shmem.c 77.9 KB
Newer Older
L
Linus Torvalds 已提交
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.
L
Linus Torvalds 已提交
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>
 *
L
Linus Torvalds 已提交
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>
A
Andrew Morton 已提交
28
#include <linux/ramfs.h>
H
Hugh Dickins 已提交
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
L
Linus Torvalds 已提交
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>
C
Christoph Hellwig 已提交
48
#include <linux/posix_acl_xattr.h>
L
Linus Torvalds 已提交
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>
L
Linus Torvalds 已提交
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>
M
Mimi Zohar 已提交
68
#include <linux/magic.h>
69

L
Linus Torvalds 已提交
70 71 72
#include <asm/uaccess.h>
#include <asm/pgtable.h>

H
Hugh Dickins 已提交
73
#define BLOCKS_PER_PAGE  (PAGE_CACHE_SIZE/512)
L
Linus Torvalds 已提交
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 */
L
Linus Torvalds 已提交
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 */
L
Linus Torvalds 已提交
101 102
};

A
Andrew Morton 已提交
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);
}
A
Andrew Morton 已提交
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);
}
L
Linus Torvalds 已提交
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));
L
Linus Torvalds 已提交
143 144 145 146
}

static inline void shmem_unacct_size(unsigned long flags, loff_t size)
{
147
	if (!(flags & VM_NORESERVE))
L
Linus Torvalds 已提交
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;
L
Linus Torvalds 已提交
161 162 163 164
}

static inline void shmem_unacct_blocks(unsigned long flags, long pages)
{
165
	if (flags & VM_NORESERVE)
L
Linus Torvalds 已提交
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;
L
Linus Torvalds 已提交
176

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

static LIST_HEAD(shmem_swaplist);
183
static DEFINE_MUTEX(shmem_swaplist_mutex);
L
Linus Torvalds 已提交
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
L
Linus Torvalds 已提交
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);
L
Linus Torvalds 已提交
232
		info->alloced -= freed;
233
		inode->i_blocks -= freed * BLOCKS_PER_PAGE;
L
Linus Torvalds 已提交
234 235 236 237
		shmem_unacct_blocks(info->flags, freed);
	}
}

238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
/*
 * Replace item expected in radix tree by a new item, while holding tree lock.
 */
static int shmem_radix_tree_replace(struct address_space *mapping,
			pgoff_t index, void *expected, void *replacement)
{
	void **pslot;
	void *item = NULL;

	VM_BUG_ON(!expected);
	pslot = radix_tree_lookup_slot(&mapping->page_tree, index);
	if (pslot)
		item = radix_tree_deref_slot_protected(pslot,
							&mapping->tree_lock);
	if (item != expected)
		return -ENOENT;
	if (replacement)
		radix_tree_replace_slot(pslot, replacement);
	else
		radix_tree_delete(&mapping->page_tree, index);
	return 0;
}

261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
/*
 * Sometimes, before we decide whether to proceed or to fail, we must check
 * that an entry was not already brought back from swap by a racing thread.
 *
 * Checking page is not enough: by the time a SwapCache page is locked, it
 * might be reused, and again be SwapCache, using the same swap as before.
 */
static bool shmem_confirm_swap(struct address_space *mapping,
			       pgoff_t index, swp_entry_t swap)
{
	void *item;

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

279 280 281 282 283 284 285
/*
 * Like add_to_page_cache_locked, but error if expected item has gone.
 */
static int shmem_add_to_page_cache(struct page *page,
				   struct address_space *mapping,
				   pgoff_t index, gfp_t gfp, void *expected)
{
286
	int error;
287

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

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

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

314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332
/*
 * Like delete_from_page_cache, but substitutes swap for page.
 */
static void shmem_delete_from_page_cache(struct page *page, void *radswap)
{
	struct address_space *mapping = page->mapping;
	int error;

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

333 334 335 336 337 338 339
/*
 * Like find_get_pages, but collecting swap entries as well as pages.
 */
static unsigned shmem_find_get_pages_and_swap(struct address_space *mapping,
					pgoff_t start, unsigned int nr_pages,
					struct page **pages, pgoff_t *indices)
{
340 341 342 343 344 345
	void **slot;
	unsigned int ret = 0;
	struct radix_tree_iter iter;

	if (!nr_pages)
		return 0;
346 347 348

	rcu_read_lock();
restart:
349
	radix_tree_for_each_slot(slot, &mapping->page_tree, &iter, start) {
350 351
		struct page *page;
repeat:
352
		page = radix_tree_deref_slot(slot);
353 354 355
		if (unlikely(!page))
			continue;
		if (radix_tree_exception(page)) {
356 357 358 359 360 361 362 363
			if (radix_tree_deref_retry(page))
				goto restart;
			/*
			 * Otherwise, we must be storing a swap entry
			 * here as an exceptional entry: so return it
			 * without attempting to raise page count.
			 */
			goto export;
364 365 366 367 368
		}
		if (!page_cache_get_speculative(page))
			goto repeat;

		/* Has the page moved? */
369
		if (unlikely(page != *slot)) {
370 371 372 373
			page_cache_release(page);
			goto repeat;
		}
export:
374
		indices[ret] = iter.index;
375
		pages[ret] = page;
376 377
		if (++ret == nr_pages)
			break;
378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401
	}
	rcu_read_unlock();
	return ret;
}

/*
 * Remove swap entry from radix tree, free the swap and its page cache.
 */
static int shmem_free_swap(struct address_space *mapping,
			   pgoff_t index, void *radswap)
{
	int error;

	spin_lock_irq(&mapping->tree_lock);
	error = shmem_radix_tree_replace(mapping, index, radswap, NULL);
	spin_unlock_irq(&mapping->tree_lock);
	if (!error)
		free_swap_and_cache(radix_to_swp_entry(radswap));
	return error;
}

/*
 * Pagevec may contain swap entries, so shuffle up pages before releasing.
 */
402
static void shmem_deswap_pagevec(struct pagevec *pvec)
403 404 405 406 407 408 409 410 411
{
	int i, j;

	for (i = 0, j = 0; i < pagevec_count(pvec); i++) {
		struct page *page = pvec->pages[i];
		if (!radix_tree_exceptional_entry(page))
			pvec->pages[j++] = page;
	}
	pvec->nr = j;
412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441
}

/*
 * SysV IPC SHM_UNLOCK restore Unevictable pages to their evictable lists.
 */
void shmem_unlock_mapping(struct address_space *mapping)
{
	struct pagevec pvec;
	pgoff_t indices[PAGEVEC_SIZE];
	pgoff_t index = 0;

	pagevec_init(&pvec, 0);
	/*
	 * Minor point, but we might as well stop if someone else SHM_LOCKs it.
	 */
	while (!mapping_unevictable(mapping)) {
		/*
		 * Avoid pagevec_lookup(): find_get_pages() returns 0 as if it
		 * has finished, if it hits a row of PAGEVEC_SIZE swap entries.
		 */
		pvec.nr = shmem_find_get_pages_and_swap(mapping, index,
					PAGEVEC_SIZE, pvec.pages, indices);
		if (!pvec.nr)
			break;
		index = indices[pvec.nr - 1] + 1;
		shmem_deswap_pagevec(&pvec);
		check_move_unevictable_pages(pvec.pages, pvec.nr);
		pagevec_release(&pvec);
		cond_resched();
	}
442 443 444 445
}

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

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

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

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

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

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

507
	if (partial_start) {
508 509 510
		struct page *page = NULL;
		shmem_getpage(inode, start - 1, &page, SGP_READ, NULL);
		if (page) {
511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526
			unsigned int top = PAGE_CACHE_SIZE;
			if (start > end) {
				top = partial_end;
				partial_end = 0;
			}
			zero_user_segment(page, partial_start, top);
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}
	if (partial_end) {
		struct page *page = NULL;
		shmem_getpage(inode, end, &page, SGP_READ, NULL);
		if (page) {
			zero_user_segment(page, 0, partial_end);
527 528 529 530 531
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}
532 533
	if (start >= end)
		return;
534 535 536 537

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

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

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

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

L
Linus Torvalds 已提交
583
	spin_lock(&info->lock);
584
	info->swapped -= nr_swaps_freed;
L
Linus Torvalds 已提交
585 586
	shmem_recalc_inode(inode);
	spin_unlock(&info->lock);
587
}
L
Linus Torvalds 已提交
588

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

596
static int shmem_setattr(struct dentry *dentry, struct iattr *attr)
L
Linus Torvalds 已提交
597 598 599 600
{
	struct inode *inode = dentry->d_inode;
	int error;

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

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

609 610 611 612 613 614 615 616 617 618 619
		if (newsize != oldsize) {
			i_size_write(inode, newsize);
			inode->i_ctime = inode->i_mtime = CURRENT_TIME;
		}
		if (newsize < oldsize) {
			loff_t holebegin = round_up(newsize, PAGE_SIZE);
			unmap_mapping_range(inode->i_mapping, holebegin, 0, 1);
			shmem_truncate_range(inode, newsize, (loff_t)-1);
			/* unmap again to remove racily COWed private pages */
			unmap_mapping_range(inode->i_mapping, holebegin, 0, 1);
		}
L
Linus Torvalds 已提交
620 621
	}

622 623
	setattr_copy(inode, attr);
	if (attr->ia_valid & ATTR_MODE)
C
Christoph Hellwig 已提交
624
		error = posix_acl_chmod(inode, inode->i_mode);
L
Linus Torvalds 已提交
625 626 627
	return error;
}

A
Al Viro 已提交
628
static void shmem_evict_inode(struct inode *inode)
L
Linus Torvalds 已提交
629 630 631
{
	struct shmem_inode_info *info = SHMEM_I(inode);

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

644
	simple_xattrs_free(&info->xattrs);
645
	WARN_ON(inode->i_blocks);
646
	shmem_free_inode(inode->i_sb);
647
	clear_inode(inode);
L
Linus Torvalds 已提交
648 649
}

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

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

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

676 677 678 679 680 681 682
	gfp = mapping_gfp_mask(mapping);
	if (shmem_should_replace_page(*pagep, gfp)) {
		mutex_unlock(&shmem_swaplist_mutex);
		error = shmem_replace_page(pagep, gfp, info, index);
		mutex_lock(&shmem_swaplist_mutex);
		/*
		 * We needed to drop mutex to make that restrictive page
683 684 685 686 687 688 689 690 691 692 693 694 695 696 697
		 * allocation, but the inode might have been freed while we
		 * dropped it: although a racing shmem_evict_inode() cannot
		 * complete without emptying the radix_tree, our page lock
		 * on this swapcache page is not enough to prevent that -
		 * free_swap_and_cache() of our swap entry will only
		 * trylock_page(), removing swap from radix_tree whatever.
		 *
		 * We must not proceed to shmem_add_to_page_cache() if the
		 * inode has been freed, but of course we cannot rely on
		 * inode or mapping or info to check that.  However, we can
		 * safely check if our swap entry is still in use (and here
		 * it can't have got reused for another page): if it's still
		 * in use, then the inode cannot have been freed yet, and we
		 * can safely proceed (if it's no longer in use, that tells
		 * nothing about the inode, but we don't need to unuse swap).
698 699 700 701 702
		 */
		if (!page_swapcount(*pagep))
			error = -ENOENT;
	}

K
KAMEZAWA Hiroyuki 已提交
703
	/*
704 705 706
	 * We rely on shmem_swaplist_mutex, not only to protect the swaplist,
	 * but also to hold up shmem_evict_inode(): so inode cannot be freed
	 * beneath us (pagelock doesn't help until the page is in pagecache).
K
KAMEZAWA Hiroyuki 已提交
707
	 */
708 709
	if (!error)
		error = shmem_add_to_page_cache(*pagep, mapping, index,
710
						GFP_NOWAIT, radswap);
711
	if (error != -ENOMEM) {
712 713 714 715
		/*
		 * Truncation and eviction use free_swap_and_cache(), which
		 * only does trylock page: if we raced, best clean up here.
		 */
716 717
		delete_from_swap_cache(*pagep);
		set_page_dirty(*pagep);
718 719 720 721 722 723
		if (!error) {
			spin_lock(&info->lock);
			info->swapped--;
			spin_unlock(&info->lock);
			swap_free(swap);
		}
724
		error = 1;	/* not an error, but entry was found */
L
Linus Torvalds 已提交
725
	}
726
	return error;
L
Linus Torvalds 已提交
727 728 729
}

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

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

	/*
	 * Charge page using GFP_KERNEL while we can wait, before taking
	 * the shmem_swaplist_mutex which might hold up shmem_writepage().
	 * Charged back to the user (not to caller) when swap account is used.
	 */
	error = mem_cgroup_cache_charge(page, current->mm, GFP_KERNEL);
	if (error)
		goto out;
754
	/* No radix_tree_preload: swap entry keeps a place for page in tree */
L
Linus Torvalds 已提交
755

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

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

/*
 * Move the page from the page cache to the swap cache.
 */
static int shmem_writepage(struct page *page, struct writeback_control *wbc)
{
	struct shmem_inode_info *info;
	struct address_space *mapping;
	struct inode *inode;
785 786
	swp_entry_t swap;
	pgoff_t index;
L
Linus Torvalds 已提交
787 788 789 790 791 792 793 794

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

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

	/*
	 * This is somewhat ridiculous, but without plumbing a SWAP_MAP_FALLOC
	 * value into swapfile.c, the only way we can correctly account for a
	 * fallocated page arriving here is now to initialize it and write it.
814 815 816 817 818 819
	 *
	 * That's okay for a page already fallocated earlier, but if we have
	 * not yet completed the fallocation, then (a) we want to keep track
	 * of this page in case we have to undo it, and (b) it may not be a
	 * good idea to continue anyway, once we're pushing into swap.  So
	 * reactivate the page, and let shmem_fallocate() quit when too many.
820 821
	 */
	if (!PageUptodate(page)) {
822 823 824 825 826 827 828 829 830 831 832 833 834 835
		if (inode->i_private) {
			struct shmem_falloc *shmem_falloc;
			spin_lock(&inode->i_lock);
			shmem_falloc = inode->i_private;
			if (shmem_falloc &&
			    index >= shmem_falloc->start &&
			    index < shmem_falloc->next)
				shmem_falloc->nr_unswapped++;
			else
				shmem_falloc = NULL;
			spin_unlock(&inode->i_lock);
			if (shmem_falloc)
				goto redirty;
		}
836 837 838 839 840
		clear_highpage(page);
		flush_dcache_page(page);
		SetPageUptodate(page);
	}

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

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

857
	if (add_to_swap_cache(page, swap, GFP_ATOMIC) == 0) {
H
Hugh Dickins 已提交
858
		swap_shmem_alloc(swap);
859 860 861 862 863
		shmem_delete_from_page_cache(page, swp_to_radix_entry(swap));

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

		mutex_unlock(&shmem_swaplist_mutex);
867
		BUG_ON(page_mapped(page));
H
Hugh Dickins 已提交
868
		swap_writepage(page, wbc);
L
Linus Torvalds 已提交
869 870 871
		return 0;
	}

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

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

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

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

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

static struct mempolicy *shmem_get_sbmpol(struct shmem_sb_info *sbinfo)
{
	struct mempolicy *mpol = NULL;
	if (sbinfo->mpol) {
		spin_lock(&sbinfo->stat_lock);	/* prevent replace/use races */
		mpol = sbinfo->mpol;
		mpol_get(mpol);
		spin_unlock(&sbinfo->stat_lock);
	}
	return mpol;
}
907 908
#endif /* CONFIG_TMPFS */

909 910
static struct page *shmem_swapin(swp_entry_t swap, gfp_t gfp,
			struct shmem_inode_info *info, pgoff_t index)
L
Linus Torvalds 已提交
911 912
{
	struct vm_area_struct pvma;
M
Mel Gorman 已提交
913
	struct page *page;
914

L
Linus Torvalds 已提交
915
	/* Create a pseudo vma that just contains the policy */
916
	pvma.vm_start = 0;
917 918
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
919
	pvma.vm_ops = NULL;
M
Mel Gorman 已提交
920 921 922 923 924 925 926 927
	pvma.vm_policy = mpol_shared_policy_lookup(&info->policy, index);

	page = swapin_readahead(swap, gfp, &pvma, 0);

	/* Drop reference taken by mpol_shared_policy_lookup() */
	mpol_cond_put(pvma.vm_policy);

	return page;
L
Linus Torvalds 已提交
928 929
}

930
static struct page *shmem_alloc_page(gfp_t gfp,
931
			struct shmem_inode_info *info, pgoff_t index)
L
Linus Torvalds 已提交
932 933
{
	struct vm_area_struct pvma;
M
Mel Gorman 已提交
934
	struct page *page;
L
Linus Torvalds 已提交
935

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

M
Mel Gorman 已提交
943 944 945 946 947 948
	page = alloc_page_vma(gfp, &pvma, 0);

	/* Drop reference taken by mpol_shared_policy_lookup() */
	mpol_cond_put(pvma.vm_policy);

	return page;
L
Linus Torvalds 已提交
949
}
950 951
#else /* !CONFIG_NUMA */
#ifdef CONFIG_TMPFS
952
static inline void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
953 954 955 956
{
}
#endif /* CONFIG_TMPFS */

957 958
static inline struct page *shmem_swapin(swp_entry_t swap, gfp_t gfp,
			struct shmem_inode_info *info, pgoff_t index)
L
Linus Torvalds 已提交
959
{
960
	return swapin_readahead(swap, gfp, NULL, 0);
L
Linus Torvalds 已提交
961 962
}

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

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

977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016
/*
 * When a page is moved from swapcache to shmem filecache (either by the
 * usual swapin of shmem_getpage_gfp(), or by the less common swapoff of
 * shmem_unuse_inode()), it may have been read in earlier from swap, in
 * ignorance of the mapping it belongs to.  If that mapping has special
 * constraints (like the gma500 GEM driver, which requires RAM below 4GB),
 * we may need to copy to a suitable page before moving to filecache.
 *
 * In a future release, this may well be extended to respect cpuset and
 * NUMA mempolicy, and applied also to anonymous pages in do_swap_page();
 * but for now it is a simple matter of zone.
 */
static bool shmem_should_replace_page(struct page *page, gfp_t gfp)
{
	return page_zonenum(page) > gfp_zone(gfp);
}

static int shmem_replace_page(struct page **pagep, gfp_t gfp,
				struct shmem_inode_info *info, pgoff_t index)
{
	struct page *oldpage, *newpage;
	struct address_space *swap_mapping;
	pgoff_t swap_index;
	int error;

	oldpage = *pagep;
	swap_index = page_private(oldpage);
	swap_mapping = page_mapping(oldpage);

	/*
	 * We have arrived here because our zones are constrained, so don't
	 * limit chance of success by further cpuset and node constraints.
	 */
	gfp &= ~GFP_CONSTRAINT_MASK;
	newpage = shmem_alloc_page(gfp, info, index);
	if (!newpage)
		return -ENOMEM;

	page_cache_get(newpage);
	copy_highpage(newpage, oldpage);
1017
	flush_dcache_page(newpage);
1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031

	__set_page_locked(newpage);
	SetPageUptodate(newpage);
	SetPageSwapBacked(newpage);
	set_page_private(newpage, swap_index);
	SetPageSwapCache(newpage);

	/*
	 * Our caller will very soon move newpage out of swapcache, but it's
	 * a nice clean interface for us to replace oldpage by newpage there.
	 */
	spin_lock_irq(&swap_mapping->tree_lock);
	error = shmem_radix_tree_replace(swap_mapping, swap_index, oldpage,
								   newpage);
1032 1033 1034 1035
	if (!error) {
		__inc_zone_page_state(newpage, NR_FILE_PAGES);
		__dec_zone_page_state(oldpage, NR_FILE_PAGES);
	}
1036 1037
	spin_unlock_irq(&swap_mapping->tree_lock);

1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
	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;
	}
1050 1051 1052 1053 1054 1055 1056

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

	unlock_page(oldpage);
	page_cache_release(oldpage);
	page_cache_release(oldpage);
1057
	return error;
1058 1059
}

L
Linus Torvalds 已提交
1060
/*
1061
 * shmem_getpage_gfp - find page in cache, or get from swap, or allocate
L
Linus Torvalds 已提交
1062 1063 1064 1065 1066
 *
 * 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
 */
1067
static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
1068
	struct page **pagep, enum sgp_type sgp, gfp_t gfp, int *fault_type)
L
Linus Torvalds 已提交
1069 1070
{
	struct address_space *mapping = inode->i_mapping;
1071
	struct shmem_inode_info *info;
L
Linus Torvalds 已提交
1072
	struct shmem_sb_info *sbinfo;
H
Hugh Dickins 已提交
1073
	struct page *page;
L
Linus Torvalds 已提交
1074 1075
	swp_entry_t swap;
	int error;
1076
	int once = 0;
1077
	int alloced = 0;
L
Linus Torvalds 已提交
1078

1079
	if (index > (MAX_LFS_FILESIZE >> PAGE_CACHE_SHIFT))
L
Linus Torvalds 已提交
1080 1081
		return -EFBIG;
repeat:
1082
	swap.val = 0;
1083
	page = find_lock_page(mapping, index);
1084 1085 1086 1087 1088
	if (radix_tree_exceptional_entry(page)) {
		swap = radix_to_swp_entry(page);
		page = NULL;
	}

1089
	if (sgp != SGP_WRITE && sgp != SGP_FALLOC &&
1090 1091 1092 1093 1094
	    ((loff_t)index << PAGE_CACHE_SHIFT) >= i_size_read(inode)) {
		error = -EINVAL;
		goto failed;
	}

1095 1096 1097 1098 1099 1100 1101 1102
	/* fallocated page? */
	if (page && !PageUptodate(page)) {
		if (sgp != SGP_READ)
			goto clear;
		unlock_page(page);
		page_cache_release(page);
		page = NULL;
	}
1103 1104 1105
	if (page || (sgp == SGP_READ && !swap.val)) {
		*pagep = page;
		return 0;
H
Hugh Dickins 已提交
1106 1107 1108
	}

	/*
1109 1110
	 * Fast cache lookup did not find it:
	 * bring it back from swap or allocate.
H
Hugh Dickins 已提交
1111
	 */
1112 1113
	info = SHMEM_I(inode);
	sbinfo = SHMEM_SB(inode->i_sb);
L
Linus Torvalds 已提交
1114 1115 1116

	if (swap.val) {
		/* Look it up and read it in.. */
H
Hugh Dickins 已提交
1117 1118
		page = lookup_swap_cache(swap);
		if (!page) {
L
Linus Torvalds 已提交
1119
			/* here we actually do the io */
1120 1121
			if (fault_type)
				*fault_type |= VM_FAULT_MAJOR;
1122
			page = shmem_swapin(swap, gfp, info, index);
H
Hugh Dickins 已提交
1123
			if (!page) {
1124 1125
				error = -ENOMEM;
				goto failed;
L
Linus Torvalds 已提交
1126 1127 1128 1129
			}
		}

		/* We have to do this with page locked to prevent races */
1130
		lock_page(page);
1131
		if (!PageSwapCache(page) || page_private(page) != swap.val ||
1132
		    !shmem_confirm_swap(mapping, index, swap)) {
1133
			error = -EEXIST;	/* try again */
1134
			goto unlock;
1135
		}
H
Hugh Dickins 已提交
1136
		if (!PageUptodate(page)) {
L
Linus Torvalds 已提交
1137
			error = -EIO;
1138
			goto failed;
L
Linus Torvalds 已提交
1139
		}
1140 1141
		wait_on_page_writeback(page);

1142 1143 1144 1145
		if (shmem_should_replace_page(page, gfp)) {
			error = shmem_replace_page(&page, gfp, info, index);
			if (error)
				goto failed;
L
Linus Torvalds 已提交
1146
		}
H
Hugh Dickins 已提交
1147

1148 1149
		error = mem_cgroup_cache_charge(page, current->mm,
						gfp & GFP_RECLAIM_MASK);
1150
		if (!error) {
1151 1152
			error = shmem_add_to_page_cache(page, mapping, index,
						gfp, swp_to_radix_entry(swap));
1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166
			/*
			 * We already confirmed swap under page lock, and make
			 * no memory allocation here, so usually no possibility
			 * of error; but free_swap_and_cache() only trylocks a
			 * page, so it is just possible that the entry has been
			 * truncated or holepunched since swap was confirmed.
			 * shmem_undo_range() will have done some of the
			 * unaccounting, now delete_from_swap_cache() will do
			 * the rest (including mem_cgroup_uncharge_swapcache).
			 * Reset swap.val? No, leave it so "failed" goes back to
			 * "repeat": reading a hole and writing should succeed.
			 */
			if (error)
				delete_from_swap_cache(page);
1167
		}
1168 1169 1170 1171
		if (error)
			goto failed;

		spin_lock(&info->lock);
1172
		info->swapped--;
1173
		shmem_recalc_inode(inode);
H
Hugh Dickins 已提交
1174
		spin_unlock(&info->lock);
1175 1176

		delete_from_swap_cache(page);
H
Hugh Dickins 已提交
1177 1178 1179
		set_page_dirty(page);
		swap_free(swap);

1180 1181 1182 1183
	} else {
		if (shmem_acct_block(info->flags)) {
			error = -ENOSPC;
			goto failed;
L
Linus Torvalds 已提交
1184
		}
1185
		if (sbinfo->max_blocks) {
1186
			if (percpu_counter_compare(&sbinfo->used_blocks,
1187 1188 1189 1190
						sbinfo->max_blocks) >= 0) {
				error = -ENOSPC;
				goto unacct;
			}
1191
			percpu_counter_inc(&sbinfo->used_blocks);
1192
		}
L
Linus Torvalds 已提交
1193

1194 1195 1196 1197
		page = shmem_alloc_page(gfp, info, index);
		if (!page) {
			error = -ENOMEM;
			goto decused;
L
Linus Torvalds 已提交
1198 1199
		}

1200 1201
		SetPageSwapBacked(page);
		__set_page_locked(page);
1202 1203
		error = mem_cgroup_cache_charge(page, current->mm,
						gfp & GFP_RECLAIM_MASK);
1204 1205
		if (error)
			goto decused;
1206
		error = radix_tree_maybe_preload(gfp & GFP_RECLAIM_MASK);
1207 1208 1209 1210 1211 1212 1213 1214 1215
		if (!error) {
			error = shmem_add_to_page_cache(page, mapping, index,
							gfp, NULL);
			radix_tree_preload_end();
		}
		if (error) {
			mem_cgroup_uncharge_cache_page(page);
			goto decused;
		}
1216 1217 1218
		lru_cache_add_anon(page);

		spin_lock(&info->lock);
L
Linus Torvalds 已提交
1219
		info->alloced++;
1220 1221
		inode->i_blocks += BLOCKS_PER_PAGE;
		shmem_recalc_inode(inode);
L
Linus Torvalds 已提交
1222
		spin_unlock(&info->lock);
1223
		alloced = true;
1224

1225
		/*
1226 1227 1228 1229 1230 1231 1232 1233 1234
		 * Let SGP_FALLOC use the SGP_WRITE optimization on a new page.
		 */
		if (sgp == SGP_FALLOC)
			sgp = SGP_WRITE;
clear:
		/*
		 * Let SGP_WRITE caller clear ends if write does not fill page;
		 * but SGP_FALLOC on a page fallocated earlier must initialize
		 * it now, lest undo on failure cancel our earlier guarantee.
1235 1236 1237 1238 1239 1240
		 */
		if (sgp != SGP_WRITE) {
			clear_highpage(page);
			flush_dcache_page(page);
			SetPageUptodate(page);
		}
1241
		if (sgp == SGP_DIRTY)
H
Hugh Dickins 已提交
1242
			set_page_dirty(page);
L
Linus Torvalds 已提交
1243
	}
1244

1245
	/* Perhaps the file has been truncated since we checked */
1246
	if (sgp != SGP_WRITE && sgp != SGP_FALLOC &&
1247 1248
	    ((loff_t)index << PAGE_CACHE_SHIFT) >= i_size_read(inode)) {
		error = -EINVAL;
1249 1250 1251 1252
		if (alloced)
			goto trunc;
		else
			goto failed;
H
Hugh Dickins 已提交
1253
	}
1254 1255
	*pagep = page;
	return 0;
L
Linus Torvalds 已提交
1256

1257
	/*
1258
	 * Error recovery.
1259
	 */
1260
trunc:
1261
	info = SHMEM_I(inode);
1262 1263 1264 1265 1266
	ClearPageDirty(page);
	delete_from_page_cache(page);
	spin_lock(&info->lock);
	info->alloced--;
	inode->i_blocks -= BLOCKS_PER_PAGE;
1267
	spin_unlock(&info->lock);
1268
decused:
1269
	sbinfo = SHMEM_SB(inode->i_sb);
1270 1271 1272 1273 1274
	if (sbinfo->max_blocks)
		percpu_counter_add(&sbinfo->used_blocks, -1);
unacct:
	shmem_unacct_blocks(info->flags, 1);
failed:
1275 1276 1277 1278
	if (swap.val && error != -EINVAL &&
	    !shmem_confirm_swap(mapping, index, swap))
		error = -EEXIST;
unlock:
H
Hugh Dickins 已提交
1279
	if (page) {
1280
		unlock_page(page);
H
Hugh Dickins 已提交
1281
		page_cache_release(page);
1282 1283 1284 1285 1286 1287
	}
	if (error == -ENOSPC && !once++) {
		info = SHMEM_I(inode);
		spin_lock(&info->lock);
		shmem_recalc_inode(inode);
		spin_unlock(&info->lock);
H
Hugh Dickins 已提交
1288
		goto repeat;
1289
	}
1290
	if (error == -EEXIST)	/* from above or from radix_tree_insert */
1291 1292
		goto repeat;
	return error;
L
Linus Torvalds 已提交
1293 1294
}

N
Nick Piggin 已提交
1295
static int shmem_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
L
Linus Torvalds 已提交
1296
{
A
Al Viro 已提交
1297
	struct inode *inode = file_inode(vma->vm_file);
L
Linus Torvalds 已提交
1298
	int error;
1299
	int ret = VM_FAULT_LOCKED;
L
Linus Torvalds 已提交
1300

1301
	error = shmem_getpage(inode, vmf->pgoff, &vmf->page, SGP_CACHE, &ret);
N
Nick Piggin 已提交
1302 1303
	if (error)
		return ((error == -ENOMEM) ? VM_FAULT_OOM : VM_FAULT_SIGBUS);
1304

1305 1306 1307 1308
	if (ret & VM_FAULT_MAJOR) {
		count_vm_event(PGMAJFAULT);
		mem_cgroup_count_vm_event(vma->vm_mm, PGMAJFAULT);
	}
1309
	return ret;
L
Linus Torvalds 已提交
1310 1311 1312
}

#ifdef CONFIG_NUMA
1313
static int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *mpol)
L
Linus Torvalds 已提交
1314
{
A
Al Viro 已提交
1315
	struct inode *inode = file_inode(vma->vm_file);
1316
	return mpol_set_shared_policy(&SHMEM_I(inode)->policy, vma, mpol);
L
Linus Torvalds 已提交
1317 1318
}

A
Adrian Bunk 已提交
1319 1320
static struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
					  unsigned long addr)
L
Linus Torvalds 已提交
1321
{
A
Al Viro 已提交
1322
	struct inode *inode = file_inode(vma->vm_file);
1323
	pgoff_t index;
L
Linus Torvalds 已提交
1324

1325 1326
	index = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
	return mpol_shared_policy_lookup(&SHMEM_I(inode)->policy, index);
L
Linus Torvalds 已提交
1327 1328 1329 1330 1331
}
#endif

int shmem_lock(struct file *file, int lock, struct user_struct *user)
{
A
Al Viro 已提交
1332
	struct inode *inode = file_inode(file);
L
Linus Torvalds 已提交
1333 1334 1335 1336 1337 1338 1339 1340
	struct shmem_inode_info *info = SHMEM_I(inode);
	int retval = -ENOMEM;

	spin_lock(&info->lock);
	if (lock && !(info->flags & VM_LOCKED)) {
		if (!user_shm_lock(inode->i_size, user))
			goto out_nomem;
		info->flags |= VM_LOCKED;
1341
		mapping_set_unevictable(file->f_mapping);
L
Linus Torvalds 已提交
1342 1343 1344 1345
	}
	if (!lock && (info->flags & VM_LOCKED) && user) {
		user_shm_unlock(inode->i_size, user);
		info->flags &= ~VM_LOCKED;
1346
		mapping_clear_unevictable(file->f_mapping);
L
Linus Torvalds 已提交
1347 1348
	}
	retval = 0;
1349

L
Linus Torvalds 已提交
1350 1351 1352 1353 1354
out_nomem:
	spin_unlock(&info->lock);
	return retval;
}

1355
static int shmem_mmap(struct file *file, struct vm_area_struct *vma)
L
Linus Torvalds 已提交
1356 1357 1358 1359 1360 1361
{
	file_accessed(file);
	vma->vm_ops = &shmem_vm_ops;
	return 0;
}

1362
static struct inode *shmem_get_inode(struct super_block *sb, const struct inode *dir,
1363
				     umode_t mode, dev_t dev, unsigned long flags)
L
Linus Torvalds 已提交
1364 1365 1366 1367 1368
{
	struct inode *inode;
	struct shmem_inode_info *info;
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);

1369 1370
	if (shmem_reserve_inode(sb))
		return NULL;
L
Linus Torvalds 已提交
1371 1372 1373

	inode = new_inode(sb);
	if (inode) {
1374
		inode->i_ino = get_next_ino();
1375
		inode_init_owner(inode, dir, mode);
L
Linus Torvalds 已提交
1376 1377 1378
		inode->i_blocks = 0;
		inode->i_mapping->backing_dev_info = &shmem_backing_dev_info;
		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1379
		inode->i_generation = get_seconds();
L
Linus Torvalds 已提交
1380 1381 1382
		info = SHMEM_I(inode);
		memset(info, 0, (char *)inode - (char *)info);
		spin_lock_init(&info->lock);
1383
		info->flags = flags & VM_NORESERVE;
L
Linus Torvalds 已提交
1384
		INIT_LIST_HEAD(&info->swaplist);
1385
		simple_xattrs_init(&info->xattrs);
1386
		cache_no_acl(inode);
L
Linus Torvalds 已提交
1387 1388 1389

		switch (mode & S_IFMT) {
		default:
1390
			inode->i_op = &shmem_special_inode_operations;
L
Linus Torvalds 已提交
1391 1392 1393
			init_special_inode(inode, mode, dev);
			break;
		case S_IFREG:
1394
			inode->i_mapping->a_ops = &shmem_aops;
L
Linus Torvalds 已提交
1395 1396
			inode->i_op = &shmem_inode_operations;
			inode->i_fop = &shmem_file_operations;
1397 1398
			mpol_shared_policy_init(&info->policy,
						 shmem_get_sbmpol(sbinfo));
L
Linus Torvalds 已提交
1399 1400
			break;
		case S_IFDIR:
1401
			inc_nlink(inode);
L
Linus Torvalds 已提交
1402 1403 1404 1405 1406 1407 1408 1409 1410 1411
			/* Some things misbehave if size == 0 on a directory */
			inode->i_size = 2 * BOGO_DIRENT_SIZE;
			inode->i_op = &shmem_dir_inode_operations;
			inode->i_fop = &simple_dir_operations;
			break;
		case S_IFLNK:
			/*
			 * Must not load anything in the rbtree,
			 * mpol_free_shared_policy will not be called.
			 */
1412
			mpol_shared_policy_init(&info->policy, NULL);
L
Linus Torvalds 已提交
1413 1414
			break;
		}
1415 1416
	} else
		shmem_free_inode(sb);
L
Linus Torvalds 已提交
1417 1418 1419 1420
	return inode;
}

#ifdef CONFIG_TMPFS
1421
static const struct inode_operations shmem_symlink_inode_operations;
1422
static const struct inode_operations shmem_short_symlink_operations;
L
Linus Torvalds 已提交
1423

1424 1425 1426 1427 1428 1429
#ifdef CONFIG_TMPFS_XATTR
static int shmem_initxattrs(struct inode *, const struct xattr *, void *);
#else
#define shmem_initxattrs NULL
#endif

L
Linus Torvalds 已提交
1430
static int
N
Nick Piggin 已提交
1431 1432 1433
shmem_write_begin(struct file *file, struct address_space *mapping,
			loff_t pos, unsigned len, unsigned flags,
			struct page **pagep, void **fsdata)
L
Linus Torvalds 已提交
1434
{
N
Nick Piggin 已提交
1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446
	struct inode *inode = mapping->host;
	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
	return shmem_getpage(inode, index, pagep, SGP_WRITE, NULL);
}

static int
shmem_write_end(struct file *file, struct address_space *mapping,
			loff_t pos, unsigned len, unsigned copied,
			struct page *page, void *fsdata)
{
	struct inode *inode = mapping->host;

H
Hugh Dickins 已提交
1447 1448 1449
	if (pos + copied > inode->i_size)
		i_size_write(inode, pos + copied);

1450 1451 1452 1453 1454 1455 1456 1457
	if (!PageUptodate(page)) {
		if (copied < PAGE_CACHE_SIZE) {
			unsigned from = pos & (PAGE_CACHE_SIZE - 1);
			zero_user_segments(page, 0, from,
					from + copied, PAGE_CACHE_SIZE);
		}
		SetPageUptodate(page);
	}
N
Nick Piggin 已提交
1458
	set_page_dirty(page);
1459
	unlock_page(page);
N
Nick Piggin 已提交
1460 1461 1462
	page_cache_release(page);

	return copied;
L
Linus Torvalds 已提交
1463 1464
}

1465 1466
static ssize_t shmem_file_aio_read(struct kiocb *iocb,
		const struct iovec *iov, unsigned long nr_segs, loff_t pos)
L
Linus Torvalds 已提交
1467
{
1468 1469
	struct file *file = iocb->ki_filp;
	struct inode *inode = file_inode(file);
L
Linus Torvalds 已提交
1470
	struct address_space *mapping = inode->i_mapping;
1471 1472
	pgoff_t index;
	unsigned long offset;
1473
	enum sgp_type sgp = SGP_READ;
1474 1475 1476 1477 1478 1479 1480 1481 1482 1483
	int error;
	ssize_t retval;
	size_t count;
	loff_t *ppos = &iocb->ki_pos;
	struct iov_iter iter;

	retval = generic_segment_checks(iov, &nr_segs, &count, VERIFY_WRITE);
	if (retval)
		return retval;
	iov_iter_init(&iter, iov, nr_segs, count, 0);
1484 1485 1486 1487 1488 1489 1490 1491

	/*
	 * Might this read be for a stacking filesystem?  Then when reading
	 * holes of a sparse file, we actually need to allocate those pages,
	 * and even mark them dirty, so it cannot exceed the max_blocks limit.
	 */
	if (segment_eq(get_fs(), KERNEL_DS))
		sgp = SGP_DIRTY;
L
Linus Torvalds 已提交
1492 1493 1494 1495 1496 1497

	index = *ppos >> PAGE_CACHE_SHIFT;
	offset = *ppos & ~PAGE_CACHE_MASK;

	for (;;) {
		struct page *page = NULL;
1498 1499
		pgoff_t end_index;
		unsigned long nr, ret;
L
Linus Torvalds 已提交
1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510
		loff_t i_size = i_size_read(inode);

		end_index = i_size >> PAGE_CACHE_SHIFT;
		if (index > end_index)
			break;
		if (index == end_index) {
			nr = i_size & ~PAGE_CACHE_MASK;
			if (nr <= offset)
				break;
		}

1511 1512 1513 1514
		error = shmem_getpage(inode, index, &page, sgp, NULL);
		if (error) {
			if (error == -EINVAL)
				error = 0;
L
Linus Torvalds 已提交
1515 1516
			break;
		}
H
Hugh Dickins 已提交
1517 1518
		if (page)
			unlock_page(page);
L
Linus Torvalds 已提交
1519 1520 1521

		/*
		 * We must evaluate after, since reads (unlike writes)
1522
		 * are called without i_mutex protection against truncate
L
Linus Torvalds 已提交
1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
		 */
		nr = PAGE_CACHE_SIZE;
		i_size = i_size_read(inode);
		end_index = i_size >> PAGE_CACHE_SHIFT;
		if (index == end_index) {
			nr = i_size & ~PAGE_CACHE_MASK;
			if (nr <= offset) {
				if (page)
					page_cache_release(page);
				break;
			}
		}
		nr -= offset;

		if (page) {
			/*
			 * If users can be writing to this page using arbitrary
			 * virtual addresses, take care about potential aliasing
			 * before reading the page on the kernel side.
			 */
			if (mapping_writably_mapped(mapping))
				flush_dcache_page(page);
			/*
			 * Mark the page accessed if we read the beginning.
			 */
			if (!offset)
				mark_page_accessed(page);
N
Nick Piggin 已提交
1550
		} else {
L
Linus Torvalds 已提交
1551
			page = ZERO_PAGE(0);
N
Nick Piggin 已提交
1552 1553
			page_cache_get(page);
		}
L
Linus Torvalds 已提交
1554 1555 1556 1557 1558

		/*
		 * Ok, we have the page, and it's up-to-date, so
		 * now we can copy it to user space...
		 */
1559 1560
		ret = copy_page_to_iter(page, offset, nr, &iter);
		retval += ret;
L
Linus Torvalds 已提交
1561 1562 1563 1564 1565
		offset += ret;
		index += offset >> PAGE_CACHE_SHIFT;
		offset &= ~PAGE_CACHE_MASK;

		page_cache_release(page);
1566
		if (!iov_iter_count(&iter))
L
Linus Torvalds 已提交
1567
			break;
1568 1569 1570 1571
		if (ret < nr) {
			error = -EFAULT;
			break;
		}
L
Linus Torvalds 已提交
1572 1573 1574 1575
		cond_resched();
	}

	*ppos = ((loff_t) index << PAGE_CACHE_SHIFT) + offset;
1576 1577
	file_accessed(file);
	return retval ? retval : error;
L
Linus Torvalds 已提交
1578 1579
}

1580 1581 1582 1583 1584
static ssize_t shmem_file_splice_read(struct file *in, loff_t *ppos,
				struct pipe_inode_info *pipe, size_t len,
				unsigned int flags)
{
	struct address_space *mapping = in->f_mapping;
1585
	struct inode *inode = mapping->host;
1586 1587 1588 1589 1590 1591 1592 1593 1594 1595
	unsigned int loff, nr_pages, req_pages;
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
	struct page *page;
	pgoff_t index, end_index;
	loff_t isize, left;
	int error, page_nr;
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
1596
		.nr_pages_max = PIPE_DEF_BUFFERS,
1597 1598 1599 1600 1601
		.flags = flags,
		.ops = &page_cache_pipe_buf_ops,
		.spd_release = spd_release_page,
	};

1602
	isize = i_size_read(inode);
1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615
	if (unlikely(*ppos >= isize))
		return 0;

	left = isize - *ppos;
	if (unlikely(left < len))
		len = left;

	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

	index = *ppos >> PAGE_CACHE_SHIFT;
	loff = *ppos & ~PAGE_CACHE_MASK;
	req_pages = (len + loff + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1616
	nr_pages = min(req_pages, spd.nr_pages_max);
1617 1618 1619 1620 1621 1622

	spd.nr_pages = find_get_pages_contig(mapping, index,
						nr_pages, spd.pages);
	index += spd.nr_pages;
	error = 0;

1623 1624 1625 1626 1627
	while (spd.nr_pages < nr_pages) {
		error = shmem_getpage(inode, index, &page, SGP_CACHE, NULL);
		if (error)
			break;
		unlock_page(page);
1628 1629 1630 1631 1632 1633 1634
		spd.pages[spd.nr_pages++] = page;
		index++;
	}

	index = *ppos >> PAGE_CACHE_SHIFT;
	nr_pages = spd.nr_pages;
	spd.nr_pages = 0;
1635

1636 1637 1638 1639 1640 1641 1642 1643 1644
	for (page_nr = 0; page_nr < nr_pages; page_nr++) {
		unsigned int this_len;

		if (!len)
			break;

		this_len = min_t(unsigned long, len, PAGE_CACHE_SIZE - loff);
		page = spd.pages[page_nr];

1645 1646 1647 1648
		if (!PageUptodate(page) || page->mapping != mapping) {
			error = shmem_getpage(inode, index, &page,
							SGP_CACHE, NULL);
			if (error)
1649
				break;
1650 1651 1652
			unlock_page(page);
			page_cache_release(spd.pages[page_nr]);
			spd.pages[page_nr] = page;
1653
		}
1654 1655

		isize = i_size_read(inode);
1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684
		end_index = (isize - 1) >> PAGE_CACHE_SHIFT;
		if (unlikely(!isize || index > end_index))
			break;

		if (end_index == index) {
			unsigned int plen;

			plen = ((isize - 1) & ~PAGE_CACHE_MASK) + 1;
			if (plen <= loff)
				break;

			this_len = min(this_len, plen - loff);
			len = this_len;
		}

		spd.partial[page_nr].offset = loff;
		spd.partial[page_nr].len = this_len;
		len -= this_len;
		loff = 0;
		spd.nr_pages++;
		index++;
	}

	while (page_nr < nr_pages)
		page_cache_release(spd.pages[page_nr++]);

	if (spd.nr_pages)
		error = splice_to_pipe(pipe, &spd);

1685
	splice_shrink_spd(&spd);
1686 1687 1688 1689 1690 1691 1692 1693

	if (error > 0) {
		*ppos += error;
		file_accessed(in);
	}
	return error;
}

1694 1695 1696 1697
/*
 * llseek SEEK_DATA or SEEK_HOLE through the radix_tree.
 */
static pgoff_t shmem_seek_hole_data(struct address_space *mapping,
1698
				    pgoff_t index, pgoff_t end, int whence)
1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711
{
	struct page *page;
	struct pagevec pvec;
	pgoff_t indices[PAGEVEC_SIZE];
	bool done = false;
	int i;

	pagevec_init(&pvec, 0);
	pvec.nr = 1;		/* start small: we may be there already */
	while (!done) {
		pvec.nr = shmem_find_get_pages_and_swap(mapping, index,
					pvec.nr, pvec.pages, indices);
		if (!pvec.nr) {
1712
			if (whence == SEEK_DATA)
1713 1714 1715 1716 1717
				index = end;
			break;
		}
		for (i = 0; i < pvec.nr; i++, index++) {
			if (index < indices[i]) {
1718
				if (whence == SEEK_HOLE) {
1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729
					done = true;
					break;
				}
				index = indices[i];
			}
			page = pvec.pages[i];
			if (page && !radix_tree_exceptional_entry(page)) {
				if (!PageUptodate(page))
					page = NULL;
			}
			if (index >= end ||
1730 1731
			    (page && whence == SEEK_DATA) ||
			    (!page && whence == SEEK_HOLE)) {
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743
				done = true;
				break;
			}
		}
		shmem_deswap_pagevec(&pvec);
		pagevec_release(&pvec);
		pvec.nr = PAGEVEC_SIZE;
		cond_resched();
	}
	return index;
}

1744
static loff_t shmem_file_llseek(struct file *file, loff_t offset, int whence)
1745 1746 1747 1748 1749 1750
{
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
	pgoff_t start, end;
	loff_t new_offset;

1751 1752
	if (whence != SEEK_DATA && whence != SEEK_HOLE)
		return generic_file_llseek_size(file, offset, whence,
1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763
					MAX_LFS_FILESIZE, i_size_read(inode));
	mutex_lock(&inode->i_mutex);
	/* We're holding i_mutex so we can access i_size directly */

	if (offset < 0)
		offset = -EINVAL;
	else if (offset >= inode->i_size)
		offset = -ENXIO;
	else {
		start = offset >> PAGE_CACHE_SHIFT;
		end = (inode->i_size + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1764
		new_offset = shmem_seek_hole_data(mapping, start, end, whence);
1765 1766 1767 1768
		new_offset <<= PAGE_CACHE_SHIFT;
		if (new_offset > offset) {
			if (new_offset < inode->i_size)
				offset = new_offset;
1769
			else if (whence == SEEK_DATA)
1770 1771 1772 1773 1774 1775
				offset = -ENXIO;
			else
				offset = inode->i_size;
		}
	}

1776 1777
	if (offset >= 0)
		offset = vfs_setpos(file, offset, MAX_LFS_FILESIZE);
1778 1779 1780 1781
	mutex_unlock(&inode->i_mutex);
	return offset;
}

1782 1783 1784
static long shmem_fallocate(struct file *file, int mode, loff_t offset,
							 loff_t len)
{
A
Al Viro 已提交
1785
	struct inode *inode = file_inode(file);
1786
	struct shmem_sb_info *sbinfo = SHMEM_SB(inode->i_sb);
1787
	struct shmem_falloc shmem_falloc;
1788 1789
	pgoff_t start, index, end;
	int error;
1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803

	mutex_lock(&inode->i_mutex);

	if (mode & FALLOC_FL_PUNCH_HOLE) {
		struct address_space *mapping = file->f_mapping;
		loff_t unmap_start = round_up(offset, PAGE_SIZE);
		loff_t unmap_end = round_down(offset + len, PAGE_SIZE) - 1;

		if ((u64)unmap_end > (u64)unmap_start)
			unmap_mapping_range(mapping, unmap_start,
					    1 + unmap_end - unmap_start, 0);
		shmem_truncate_range(inode, offset, offset + len - 1);
		/* No need to unmap again: hole-punching leaves COWed pages */
		error = 0;
1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817
		goto out;
	}

	/* We need to check rlimit even when FALLOC_FL_KEEP_SIZE */
	error = inode_newsize_ok(inode, offset + len);
	if (error)
		goto out;

	start = offset >> PAGE_CACHE_SHIFT;
	end = (offset + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
	/* Try to avoid a swapstorm if len is impossible to satisfy */
	if (sbinfo->max_blocks && end - start > sbinfo->max_blocks) {
		error = -ENOSPC;
		goto out;
1818 1819
	}

1820 1821 1822 1823 1824 1825 1826 1827
	shmem_falloc.start = start;
	shmem_falloc.next  = start;
	shmem_falloc.nr_falloced = 0;
	shmem_falloc.nr_unswapped = 0;
	spin_lock(&inode->i_lock);
	inode->i_private = &shmem_falloc;
	spin_unlock(&inode->i_lock);

1828 1829 1830 1831 1832 1833 1834 1835 1836
	for (index = start; index < end; index++) {
		struct page *page;

		/*
		 * Good, the fallocate(2) manpage permits EINTR: we may have
		 * been interrupted because we are using up too much memory.
		 */
		if (signal_pending(current))
			error = -EINTR;
1837 1838
		else if (shmem_falloc.nr_unswapped > shmem_falloc.nr_falloced)
			error = -ENOMEM;
1839
		else
1840
			error = shmem_getpage(inode, index, &page, SGP_FALLOC,
1841 1842
									NULL);
		if (error) {
1843 1844 1845 1846
			/* Remove the !PageUptodate pages we added */
			shmem_undo_range(inode,
				(loff_t)start << PAGE_CACHE_SHIFT,
				(loff_t)index << PAGE_CACHE_SHIFT, true);
1847
			goto undone;
1848 1849
		}

1850 1851 1852 1853 1854 1855 1856 1857
		/*
		 * Inform shmem_writepage() how far we have reached.
		 * No need for lock or barrier: we have the page lock.
		 */
		shmem_falloc.next++;
		if (!PageUptodate(page))
			shmem_falloc.nr_falloced++;

1858
		/*
1859 1860 1861
		 * If !PageUptodate, leave it that way so that freeable pages
		 * can be recognized if we need to rollback on error later.
		 * But set_page_dirty so that memory pressure will swap rather
1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873
		 * than free the pages we are allocating (and SGP_CACHE pages
		 * might still be clean: we now need to mark those dirty too).
		 */
		set_page_dirty(page);
		unlock_page(page);
		page_cache_release(page);
		cond_resched();
	}

	if (!(mode & FALLOC_FL_KEEP_SIZE) && offset + len > inode->i_size)
		i_size_write(inode, offset + len);
	inode->i_ctime = CURRENT_TIME;
1874 1875 1876 1877
undone:
	spin_lock(&inode->i_lock);
	inode->i_private = NULL;
	spin_unlock(&inode->i_lock);
1878
out:
1879 1880 1881 1882
	mutex_unlock(&inode->i_mutex);
	return error;
}

1883
static int shmem_statfs(struct dentry *dentry, struct kstatfs *buf)
L
Linus Torvalds 已提交
1884
{
1885
	struct shmem_sb_info *sbinfo = SHMEM_SB(dentry->d_sb);
L
Linus Torvalds 已提交
1886 1887 1888 1889

	buf->f_type = TMPFS_MAGIC;
	buf->f_bsize = PAGE_CACHE_SIZE;
	buf->f_namelen = NAME_MAX;
1890
	if (sbinfo->max_blocks) {
L
Linus Torvalds 已提交
1891
		buf->f_blocks = sbinfo->max_blocks;
1892 1893 1894
		buf->f_bavail =
		buf->f_bfree  = sbinfo->max_blocks -
				percpu_counter_sum(&sbinfo->used_blocks);
1895 1896
	}
	if (sbinfo->max_inodes) {
L
Linus Torvalds 已提交
1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907
		buf->f_files = sbinfo->max_inodes;
		buf->f_ffree = sbinfo->free_inodes;
	}
	/* else leave those fields 0 like simple_statfs */
	return 0;
}

/*
 * File creation. Allocate an inode, and we're done..
 */
static int
A
Al Viro 已提交
1908
shmem_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev)
L
Linus Torvalds 已提交
1909
{
1910
	struct inode *inode;
L
Linus Torvalds 已提交
1911 1912
	int error = -ENOSPC;

1913
	inode = shmem_get_inode(dir->i_sb, dir, mode, dev, VM_NORESERVE);
L
Linus Torvalds 已提交
1914
	if (inode) {
C
Christoph Hellwig 已提交
1915 1916 1917
		error = simple_acl_create(dir, inode);
		if (error)
			goto out_iput;
1918
		error = security_inode_init_security(inode, dir,
1919
						     &dentry->d_name,
1920
						     shmem_initxattrs, NULL);
C
Christoph Hellwig 已提交
1921 1922
		if (error && error != -EOPNOTSUPP)
			goto out_iput;
1923

A
Al Viro 已提交
1924
		error = 0;
L
Linus Torvalds 已提交
1925 1926 1927 1928 1929 1930
		dir->i_size += BOGO_DIRENT_SIZE;
		dir->i_ctime = dir->i_mtime = CURRENT_TIME;
		d_instantiate(dentry, inode);
		dget(dentry); /* Extra count - pin the dentry in core */
	}
	return error;
C
Christoph Hellwig 已提交
1931 1932 1933
out_iput:
	iput(inode);
	return error;
L
Linus Torvalds 已提交
1934 1935
}

1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946
static int
shmem_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
{
	struct inode *inode;
	int error = -ENOSPC;

	inode = shmem_get_inode(dir->i_sb, dir, mode, 0, VM_NORESERVE);
	if (inode) {
		error = security_inode_init_security(inode, dir,
						     NULL,
						     shmem_initxattrs, NULL);
C
Christoph Hellwig 已提交
1947 1948 1949 1950 1951
		if (error && error != -EOPNOTSUPP)
			goto out_iput;
		error = simple_acl_create(dir, inode);
		if (error)
			goto out_iput;
1952 1953 1954
		d_tmpfile(dentry, inode);
	}
	return error;
C
Christoph Hellwig 已提交
1955 1956 1957
out_iput:
	iput(inode);
	return error;
1958 1959
}

1960
static int shmem_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
1961 1962 1963 1964 1965
{
	int error;

	if ((error = shmem_mknod(dir, dentry, mode | S_IFDIR, 0)))
		return error;
1966
	inc_nlink(dir);
L
Linus Torvalds 已提交
1967 1968 1969
	return 0;
}

A
Al Viro 已提交
1970
static int shmem_create(struct inode *dir, struct dentry *dentry, umode_t mode,
A
Al Viro 已提交
1971
		bool excl)
L
Linus Torvalds 已提交
1972 1973 1974 1975 1976 1977 1978 1979 1980 1981
{
	return shmem_mknod(dir, dentry, mode | S_IFREG, 0);
}

/*
 * Link a file..
 */
static int shmem_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
{
	struct inode *inode = old_dentry->d_inode;
1982
	int ret;
L
Linus Torvalds 已提交
1983 1984 1985 1986 1987 1988

	/*
	 * No ordinary (disk based) filesystem counts links as inodes;
	 * but each new link needs a new dentry, pinning lowmem, and
	 * tmpfs dentries cannot be pruned until they are unlinked.
	 */
1989 1990 1991
	ret = shmem_reserve_inode(inode->i_sb);
	if (ret)
		goto out;
L
Linus Torvalds 已提交
1992 1993 1994

	dir->i_size += BOGO_DIRENT_SIZE;
	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
1995
	inc_nlink(inode);
A
Al Viro 已提交
1996
	ihold(inode);	/* New dentry reference */
L
Linus Torvalds 已提交
1997 1998
	dget(dentry);		/* Extra pinning count for the created dentry */
	d_instantiate(dentry, inode);
1999 2000
out:
	return ret;
L
Linus Torvalds 已提交
2001 2002 2003 2004 2005 2006
}

static int shmem_unlink(struct inode *dir, struct dentry *dentry)
{
	struct inode *inode = dentry->d_inode;

2007 2008
	if (inode->i_nlink > 1 && !S_ISDIR(inode->i_mode))
		shmem_free_inode(inode->i_sb);
L
Linus Torvalds 已提交
2009 2010 2011

	dir->i_size -= BOGO_DIRENT_SIZE;
	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
2012
	drop_nlink(inode);
L
Linus Torvalds 已提交
2013 2014 2015 2016 2017 2018 2019 2020 2021
	dput(dentry);	/* Undo the count from "create" - this does all the work */
	return 0;
}

static int shmem_rmdir(struct inode *dir, struct dentry *dentry)
{
	if (!simple_empty(dentry))
		return -ENOTEMPTY;

2022 2023
	drop_nlink(dentry->d_inode);
	drop_nlink(dir);
L
Linus Torvalds 已提交
2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043
	return shmem_unlink(dir, dentry);
}

/*
 * The VFS layer already does all the dentry stuff for rename,
 * we just have to decrement the usage count for the target if
 * it exists so that the VFS layer correctly free's it when it
 * gets overwritten.
 */
static int shmem_rename(struct inode *old_dir, struct dentry *old_dentry, struct inode *new_dir, struct dentry *new_dentry)
{
	struct inode *inode = old_dentry->d_inode;
	int they_are_dirs = S_ISDIR(inode->i_mode);

	if (!simple_empty(new_dentry))
		return -ENOTEMPTY;

	if (new_dentry->d_inode) {
		(void) shmem_unlink(new_dir, new_dentry);
		if (they_are_dirs)
2044
			drop_nlink(old_dir);
L
Linus Torvalds 已提交
2045
	} else if (they_are_dirs) {
2046
		drop_nlink(old_dir);
2047
		inc_nlink(new_dir);
L
Linus Torvalds 已提交
2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062
	}

	old_dir->i_size -= BOGO_DIRENT_SIZE;
	new_dir->i_size += BOGO_DIRENT_SIZE;
	old_dir->i_ctime = old_dir->i_mtime =
	new_dir->i_ctime = new_dir->i_mtime =
	inode->i_ctime = CURRENT_TIME;
	return 0;
}

static int shmem_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
{
	int error;
	int len;
	struct inode *inode;
H
Hugh Dickins 已提交
2063
	struct page *page;
L
Linus Torvalds 已提交
2064 2065 2066 2067 2068 2069 2070
	char *kaddr;
	struct shmem_inode_info *info;

	len = strlen(symname) + 1;
	if (len > PAGE_CACHE_SIZE)
		return -ENAMETOOLONG;

2071
	inode = shmem_get_inode(dir->i_sb, dir, S_IFLNK|S_IRWXUGO, 0, VM_NORESERVE);
L
Linus Torvalds 已提交
2072 2073 2074
	if (!inode)
		return -ENOSPC;

2075
	error = security_inode_init_security(inode, dir, &dentry->d_name,
2076
					     shmem_initxattrs, NULL);
2077 2078 2079 2080 2081 2082 2083 2084
	if (error) {
		if (error != -EOPNOTSUPP) {
			iput(inode);
			return error;
		}
		error = 0;
	}

L
Linus Torvalds 已提交
2085 2086
	info = SHMEM_I(inode);
	inode->i_size = len-1;
2087 2088 2089 2090 2091 2092 2093
	if (len <= SHORT_SYMLINK_LEN) {
		info->symlink = kmemdup(symname, len, GFP_KERNEL);
		if (!info->symlink) {
			iput(inode);
			return -ENOMEM;
		}
		inode->i_op = &shmem_short_symlink_operations;
L
Linus Torvalds 已提交
2094 2095 2096 2097 2098 2099
	} else {
		error = shmem_getpage(inode, 0, &page, SGP_WRITE, NULL);
		if (error) {
			iput(inode);
			return error;
		}
2100
		inode->i_mapping->a_ops = &shmem_aops;
L
Linus Torvalds 已提交
2101
		inode->i_op = &shmem_symlink_inode_operations;
2102
		kaddr = kmap_atomic(page);
L
Linus Torvalds 已提交
2103
		memcpy(kaddr, symname, len);
2104
		kunmap_atomic(kaddr);
2105
		SetPageUptodate(page);
L
Linus Torvalds 已提交
2106
		set_page_dirty(page);
2107
		unlock_page(page);
L
Linus Torvalds 已提交
2108 2109 2110 2111 2112 2113 2114 2115 2116
		page_cache_release(page);
	}
	dir->i_size += BOGO_DIRENT_SIZE;
	dir->i_ctime = dir->i_mtime = CURRENT_TIME;
	d_instantiate(dentry, inode);
	dget(dentry);
	return 0;
}

2117
static void *shmem_follow_short_symlink(struct dentry *dentry, struct nameidata *nd)
L
Linus Torvalds 已提交
2118
{
2119
	nd_set_link(nd, SHMEM_I(dentry->d_inode)->symlink);
2120
	return NULL;
L
Linus Torvalds 已提交
2121 2122
}

2123
static void *shmem_follow_link(struct dentry *dentry, struct nameidata *nd)
L
Linus Torvalds 已提交
2124 2125
{
	struct page *page = NULL;
2126 2127
	int error = shmem_getpage(dentry->d_inode, 0, &page, SGP_READ, NULL);
	nd_set_link(nd, error ? ERR_PTR(error) : kmap(page));
H
Hugh Dickins 已提交
2128 2129
	if (page)
		unlock_page(page);
2130
	return page;
L
Linus Torvalds 已提交
2131 2132
}

2133
static void shmem_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie)
L
Linus Torvalds 已提交
2134 2135
{
	if (!IS_ERR(nd_get_link(nd))) {
2136
		struct page *page = cookie;
L
Linus Torvalds 已提交
2137 2138 2139 2140 2141 2142
		kunmap(page);
		mark_page_accessed(page);
		page_cache_release(page);
	}
}

2143
#ifdef CONFIG_TMPFS_XATTR
2144
/*
2145 2146
 * Superblocks without xattr inode operations may get some security.* xattr
 * support from the LSM "for free". As soon as we have any other xattrs
2147 2148 2149 2150
 * like ACLs, we also need to implement the security.* handlers at
 * filesystem level, though.
 */

2151 2152 2153 2154 2155 2156 2157 2158 2159
/*
 * Callback for security_inode_init_security() for acquiring xattrs.
 */
static int shmem_initxattrs(struct inode *inode,
			    const struct xattr *xattr_array,
			    void *fs_info)
{
	struct shmem_inode_info *info = SHMEM_I(inode);
	const struct xattr *xattr;
2160
	struct simple_xattr *new_xattr;
2161 2162 2163
	size_t len;

	for (xattr = xattr_array; xattr->name != NULL; xattr++) {
2164
		new_xattr = simple_xattr_alloc(xattr->value, xattr->value_len);
2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180
		if (!new_xattr)
			return -ENOMEM;

		len = strlen(xattr->name) + 1;
		new_xattr->name = kmalloc(XATTR_SECURITY_PREFIX_LEN + len,
					  GFP_KERNEL);
		if (!new_xattr->name) {
			kfree(new_xattr);
			return -ENOMEM;
		}

		memcpy(new_xattr->name, XATTR_SECURITY_PREFIX,
		       XATTR_SECURITY_PREFIX_LEN);
		memcpy(new_xattr->name + XATTR_SECURITY_PREFIX_LEN,
		       xattr->name, len);

2181
		simple_xattr_list_add(&info->xattrs, new_xattr);
2182 2183 2184 2185 2186
	}

	return 0;
}

2187
static const struct xattr_handler *shmem_xattr_handlers[] = {
2188
#ifdef CONFIG_TMPFS_POSIX_ACL
C
Christoph Hellwig 已提交
2189 2190
	&posix_acl_access_xattr_handler,
	&posix_acl_default_xattr_handler,
2191
#endif
2192 2193
	NULL
};
2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216

static int shmem_xattr_validate(const char *name)
{
	struct { const char *prefix; size_t len; } arr[] = {
		{ XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN },
		{ XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN }
	};
	int i;

	for (i = 0; i < ARRAY_SIZE(arr); i++) {
		size_t preflen = arr[i].len;
		if (strncmp(name, arr[i].prefix, preflen) == 0) {
			if (!name[preflen])
				return -EINVAL;
			return 0;
		}
	}
	return -EOPNOTSUPP;
}

static ssize_t shmem_getxattr(struct dentry *dentry, const char *name,
			      void *buffer, size_t size)
{
2217
	struct shmem_inode_info *info = SHMEM_I(dentry->d_inode);
2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231
	int err;

	/*
	 * If this is a request for a synthetic attribute in the system.*
	 * namespace use the generic infrastructure to resolve a handler
	 * for it via sb->s_xattr.
	 */
	if (!strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
		return generic_getxattr(dentry, name, buffer, size);

	err = shmem_xattr_validate(name);
	if (err)
		return err;

2232
	return simple_xattr_get(&info->xattrs, name, buffer, size);
2233 2234 2235 2236 2237
}

static int shmem_setxattr(struct dentry *dentry, const char *name,
			  const void *value, size_t size, int flags)
{
2238
	struct shmem_inode_info *info = SHMEM_I(dentry->d_inode);
2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252
	int err;

	/*
	 * If this is a request for a synthetic attribute in the system.*
	 * namespace use the generic infrastructure to resolve a handler
	 * for it via sb->s_xattr.
	 */
	if (!strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
		return generic_setxattr(dentry, name, value, size, flags);

	err = shmem_xattr_validate(name);
	if (err)
		return err;

2253
	return simple_xattr_set(&info->xattrs, name, value, size, flags);
2254 2255 2256 2257
}

static int shmem_removexattr(struct dentry *dentry, const char *name)
{
2258
	struct shmem_inode_info *info = SHMEM_I(dentry->d_inode);
2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272
	int err;

	/*
	 * If this is a request for a synthetic attribute in the system.*
	 * namespace use the generic infrastructure to resolve a handler
	 * for it via sb->s_xattr.
	 */
	if (!strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
		return generic_removexattr(dentry, name);

	err = shmem_xattr_validate(name);
	if (err)
		return err;

2273
	return simple_xattr_remove(&info->xattrs, name);
2274 2275 2276 2277
}

static ssize_t shmem_listxattr(struct dentry *dentry, char *buffer, size_t size)
{
2278 2279
	struct shmem_inode_info *info = SHMEM_I(dentry->d_inode);
	return simple_xattr_list(&info->xattrs, buffer, size);
2280 2281 2282
}
#endif /* CONFIG_TMPFS_XATTR */

2283
static const struct inode_operations shmem_short_symlink_operations = {
2284
	.readlink	= generic_readlink,
2285
	.follow_link	= shmem_follow_short_symlink,
2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
#endif
};

static const struct inode_operations shmem_symlink_inode_operations = {
	.readlink	= generic_readlink,
	.follow_link	= shmem_follow_link,
	.put_link	= shmem_put_link,
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
2303
#endif
2304
};
2305

2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318
static struct dentry *shmem_get_parent(struct dentry *child)
{
	return ERR_PTR(-ESTALE);
}

static int shmem_match(struct inode *ino, void *vfh)
{
	__u32 *fh = vfh;
	__u64 inum = fh[2];
	inum = (inum << 32) | fh[1];
	return ino->i_ino == inum && fh[0] == ino->i_generation;
}

C
Christoph Hellwig 已提交
2319 2320
static struct dentry *shmem_fh_to_dentry(struct super_block *sb,
		struct fid *fid, int fh_len, int fh_type)
2321 2322
{
	struct inode *inode;
C
Christoph Hellwig 已提交
2323
	struct dentry *dentry = NULL;
2324
	u64 inum;
C
Christoph Hellwig 已提交
2325 2326 2327

	if (fh_len < 3)
		return NULL;
2328

2329 2330 2331
	inum = fid->raw[2];
	inum = (inum << 32) | fid->raw[1];

C
Christoph Hellwig 已提交
2332 2333
	inode = ilookup5(sb, (unsigned long)(inum + fid->raw[0]),
			shmem_match, fid->raw);
2334
	if (inode) {
C
Christoph Hellwig 已提交
2335
		dentry = d_find_alias(inode);
2336 2337 2338
		iput(inode);
	}

C
Christoph Hellwig 已提交
2339
	return dentry;
2340 2341
}

A
Al Viro 已提交
2342 2343
static int shmem_encode_fh(struct inode *inode, __u32 *fh, int *len,
				struct inode *parent)
2344
{
2345 2346
	if (*len < 3) {
		*len = 3;
2347
		return FILEID_INVALID;
2348
	}
2349

A
Al Viro 已提交
2350
	if (inode_unhashed(inode)) {
2351 2352 2353 2354 2355 2356 2357
		/* Unfortunately insert_inode_hash is not idempotent,
		 * so as we hash inodes here rather than at creation
		 * time, we need a lock to ensure we only try
		 * to do it once
		 */
		static DEFINE_SPINLOCK(lock);
		spin_lock(&lock);
A
Al Viro 已提交
2358
		if (inode_unhashed(inode))
2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371
			__insert_inode_hash(inode,
					    inode->i_ino + inode->i_generation);
		spin_unlock(&lock);
	}

	fh[0] = inode->i_generation;
	fh[1] = inode->i_ino;
	fh[2] = ((__u64)inode->i_ino) >> 32;

	*len = 3;
	return 1;
}

2372
static const struct export_operations shmem_export_ops = {
2373 2374
	.get_parent     = shmem_get_parent,
	.encode_fh      = shmem_encode_fh,
C
Christoph Hellwig 已提交
2375
	.fh_to_dentry	= shmem_fh_to_dentry,
2376 2377
};

2378 2379
static int shmem_parse_options(char *options, struct shmem_sb_info *sbinfo,
			       bool remount)
L
Linus Torvalds 已提交
2380 2381
{
	char *this_char, *value, *rest;
G
Greg Thelen 已提交
2382
	struct mempolicy *mpol = NULL;
2383 2384
	uid_t uid;
	gid_t gid;
L
Linus Torvalds 已提交
2385

2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402
	while (options != NULL) {
		this_char = options;
		for (;;) {
			/*
			 * NUL-terminate this option: unfortunately,
			 * mount options form a comma-separated list,
			 * but mpol's nodelist may also contain commas.
			 */
			options = strchr(options, ',');
			if (options == NULL)
				break;
			options++;
			if (!isdigit(*options)) {
				options[-1] = '\0';
				break;
			}
		}
L
Linus Torvalds 已提交
2403 2404 2405 2406 2407 2408 2409 2410
		if (!*this_char)
			continue;
		if ((value = strchr(this_char,'=')) != NULL) {
			*value++ = 0;
		} else {
			printk(KERN_ERR
			    "tmpfs: No value for mount option '%s'\n",
			    this_char);
G
Greg Thelen 已提交
2411
			goto error;
L
Linus Torvalds 已提交
2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424
		}

		if (!strcmp(this_char,"size")) {
			unsigned long long size;
			size = memparse(value,&rest);
			if (*rest == '%') {
				size <<= PAGE_SHIFT;
				size *= totalram_pages;
				do_div(size, 100);
				rest++;
			}
			if (*rest)
				goto bad_val;
2425 2426
			sbinfo->max_blocks =
				DIV_ROUND_UP(size, PAGE_CACHE_SIZE);
L
Linus Torvalds 已提交
2427
		} else if (!strcmp(this_char,"nr_blocks")) {
2428
			sbinfo->max_blocks = memparse(value, &rest);
L
Linus Torvalds 已提交
2429 2430 2431
			if (*rest)
				goto bad_val;
		} else if (!strcmp(this_char,"nr_inodes")) {
2432
			sbinfo->max_inodes = memparse(value, &rest);
L
Linus Torvalds 已提交
2433 2434 2435
			if (*rest)
				goto bad_val;
		} else if (!strcmp(this_char,"mode")) {
2436
			if (remount)
L
Linus Torvalds 已提交
2437
				continue;
2438
			sbinfo->mode = simple_strtoul(value, &rest, 8) & 07777;
L
Linus Torvalds 已提交
2439 2440 2441
			if (*rest)
				goto bad_val;
		} else if (!strcmp(this_char,"uid")) {
2442
			if (remount)
L
Linus Torvalds 已提交
2443
				continue;
2444
			uid = simple_strtoul(value, &rest, 0);
L
Linus Torvalds 已提交
2445 2446
			if (*rest)
				goto bad_val;
2447 2448 2449
			sbinfo->uid = make_kuid(current_user_ns(), uid);
			if (!uid_valid(sbinfo->uid))
				goto bad_val;
L
Linus Torvalds 已提交
2450
		} else if (!strcmp(this_char,"gid")) {
2451
			if (remount)
L
Linus Torvalds 已提交
2452
				continue;
2453
			gid = simple_strtoul(value, &rest, 0);
L
Linus Torvalds 已提交
2454 2455
			if (*rest)
				goto bad_val;
2456 2457 2458
			sbinfo->gid = make_kgid(current_user_ns(), gid);
			if (!gid_valid(sbinfo->gid))
				goto bad_val;
2459
		} else if (!strcmp(this_char,"mpol")) {
G
Greg Thelen 已提交
2460 2461 2462
			mpol_put(mpol);
			mpol = NULL;
			if (mpol_parse_str(value, &mpol))
2463
				goto bad_val;
L
Linus Torvalds 已提交
2464 2465 2466
		} else {
			printk(KERN_ERR "tmpfs: Bad mount option %s\n",
			       this_char);
G
Greg Thelen 已提交
2467
			goto error;
L
Linus Torvalds 已提交
2468 2469
		}
	}
G
Greg Thelen 已提交
2470
	sbinfo->mpol = mpol;
L
Linus Torvalds 已提交
2471 2472 2473 2474 2475
	return 0;

bad_val:
	printk(KERN_ERR "tmpfs: Bad value '%s' for mount option '%s'\n",
	       value, this_char);
G
Greg Thelen 已提交
2476 2477
error:
	mpol_put(mpol);
L
Linus Torvalds 已提交
2478 2479 2480 2481 2482 2483 2484
	return 1;

}

static int shmem_remount_fs(struct super_block *sb, int *flags, char *data)
{
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);
2485
	struct shmem_sb_info config = *sbinfo;
2486 2487 2488
	unsigned long inodes;
	int error = -EINVAL;

2489
	config.mpol = NULL;
2490
	if (shmem_parse_options(data, &config, true))
2491
		return error;
L
Linus Torvalds 已提交
2492

2493 2494
	spin_lock(&sbinfo->stat_lock);
	inodes = sbinfo->max_inodes - sbinfo->free_inodes;
2495
	if (percpu_counter_compare(&sbinfo->used_blocks, config.max_blocks) > 0)
2496
		goto out;
2497
	if (config.max_inodes < inodes)
2498 2499
		goto out;
	/*
2500
	 * Those tests disallow limited->unlimited while any are in use;
2501 2502 2503
	 * but we must separately disallow unlimited->limited, because
	 * in that case we have no record of how much is already in use.
	 */
2504
	if (config.max_blocks && !sbinfo->max_blocks)
2505
		goto out;
2506
	if (config.max_inodes && !sbinfo->max_inodes)
2507 2508 2509
		goto out;

	error = 0;
2510 2511 2512
	sbinfo->max_blocks  = config.max_blocks;
	sbinfo->max_inodes  = config.max_inodes;
	sbinfo->free_inodes = config.max_inodes - inodes;
2513

2514 2515 2516 2517 2518 2519 2520
	/*
	 * Preserve previous mempolicy unless mpol remount option was specified.
	 */
	if (config.mpol) {
		mpol_put(sbinfo->mpol);
		sbinfo->mpol = config.mpol;	/* transfers initial ref */
	}
2521 2522 2523
out:
	spin_unlock(&sbinfo->stat_lock);
	return error;
L
Linus Torvalds 已提交
2524
}
2525

2526
static int shmem_show_options(struct seq_file *seq, struct dentry *root)
2527
{
2528
	struct shmem_sb_info *sbinfo = SHMEM_SB(root->d_sb);
2529 2530 2531 2532 2533 2534 2535

	if (sbinfo->max_blocks != shmem_default_max_blocks())
		seq_printf(seq, ",size=%luk",
			sbinfo->max_blocks << (PAGE_CACHE_SHIFT - 10));
	if (sbinfo->max_inodes != shmem_default_max_inodes())
		seq_printf(seq, ",nr_inodes=%lu", sbinfo->max_inodes);
	if (sbinfo->mode != (S_IRWXUGO | S_ISVTX))
2536
		seq_printf(seq, ",mode=%03ho", sbinfo->mode);
2537 2538 2539 2540 2541 2542
	if (!uid_eq(sbinfo->uid, GLOBAL_ROOT_UID))
		seq_printf(seq, ",uid=%u",
				from_kuid_munged(&init_user_ns, sbinfo->uid));
	if (!gid_eq(sbinfo->gid, GLOBAL_ROOT_GID))
		seq_printf(seq, ",gid=%u",
				from_kgid_munged(&init_user_ns, sbinfo->gid));
2543
	shmem_show_mpol(seq, sbinfo->mpol);
2544 2545 2546
	return 0;
}
#endif /* CONFIG_TMPFS */
L
Linus Torvalds 已提交
2547 2548 2549

static void shmem_put_super(struct super_block *sb)
{
2550 2551 2552
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);

	percpu_counter_destroy(&sbinfo->used_blocks);
G
Greg Thelen 已提交
2553
	mpol_put(sbinfo->mpol);
2554
	kfree(sbinfo);
L
Linus Torvalds 已提交
2555 2556 2557
	sb->s_fs_info = NULL;
}

2558
int shmem_fill_super(struct super_block *sb, void *data, int silent)
L
Linus Torvalds 已提交
2559 2560
{
	struct inode *inode;
2561
	struct shmem_sb_info *sbinfo;
2562 2563 2564
	int err = -ENOMEM;

	/* Round up to L1_CACHE_BYTES to resist false sharing */
2565
	sbinfo = kzalloc(max((int)sizeof(struct shmem_sb_info),
2566 2567 2568 2569 2570
				L1_CACHE_BYTES), GFP_KERNEL);
	if (!sbinfo)
		return -ENOMEM;

	sbinfo->mode = S_IRWXUGO | S_ISVTX;
2571 2572
	sbinfo->uid = current_fsuid();
	sbinfo->gid = current_fsgid();
2573
	sb->s_fs_info = sbinfo;
L
Linus Torvalds 已提交
2574

2575
#ifdef CONFIG_TMPFS
L
Linus Torvalds 已提交
2576 2577 2578 2579 2580
	/*
	 * Per default we only allow half of the physical ram per
	 * tmpfs instance, limiting inodes to one per page of lowmem;
	 * but the internal instance is left unlimited.
	 */
2581
	if (!(sb->s_flags & MS_KERNMOUNT)) {
2582 2583 2584 2585 2586 2587
		sbinfo->max_blocks = shmem_default_max_blocks();
		sbinfo->max_inodes = shmem_default_max_inodes();
		if (shmem_parse_options(data, sbinfo, false)) {
			err = -EINVAL;
			goto failed;
		}
2588 2589
	} else {
		sb->s_flags |= MS_NOUSER;
L
Linus Torvalds 已提交
2590
	}
2591
	sb->s_export_op = &shmem_export_ops;
H
Hugh Dickins 已提交
2592
	sb->s_flags |= MS_NOSEC;
L
Linus Torvalds 已提交
2593 2594 2595 2596
#else
	sb->s_flags |= MS_NOUSER;
#endif

2597
	spin_lock_init(&sbinfo->stat_lock);
2598 2599
	if (percpu_counter_init(&sbinfo->used_blocks, 0))
		goto failed;
2600
	sbinfo->free_inodes = sbinfo->max_inodes;
2601

2602
	sb->s_maxbytes = MAX_LFS_FILESIZE;
L
Linus Torvalds 已提交
2603 2604 2605 2606
	sb->s_blocksize = PAGE_CACHE_SIZE;
	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
	sb->s_magic = TMPFS_MAGIC;
	sb->s_op = &shmem_ops;
2607
	sb->s_time_gran = 1;
2608
#ifdef CONFIG_TMPFS_XATTR
2609
	sb->s_xattr = shmem_xattr_handlers;
2610 2611
#endif
#ifdef CONFIG_TMPFS_POSIX_ACL
2612 2613
	sb->s_flags |= MS_POSIXACL;
#endif
2614

2615
	inode = shmem_get_inode(sb, NULL, S_IFDIR | sbinfo->mode, 0, VM_NORESERVE);
L
Linus Torvalds 已提交
2616 2617
	if (!inode)
		goto failed;
2618 2619
	inode->i_uid = sbinfo->uid;
	inode->i_gid = sbinfo->gid;
A
Al Viro 已提交
2620 2621
	sb->s_root = d_make_root(inode);
	if (!sb->s_root)
2622
		goto failed;
L
Linus Torvalds 已提交
2623 2624 2625 2626 2627 2628 2629
	return 0;

failed:
	shmem_put_super(sb);
	return err;
}

2630
static struct kmem_cache *shmem_inode_cachep;
L
Linus Torvalds 已提交
2631 2632 2633

static struct inode *shmem_alloc_inode(struct super_block *sb)
{
2634 2635 2636
	struct shmem_inode_info *info;
	info = kmem_cache_alloc(shmem_inode_cachep, GFP_KERNEL);
	if (!info)
L
Linus Torvalds 已提交
2637
		return NULL;
2638
	return &info->vfs_inode;
L
Linus Torvalds 已提交
2639 2640
}

2641
static void shmem_destroy_callback(struct rcu_head *head)
N
Nick Piggin 已提交
2642 2643 2644 2645 2646
{
	struct inode *inode = container_of(head, struct inode, i_rcu);
	kmem_cache_free(shmem_inode_cachep, SHMEM_I(inode));
}

L
Linus Torvalds 已提交
2647 2648
static void shmem_destroy_inode(struct inode *inode)
{
2649
	if (S_ISREG(inode->i_mode))
L
Linus Torvalds 已提交
2650
		mpol_free_shared_policy(&SHMEM_I(inode)->policy);
2651
	call_rcu(&inode->i_rcu, shmem_destroy_callback);
L
Linus Torvalds 已提交
2652 2653
}

2654
static void shmem_init_inode(void *foo)
L
Linus Torvalds 已提交
2655
{
2656 2657
	struct shmem_inode_info *info = foo;
	inode_init_once(&info->vfs_inode);
L
Linus Torvalds 已提交
2658 2659
}

2660
static int shmem_init_inodecache(void)
L
Linus Torvalds 已提交
2661 2662 2663
{
	shmem_inode_cachep = kmem_cache_create("shmem_inode_cache",
				sizeof(struct shmem_inode_info),
2664
				0, SLAB_PANIC, shmem_init_inode);
L
Linus Torvalds 已提交
2665 2666 2667
	return 0;
}

2668
static void shmem_destroy_inodecache(void)
L
Linus Torvalds 已提交
2669
{
2670
	kmem_cache_destroy(shmem_inode_cachep);
L
Linus Torvalds 已提交
2671 2672
}

2673
static const struct address_space_operations shmem_aops = {
L
Linus Torvalds 已提交
2674
	.writepage	= shmem_writepage,
2675
	.set_page_dirty	= __set_page_dirty_no_writeback,
L
Linus Torvalds 已提交
2676
#ifdef CONFIG_TMPFS
N
Nick Piggin 已提交
2677 2678
	.write_begin	= shmem_write_begin,
	.write_end	= shmem_write_end,
L
Linus Torvalds 已提交
2679
#endif
2680
	.migratepage	= migrate_page,
2681
	.error_remove_page = generic_error_remove_page,
L
Linus Torvalds 已提交
2682 2683
};

2684
static const struct file_operations shmem_file_operations = {
L
Linus Torvalds 已提交
2685 2686
	.mmap		= shmem_mmap,
#ifdef CONFIG_TMPFS
2687
	.llseek		= shmem_file_llseek,
H
Hugh Dickins 已提交
2688
	.read		= do_sync_read,
H
Hugh Dickins 已提交
2689
	.write		= do_sync_write,
H
Hugh Dickins 已提交
2690
	.aio_read	= shmem_file_aio_read,
H
Hugh Dickins 已提交
2691
	.aio_write	= generic_file_aio_write,
2692
	.fsync		= noop_fsync,
2693
	.splice_read	= shmem_file_splice_read,
2694
	.splice_write	= generic_file_splice_write,
2695
	.fallocate	= shmem_fallocate,
L
Linus Torvalds 已提交
2696 2697 2698
#endif
};

2699
static const struct inode_operations shmem_inode_operations = {
2700
	.setattr	= shmem_setattr,
2701 2702 2703 2704 2705
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
C
Christoph Hellwig 已提交
2706
	.set_acl	= simple_set_acl,
2707
#endif
L
Linus Torvalds 已提交
2708 2709
};

2710
static const struct inode_operations shmem_dir_inode_operations = {
L
Linus Torvalds 已提交
2711 2712 2713 2714 2715 2716 2717 2718 2719 2720
#ifdef CONFIG_TMPFS
	.create		= shmem_create,
	.lookup		= simple_lookup,
	.link		= shmem_link,
	.unlink		= shmem_unlink,
	.symlink	= shmem_symlink,
	.mkdir		= shmem_mkdir,
	.rmdir		= shmem_rmdir,
	.mknod		= shmem_mknod,
	.rename		= shmem_rename,
2721
	.tmpfile	= shmem_tmpfile,
L
Linus Torvalds 已提交
2722
#endif
2723 2724 2725 2726 2727 2728
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
#endif
2729
#ifdef CONFIG_TMPFS_POSIX_ACL
2730
	.setattr	= shmem_setattr,
C
Christoph Hellwig 已提交
2731
	.set_acl	= simple_set_acl,
2732 2733 2734
#endif
};

2735
static const struct inode_operations shmem_special_inode_operations = {
2736 2737 2738 2739 2740 2741
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
#endif
2742
#ifdef CONFIG_TMPFS_POSIX_ACL
2743
	.setattr	= shmem_setattr,
C
Christoph Hellwig 已提交
2744
	.set_acl	= simple_set_acl,
2745
#endif
L
Linus Torvalds 已提交
2746 2747
};

2748
static const struct super_operations shmem_ops = {
L
Linus Torvalds 已提交
2749 2750 2751 2752 2753
	.alloc_inode	= shmem_alloc_inode,
	.destroy_inode	= shmem_destroy_inode,
#ifdef CONFIG_TMPFS
	.statfs		= shmem_statfs,
	.remount_fs	= shmem_remount_fs,
2754
	.show_options	= shmem_show_options,
L
Linus Torvalds 已提交
2755
#endif
A
Al Viro 已提交
2756
	.evict_inode	= shmem_evict_inode,
L
Linus Torvalds 已提交
2757 2758 2759 2760
	.drop_inode	= generic_delete_inode,
	.put_super	= shmem_put_super,
};

2761
static const struct vm_operations_struct shmem_vm_ops = {
2762
	.fault		= shmem_fault,
L
Linus Torvalds 已提交
2763 2764 2765 2766
#ifdef CONFIG_NUMA
	.set_policy     = shmem_set_policy,
	.get_policy     = shmem_get_policy,
#endif
2767
	.remap_pages	= generic_file_remap_pages,
L
Linus Torvalds 已提交
2768 2769
};

A
Al Viro 已提交
2770 2771
static struct dentry *shmem_mount(struct file_system_type *fs_type,
	int flags, const char *dev_name, void *data)
L
Linus Torvalds 已提交
2772
{
A
Al Viro 已提交
2773
	return mount_nodev(fs_type, flags, data, shmem_fill_super);
L
Linus Torvalds 已提交
2774 2775
}

2776
static struct file_system_type shmem_fs_type = {
L
Linus Torvalds 已提交
2777 2778
	.owner		= THIS_MODULE,
	.name		= "tmpfs",
A
Al Viro 已提交
2779
	.mount		= shmem_mount,
L
Linus Torvalds 已提交
2780
	.kill_sb	= kill_litter_super,
2781
	.fs_flags	= FS_USERNS_MOUNT,
L
Linus Torvalds 已提交
2782 2783
};

2784
int __init shmem_init(void)
L
Linus Torvalds 已提交
2785 2786 2787
{
	int error;

2788 2789 2790 2791
	/* If rootfs called this, don't re-init */
	if (shmem_inode_cachep)
		return 0;

P
Peter Zijlstra 已提交
2792 2793 2794 2795
	error = bdi_init(&shmem_backing_dev_info);
	if (error)
		goto out4;

2796
	error = shmem_init_inodecache();
L
Linus Torvalds 已提交
2797 2798 2799
	if (error)
		goto out3;

2800
	error = register_filesystem(&shmem_fs_type);
L
Linus Torvalds 已提交
2801 2802 2803 2804
	if (error) {
		printk(KERN_ERR "Could not register tmpfs\n");
		goto out2;
	}
2805

2806
	shm_mnt = kern_mount(&shmem_fs_type);
L
Linus Torvalds 已提交
2807 2808 2809 2810 2811 2812 2813 2814
	if (IS_ERR(shm_mnt)) {
		error = PTR_ERR(shm_mnt);
		printk(KERN_ERR "Could not kern_mount tmpfs\n");
		goto out1;
	}
	return 0;

out1:
2815
	unregister_filesystem(&shmem_fs_type);
L
Linus Torvalds 已提交
2816
out2:
2817
	shmem_destroy_inodecache();
L
Linus Torvalds 已提交
2818
out3:
P
Peter Zijlstra 已提交
2819 2820
	bdi_destroy(&shmem_backing_dev_info);
out4:
L
Linus Torvalds 已提交
2821 2822 2823
	shm_mnt = ERR_PTR(error);
	return error;
}
2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835

#else /* !CONFIG_SHMEM */

/*
 * tiny-shmem: simple shmemfs and tmpfs using ramfs code
 *
 * This is intended for small system where the benefits of the full
 * shmem code (swap-backed and resource-limited) are outweighed by
 * their complexity. On systems without swap this code should be
 * effectively equivalent, but much lighter weight.
 */

2836
static struct file_system_type shmem_fs_type = {
2837
	.name		= "tmpfs",
A
Al Viro 已提交
2838
	.mount		= ramfs_mount,
2839
	.kill_sb	= kill_litter_super,
2840
	.fs_flags	= FS_USERNS_MOUNT,
2841 2842
};

2843
int __init shmem_init(void)
2844
{
2845
	BUG_ON(register_filesystem(&shmem_fs_type) != 0);
2846

2847
	shm_mnt = kern_mount(&shmem_fs_type);
2848 2849 2850 2851 2852
	BUG_ON(IS_ERR(shm_mnt));

	return 0;
}

2853
int shmem_unuse(swp_entry_t swap, struct page *page)
2854 2855 2856 2857
{
	return 0;
}

H
Hugh Dickins 已提交
2858 2859 2860 2861 2862
int shmem_lock(struct file *file, int lock, struct user_struct *user)
{
	return 0;
}

2863 2864 2865 2866
void shmem_unlock_mapping(struct address_space *mapping)
{
}

2867
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
2868
{
2869
	truncate_inode_pages_range(inode->i_mapping, lstart, lend);
2870 2871 2872
}
EXPORT_SYMBOL_GPL(shmem_truncate_range);

2873 2874
#define shmem_vm_ops				generic_file_vm_ops
#define shmem_file_operations			ramfs_file_operations
2875
#define shmem_get_inode(sb, dir, mode, dev, flags)	ramfs_get_inode(sb, dir, mode, dev)
2876 2877
#define shmem_acct_size(flags, size)		0
#define shmem_unacct_size(flags, size)		do {} while (0)
2878 2879 2880 2881

#endif /* CONFIG_SHMEM */

/* common code */
L
Linus Torvalds 已提交
2882

2883
static struct dentry_operations anon_ops = {
2884
	.d_dname = simple_dname
2885 2886
};

2887 2888
static struct file *__shmem_file_setup(const char *name, loff_t size,
				       unsigned long flags, unsigned int i_flags)
L
Linus Torvalds 已提交
2889
{
A
Al Viro 已提交
2890
	struct file *res;
L
Linus Torvalds 已提交
2891
	struct inode *inode;
2892
	struct path path;
2893
	struct super_block *sb;
L
Linus Torvalds 已提交
2894 2895 2896
	struct qstr this;

	if (IS_ERR(shm_mnt))
A
Al Viro 已提交
2897
		return ERR_CAST(shm_mnt);
L
Linus Torvalds 已提交
2898

2899
	if (size < 0 || size > MAX_LFS_FILESIZE)
L
Linus Torvalds 已提交
2900 2901 2902 2903 2904
		return ERR_PTR(-EINVAL);

	if (shmem_acct_size(flags, size))
		return ERR_PTR(-ENOMEM);

A
Al Viro 已提交
2905
	res = ERR_PTR(-ENOMEM);
L
Linus Torvalds 已提交
2906 2907 2908
	this.name = name;
	this.len = strlen(name);
	this.hash = 0; /* will go */
2909 2910
	sb = shm_mnt->mnt_sb;
	path.dentry = d_alloc_pseudo(sb, &this);
2911
	if (!path.dentry)
L
Linus Torvalds 已提交
2912
		goto put_memory;
2913
	d_set_d_op(path.dentry, &anon_ops);
2914
	path.mnt = mntget(shm_mnt);
L
Linus Torvalds 已提交
2915

A
Al Viro 已提交
2916
	res = ERR_PTR(-ENOSPC);
2917
	inode = shmem_get_inode(sb, NULL, S_IFREG | S_IRWXUGO, 0, flags);
L
Linus Torvalds 已提交
2918
	if (!inode)
2919
		goto put_dentry;
L
Linus Torvalds 已提交
2920

2921
	inode->i_flags |= i_flags;
2922
	d_instantiate(path.dentry, inode);
L
Linus Torvalds 已提交
2923
	inode->i_size = size;
2924
	clear_nlink(inode);	/* It is unlinked */
A
Al Viro 已提交
2925 2926
	res = ERR_PTR(ramfs_nommu_expand_for_mapping(inode, size));
	if (IS_ERR(res))
2927 2928
		goto put_dentry;

A
Al Viro 已提交
2929
	res = alloc_file(&path, FMODE_WRITE | FMODE_READ,
2930
		  &shmem_file_operations);
A
Al Viro 已提交
2931
	if (IS_ERR(res))
2932 2933
		goto put_dentry;

A
Al Viro 已提交
2934
	return res;
L
Linus Torvalds 已提交
2935 2936

put_dentry:
2937
	path_put(&path);
L
Linus Torvalds 已提交
2938 2939
put_memory:
	shmem_unacct_size(flags, size);
A
Al Viro 已提交
2940
	return res;
L
Linus Torvalds 已提交
2941
}
2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967

/**
 * shmem_kernel_file_setup - get an unlinked file living in tmpfs which must be
 * 	kernel internal.  There will be NO LSM permission checks against the
 * 	underlying inode.  So users of this interface must do LSM checks at a
 * 	higher layer.  The one user is the big_key implementation.  LSM checks
 * 	are provided at the key level rather than the inode level.
 * @name: name for dentry (to be seen in /proc/<pid>/maps
 * @size: size to be set for the file
 * @flags: VM_NORESERVE suppresses pre-accounting of the entire object size
 */
struct file *shmem_kernel_file_setup(const char *name, loff_t size, unsigned long flags)
{
	return __shmem_file_setup(name, size, flags, S_PRIVATE);
}

/**
 * shmem_file_setup - get an unlinked file living in tmpfs
 * @name: name for dentry (to be seen in /proc/<pid>/maps
 * @size: size to be set for the file
 * @flags: VM_NORESERVE suppresses pre-accounting of the entire object size
 */
struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags)
{
	return __shmem_file_setup(name, size, flags, 0);
}
2968
EXPORT_SYMBOL_GPL(shmem_file_setup);
L
Linus Torvalds 已提交
2969

2970
/**
L
Linus Torvalds 已提交
2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988
 * shmem_zero_setup - setup a shared anonymous mapping
 * @vma: the vma to be mmapped is prepared by do_mmap_pgoff
 */
int shmem_zero_setup(struct vm_area_struct *vma)
{
	struct file *file;
	loff_t size = vma->vm_end - vma->vm_start;

	file = shmem_file_setup("dev/zero", size, vma->vm_flags);
	if (IS_ERR(file))
		return PTR_ERR(file);

	if (vma->vm_file)
		fput(vma->vm_file);
	vma->vm_file = file;
	vma->vm_ops = &shmem_vm_ops;
	return 0;
}
2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001

/**
 * shmem_read_mapping_page_gfp - read into page cache, using specified page allocation flags.
 * @mapping:	the page's address_space
 * @index:	the page index
 * @gfp:	the page allocator flags to use if allocating
 *
 * This behaves as a tmpfs "read_cache_page_gfp(mapping, index, gfp)",
 * with any new page allocations done using the specified allocation flags.
 * But read_cache_page_gfp() uses the ->readpage() method: which does not
 * suit tmpfs, since it may have pages in swapcache, and needs to find those
 * for itself; although drivers/gpu/drm i915 and ttm rely upon this support.
 *
3002 3003
 * i915_gem_object_get_pages_gtt() mixes __GFP_NORETRY | __GFP_NOWARN in
 * with the mapping_gfp_mask(), to avoid OOMing the machine unnecessarily.
3004 3005 3006 3007
 */
struct page *shmem_read_mapping_page_gfp(struct address_space *mapping,
					 pgoff_t index, gfp_t gfp)
{
3008 3009
#ifdef CONFIG_SHMEM
	struct inode *inode = mapping->host;
H
Hugh Dickins 已提交
3010
	struct page *page;
3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023
	int error;

	BUG_ON(mapping->a_ops != &shmem_aops);
	error = shmem_getpage_gfp(inode, index, &page, SGP_CACHE, gfp, NULL);
	if (error)
		page = ERR_PTR(error);
	else
		unlock_page(page);
	return page;
#else
	/*
	 * The tiny !SHMEM case uses ramfs without swap
	 */
3024
	return read_cache_page_gfp(mapping, index, gfp);
3025
#endif
3026 3027
}
EXPORT_SYMBOL_GPL(shmem_read_mapping_page_gfp);