON(!replacement);
263
	pslot = radix_tree_lookup_slot(&mapping->page_tree, index);
264 265 266
	if (!pslot)
		return -ENOENT;
	item = radix_tree_deref_slot_protected(pslot, &mapping->tree_lock);
267 268
	if (item != expected)
		return -ENOENT;
269
	radix_tree_replace_slot(pslot, replacement);
270 271 272
	return 0;
}

273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290
/*
 * 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);
}

291 292 293 294 295 296 297
/*
 * 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)
{
298
	int error;
299

300 301
	VM_BUG_ON_PAGE(!PageLocked(page), page);
	VM_BUG_ON_PAGE(!PageSwapBacked(page), page);
302

303 304 305 306 307
	page_cache_get(page);
	page->mapping = mapping;
	page->index = index;

	spin_lock_irq(&mapping->tree_lock);
308
	if (!expected)
309 310 311 312
		error = radix_tree_insert(&mapping->page_tree, index, page);
	else
		error = shmem_radix_tree_replace(mapping, index, expected,
								 page);
313
	if (!error) {
314 315 316 317 318 319 320 321
		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);
322 323 324 325
	}
	return error;
}

326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344
/*
 * 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);
}

345 346 347 348 349 350
/*
 * 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)
{
351
	void *old;
352 353

	spin_lock_irq(&mapping->tree_lock);
354
	old = radix_tree_delete_item(&mapping->page_tree, index, radswap);
355
	spin_unlock_irq(&mapping->tree_lock);
356 357 358 359
	if (old != radswap)
		return -ENOENT;
	free_swap_and_cache(radix_to_swp_entry(radswap));
	return 0;
360 361
}

362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379
/*
 * 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.
		 */
380 381
		pvec.nr = find_get_entries(mapping, index,
					   PAGEVEC_SIZE, pvec.pages, indices);
382 383 384
		if (!pvec.nr)
			break;
		index = indices[pvec.nr - 1] + 1;
385
		pagevec_remove_exceptionals(&pvec);
386 387 388 389
		check_move_unevictable_pages(pvec.pages, pvec.nr);
		pagevec_release(&pvec);
		cond_resched();
	}
390 391 392 393
}

/*
 * Remove range of pages and swap entries from radix tree, and free them.
394
 * If !unfalloc, truncate or punch hole; if unfalloc, undo failed fallocate.
395
 */
396 397
static void shmem_undo_range(struct inode *inode, loff_t lstart, loff_t lend,
								 bool unfalloc)
Linus Torvalds's avatar
Linus Torvalds committed
398
{
399
	struct address_space *mapping = inode->i_mapping;
Linus Torvalds's avatar
Linus Torvalds committed
400
	struct shmem_inode_info *info = SHMEM_I(inode);
401
	pgoff_t start = (lstart + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
402 403 404
	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);
405
	struct pagevec pvec;
406 407
	pgoff_t indices[PAGEVEC_SIZE];
	long nr_swaps_freed = 0;
408
	pgoff_t index;
409 410
	int i;

411 412
	if (lend == -1)
		end = -1;	/* unsigned, so actually very big */
413 414 415

	pagevec_init(&pvec, 0);
	index = start;
416
	while (index < end) {
417 418 419
		pvec.nr = find_get_entries(mapping, index,
			min(end - index, (pgoff_t)PAGEVEC_SIZE),
			pvec.pages, indices);
420 421
		if (!pvec.nr)
			break;
422 423 424 425
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

426
			index = indices[i];
427
			if (index >= end)
428 429
				break;

430
			if (radix_tree_exceptional_entry(page)) {
431 432
				if (unfalloc)
					continue;
433 434
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
435
				continue;
436 437 438
			}

			if (!trylock_page(page))
439
				continue;
440 441
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
442
					VM_BUG_ON_PAGE(PageWriteback(page), page);
443 444
					truncate_inode_page(mapping, page);
				}
445 446 447
			}
			unlock_page(page);
		}
448
		pagevec_remove_exceptionals(&pvec);
449
		pagevec_release(&pvec);
450 451 452 453
		mem_cgroup_uncharge_end();
		cond_resched();
		index++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
454

455
	if (partial_start) {
456 457 458
		struct page *page = NULL;
		shmem_getpage(inode, start - 1, &page, SGP_READ, NULL);
		if (page) {
459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
			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);
475 476 477 478 479
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}
480 481
	if (start >= end)
		return;
482 483

	index = start;
484
	while (index < end) {
485
		cond_resched();
486 487

		pvec.nr = find_get_entries(mapping, index,
488
				min(end - index, (pgoff_t)PAGEVEC_SIZE),
489
				pvec.pages, indices);
490
		if (!pvec.nr) {
491 492
			/* If all gone or hole-punch or unfalloc, we're done */
			if (index == start || end != -1)
493
				break;
494
			/* But if truncating, restart to make sure all gone */
495 496 497 498 499 500 501
			index = start;
			continue;
		}
		mem_cgroup_uncharge_start();
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

502
			index = indices[i];
503
			if (index >= end)
504 505
				break;

506
			if (radix_tree_exceptional_entry(page)) {
507 508
				if (unfalloc)
					continue;
509 510 511 512 513 514
				if (shmem_free_swap(mapping, index, page)) {
					/* Swap was replaced by page: retry */
					index--;
					break;
				}
				nr_swaps_freed++;
515 516 517
				continue;
			}

518
			lock_page(page);
519 520
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
521
					VM_BUG_ON_PAGE(PageWriteback(page), page);
522
					truncate_inode_page(mapping, page);
523 524 525 526 527
				} else {
					/* Page was replaced by swap: retry */
					unlock_page(page);
					index--;
					break;
528
				}
529
			}
530 531
			unlock_page(page);
		}
532
		pagevec_remove_exceptionals(&pvec);
533
		pagevec_release(&pvec);
534 535 536
		mem_cgroup_uncharge_end();
		index++;
	}
537

Linus Torvalds's avatar
Linus Torvalds committed
538
	spin_lock(&info->lock);
539
	info->swapped -= nr_swaps_freed;
Linus Torvalds's avatar
Linus Torvalds committed
540 541
	shmem_recalc_inode(inode);
	spin_unlock(&info->lock);
542
}
Linus Torvalds's avatar
Linus Torvalds committed
543

544 545 546
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
{
	shmem_undo_range(inode, lstart, lend, false);
547
	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
Linus Torvalds's avatar
Linus Torvalds committed
548
}
549
EXPORT_SYMBOL_GPL(shmem_truncate_range);
Linus Torvalds's avatar
Linus Torvalds committed
550

551
static int shmem_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds's avatar
Linus Torvalds committed
552 553 554 555
{
	struct inode *inode = dentry->d_inode;
	int error;

556 557 558 559
	error = inode_change_ok(inode, attr);
	if (error)
		return error;

560 561 562
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
		loff_t oldsize = inode->i_size;
		loff_t newsize = attr->ia_size;
563

564
		if (newsize != oldsize) {
565 566 567 568
			error = shmem_reacct_size(SHMEM_I(inode)->flags,
					oldsize, newsize);
			if (error)
				return error;
569 570 571 572 573 574 575 576 577 578
			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
579 580
	}

581 582
	setattr_copy(inode, attr);
	if (attr->ia_valid & ATTR_MODE)
Christoph Hellwig's avatar
Christoph Hellwig committed
583
		error = posix_acl_chmod(inode, inode->i_mode);
Linus Torvalds's avatar
Linus Torvalds committed
584 585 586
	return error;
}

Al Viro's avatar
Al Viro committed
587
static void shmem_evict_inode(struct inode *inode)
Linus Torvalds's avatar
Linus Torvalds committed
588 589 590
{
	struct shmem_inode_info *info = SHMEM_I(inode);

591
	if (inode->i_mapping->a_ops == &shmem_aops) {
Linus Torvalds's avatar
Linus Torvalds committed
592 593
		shmem_unacct_size(info->flags, inode->i_size);
		inode->i_size = 0;
594
		shmem_truncate_range(inode, 0, (loff_t)-1);
Linus Torvalds's avatar
Linus Torvalds committed
595
		if (!list_empty(&info->swaplist)) {
596
			mutex_lock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
597
			list_del_init(&info->swaplist);
598
			mutex_unlock(&shmem_swaplist_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
599
		}
600 601
	} else
		kfree(info->symlink);
602

603
	simple_xattrs_free(&info->xattrs);
604
	WARN_ON(inode->i_blocks);
605
	shmem_free_inode(inode->i_sb);
606
	clear_inode(inode);
Linus Torvalds's avatar
Linus Torvalds committed
607 608
}

609 610 611
/*
 * If swap found in inode, free it and move page from swapcache to filecache.
 */
612
static int shmem_unuse_inode(struct shmem_inode_info *info,
613
			     swp_entry_t swap, struct page **pagep)
Linus Torvalds's avatar
Linus Torvalds committed
614
{
615
	struct address_space *mapping = info->vfs_inode.i_mapping;
616
	void *radswap;
617
	pgoff_t index;
618 619
	gfp_t gfp;
	int error = 0;
Linus Torvalds's avatar
Linus Torvalds committed
620

621
	radswap = swp_to_radix_entry(swap);
622
	index = radix_tree_locate_item(&mapping->page_tree, radswap);
623
	if (index == -1)
624
		return 0;
625

Hugh Dickins's avatar
Hugh Dickins committed
626 627
	/*
	 * Move _head_ to start search for next from here.
Al Viro's avatar
Al Viro committed
628
	 * But be careful: shmem_evict_inode checks list_empty without taking
Hugh Dickins's avatar
Hugh Dickins committed
629
	 * mutex, and there's an instant in list_move_tail when info->swaplist
630
	 * would appear empty, if it were the only one on shmem_swaplist.
Hugh Dickins's avatar
Hugh Dickins committed
631 632 633
	 */
	if (shmem_swaplist.next != &info->swaplist)
		list_move_tail(&shmem_swaplist, &info->swaplist);
634

635 636 637 638 639 640 641
	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
642 643 644 645 646 647 648 649 650 651 652 653 654 655 656
		 * 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).
657 658 659 660 661
		 */
		if (!page_swapcount(*pagep))
			error = -ENOENT;
	}

KAMEZAWA Hiroyuki's avatar
KAMEZAWA Hiroyuki committed
662
	/*
663 664 665
	 * 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
666
	 */
667 668
	if (!error)
		error = shmem_add_to_page_cache(*pagep, mapping, index,
669
						GFP_NOWAIT, radswap);
670
	if (error != -ENOMEM) {
671 672 673 674
		/*
		 * Truncation and eviction use free_swap_and_cache(), which
		 * only does trylock page: if we raced, best clean up here.
		 */
675 676
		delete_from_swap_cache(*pagep);
		set_page_dirty(*pagep);
677 678 679 680 681 682
		if (!error) {
			spin_lock(&info->lock);
			info->swapped--;
			spin_unlock(&info->lock);
			swap_free(swap);
		}
683
		error = 1;	/* not an error, but entry was found */
Linus Torvalds's avatar
Linus Torvalds committed
684
	}
685
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
686 687 688
}

/*
689
 * Search through swapped inodes to find and replace swap by page.
Linus Torvalds's avatar
Linus Torvalds committed
690
 */
691
int shmem_unuse(swp_entry_t swap, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
692
{
693
	struct list_head *this, *next;
Linus Torvalds's avatar
Linus Torvalds committed
694 695
	struct shmem_inode_info *info;
	int found = 0;
696 697 698 699
	int error = 0;

	/*
	 * There's a faint possibility that swap page was replaced before
700
	 * caller locked it: caller will come back later with the right page.
701
	 */
702
	if (unlikely(!PageSwapCache(page) || page_private(page) != swap.val))
703
		goto out;
704 705 706 707 708 709

	/*
	 * 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.
	 */
710
	error = mem_cgroup_charge_file(page, current->mm, GFP_KERNEL);
711 712
	if (error)
		goto out;
713
	/* No radix_tree_preload: swap entry keeps a place for page in tree */
Linus Torvalds's avatar
Linus Torvalds committed
714

715
	mutex_lock(&shmem_swaplist_mutex);
716 717
	list_for_each_safe(this, next, &shmem_swaplist) {
		info = list_entry(this, struct shmem_inode_info, swaplist);
718
		if (info->swapped)
719
			found = shmem_unuse_inode(info, swap, &page);
720 721
		else
			list_del_init(&info->swaplist);
722
		cond_resched();
723
		if (found)
724
			break;
Linus Torvalds's avatar
Linus Torvalds committed
725
	}
726
	mutex_unlock(&shmem_swaplist_mutex);
727 728 729 730

	if (found < 0)
		error = found;
out:
Hugh Dickins's avatar
Hugh Dickins committed
731 732
	unlock_page(page);
	page_cache_release(page);
733
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
734 735 736 737 738 739 740 741 742 743
}

/*
 * 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;
744 745
	swp_entry_t swap;
	pgoff_t index;
Linus Torvalds's avatar
Linus Torvalds committed
746 747 748 749 750 751 752 753

	BUG_ON(!PageLocked(page));
	mapping = page->mapping;
	index = page->index;
	inode = mapping->host;
	info = SHMEM_I(inode);
	if (info->flags & VM_LOCKED)
		goto redirty;
754
	if (!total_swap_pages)
Linus Torvalds's avatar
Linus Torvalds committed
755 756
		goto redirty;

757 758 759
	/*
	 * shmem_backing_dev_info's capabilities prevent regular writeback or
	 * sync from ever calling shmem_writepage; but a stacking filesystem
760
	 * might use ->writepage of its underlying filesystem, in which case
761
	 * tmpfs should write out to swap only in response to memory pressure,
762
	 * and not for the writeback threads or sync.
763
	 */
764 765 766 767
	if (!wbc->for_reclaim) {
		WARN_ON_ONCE(1);	/* Still happens? Tell us about it! */
		goto redirty;
	}
768 769 770 771 772

	/*
	 * 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.
773 774 775 776 777 778
	 *
	 * 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.
779 780
	 */
	if (!PageUptodate(page)) {
781 782 783 784 785
		if (inode->i_private) {
			struct shmem_falloc *shmem_falloc;
			spin_lock(&inode->i_lock);
			shmem_falloc = inode->i_private;
			if (shmem_falloc &&
786
			    !shmem_falloc->waitq &&
787 788 789 790 791 792 793 794 795
			    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;
		}
796 797 798 799 800
		clear_highpage(page);
		flush_dcache_page(page);
		SetPageUptodate(page);
	}

801 802 803
	swap = get_swap_page();
	if (!swap.val)
		goto redirty;
804

805 806
	/*
	 * Add inode to shmem_unuse()'s list of swapped-out inodes,
807 808
	 * if it's not already there.  Do it now before the page is
	 * moved to swap cache, when its pagelock no longer protects
809
	 * the inode from eviction.  But don't unlock the mutex until
810 811
	 * we've incremented swapped, because shmem_unuse_inode() will
	 * prune a !swapped inode from the swaplist under this mutex.
812
	 */
813 814 815
	mutex_lock(&shmem_swaplist_mutex);
	if (list_empty(&info->swaplist))
		list_add_tail(&info->swaplist, &shmem_swaplist);
816

817
	if (add_to_swap_cache(page, swap, GFP_ATOMIC) == 0) {
Hugh Dickins's avatar
Hugh Dickins committed
818
		swap_shmem_alloc(swap);
819 820 821 822 823
		shmem_delete_from_page_cache(page, swp_to_radix_entry(swap));

		spin_lock(&info->lock);
		info->swapped++;
		shmem_recalc_inode(inode);
824
		spin_unlock(&info->lock);
825 826

		mutex_unlock(&shmem_swaplist_mutex);
827
		BUG_ON(page_mapped(page));
828
		swap_writepage(page, wbc);
Linus Torvalds's avatar
Linus Torvalds committed
829 830 831
		return 0;
	}

832
	mutex_unlock(&shmem_swaplist_mutex);
833
	swapcache_free(swap, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
834 835
redirty:
	set_page_dirty(page);
836 837 838 839
	if (wbc->for_reclaim)
		return AOP_WRITEPAGE_ACTIVATE;	/* Return with page locked */
	unlock_page(page);
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
840 841 842
}

#ifdef CONFIG_NUMA
843
#ifdef CONFIG_TMPFS
844
static void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
845
{
846
	char buffer[64];
847

848
	if (!mpol || mpol->mode == MPOL_DEFAULT)
849
		return;		/* show nothing */
850

851
	mpol_to_str(buffer, sizeof(buffer), mpol);
852 853

	seq_printf(seq, ",mpol=%s", buffer);
854
}
855 856 857 858 859 860 861 862 863 864 865 866

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;
}
867 868
#endif /* CONFIG_TMPFS */

869 870
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
871 872
{
	struct vm_area_struct pvma;
Mel Gorman's avatar
Mel Gorman committed
873
	struct page *page;
874

Linus Torvalds's avatar
Linus Torvalds committed
875
	/* Create a pseudo vma that just contains the policy */
876
	pvma.vm_start = 0;
877 878
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
879
	pvma.vm_ops = NULL;
Mel Gorman's avatar
Mel Gorman committed
880 881 882 883 884 885 886 887
	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
888 889
}

890
static struct page *shmem_alloc_page(gfp_t gfp,
891
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
892 893
{
	struct vm_area_struct pvma;
Mel Gorman's avatar
Mel Gorman committed
894
	struct page *page;
Linus Torvalds's avatar
Linus Torvalds committed
895

896 897
	/* Create a pseudo vma that just contains the policy */
	pvma.vm_start = 0;
898 899
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
900
	pvma.vm_ops = NULL;
901
	pvma.vm_policy = mpol_shared_policy_lookup(&info->policy, index);
902

Mel Gorman's avatar
Mel Gorman committed
903 904 905 906 907 908
	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
909
}
910 911
#else /* !CONFIG_NUMA */
#ifdef CONFIG_TMPFS
912
static inline void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
913 914 915 916
{
}
#endif /* CONFIG_TMPFS */

917 918
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
919
{
920
	return swapin_readahead(swap, gfp, NULL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
921 922
}

923
static inline struct page *shmem_alloc_page(gfp_t gfp,
924
			struct shmem_inode_info *info, pgoff_t index)
Linus Torvalds's avatar
Linus Torvalds committed
925
{
926
	return alloc_page(gfp);
Linus Torvalds's avatar
Linus Torvalds committed
927
}
928
#endif /* CONFIG_NUMA */
Linus Torvalds's avatar
Linus Torvalds committed
929

930 931 932 933 934 935 936
#if !defined(CONFIG_NUMA) || !defined(CONFIG_TMPFS)
static inline struct mempolicy *shmem_get_sbmpol(struct shmem_sb_info *sbinfo)
{
	return NULL;
}
#endif

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
/*
 * 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);
977
	flush_dcache_page(newpage);