shmem.c 89.6 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>
D
David Herrmann 已提交
69
#include <linux/syscalls.h>
D
David Herrmann 已提交
70
#include <linux/fcntl.h>
D
David Herrmann 已提交
71
#include <uapi/linux/memfd.h>
72

L
Linus Torvalds 已提交
73 74 75
#include <asm/uaccess.h>
#include <asm/pgtable.h>

H
Hugh Dickins 已提交
76
#define BLOCKS_PER_PAGE  (PAGE_CACHE_SIZE/512)
L
Linus Torvalds 已提交
77 78 79 80 81
#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

82 83 84
/* Symlink up to this size is kmalloc'ed instead of using a swappable page */
#define SHORT_SYMLINK_LEN 128

85
/*
86 87 88
 * shmem_fallocate communicates with shmem_fault or shmem_writepage via
 * inode->i_private (with i_mutex making sure that it has only one user at
 * a time): we would prefer not to enlarge the shmem inode just for that.
89 90
 */
struct shmem_falloc {
91
	wait_queue_head_t *waitq; /* faults into hole wait for punch to end */
92 93 94 95 96 97
	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 */
};

98
/* Flag allocation requirements to shmem_getpage */
L
Linus Torvalds 已提交
99 100 101
enum sgp_type {
	SGP_READ,	/* don't exceed i_size, don't allocate page */
	SGP_CACHE,	/* don't exceed i_size, may allocate page */
102
	SGP_DIRTY,	/* like SGP_CACHE, but set new page dirty */
103 104
	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
105 106
};

A
Andrew Morton 已提交
107
#ifdef CONFIG_TMPFS
108 109 110 111 112 113 114 115 116
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
117
#endif
118

119 120 121
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);
122 123 124 125 126 127 128 129 130
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
131 132 133 134 135 136 137 138 139 140 141 142 143 144

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)
{
145
	return (flags & VM_NORESERVE) ?
146
		0 : security_vm_enough_memory_mm(current->mm, VM_ACCT(size));
L
Linus Torvalds 已提交
147 148 149 150
}

static inline void shmem_unacct_size(unsigned long flags, loff_t size)
{
151
	if (!(flags & VM_NORESERVE))
L
Linus Torvalds 已提交
152 153 154
		vm_unacct_memory(VM_ACCT(size));
}

155 156 157 158 159 160 161 162 163 164 165 166 167
static inline int shmem_reacct_size(unsigned long flags,
		loff_t oldsize, loff_t newsize)
{
	if (!(flags & VM_NORESERVE)) {
		if (VM_ACCT(newsize) > VM_ACCT(oldsize))
			return security_vm_enough_memory_mm(current->mm,
					VM_ACCT(newsize) - VM_ACCT(oldsize));
		else if (VM_ACCT(newsize) < VM_ACCT(oldsize))
			vm_unacct_memory(VM_ACCT(oldsize) - VM_ACCT(newsize));
	}
	return 0;
}

L
Linus Torvalds 已提交
168 169 170 171 172 173 174 175
/*
 * ... 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)
{
176
	return (flags & VM_NORESERVE) ?
177
		security_vm_enough_memory_mm(current->mm, VM_ACCT(PAGE_CACHE_SIZE)) : 0;
L
Linus Torvalds 已提交
178 179 180 181
}

static inline void shmem_unacct_blocks(unsigned long flags, long pages)
{
182
	if (flags & VM_NORESERVE)
L
Linus Torvalds 已提交
183 184 185
		vm_unacct_memory(pages * VM_ACCT(PAGE_CACHE_SIZE));
}

186
static const struct super_operations shmem_ops;
187
static const struct address_space_operations shmem_aops;
188
static const struct file_operations shmem_file_operations;
189 190 191
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;
192
static const struct vm_operations_struct shmem_vm_ops;
L
Linus Torvalds 已提交
193

194
static struct backing_dev_info shmem_backing_dev_info  __read_mostly = {
L
Linus Torvalds 已提交
195
	.ra_pages	= 0,	/* No readahead */
196
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK | BDI_CAP_SWAP_BACKED,
L
Linus Torvalds 已提交
197 198 199
};

static LIST_HEAD(shmem_swaplist);
200
static DEFINE_MUTEX(shmem_swaplist_mutex);
L
Linus Torvalds 已提交
201

202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226
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);
	}
}

227
/**
228
 * shmem_recalc_inode - recalculate the block usage of an inode
L
Linus Torvalds 已提交
229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245
 * @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) {
246 247 248
		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
249
		info->alloced -= freed;
250
		inode->i_blocks -= freed * BLOCKS_PER_PAGE;
L
Linus Torvalds 已提交
251 252 253 254
		shmem_unacct_blocks(info->flags, freed);
	}
}

255 256 257 258 259 260 261
/*
 * 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;
262
	void *item;
263 264

	VM_BUG_ON(!expected);
265
	VM_BUG_ON(!replacement);
266
	pslot = radix_tree_lookup_slot(&mapping->page_tree, index);
267 268 269
	if (!pslot)
		return -ENOENT;
	item = radix_tree_deref_slot_protected(pslot, &mapping->tree_lock);
270 271
	if (item != expected)
		return -ENOENT;
272
	radix_tree_replace_slot(pslot, replacement);
273 274 275
	return 0;
}

276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293
/*
 * 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);
}

294 295 296 297 298
/*
 * 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,
299
				   pgoff_t index, void *expected)
300
{
301
	int error;
302

303 304
	VM_BUG_ON_PAGE(!PageLocked(page), page);
	VM_BUG_ON_PAGE(!PageSwapBacked(page), page);
305

306 307 308 309 310
	page_cache_get(page);
	page->mapping = mapping;
	page->index = index;

	spin_lock_irq(&mapping->tree_lock);
311
	if (!expected)
312 313 314 315
		error = radix_tree_insert(&mapping->page_tree, index, page);
	else
		error = shmem_radix_tree_replace(mapping, index, expected,
								 page);
316
	if (!error) {
317 318 319 320 321 322 323 324
		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);
325 326 327 328
	}
	return error;
}

329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347
/*
 * 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);
}

348 349 350 351 352 353
/*
 * 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)
{
354
	void *old;
355 356

	spin_lock_irq(&mapping->tree_lock);
357
	old = radix_tree_delete_item(&mapping->page_tree, index, radswap);
358
	spin_unlock_irq(&mapping->tree_lock);
359 360 361 362
	if (old != radswap)
		return -ENOENT;
	free_swap_and_cache(radix_to_swp_entry(radswap));
	return 0;
363 364
}

365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382
/*
 * 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.
		 */
383 384
		pvec.nr = find_get_entries(mapping, index,
					   PAGEVEC_SIZE, pvec.pages, indices);
385 386 387
		if (!pvec.nr)
			break;
		index = indices[pvec.nr - 1] + 1;
388
		pagevec_remove_exceptionals(&pvec);
389 390 391 392
		check_move_unevictable_pages(pvec.pages, pvec.nr);
		pagevec_release(&pvec);
		cond_resched();
	}
393 394 395 396
}

/*
 * Remove range of pages and swap entries from radix tree, and free them.
397
 * If !unfalloc, truncate or punch hole; if unfalloc, undo failed fallocate.
398
 */
399 400
static void shmem_undo_range(struct inode *inode, loff_t lstart, loff_t lend,
								 bool unfalloc)
L
Linus Torvalds 已提交
401
{
402
	struct address_space *mapping = inode->i_mapping;
L
Linus Torvalds 已提交
403
	struct shmem_inode_info *info = SHMEM_I(inode);
404
	pgoff_t start = (lstart + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
405 406 407
	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);
408
	struct pagevec pvec;
409 410
	pgoff_t indices[PAGEVEC_SIZE];
	long nr_swaps_freed = 0;
411
	pgoff_t index;
412 413
	int i;

414 415
	if (lend == -1)
		end = -1;	/* unsigned, so actually very big */
416 417 418

	pagevec_init(&pvec, 0);
	index = start;
419
	while (index < end) {
420 421 422
		pvec.nr = find_get_entries(mapping, index,
			min(end - index, (pgoff_t)PAGEVEC_SIZE),
			pvec.pages, indices);
423 424
		if (!pvec.nr)
			break;
425 426 427
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

428
			index = indices[i];
429
			if (index >= end)
430 431
				break;

432
			if (radix_tree_exceptional_entry(page)) {
433 434
				if (unfalloc)
					continue;
435 436
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
437
				continue;
438 439 440
			}

			if (!trylock_page(page))
441
				continue;
442 443
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
444
					VM_BUG_ON_PAGE(PageWriteback(page), page);
445 446
					truncate_inode_page(mapping, page);
				}
447 448 449
			}
			unlock_page(page);
		}
450
		pagevec_remove_exceptionals(&pvec);
451
		pagevec_release(&pvec);
452 453 454
		cond_resched();
		index++;
	}
L
Linus Torvalds 已提交
455

456
	if (partial_start) {
457 458 459
		struct page *page = NULL;
		shmem_getpage(inode, start - 1, &page, SGP_READ, NULL);
		if (page) {
460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475
			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);
476 477 478 479 480
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}
481 482
	if (start >= end)
		return;
483 484

	index = start;
485
	while (index < end) {
486
		cond_resched();
487 488

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

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

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

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

L
Linus Torvalds 已提交
537
	spin_lock(&info->lock);
538
	info->swapped -= nr_swaps_freed;
L
Linus Torvalds 已提交
539 540
	shmem_recalc_inode(inode);
	spin_unlock(&info->lock);
541
}
L
Linus Torvalds 已提交
542

543 544 545
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
{
	shmem_undo_range(inode, lstart, lend, false);
546
	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
L
Linus Torvalds 已提交
547
}
548
EXPORT_SYMBOL_GPL(shmem_truncate_range);
L
Linus Torvalds 已提交
549

550
static int shmem_setattr(struct dentry *dentry, struct iattr *attr)
L
Linus Torvalds 已提交
551 552
{
	struct inode *inode = dentry->d_inode;
D
David Herrmann 已提交
553
	struct shmem_inode_info *info = SHMEM_I(inode);
L
Linus Torvalds 已提交
554 555
	int error;

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

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

D
David Herrmann 已提交
564 565 566 567 568
		/* protected by i_mutex */
		if ((newsize < oldsize && (info->seals & F_SEAL_SHRINK)) ||
		    (newsize > oldsize && (info->seals & F_SEAL_GROW)))
			return -EPERM;

569
		if (newsize != oldsize) {
570 571 572 573
			error = shmem_reacct_size(SHMEM_I(inode)->flags,
					oldsize, newsize);
			if (error)
				return error;
574 575 576 577 578 579 580 581 582 583
			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
584 585
	}

586 587
	setattr_copy(inode, attr);
	if (attr->ia_valid & ATTR_MODE)
C
Christoph Hellwig 已提交
588
		error = posix_acl_chmod(inode, inode->i_mode);
L
Linus Torvalds 已提交
589 590 591
	return error;
}

A
Al Viro 已提交
592
static void shmem_evict_inode(struct inode *inode)
L
Linus Torvalds 已提交
593 594 595
{
	struct shmem_inode_info *info = SHMEM_I(inode);

596
	if (inode->i_mapping->a_ops == &shmem_aops) {
L
Linus Torvalds 已提交
597 598
		shmem_unacct_size(info->flags, inode->i_size);
		inode->i_size = 0;
599
		shmem_truncate_range(inode, 0, (loff_t)-1);
L
Linus Torvalds 已提交
600
		if (!list_empty(&info->swaplist)) {
601
			mutex_lock(&shmem_swaplist_mutex);
L
Linus Torvalds 已提交
602
			list_del_init(&info->swaplist);
603
			mutex_unlock(&shmem_swaplist_mutex);
L
Linus Torvalds 已提交
604
		}
605 606
	} else
		kfree(info->symlink);
607

608
	simple_xattrs_free(&info->xattrs);
609
	WARN_ON(inode->i_blocks);
610
	shmem_free_inode(inode->i_sb);
611
	clear_inode(inode);
L
Linus Torvalds 已提交
612 613
}

614 615 616
/*
 * If swap found in inode, free it and move page from swapcache to filecache.
 */
617
static int shmem_unuse_inode(struct shmem_inode_info *info,
618
			     swp_entry_t swap, struct page **pagep)
L
Linus Torvalds 已提交
619
{
620
	struct address_space *mapping = info->vfs_inode.i_mapping;
621
	void *radswap;
622
	pgoff_t index;
623 624
	gfp_t gfp;
	int error = 0;
L
Linus Torvalds 已提交
625

626
	radswap = swp_to_radix_entry(swap);
627
	index = radix_tree_locate_item(&mapping->page_tree, radswap);
628
	if (index == -1)
629
		return -EAGAIN;	/* tell shmem_unuse we found nothing */
630

H
Hugh Dickins 已提交
631 632
	/*
	 * Move _head_ to start search for next from here.
A
Al Viro 已提交
633
	 * But be careful: shmem_evict_inode checks list_empty without taking
H
Hugh Dickins 已提交
634
	 * mutex, and there's an instant in list_move_tail when info->swaplist
635
	 * would appear empty, if it were the only one on shmem_swaplist.
H
Hugh Dickins 已提交
636 637 638
	 */
	if (shmem_swaplist.next != &info->swaplist)
		list_move_tail(&shmem_swaplist, &info->swaplist);
639

640 641 642 643 644 645 646
	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
647 648 649 650 651 652 653 654 655 656 657 658 659 660 661
		 * 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).
662 663 664 665 666
		 */
		if (!page_swapcount(*pagep))
			error = -ENOENT;
	}

K
KAMEZAWA Hiroyuki 已提交
667
	/*
668 669 670
	 * 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
671
	 */
672 673
	if (!error)
		error = shmem_add_to_page_cache(*pagep, mapping, index,
674
						radswap);
675
	if (error != -ENOMEM) {
676 677 678 679
		/*
		 * Truncation and eviction use free_swap_and_cache(), which
		 * only does trylock page: if we raced, best clean up here.
		 */
680 681
		delete_from_swap_cache(*pagep);
		set_page_dirty(*pagep);
682 683 684 685 686 687
		if (!error) {
			spin_lock(&info->lock);
			info->swapped--;
			spin_unlock(&info->lock);
			swap_free(swap);
		}
L
Linus Torvalds 已提交
688
	}
689
	return error;
L
Linus Torvalds 已提交
690 691 692
}

/*
693
 * Search through swapped inodes to find and replace swap by page.
L
Linus Torvalds 已提交
694
 */
695
int shmem_unuse(swp_entry_t swap, struct page *page)
L
Linus Torvalds 已提交
696
{
697
	struct list_head *this, *next;
L
Linus Torvalds 已提交
698
	struct shmem_inode_info *info;
699
	struct mem_cgroup *memcg;
700 701 702 703
	int error = 0;

	/*
	 * There's a faint possibility that swap page was replaced before
704
	 * caller locked it: caller will come back later with the right page.
705
	 */
706
	if (unlikely(!PageSwapCache(page) || page_private(page) != swap.val))
707
		goto out;
708 709 710 711 712 713

	/*
	 * 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.
	 */
714
	error = mem_cgroup_try_charge(page, current->mm, GFP_KERNEL, &memcg);
715 716
	if (error)
		goto out;
717
	/* No radix_tree_preload: swap entry keeps a place for page in tree */
718
	error = -EAGAIN;
L
Linus Torvalds 已提交
719

720
	mutex_lock(&shmem_swaplist_mutex);
721 722
	list_for_each_safe(this, next, &shmem_swaplist) {
		info = list_entry(this, struct shmem_inode_info, swaplist);
723
		if (info->swapped)
724
			error = shmem_unuse_inode(info, swap, &page);
725 726
		else
			list_del_init(&info->swaplist);
727
		cond_resched();
728
		if (error != -EAGAIN)
729
			break;
730
		/* found nothing in this: move on to search the next */
L
Linus Torvalds 已提交
731
	}
732
	mutex_unlock(&shmem_swaplist_mutex);
733

734 735 736 737 738 739
	if (error) {
		if (error != -ENOMEM)
			error = 0;
		mem_cgroup_cancel_charge(page, memcg);
	} else
		mem_cgroup_commit_charge(page, memcg, true);
740
out:
H
Hugh Dickins 已提交
741 742
	unlock_page(page);
	page_cache_release(page);
743
	return error;
L
Linus Torvalds 已提交
744 745 746 747 748 749 750 751 752 753
}

/*
 * 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;
754 755
	swp_entry_t swap;
	pgoff_t index;
L
Linus Torvalds 已提交
756 757 758 759 760 761 762 763

	BUG_ON(!PageLocked(page));
	mapping = page->mapping;
	index = page->index;
	inode = mapping->host;
	info = SHMEM_I(inode);
	if (info->flags & VM_LOCKED)
		goto redirty;
764
	if (!total_swap_pages)
L
Linus Torvalds 已提交
765 766
		goto redirty;

767 768 769
	/*
	 * shmem_backing_dev_info's capabilities prevent regular writeback or
	 * sync from ever calling shmem_writepage; but a stacking filesystem
770
	 * might use ->writepage of its underlying filesystem, in which case
771
	 * tmpfs should write out to swap only in response to memory pressure,
772
	 * and not for the writeback threads or sync.
773
	 */
774 775 776 777
	if (!wbc->for_reclaim) {
		WARN_ON_ONCE(1);	/* Still happens? Tell us about it! */
		goto redirty;
	}
778 779 780 781 782

	/*
	 * 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.
783 784 785 786 787 788
	 *
	 * 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.
789 790
	 */
	if (!PageUptodate(page)) {
791 792 793 794 795
		if (inode->i_private) {
			struct shmem_falloc *shmem_falloc;
			spin_lock(&inode->i_lock);
			shmem_falloc = inode->i_private;
			if (shmem_falloc &&
796
			    !shmem_falloc->waitq &&
797 798 799 800 801 802 803 804 805
			    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;
		}
806 807 808 809 810
		clear_highpage(page);
		flush_dcache_page(page);
		SetPageUptodate(page);
	}

811 812 813
	swap = get_swap_page();
	if (!swap.val)
		goto redirty;
814

815 816
	/*
	 * Add inode to shmem_unuse()'s list of swapped-out inodes,
817 818
	 * if it's not already there.  Do it now before the page is
	 * moved to swap cache, when its pagelock no longer protects
819
	 * the inode from eviction.  But don't unlock the mutex until
820 821
	 * we've incremented swapped, because shmem_unuse_inode() will
	 * prune a !swapped inode from the swaplist under this mutex.
822
	 */
823 824 825
	mutex_lock(&shmem_swaplist_mutex);
	if (list_empty(&info->swaplist))
		list_add_tail(&info->swaplist, &shmem_swaplist);
826

827
	if (add_to_swap_cache(page, swap, GFP_ATOMIC) == 0) {
H
Hugh Dickins 已提交
828
		swap_shmem_alloc(swap);
829 830 831 832 833
		shmem_delete_from_page_cache(page, swp_to_radix_entry(swap));

		spin_lock(&info->lock);
		info->swapped++;
		shmem_recalc_inode(inode);
834
		spin_unlock(&info->lock);
835 836

		mutex_unlock(&shmem_swaplist_mutex);
837
		BUG_ON(page_mapped(page));
H
Hugh Dickins 已提交
838
		swap_writepage(page, wbc);
L
Linus Torvalds 已提交
839 840 841
		return 0;
	}

842
	mutex_unlock(&shmem_swaplist_mutex);
843
	swapcache_free(swap);
L
Linus Torvalds 已提交
844 845
redirty:
	set_page_dirty(page);
846 847 848 849
	if (wbc->for_reclaim)
		return AOP_WRITEPAGE_ACTIVATE;	/* Return with page locked */
	unlock_page(page);
	return 0;
L
Linus Torvalds 已提交
850 851 852
}

#ifdef CONFIG_NUMA
853
#ifdef CONFIG_TMPFS
854
static void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
855
{
856
	char buffer[64];
857

858
	if (!mpol || mpol->mode == MPOL_DEFAULT)
859
		return;		/* show nothing */
860

861
	mpol_to_str(buffer, sizeof(buffer), mpol);
862 863

	seq_printf(seq, ",mpol=%s", buffer);
864
}
865 866 867 868 869 870 871 872 873 874 875 876

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;
}
877 878
#endif /* CONFIG_TMPFS */

879 880
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
881 882
{
	struct vm_area_struct pvma;
M
Mel Gorman 已提交
883
	struct page *page;
884

L
Linus Torvalds 已提交
885
	/* Create a pseudo vma that just contains the policy */
886
	pvma.vm_start = 0;
887 888
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
889
	pvma.vm_ops = NULL;
M
Mel Gorman 已提交
890 891 892 893 894 895 896 897
	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
898 899
}

900
static struct page *shmem_alloc_page(gfp_t gfp,
901
			struct shmem_inode_info *info, pgoff_t index)
L
Linus Torvalds 已提交
902 903
{
	struct vm_area_struct pvma;
M
Mel Gorman 已提交
904
	struct page *page;
L
Linus Torvalds 已提交
905

906 907
	/* Create a pseudo vma that just contains the policy */
	pvma.vm_start = 0;
908 909
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
910
	pvma.vm_ops = NULL;
911
	pvma.vm_policy = mpol_shared_policy_lookup(&info->policy, index);
912

M
Mel Gorman 已提交
913 914 915 916 917 918
	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
919
}
920 921
#else /* !CONFIG_NUMA */
#ifdef CONFIG_TMPFS
922
static inline void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
923 924 925 926
{
}
#endif /* CONFIG_TMPFS */

927 928
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
929
{
930
	return swapin_readahead(swap, gfp, NULL, 0);
L
Linus Torvalds 已提交
931 932
}

933
static inline struct page *shmem_alloc_page(gfp_t gfp,
934
			struct shmem_inode_info *info, pgoff_t index)
L
Linus Torvalds 已提交
935
{
936
	return alloc_page(gfp);
L
Linus Torvalds 已提交
937
}
938
#endif /* CONFIG_NUMA */
L
Linus Torvalds 已提交
939

940 941 942 943 944 945 946
#if !defined(CONFIG_NUMA) || !defined(CONFIG_TMPFS)
static inline struct mempolicy *shmem_get_sbmpol(struct shmem_sb_info *sbinfo)
{
	return NULL;
}
#endif

947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986
/*
 * 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);
987
	flush_dcache_page(newpage);
988 989 990 991 992 993 994 995 996 997 998 999 1000 1001

	__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);
1002 1003 1004 1005
	if (!error) {
		__inc_zone_page_state(newpage, NR_FILE_PAGES);
		__dec_zone_page_state(oldpage, NR_FILE_PAGES);
	}
1006 1007
	spin_unlock_irq(&swap_mapping->tree_lock);

1008 1009 1010 1011 1012 1013 1014 1015
	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 {
1016
		mem_cgroup_migrate(oldpage, newpage, true);
1017 1018 1019
		lru_cache_add_anon(newpage);
		*pagep = newpage;
	}
1020 1021 1022 1023 1024 1025 1026

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

	unlock_page(oldpage);
	page_cache_release(oldpage);
	page_cache_release(oldpage);
1027
	return error;
1028 1029
}

L
Linus Torvalds 已提交
1030
/*
1031
 * shmem_getpage_gfp - find page in cache, or get from swap, or allocate
L
Linus Torvalds 已提交
1032 1033 1034 1035 1036
 *
 * 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
 */
1037
static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
1038
	struct page **pagep, enum sgp_type sgp, gfp_t gfp, int *fault_type)
L
Linus Torvalds 已提交
1039 1040
{
	struct address_space *mapping = inode->i_mapping;
1041
	struct shmem_inode_info *info;
L
Linus Torvalds 已提交
1042
	struct shmem_sb_info *sbinfo;
1043
	struct mem_cgroup *memcg;
H
Hugh Dickins 已提交
1044
	struct page *page;
L
Linus Torvalds 已提交
1045 1046
	swp_entry_t swap;
	int error;
1047
	int once = 0;
1048
	int alloced = 0;
L
Linus Torvalds 已提交
1049

1050
	if (index > (MAX_LFS_FILESIZE >> PAGE_CACHE_SHIFT))
L
Linus Torvalds 已提交
1051 1052
		return -EFBIG;
repeat:
1053
	swap.val = 0;
1054
	page = find_lock_entry(mapping, index);
1055 1056 1057 1058 1059
	if (radix_tree_exceptional_entry(page)) {
		swap = radix_to_swp_entry(page);
		page = NULL;
	}

1060
	if (sgp != SGP_WRITE && sgp != SGP_FALLOC &&
1061 1062 1063 1064 1065
	    ((loff_t)index << PAGE_CACHE_SHIFT) >= i_size_read(inode)) {
		error = -EINVAL;
		goto failed;
	}

1066 1067 1068
	if (page && sgp == SGP_WRITE)
		mark_page_accessed(page);

1069 1070 1071 1072 1073 1074 1075 1076
	/* fallocated page? */
	if (page && !PageUptodate(page)) {
		if (sgp != SGP_READ)
			goto clear;
		unlock_page(page);
		page_cache_release(page);
		page = NULL;
	}
1077 1078 1079
	if (page || (sgp == SGP_READ && !swap.val)) {
		*pagep = page;
		return 0;
H
Hugh Dickins 已提交
1080 1081 1082
	}

	/*
1083 1084
	 * Fast cache lookup did not find it:
	 * bring it back from swap or allocate.
H
Hugh Dickins 已提交
1085
	 */
1086 1087
	info = SHMEM_I(inode);
	sbinfo = SHMEM_SB(inode->i_sb);
L
Linus Torvalds 已提交
1088 1089 1090

	if (swap.val) {
		/* Look it up and read it in.. */
H
Hugh Dickins 已提交
1091 1092
		page = lookup_swap_cache(swap);
		if (!page) {
L
Linus Torvalds 已提交
1093
			/* here we actually do the io */
1094 1095
			if (fault_type)
				*fault_type |= VM_FAULT_MAJOR;
1096
			page = shmem_swapin(swap, gfp, info, index);
H
Hugh Dickins 已提交
1097
			if (!page) {
1098 1099
				error = -ENOMEM;
				goto failed;
L
Linus Torvalds 已提交
1100 1101 1102 1103
			}
		}

		/* We have to do this with page locked to prevent races */
1104
		lock_page(page);
1105
		if (!PageSwapCache(page) || page_private(page) != swap.val ||
1106
		    !shmem_confirm_swap(mapping, index, swap)) {
1107
			error = -EEXIST;	/* try again */
1108
			goto unlock;
1109
		}
H
Hugh Dickins 已提交
1110
		if (!PageUptodate(page)) {
L
Linus Torvalds 已提交
1111
			error = -EIO;
1112
			goto failed;
L
Linus Torvalds 已提交
1113
		}
1114 1115
		wait_on_page_writeback(page);

1116 1117 1118 1119
		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
1120
		}
H
Hugh Dickins 已提交
1121

1122
		error = mem_cgroup_try_charge(page, current->mm, gfp, &memcg);
1123
		if (!error) {
1124
			error = shmem_add_to_page_cache(page, mapping, index,
1125
						swp_to_radix_entry(swap));
1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
			/*
			 * 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.
			 */
1138 1139
			if (error) {
				mem_cgroup_cancel_charge(page, memcg);
1140
				delete_from_swap_cache(page);
1141
			}
1142
		}
1143 1144 1145
		if (error)
			goto failed;

1146 1147
		mem_cgroup_commit_charge(page, memcg, true);

1148
		spin_lock(&info->lock);
1149
		info->swapped--;
1150
		shmem_recalc_inode(inode);
H
Hugh Dickins 已提交
1151
		spin_unlock(&info->lock);
1152

1153 1154 1155
		if (sgp == SGP_WRITE)
			mark_page_accessed(page);

1156
		delete_from_swap_cache(page);
H
Hugh Dickins 已提交
1157 1158 1159
		set_page_dirty(page);
		swap_free(swap);

1160 1161 1162 1163
	} else {
		if (shmem_acct_block(info->flags)) {
			error = -ENOSPC;
			goto failed;
L
Linus Torvalds 已提交
1164
		}
1165
		if (sbinfo->max_blocks) {
1166
			if (percpu_counter_compare(&sbinfo->used_blocks,
1167 1168 1169 1170
						sbinfo->max_blocks) >= 0) {
				error = -ENOSPC;
				goto unacct;
			}
1171
			percpu_counter_inc(&sbinfo->used_blocks);
1172
		}
L
Linus Torvalds 已提交
1173

1174 1175 1176 1177
		page = shmem_alloc_page(gfp, info, index);
		if (!page) {
			error = -ENOMEM;
			goto decused;
L
Linus Torvalds 已提交
1178 1179
		}

1180
		__SetPageSwapBacked(page);
1181
		__set_page_locked(page);
1182
		if (sgp == SGP_WRITE)
1183
			__SetPageReferenced(page);
1184

1185
		error = mem_cgroup_try_charge(page, current->mm, gfp, &memcg);
1186 1187
		if (error)
			goto decused;
1188
		error = radix_tree_maybe_preload(gfp & GFP_RECLAIM_MASK);
1189 1190
		if (!error) {
			error = shmem_add_to_page_cache(page, mapping, index,
1191
							NULL);
1192 1193 1194
			radix_tree_preload_end();
		}
		if (error) {
1195
			mem_cgroup_cancel_charge(page, memcg);
1196 1197
			goto decused;
		}
1198
		mem_cgroup_commit_charge(page, memcg, false);
1199 1200 1201
		lru_cache_add_anon(page);

		spin_lock(&info->lock);
L
Linus Torvalds 已提交
1202
		info->alloced++;
1203 1204
		inode->i_blocks += BLOCKS_PER_PAGE;
		shmem_recalc_inode(inode);
L
Linus Torvalds 已提交
1205
		spin_unlock(&info->lock);
1206
		alloced = true;
1207

1208
		/*
1209 1210 1211 1212 1213 1214 1215 1216 1217
		 * 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.
1218 1219 1220 1221 1222 1223
		 */
		if (sgp != SGP_WRITE) {
			clear_highpage(page);
			flush_dcache_page(page);
			SetPageUptodate(page);
		}
1224
		if (sgp == SGP_DIRTY)
H
Hugh Dickins 已提交
1225
			set_page_dirty(page);
L
Linus Torvalds 已提交
1226
	}
1227

1228
	/* Perhaps the file has been truncated since we checked */
1229
	if (sgp != SGP_WRITE && sgp != SGP_FALLOC &&
1230 1231
	    ((loff_t)index << PAGE_CACHE_SHIFT) >= i_size_read(inode)) {
		error = -EINVAL;
1232 1233 1234 1235
		if (alloced)
			goto trunc;
		else
			goto failed;
H
Hugh Dickins 已提交
1236
	}
1237 1238
	*pagep = page;
	return 0;
L
Linus Torvalds 已提交
1239

1240
	/*
1241
	 * Error recovery.
1242
	 */
1243
trunc:
1244
	info = SHMEM_I(inode);
1245 1246 1247 1248 1249
	ClearPageDirty(page);
	delete_from_page_cache(page);
	spin_lock(&info->lock);
	info->alloced--;
	inode->i_blocks -= BLOCKS_PER_PAGE;
1250
	spin_unlock(&info->lock);
1251
decused:
1252
	sbinfo = SHMEM_SB(inode->i_sb);
1253 1254 1255 1256 1257
	if (sbinfo->max_blocks)
		percpu_counter_add(&sbinfo->used_blocks, -1);
unacct:
	shmem_unacct_blocks(info->flags, 1);
failed:
1258 1259 1260 1261
	if (swap.val && error != -EINVAL &&
	    !shmem_confirm_swap(mapping, index, swap))
		error = -EEXIST;
unlock:
H
Hugh Dickins 已提交
1262
	if (page) {
1263
		unlock_page(page);
H
Hugh Dickins 已提交
1264
		page_cache_release(page);
1265 1266 1267 1268 1269 1270
	}
	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
1271
		goto repeat;
1272
	}
1273
	if (error == -EEXIST)	/* from above or from radix_tree_insert */
1274 1275
		goto repeat;
	return error;
L
Linus Torvalds 已提交
1276 1277
}

N
Nick Piggin 已提交
1278
static int shmem_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
L
Linus Torvalds 已提交
1279
{
A
Al Viro 已提交
1280
	struct inode *inode = file_inode(vma->vm_file);
L
Linus Torvalds 已提交
1281
	int error;
1282
	int ret = VM_FAULT_LOCKED;
L
Linus Torvalds 已提交
1283

1284 1285 1286 1287
	/*
	 * Trinity finds that probing a hole which tmpfs is punching can
	 * prevent the hole-punch from ever completing: which in turn
	 * locks writers out with its hold on i_mutex.  So refrain from
1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299
	 * faulting pages into the hole while it's being punched.  Although
	 * shmem_undo_range() does remove the additions, it may be unable to
	 * keep up, as each new page needs its own unmap_mapping_range() call,
	 * and the i_mmap tree grows ever slower to scan if new vmas are added.
	 *
	 * It does not matter if we sometimes reach this check just before the
	 * hole-punch begins, so that one fault then races with the punch:
	 * we just need to make racing faults a rare case.
	 *
	 * The implementation below would be much simpler if we just used a
	 * standard mutex or completion: but we cannot take i_mutex in fault,
	 * and bloating every shmem inode for this unlikely case would be sad.
1300 1301 1302 1303 1304 1305
	 */
	if (unlikely(inode->i_private)) {
		struct shmem_falloc *shmem_falloc;

		spin_lock(&inode->i_lock);
		shmem_falloc = inode->i_private;
1306 1307 1308 1309 1310 1311 1312 1313
		if (shmem_falloc &&
		    shmem_falloc->waitq &&
		    vmf->pgoff >= shmem_falloc->start &&
		    vmf->pgoff < shmem_falloc->next) {
			wait_queue_head_t *shmem_falloc_waitq;
			DEFINE_WAIT(shmem_fault_wait);

			ret = VM_FAULT_NOPAGE;
1314 1315
			if ((vmf->flags & FAULT_FLAG_ALLOW_RETRY) &&
			   !(vmf->flags & FAULT_FLAG_RETRY_NOWAIT)) {
1316
				/* It's polite to up mmap_sem if we can */
1317
				up_read(&vma->vm_mm->mmap_sem);
1318
				ret = VM_FAULT_RETRY;
1319
			}
1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337

			shmem_falloc_waitq = shmem_falloc->waitq;
			prepare_to_wait(shmem_falloc_waitq, &shmem_fault_wait,
					TASK_UNINTERRUPTIBLE);
			spin_unlock(&inode->i_lock);
			schedule();

			/*
			 * shmem_falloc_waitq points into the shmem_fallocate()
			 * stack of the hole-punching task: shmem_falloc_waitq
			 * is usually invalid by the time we reach here, but
			 * finish_wait() does not dereference it in that case;
			 * though i_lock needed lest racing with wake_up_all().
			 */
			spin_lock(&inode->i_lock);
			finish_wait(shmem_falloc_waitq, &shmem_fault_wait);
			spin_unlock(&inode->i_lock);
			return ret;
1338
		}
1339
		spin_unlock(&inode->i_lock);
1340 1341
	}

1342
	error = shmem_getpage(inode, vmf->pgoff, &vmf->page, SGP_CACHE, &ret);
N
Nick Piggin 已提交
1343 1344
	if (error)
		return ((error == -ENOMEM) ? VM_FAULT_OOM : VM_FAULT_SIGBUS);
1345

1346 1347 1348 1349
	if (ret & VM_FAULT_MAJOR) {
		count_vm_event(PGMAJFAULT);
		mem_cgroup_count_vm_event(vma->vm_mm, PGMAJFAULT);
	}
1350
	return ret;
L
Linus Torvalds 已提交
1351 1352 1353
}

#ifdef CONFIG_NUMA
1354
static int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *mpol)
L
Linus Torvalds 已提交
1355
{
A
Al Viro 已提交
1356
	struct inode *inode = file_inode(vma->vm_file);
1357
	return mpol_set_shared_policy(&SHMEM_I(inode)->policy, vma, mpol);
L
Linus Torvalds 已提交
1358 1359
}

A
Adrian Bunk 已提交
1360 1361
static struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
					  unsigned long addr)
L
Linus Torvalds 已提交
1362
{
A
Al Viro 已提交
1363
	struct inode *inode = file_inode(vma->vm_file);
1364
	pgoff_t index;
L
Linus Torvalds 已提交
1365

1366 1367
	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
1368 1369 1370 1371 1372
}
#endif

int shmem_lock(struct file *file, int lock, struct user_struct *user)
{
A
Al Viro 已提交
1373
	struct inode *inode = file_inode(file);
L
Linus Torvalds 已提交
1374 1375 1376 1377 1378 1379 1380 1381
	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;
1382
		mapping_set_unevictable(file->f_mapping);
L
Linus Torvalds 已提交
1383 1384 1385 1386
	}
	if (!lock && (info->flags & VM_LOCKED) && user) {
		user_shm_unlock(inode->i_size, user);
		info->flags &= ~VM_LOCKED;
1387
		mapping_clear_unevictable(file->f_mapping);
L
Linus Torvalds 已提交
1388 1389
	}
	retval = 0;
1390

L
Linus Torvalds 已提交
1391 1392 1393 1394 1395
out_nomem:
	spin_unlock(&info->lock);
	return retval;
}

1396
static int shmem_mmap(struct file *file, struct vm_area_struct *vma)
L
Linus Torvalds 已提交
1397 1398 1399 1400 1401 1402
{
	file_accessed(file);
	vma->vm_ops = &shmem_vm_ops;
	return 0;
}

1403
static struct inode *shmem_get_inode(struct super_block *sb, const struct inode *dir,
1404
				     umode_t mode, dev_t dev, unsigned long flags)
L
Linus Torvalds 已提交
1405 1406 1407 1408 1409
{
	struct inode *inode;
	struct shmem_inode_info *info;
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);

1410 1411
	if (shmem_reserve_inode(sb))
		return NULL;
L
Linus Torvalds 已提交
1412 1413 1414

	inode = new_inode(sb);
	if (inode) {
1415
		inode->i_ino = get_next_ino();
1416
		inode_init_owner(inode, dir, mode);
L
Linus Torvalds 已提交
1417 1418 1419
		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;
1420
		inode->i_generation = get_seconds();
L
Linus Torvalds 已提交
1421 1422 1423
		info = SHMEM_I(inode);
		memset(info, 0, (char *)inode - (char *)info);
		spin_lock_init(&info->lock);
D
David Herrmann 已提交
1424
		info->seals = F_SEAL_SEAL;
1425
		info->flags = flags & VM_NORESERVE;
L
Linus Torvalds 已提交
1426
		INIT_LIST_HEAD(&info->swaplist);
1427
		simple_xattrs_init(&info->xattrs);
1428
		cache_no_acl(inode);
L
Linus Torvalds 已提交
1429 1430 1431

		switch (mode & S_IFMT) {
		default:
1432
			inode->i_op = &shmem_special_inode_operations;
L
Linus Torvalds 已提交
1433 1434 1435
			init_special_inode(inode, mode, dev);
			break;
		case S_IFREG:
1436
			inode->i_mapping->a_ops = &shmem_aops;
L
Linus Torvalds 已提交
1437 1438
			inode->i_op = &shmem_inode_operations;
			inode->i_fop = &shmem_file_operations;
1439 1440
			mpol_shared_policy_init(&info->policy,
						 shmem_get_sbmpol(sbinfo));
L
Linus Torvalds 已提交
1441 1442
			break;
		case S_IFDIR:
1443
			inc_nlink(inode);
L
Linus Torvalds 已提交
1444 1445 1446 1447 1448 1449 1450 1451 1452 1453
			/* 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.
			 */
1454
			mpol_shared_policy_init(&info->policy, NULL);
L
Linus Torvalds 已提交
1455 1456
			break;
		}
1457 1458
	} else
		shmem_free_inode(sb);
L
Linus Torvalds 已提交
1459 1460 1461
	return inode;
}

1462 1463 1464 1465 1466
bool shmem_mapping(struct address_space *mapping)
{
	return mapping->backing_dev_info == &shmem_backing_dev_info;
}

L
Linus Torvalds 已提交
1467
#ifdef CONFIG_TMPFS
1468
static const struct inode_operations shmem_symlink_inode_operations;
1469
static const struct inode_operations shmem_short_symlink_operations;
L
Linus Torvalds 已提交
1470

1471 1472 1473 1474 1475 1476
#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
1477
static int
N
Nick Piggin 已提交
1478 1479 1480
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
1481
{
N
Nick Piggin 已提交
1482
	struct inode *inode = mapping->host;
D
David Herrmann 已提交
1483
	struct shmem_inode_info *info = SHMEM_I(inode);
N
Nick Piggin 已提交
1484
	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
D
David Herrmann 已提交
1485 1486 1487 1488 1489 1490 1491 1492 1493

	/* i_mutex is held by caller */
	if (unlikely(info->seals)) {
		if (info->seals & F_SEAL_WRITE)
			return -EPERM;
		if ((info->seals & F_SEAL_GROW) && pos + len > inode->i_size)
			return -EPERM;
	}

1494
	return shmem_getpage(inode, index, pagep, SGP_WRITE, NULL);
N
Nick Piggin 已提交
1495 1496 1497 1498 1499 1500 1501 1502 1503
}

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
1504 1505 1506
	if (pos + copied > inode->i_size)
		i_size_write(inode, pos + copied);

1507 1508 1509 1510 1511 1512 1513 1514
	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
1515
	set_page_dirty(page);
1516
	unlock_page(page);
N
Nick Piggin 已提交
1517 1518 1519
	page_cache_release(page);

	return copied;
L
Linus Torvalds 已提交
1520 1521
}

A
Al Viro 已提交
1522
static ssize_t shmem_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
L
Linus Torvalds 已提交
1523
{
1524 1525
	struct file *file = iocb->ki_filp;
	struct inode *inode = file_inode(file);
L
Linus Torvalds 已提交
1526
	struct address_space *mapping = inode->i_mapping;
1527 1528
	pgoff_t index;
	unsigned long offset;
1529
	enum sgp_type sgp = SGP_READ;
1530
	int error = 0;
A
Al Viro 已提交
1531
	ssize_t retval = 0;
1532
	loff_t *ppos = &iocb->ki_pos;
1533 1534 1535 1536 1537 1538

	/*
	 * 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.
	 */
A
Al Viro 已提交
1539
	if (!iter_is_iovec(to))
1540
		sgp = SGP_DIRTY;
L
Linus Torvalds 已提交
1541 1542 1543 1544 1545 1546

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

	for (;;) {
		struct page *page = NULL;
1547 1548
		pgoff_t end_index;
		unsigned long nr, ret;
L
Linus Torvalds 已提交
1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559
		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;
		}

1560 1561 1562 1563
		error = shmem_getpage(inode, index, &page, sgp, NULL);
		if (error) {
			if (error == -EINVAL)
				error = 0;
L
Linus Torvalds 已提交
1564 1565
			break;
		}
H
Hugh Dickins 已提交
1566 1567
		if (page)
			unlock_page(page);
L
Linus Torvalds 已提交
1568 1569 1570

		/*
		 * We must evaluate after, since reads (unlike writes)
1571
		 * are called without i_mutex protection against truncate
L
Linus Torvalds 已提交
1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598
		 */
		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
1599
		} else {
L
Linus Torvalds 已提交
1600
			page = ZERO_PAGE(0);
N
Nick Piggin 已提交
1601 1602
			page_cache_get(page);
		}
L
Linus Torvalds 已提交
1603 1604 1605 1606 1607

		/*
		 * Ok, we have the page, and it's up-to-date, so
		 * now we can copy it to user space...
		 */
A
Al Viro 已提交
1608
		ret = copy_page_to_iter(page, offset, nr, to);
1609
		retval += ret;
L
Linus Torvalds 已提交
1610 1611 1612 1613 1614
		offset += ret;
		index += offset >> PAGE_CACHE_SHIFT;
		offset &= ~PAGE_CACHE_MASK;

		page_cache_release(page);
A
Al Viro 已提交
1615
		if (!iov_iter_count(to))
L
Linus Torvalds 已提交
1616
			break;
1617 1618 1619 1620
		if (ret < nr) {
			error = -EFAULT;
			break;
		}
L
Linus Torvalds 已提交
1621 1622 1623 1624
		cond_resched();
	}

	*ppos = ((loff_t) index << PAGE_CACHE_SHIFT) + offset;
1625 1626
	file_accessed(file);
	return retval ? retval : error;
L
Linus Torvalds 已提交
1627 1628
}

1629 1630 1631 1632 1633
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;
1634
	struct inode *inode = mapping->host;
1635 1636 1637 1638 1639 1640 1641 1642 1643 1644
	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,
1645
		.nr_pages_max = PIPE_DEF_BUFFERS,
1646 1647 1648 1649 1650
		.flags = flags,
		.ops = &page_cache_pipe_buf_ops,
		.spd_release = spd_release_page,
	};

1651
	isize = i_size_read(inode);
1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664
	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;
1665
	nr_pages = min(req_pages, spd.nr_pages_max);
1666 1667 1668 1669 1670 1671

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

1672 1673 1674 1675 1676
	while (spd.nr_pages < nr_pages) {
		error = shmem_getpage(inode, index, &page, SGP_CACHE, NULL);
		if (error)
			break;
		unlock_page(page);
1677 1678 1679 1680 1681 1682 1683
		spd.pages[spd.nr_pages++] = page;
		index++;
	}

	index = *ppos >> PAGE_CACHE_SHIFT;
	nr_pages = spd.nr_pages;
	spd.nr_pages = 0;
1684

1685 1686 1687 1688 1689 1690 1691 1692 1693
	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];

1694 1695 1696 1697
		if (!PageUptodate(page) || page->mapping != mapping) {
			error = shmem_getpage(inode, index, &page,
							SGP_CACHE, NULL);
			if (error)
1698
				break;
1699 1700 1701
			unlock_page(page);
			page_cache_release(spd.pages[page_nr]);
			spd.pages[page_nr] = page;
1702
		}
1703 1704

		isize = i_size_read(inode);
1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733
		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);

1734
	splice_shrink_spd(&spd);
1735 1736 1737 1738 1739 1740 1741 1742

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

1743 1744 1745 1746
/*
 * llseek SEEK_DATA or SEEK_HOLE through the radix_tree.
 */
static pgoff_t shmem_seek_hole_data(struct address_space *mapping,
1747
				    pgoff_t index, pgoff_t end, int whence)
1748 1749 1750 1751 1752 1753 1754 1755 1756 1757
{
	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) {
1758
		pvec.nr = find_get_entries(mapping, index,
1759 1760
					pvec.nr, pvec.pages, indices);
		if (!pvec.nr) {
1761
			if (whence == SEEK_DATA)
1762 1763 1764 1765 1766
				index = end;
			break;
		}
		for (i = 0; i < pvec.nr; i++, index++) {
			if (index < indices[i]) {
1767
				if (whence == SEEK_HOLE) {
1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778
					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 ||
1779 1780
			    (page && whence == SEEK_DATA) ||
			    (!page && whence == SEEK_HOLE)) {
1781 1782 1783 1784
				done = true;
				break;
			}
		}
1785
		pagevec_remove_exceptionals(&pvec);
1786 1787 1788 1789 1790 1791 1792
		pagevec_release(&pvec);
		pvec.nr = PAGEVEC_SIZE;
		cond_resched();
	}
	return index;
}

1793
static loff_t shmem_file_llseek(struct file *file, loff_t offset, int whence)
1794 1795 1796 1797 1798 1799
{
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
	pgoff_t start, end;
	loff_t new_offset;

1800 1801
	if (whence != SEEK_DATA && whence != SEEK_HOLE)
		return generic_file_llseek_size(file, offset, whence,
1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812
					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;
1813
		new_offset = shmem_seek_hole_data(mapping, start, end, whence);
1814 1815 1816 1817
		new_offset <<= PAGE_CACHE_SHIFT;
		if (new_offset > offset) {
			if (new_offset < inode->i_size)
				offset = new_offset;
1818
			else if (whence == SEEK_DATA)
1819 1820 1821 1822 1823 1824
				offset = -ENXIO;
			else
				offset = inode->i_size;
		}
	}

1825 1826
	if (offset >= 0)
		offset = vfs_setpos(file, offset, MAX_LFS_FILESIZE);
1827 1828 1829 1830
	mutex_unlock(&inode->i_mutex);
	return offset;
}

1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879
/*
 * We need a tag: a new tag would expand every radix_tree_node by 8 bytes,
 * so reuse a tag which we firmly believe is never set or cleared on shmem.
 */
#define SHMEM_TAG_PINNED        PAGECACHE_TAG_TOWRITE
#define LAST_SCAN               4       /* about 150ms max */

static void shmem_tag_pins(struct address_space *mapping)
{
	struct radix_tree_iter iter;
	void **slot;
	pgoff_t start;
	struct page *page;

	lru_add_drain();
	start = 0;
	rcu_read_lock();

restart:
	radix_tree_for_each_slot(slot, &mapping->page_tree, &iter, start) {
		page = radix_tree_deref_slot(slot);
		if (!page || radix_tree_exception(page)) {
			if (radix_tree_deref_retry(page))
				goto restart;
		} else if (page_count(page) - page_mapcount(page) > 1) {
			spin_lock_irq(&mapping->tree_lock);
			radix_tree_tag_set(&mapping->page_tree, iter.index,
					   SHMEM_TAG_PINNED);
			spin_unlock_irq(&mapping->tree_lock);
		}

		if (need_resched()) {
			cond_resched_rcu();
			start = iter.index + 1;
			goto restart;
		}
	}
	rcu_read_unlock();
}

/*
 * Setting SEAL_WRITE requires us to verify there's no pending writer. However,
 * via get_user_pages(), drivers might have some pending I/O without any active
 * user-space mappings (eg., direct-IO, AIO). Therefore, we look at all pages
 * and see whether it has an elevated ref-count. If so, we tag them and wait for
 * them to be dropped.
 * The caller must guarantee that no new user will acquire writable references
 * to those pages to avoid races.
 */
D
David Herrmann 已提交
1880 1881
static int shmem_wait_for_pins(struct address_space *mapping)
{
1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941
	struct radix_tree_iter iter;
	void **slot;
	pgoff_t start;
	struct page *page;
	int error, scan;

	shmem_tag_pins(mapping);

	error = 0;
	for (scan = 0; scan <= LAST_SCAN; scan++) {
		if (!radix_tree_tagged(&mapping->page_tree, SHMEM_TAG_PINNED))
			break;

		if (!scan)
			lru_add_drain_all();
		else if (schedule_timeout_killable((HZ << scan) / 200))
			scan = LAST_SCAN;

		start = 0;
		rcu_read_lock();
restart:
		radix_tree_for_each_tagged(slot, &mapping->page_tree, &iter,
					   start, SHMEM_TAG_PINNED) {

			page = radix_tree_deref_slot(slot);
			if (radix_tree_exception(page)) {
				if (radix_tree_deref_retry(page))
					goto restart;

				page = NULL;
			}

			if (page &&
			    page_count(page) - page_mapcount(page) != 1) {
				if (scan < LAST_SCAN)
					goto continue_resched;

				/*
				 * On the last scan, we clean up all those tags
				 * we inserted; but make a note that we still
				 * found pages pinned.
				 */
				error = -EBUSY;
			}

			spin_lock_irq(&mapping->tree_lock);
			radix_tree_tag_clear(&mapping->page_tree,
					     iter.index, SHMEM_TAG_PINNED);
			spin_unlock_irq(&mapping->tree_lock);
continue_resched:
			if (need_resched()) {
				cond_resched_rcu();
				start = iter.index + 1;
				goto restart;
			}
		}
		rcu_read_unlock();
	}

	return error;
D
David Herrmann 已提交
1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051
}

#define F_ALL_SEALS (F_SEAL_SEAL | \
		     F_SEAL_SHRINK | \
		     F_SEAL_GROW | \
		     F_SEAL_WRITE)

int shmem_add_seals(struct file *file, unsigned int seals)
{
	struct inode *inode = file_inode(file);
	struct shmem_inode_info *info = SHMEM_I(inode);
	int error;

	/*
	 * SEALING
	 * Sealing allows multiple parties to share a shmem-file but restrict
	 * access to a specific subset of file operations. Seals can only be
	 * added, but never removed. This way, mutually untrusted parties can
	 * share common memory regions with a well-defined policy. A malicious
	 * peer can thus never perform unwanted operations on a shared object.
	 *
	 * Seals are only supported on special shmem-files and always affect
	 * the whole underlying inode. Once a seal is set, it may prevent some
	 * kinds of access to the file. Currently, the following seals are
	 * defined:
	 *   SEAL_SEAL: Prevent further seals from being set on this file
	 *   SEAL_SHRINK: Prevent the file from shrinking
	 *   SEAL_GROW: Prevent the file from growing
	 *   SEAL_WRITE: Prevent write access to the file
	 *
	 * As we don't require any trust relationship between two parties, we
	 * must prevent seals from being removed. Therefore, sealing a file
	 * only adds a given set of seals to the file, it never touches
	 * existing seals. Furthermore, the "setting seals"-operation can be
	 * sealed itself, which basically prevents any further seal from being
	 * added.
	 *
	 * Semantics of sealing are only defined on volatile files. Only
	 * anonymous shmem files support sealing. More importantly, seals are
	 * never written to disk. Therefore, there's no plan to support it on
	 * other file types.
	 */

	if (file->f_op != &shmem_file_operations)
		return -EINVAL;
	if (!(file->f_mode & FMODE_WRITE))
		return -EPERM;
	if (seals & ~(unsigned int)F_ALL_SEALS)
		return -EINVAL;

	mutex_lock(&inode->i_mutex);

	if (info->seals & F_SEAL_SEAL) {
		error = -EPERM;
		goto unlock;
	}

	if ((seals & F_SEAL_WRITE) && !(info->seals & F_SEAL_WRITE)) {
		error = mapping_deny_writable(file->f_mapping);
		if (error)
			goto unlock;

		error = shmem_wait_for_pins(file->f_mapping);
		if (error) {
			mapping_allow_writable(file->f_mapping);
			goto unlock;
		}
	}

	info->seals |= seals;
	error = 0;

unlock:
	mutex_unlock(&inode->i_mutex);
	return error;
}
EXPORT_SYMBOL_GPL(shmem_add_seals);

int shmem_get_seals(struct file *file)
{
	if (file->f_op != &shmem_file_operations)
		return -EINVAL;

	return SHMEM_I(file_inode(file))->seals;
}
EXPORT_SYMBOL_GPL(shmem_get_seals);

long shmem_fcntl(struct file *file, unsigned int cmd, unsigned long arg)
{
	long error;

	switch (cmd) {
	case F_ADD_SEALS:
		/* disallow upper 32bit */
		if (arg > UINT_MAX)
			return -EINVAL;

		error = shmem_add_seals(file, arg);
		break;
	case F_GET_SEALS:
		error = shmem_get_seals(file);
		break;
	default:
		error = -EINVAL;
		break;
	}

	return error;
}

2052 2053 2054
static long shmem_fallocate(struct file *file, int mode, loff_t offset,
							 loff_t len)
{
A
Al Viro 已提交
2055
	struct inode *inode = file_inode(file);
2056
	struct shmem_sb_info *sbinfo = SHMEM_SB(inode->i_sb);
D
David Herrmann 已提交
2057
	struct shmem_inode_info *info = SHMEM_I(inode);
2058
	struct shmem_falloc shmem_falloc;
2059 2060
	pgoff_t start, index, end;
	int error;
2061

2062 2063 2064
	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
		return -EOPNOTSUPP;

2065 2066 2067 2068 2069 2070
	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;
2071
		DECLARE_WAIT_QUEUE_HEAD_ONSTACK(shmem_falloc_waitq);
2072

D
David Herrmann 已提交
2073 2074 2075 2076 2077 2078
		/* protected by i_mutex */
		if (info->seals & F_SEAL_WRITE) {
			error = -EPERM;
			goto out;
		}

2079
		shmem_falloc.waitq = &shmem_falloc_waitq;
2080 2081 2082 2083 2084 2085
		shmem_falloc.start = unmap_start >> PAGE_SHIFT;
		shmem_falloc.next = (unmap_end + 1) >> PAGE_SHIFT;
		spin_lock(&inode->i_lock);
		inode->i_private = &shmem_falloc;
		spin_unlock(&inode->i_lock);

2086 2087 2088 2089 2090
		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 */
2091 2092 2093 2094 2095

		spin_lock(&inode->i_lock);
		inode->i_private = NULL;
		wake_up_all(&shmem_falloc_waitq);
		spin_unlock(&inode->i_lock);
2096
		error = 0;
2097
		goto out;
2098 2099 2100 2101 2102 2103 2104
	}

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

D
David Herrmann 已提交
2105 2106 2107 2108 2109
	if ((info->seals & F_SEAL_GROW) && offset + len > inode->i_size) {
		error = -EPERM;
		goto out;
	}

2110 2111 2112 2113 2114 2115
	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;
2116 2117
	}

2118
	shmem_falloc.waitq = NULL;
2119 2120 2121 2122 2123 2124 2125 2126
	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);

2127 2128 2129 2130 2131 2132 2133 2134 2135
	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;
2136 2137
		else if (shmem_falloc.nr_unswapped > shmem_falloc.nr_falloced)
			error = -ENOMEM;
2138
		else
2139
			error = shmem_getpage(inode, index, &page, SGP_FALLOC,
2140 2141
									NULL);
		if (error) {
2142 2143 2144 2145
			/* Remove the !PageUptodate pages we added */
			shmem_undo_range(inode,
				(loff_t)start << PAGE_CACHE_SHIFT,
				(loff_t)index << PAGE_CACHE_SHIFT, true);
2146
			goto undone;
2147 2148
		}

2149 2150 2151 2152 2153 2154 2155 2156
		/*
		 * 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++;

2157
		/*
2158 2159 2160
		 * 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
2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172
		 * 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;
2173 2174 2175 2176
undone:
	spin_lock(&inode->i_lock);
	inode->i_private = NULL;
	spin_unlock(&inode->i_lock);
2177
out:
2178 2179 2180 2181
	mutex_unlock(&inode->i_mutex);
	return error;
}

2182
static int shmem_statfs(struct dentry *dentry, struct kstatfs *buf)
L
Linus Torvalds 已提交
2183
{
2184
	struct shmem_sb_info *sbinfo = SHMEM_SB(dentry->d_sb);
L
Linus Torvalds 已提交
2185 2186 2187 2188

	buf->f_type = TMPFS_MAGIC;
	buf->f_bsize = PAGE_CACHE_SIZE;
	buf->f_namelen = NAME_MAX;
2189
	if (sbinfo->max_blocks) {
L
Linus Torvalds 已提交
2190
		buf->f_blocks = sbinfo->max_blocks;
2191 2192 2193
		buf->f_bavail =
		buf->f_bfree  = sbinfo->max_blocks -
				percpu_counter_sum(&sbinfo->used_blocks);
2194 2195
	}
	if (sbinfo->max_inodes) {
L
Linus Torvalds 已提交
2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206
		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
2207
shmem_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev)
L
Linus Torvalds 已提交
2208
{
2209
	struct inode *inode;
L
Linus Torvalds 已提交
2210 2211
	int error = -ENOSPC;

2212
	inode = shmem_get_inode(dir->i_sb, dir, mode, dev, VM_NORESERVE);
L
Linus Torvalds 已提交
2213
	if (inode) {
C
Christoph Hellwig 已提交
2214 2215 2216
		error = simple_acl_create(dir, inode);
		if (error)
			goto out_iput;
2217
		error = security_inode_init_security(inode, dir,
2218
						     &dentry->d_name,
2219
						     shmem_initxattrs, NULL);
C
Christoph Hellwig 已提交
2220 2221
		if (error && error != -EOPNOTSUPP)
			goto out_iput;
2222

A
Al Viro 已提交
2223
		error = 0;
L
Linus Torvalds 已提交
2224 2225 2226 2227 2228 2229
		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
2230 2231 2232
out_iput:
	iput(inode);
	return error;
L
Linus Torvalds 已提交
2233 2234
}

2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245
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
2246 2247 2248 2249 2250
		if (error && error != -EOPNOTSUPP)
			goto out_iput;
		error = simple_acl_create(dir, inode);
		if (error)
			goto out_iput;
2251 2252 2253
		d_tmpfile(dentry, inode);
	}
	return error;
C
Christoph Hellwig 已提交
2254 2255 2256
out_iput:
	iput(inode);
	return error;
2257 2258
}

2259
static int shmem_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
2260 2261 2262 2263 2264
{
	int error;

	if ((error = shmem_mknod(dir, dentry, mode | S_IFDIR, 0)))
		return error;
2265
	inc_nlink(dir);
L
Linus Torvalds 已提交
2266 2267 2268
	return 0;
}

A
Al Viro 已提交
2269
static int shmem_create(struct inode *dir, struct dentry *dentry, umode_t mode,
A
Al Viro 已提交
2270
		bool excl)
L
Linus Torvalds 已提交
2271 2272 2273 2274 2275 2276 2277 2278 2279 2280
{
	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;
2281
	int ret;
L
Linus Torvalds 已提交
2282 2283 2284 2285 2286 2287

	/*
	 * 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.
	 */
2288 2289 2290
	ret = shmem_reserve_inode(inode->i_sb);
	if (ret)
		goto out;
L
Linus Torvalds 已提交
2291 2292 2293

	dir->i_size += BOGO_DIRENT_SIZE;
	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
2294
	inc_nlink(inode);
A
Al Viro 已提交
2295
	ihold(inode);	/* New dentry reference */
L
Linus Torvalds 已提交
2296 2297
	dget(dentry);		/* Extra pinning count for the created dentry */
	d_instantiate(dentry, inode);
2298 2299
out:
	return ret;
L
Linus Torvalds 已提交
2300 2301 2302 2303 2304 2305
}

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

2306 2307
	if (inode->i_nlink > 1 && !S_ISDIR(inode->i_mode))
		shmem_free_inode(inode->i_sb);
L
Linus Torvalds 已提交
2308 2309 2310

	dir->i_size -= BOGO_DIRENT_SIZE;
	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
2311
	drop_nlink(inode);
L
Linus Torvalds 已提交
2312 2313 2314 2315 2316 2317 2318 2319 2320
	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;

2321 2322
	drop_nlink(dentry->d_inode);
	drop_nlink(dir);
L
Linus Torvalds 已提交
2323 2324 2325
	return shmem_unlink(dir, dentry);
}

M
Miklos Szeredi 已提交
2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347
static int shmem_exchange(struct inode *old_dir, struct dentry *old_dentry, struct inode *new_dir, struct dentry *new_dentry)
{
	bool old_is_dir = S_ISDIR(old_dentry->d_inode->i_mode);
	bool new_is_dir = S_ISDIR(new_dentry->d_inode->i_mode);

	if (old_dir != new_dir && old_is_dir != new_is_dir) {
		if (old_is_dir) {
			drop_nlink(old_dir);
			inc_nlink(new_dir);
		} else {
			drop_nlink(new_dir);
			inc_nlink(old_dir);
		}
	}
	old_dir->i_ctime = old_dir->i_mtime =
	new_dir->i_ctime = new_dir->i_mtime =
	old_dentry->d_inode->i_ctime =
	new_dentry->d_inode->i_ctime = CURRENT_TIME;

	return 0;
}

M
Miklos Szeredi 已提交
2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373
static int shmem_whiteout(struct inode *old_dir, struct dentry *old_dentry)
{
	struct dentry *whiteout;
	int error;

	whiteout = d_alloc(old_dentry->d_parent, &old_dentry->d_name);
	if (!whiteout)
		return -ENOMEM;

	error = shmem_mknod(old_dir, whiteout,
			    S_IFCHR | WHITEOUT_MODE, WHITEOUT_DEV);
	dput(whiteout);
	if (error)
		return error;

	/*
	 * Cheat and hash the whiteout while the old dentry is still in
	 * place, instead of playing games with FS_RENAME_DOES_D_MOVE.
	 *
	 * d_lookup() will consistently find one of them at this point,
	 * not sure which one, but that isn't even important.
	 */
	d_rehash(whiteout);
	return 0;
}

L
Linus Torvalds 已提交
2374 2375 2376 2377 2378 2379
/*
 * 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.
 */
M
Miklos Szeredi 已提交
2380
static int shmem_rename2(struct inode *old_dir, struct dentry *old_dentry, struct inode *new_dir, struct dentry *new_dentry, unsigned int flags)
L
Linus Torvalds 已提交
2381 2382 2383 2384
{
	struct inode *inode = old_dentry->d_inode;
	int they_are_dirs = S_ISDIR(inode->i_mode);

M
Miklos Szeredi 已提交
2385
	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT))
M
Miklos Szeredi 已提交
2386 2387
		return -EINVAL;

M
Miklos Szeredi 已提交
2388 2389 2390
	if (flags & RENAME_EXCHANGE)
		return shmem_exchange(old_dir, old_dentry, new_dir, new_dentry);

L
Linus Torvalds 已提交
2391 2392 2393
	if (!simple_empty(new_dentry))
		return -ENOTEMPTY;

M
Miklos Szeredi 已提交
2394 2395 2396 2397 2398 2399 2400 2401
	if (flags & RENAME_WHITEOUT) {
		int error;

		error = shmem_whiteout(old_dir, old_dentry);
		if (error)
			return error;
	}

L
Linus Torvalds 已提交
2402 2403
	if (new_dentry->d_inode) {
		(void) shmem_unlink(new_dir, new_dentry);
2404 2405
		if (they_are_dirs) {
			drop_nlink(new_dentry->d_inode);
2406
			drop_nlink(old_dir);
2407
		}
L
Linus Torvalds 已提交
2408
	} else if (they_are_dirs) {
2409
		drop_nlink(old_dir);
2410
		inc_nlink(new_dir);
L
Linus Torvalds 已提交
2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425
	}

	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
2426
	struct page *page;
L
Linus Torvalds 已提交
2427 2428 2429 2430 2431 2432 2433
	char *kaddr;
	struct shmem_inode_info *info;

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

2434
	inode = shmem_get_inode(dir->i_sb, dir, S_IFLNK|S_IRWXUGO, 0, VM_NORESERVE);
L
Linus Torvalds 已提交
2435 2436 2437
	if (!inode)
		return -ENOSPC;

2438
	error = security_inode_init_security(inode, dir, &dentry->d_name,
2439
					     shmem_initxattrs, NULL);
2440 2441 2442 2443 2444 2445 2446 2447
	if (error) {
		if (error != -EOPNOTSUPP) {
			iput(inode);
			return error;
		}
		error = 0;
	}

L
Linus Torvalds 已提交
2448 2449
	info = SHMEM_I(inode);
	inode->i_size = len-1;
2450 2451 2452 2453 2454 2455 2456
	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
2457 2458 2459 2460 2461 2462
	} else {
		error = shmem_getpage(inode, 0, &page, SGP_WRITE, NULL);
		if (error) {
			iput(inode);
			return error;
		}
2463
		inode->i_mapping->a_ops = &shmem_aops;
L
Linus Torvalds 已提交
2464
		inode->i_op = &shmem_symlink_inode_operations;
2465
		kaddr = kmap_atomic(page);
L
Linus Torvalds 已提交
2466
		memcpy(kaddr, symname, len);
2467
		kunmap_atomic(kaddr);
2468
		SetPageUptodate(page);
L
Linus Torvalds 已提交
2469
		set_page_dirty(page);
2470
		unlock_page(page);
L
Linus Torvalds 已提交
2471 2472 2473 2474 2475 2476 2477 2478 2479
		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;
}

2480
static void *shmem_follow_short_symlink(struct dentry *dentry, struct nameidata *nd)
L
Linus Torvalds 已提交
2481
{
2482
	nd_set_link(nd, SHMEM_I(dentry->d_inode)->symlink);
2483
	return NULL;
L
Linus Torvalds 已提交
2484 2485
}

2486
static void *shmem_follow_link(struct dentry *dentry, struct nameidata *nd)
L
Linus Torvalds 已提交
2487 2488
{
	struct page *page = NULL;
2489 2490
	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
2491 2492
	if (page)
		unlock_page(page);
2493
	return page;
L
Linus Torvalds 已提交
2494 2495
}

2496
static void shmem_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie)
L
Linus Torvalds 已提交
2497 2498
{
	if (!IS_ERR(nd_get_link(nd))) {
2499
		struct page *page = cookie;
L
Linus Torvalds 已提交
2500 2501 2502 2503 2504 2505
		kunmap(page);
		mark_page_accessed(page);
		page_cache_release(page);
	}
}

2506
#ifdef CONFIG_TMPFS_XATTR
2507
/*
2508 2509
 * Superblocks without xattr inode operations may get some security.* xattr
 * support from the LSM "for free". As soon as we have any other xattrs
2510 2511 2512 2513
 * like ACLs, we also need to implement the security.* handlers at
 * filesystem level, though.
 */

2514 2515 2516 2517 2518 2519 2520 2521 2522
/*
 * 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;
2523
	struct simple_xattr *new_xattr;
2524 2525 2526
	size_t len;

	for (xattr = xattr_array; xattr->name != NULL; xattr++) {
2527
		new_xattr = simple_xattr_alloc(xattr->value, xattr->value_len);
2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543
		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);

2544
		simple_xattr_list_add(&info->xattrs, new_xattr);
2545 2546 2547 2548 2549
	}

	return 0;
}

2550
static const struct xattr_handler *shmem_xattr_handlers[] = {
2551
#ifdef CONFIG_TMPFS_POSIX_ACL
C
Christoph Hellwig 已提交
2552 2553
	&posix_acl_access_xattr_handler,
	&posix_acl_default_xattr_handler,
2554
#endif
2555 2556
	NULL
};
2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579

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)
{
2580
	struct shmem_inode_info *info = SHMEM_I(dentry->d_inode);
2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594
	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;

2595
	return simple_xattr_get(&info->xattrs, name, buffer, size);
2596 2597 2598 2599 2600
}

static int shmem_setxattr(struct dentry *dentry, const char *name,
			  const void *value, size_t size, int flags)
{
2601
	struct shmem_inode_info *info = SHMEM_I(dentry->d_inode);
2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615
	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;

2616
	return simple_xattr_set(&info->xattrs, name, value, size, flags);
2617 2618 2619 2620
}

static int shmem_removexattr(struct dentry *dentry, const char *name)
{
2621
	struct shmem_inode_info *info = SHMEM_I(dentry->d_inode);
2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635
	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;

2636
	return simple_xattr_remove(&info->xattrs, name);
2637 2638 2639 2640
}

static ssize_t shmem_listxattr(struct dentry *dentry, char *buffer, size_t size)
{
2641 2642
	struct shmem_inode_info *info = SHMEM_I(dentry->d_inode);
	return simple_xattr_list(&info->xattrs, buffer, size);
2643 2644 2645
}
#endif /* CONFIG_TMPFS_XATTR */

2646
static const struct inode_operations shmem_short_symlink_operations = {
2647
	.readlink	= generic_readlink,
2648
	.follow_link	= shmem_follow_short_symlink,
2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665
#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,
2666
#endif
2667
};
2668

2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681
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
2682 2683
static struct dentry *shmem_fh_to_dentry(struct super_block *sb,
		struct fid *fid, int fh_len, int fh_type)
2684 2685
{
	struct inode *inode;
C
Christoph Hellwig 已提交
2686
	struct dentry *dentry = NULL;
2687
	u64 inum;
C
Christoph Hellwig 已提交
2688 2689 2690

	if (fh_len < 3)
		return NULL;
2691

2692 2693 2694
	inum = fid->raw[2];
	inum = (inum << 32) | fid->raw[1];

C
Christoph Hellwig 已提交
2695 2696
	inode = ilookup5(sb, (unsigned long)(inum + fid->raw[0]),
			shmem_match, fid->raw);
2697
	if (inode) {
C
Christoph Hellwig 已提交
2698
		dentry = d_find_alias(inode);
2699 2700 2701
		iput(inode);
	}

C
Christoph Hellwig 已提交
2702
	return dentry;
2703 2704
}

A
Al Viro 已提交
2705 2706
static int shmem_encode_fh(struct inode *inode, __u32 *fh, int *len,
				struct inode *parent)
2707
{
2708 2709
	if (*len < 3) {
		*len = 3;
2710
		return FILEID_INVALID;
2711
	}
2712

A
Al Viro 已提交
2713
	if (inode_unhashed(inode)) {
2714 2715 2716 2717 2718 2719 2720
		/* 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
2721
		if (inode_unhashed(inode))
2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734
			__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;
}

2735
static const struct export_operations shmem_export_ops = {
2736 2737
	.get_parent     = shmem_get_parent,
	.encode_fh      = shmem_encode_fh,
C
Christoph Hellwig 已提交
2738
	.fh_to_dentry	= shmem_fh_to_dentry,
2739 2740
};

2741 2742
static int shmem_parse_options(char *options, struct shmem_sb_info *sbinfo,
			       bool remount)
L
Linus Torvalds 已提交
2743 2744
{
	char *this_char, *value, *rest;
G
Greg Thelen 已提交
2745
	struct mempolicy *mpol = NULL;
2746 2747
	uid_t uid;
	gid_t gid;
L
Linus Torvalds 已提交
2748

2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765
	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
2766 2767 2768 2769 2770 2771 2772 2773
		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
2774
			goto error;
L
Linus Torvalds 已提交
2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787
		}

		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;
2788 2789
			sbinfo->max_blocks =
				DIV_ROUND_UP(size, PAGE_CACHE_SIZE);
L
Linus Torvalds 已提交
2790
		} else if (!strcmp(this_char,"nr_blocks")) {
2791
			sbinfo->max_blocks = memparse(value, &rest);
L
Linus Torvalds 已提交
2792 2793 2794
			if (*rest)
				goto bad_val;
		} else if (!strcmp(this_char,"nr_inodes")) {
2795
			sbinfo->max_inodes = memparse(value, &rest);
L
Linus Torvalds 已提交
2796 2797 2798
			if (*rest)
				goto bad_val;
		} else if (!strcmp(this_char,"mode")) {
2799
			if (remount)
L
Linus Torvalds 已提交
2800
				continue;
2801
			sbinfo->mode = simple_strtoul(value, &rest, 8) & 07777;
L
Linus Torvalds 已提交
2802 2803 2804
			if (*rest)
				goto bad_val;
		} else if (!strcmp(this_char,"uid")) {
2805
			if (remount)
L
Linus Torvalds 已提交
2806
				continue;
2807
			uid = simple_strtoul(value, &rest, 0);
L
Linus Torvalds 已提交
2808 2809
			if (*rest)
				goto bad_val;
2810 2811 2812
			sbinfo->uid = make_kuid(current_user_ns(), uid);
			if (!uid_valid(sbinfo->uid))
				goto bad_val;
L
Linus Torvalds 已提交
2813
		} else if (!strcmp(this_char,"gid")) {
2814
			if (remount)
L
Linus Torvalds 已提交
2815
				continue;
2816
			gid = simple_strtoul(value, &rest, 0);
L
Linus Torvalds 已提交
2817 2818
			if (*rest)
				goto bad_val;
2819 2820 2821
			sbinfo->gid = make_kgid(current_user_ns(), gid);
			if (!gid_valid(sbinfo->gid))
				goto bad_val;
2822
		} else if (!strcmp(this_char,"mpol")) {
G
Greg Thelen 已提交
2823 2824 2825
			mpol_put(mpol);
			mpol = NULL;
			if (mpol_parse_str(value, &mpol))
2826
				goto bad_val;
L
Linus Torvalds 已提交
2827 2828 2829
		} else {
			printk(KERN_ERR "tmpfs: Bad mount option %s\n",
			       this_char);
G
Greg Thelen 已提交
2830
			goto error;
L
Linus Torvalds 已提交
2831 2832
		}
	}
G
Greg Thelen 已提交
2833
	sbinfo->mpol = mpol;
L
Linus Torvalds 已提交
2834 2835 2836 2837 2838
	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
2839 2840
error:
	mpol_put(mpol);
L
Linus Torvalds 已提交
2841 2842 2843 2844 2845 2846 2847
	return 1;

}

static int shmem_remount_fs(struct super_block *sb, int *flags, char *data)
{
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);
2848
	struct shmem_sb_info config = *sbinfo;
2849 2850 2851
	unsigned long inodes;
	int error = -EINVAL;

2852
	config.mpol = NULL;
2853
	if (shmem_parse_options(data, &config, true))
2854
		return error;
L
Linus Torvalds 已提交
2855

2856 2857
	spin_lock(&sbinfo->stat_lock);
	inodes = sbinfo->max_inodes - sbinfo->free_inodes;
2858
	if (percpu_counter_compare(&sbinfo->used_blocks, config.max_blocks) > 0)
2859
		goto out;
2860
	if (config.max_inodes < inodes)
2861 2862
		goto out;
	/*
2863
	 * Those tests disallow limited->unlimited while any are in use;
2864 2865 2866
	 * but we must separately disallow unlimited->limited, because
	 * in that case we have no record of how much is already in use.
	 */
2867
	if (config.max_blocks && !sbinfo->max_blocks)
2868
		goto out;
2869
	if (config.max_inodes && !sbinfo->max_inodes)
2870 2871 2872
		goto out;

	error = 0;
2873 2874 2875
	sbinfo->max_blocks  = config.max_blocks;
	sbinfo->max_inodes  = config.max_inodes;
	sbinfo->free_inodes = config.max_inodes - inodes;
2876

2877 2878 2879 2880 2881 2882 2883
	/*
	 * Preserve previous mempolicy unless mpol remount option was specified.
	 */
	if (config.mpol) {
		mpol_put(sbinfo->mpol);
		sbinfo->mpol = config.mpol;	/* transfers initial ref */
	}
2884 2885 2886
out:
	spin_unlock(&sbinfo->stat_lock);
	return error;
L
Linus Torvalds 已提交
2887
}
2888

2889
static int shmem_show_options(struct seq_file *seq, struct dentry *root)
2890
{
2891
	struct shmem_sb_info *sbinfo = SHMEM_SB(root->d_sb);
2892 2893 2894 2895 2896 2897 2898

	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))
2899
		seq_printf(seq, ",mode=%03ho", sbinfo->mode);
2900 2901 2902 2903 2904 2905
	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));
2906
	shmem_show_mpol(seq, sbinfo->mpol);
2907 2908
	return 0;
}
D
David Herrmann 已提交
2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 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 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979

#define MFD_NAME_PREFIX "memfd:"
#define MFD_NAME_PREFIX_LEN (sizeof(MFD_NAME_PREFIX) - 1)
#define MFD_NAME_MAX_LEN (NAME_MAX - MFD_NAME_PREFIX_LEN)

#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING)

SYSCALL_DEFINE2(memfd_create,
		const char __user *, uname,
		unsigned int, flags)
{
	struct shmem_inode_info *info;
	struct file *file;
	int fd, error;
	char *name;
	long len;

	if (flags & ~(unsigned int)MFD_ALL_FLAGS)
		return -EINVAL;

	/* length includes terminating zero */
	len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1);
	if (len <= 0)
		return -EFAULT;
	if (len > MFD_NAME_MAX_LEN + 1)
		return -EINVAL;

	name = kmalloc(len + MFD_NAME_PREFIX_LEN, GFP_TEMPORARY);
	if (!name)
		return -ENOMEM;

	strcpy(name, MFD_NAME_PREFIX);
	if (copy_from_user(&name[MFD_NAME_PREFIX_LEN], uname, len)) {
		error = -EFAULT;
		goto err_name;
	}

	/* terminating-zero may have changed after strnlen_user() returned */
	if (name[len + MFD_NAME_PREFIX_LEN - 1]) {
		error = -EFAULT;
		goto err_name;
	}

	fd = get_unused_fd_flags((flags & MFD_CLOEXEC) ? O_CLOEXEC : 0);
	if (fd < 0) {
		error = fd;
		goto err_name;
	}

	file = shmem_file_setup(name, 0, VM_NORESERVE);
	if (IS_ERR(file)) {
		error = PTR_ERR(file);
		goto err_fd;
	}
	info = SHMEM_I(file_inode(file));
	file->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
	file->f_flags |= O_RDWR | O_LARGEFILE;
	if (flags & MFD_ALLOW_SEALING)
		info->seals &= ~F_SEAL_SEAL;

	fd_install(fd, file);
	kfree(name);
	return fd;

err_fd:
	put_unused_fd(fd);
err_name:
	kfree(name);
	return error;
}

2980
#endif /* CONFIG_TMPFS */
L
Linus Torvalds 已提交
2981 2982 2983

static void shmem_put_super(struct super_block *sb)
{
2984 2985 2986
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);

	percpu_counter_destroy(&sbinfo->used_blocks);
G
Greg Thelen 已提交
2987
	mpol_put(sbinfo->mpol);
2988
	kfree(sbinfo);
L
Linus Torvalds 已提交
2989 2990 2991
	sb->s_fs_info = NULL;
}

2992
int shmem_fill_super(struct super_block *sb, void *data, int silent)
L
Linus Torvalds 已提交
2993 2994
{
	struct inode *inode;
2995
	struct shmem_sb_info *sbinfo;
2996 2997 2998
	int err = -ENOMEM;

	/* Round up to L1_CACHE_BYTES to resist false sharing */
2999
	sbinfo = kzalloc(max((int)sizeof(struct shmem_sb_info),
3000 3001 3002 3003 3004
				L1_CACHE_BYTES), GFP_KERNEL);
	if (!sbinfo)
		return -ENOMEM;

	sbinfo->mode = S_IRWXUGO | S_ISVTX;
3005 3006
	sbinfo->uid = current_fsuid();
	sbinfo->gid = current_fsgid();
3007
	sb->s_fs_info = sbinfo;
L
Linus Torvalds 已提交
3008

3009
#ifdef CONFIG_TMPFS
L
Linus Torvalds 已提交
3010 3011 3012 3013 3014
	/*
	 * 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.
	 */
3015
	if (!(sb->s_flags & MS_KERNMOUNT)) {
3016 3017 3018 3019 3020 3021
		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;
		}
3022 3023
	} else {
		sb->s_flags |= MS_NOUSER;
L
Linus Torvalds 已提交
3024
	}
3025
	sb->s_export_op = &shmem_export_ops;
H
Hugh Dickins 已提交
3026
	sb->s_flags |= MS_NOSEC;
L
Linus Torvalds 已提交
3027 3028 3029 3030
#else
	sb->s_flags |= MS_NOUSER;
#endif

3031
	spin_lock_init(&sbinfo->stat_lock);
3032
	if (percpu_counter_init(&sbinfo->used_blocks, 0, GFP_KERNEL))
3033
		goto failed;
3034
	sbinfo->free_inodes = sbinfo->max_inodes;
3035

3036
	sb->s_maxbytes = MAX_LFS_FILESIZE;
L
Linus Torvalds 已提交
3037 3038 3039 3040
	sb->s_blocksize = PAGE_CACHE_SIZE;
	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
	sb->s_magic = TMPFS_MAGIC;
	sb->s_op = &shmem_ops;
3041
	sb->s_time_gran = 1;
3042
#ifdef CONFIG_TMPFS_XATTR
3043
	sb->s_xattr = shmem_xattr_handlers;
3044 3045
#endif
#ifdef CONFIG_TMPFS_POSIX_ACL
3046 3047
	sb->s_flags |= MS_POSIXACL;
#endif
3048

3049
	inode = shmem_get_inode(sb, NULL, S_IFDIR | sbinfo->mode, 0, VM_NORESERVE);
L
Linus Torvalds 已提交
3050 3051
	if (!inode)
		goto failed;
3052 3053
	inode->i_uid = sbinfo->uid;
	inode->i_gid = sbinfo->gid;
A
Al Viro 已提交
3054 3055
	sb->s_root = d_make_root(inode);
	if (!sb->s_root)
3056
		goto failed;
L
Linus Torvalds 已提交
3057 3058 3059 3060 3061 3062 3063
	return 0;

failed:
	shmem_put_super(sb);
	return err;
}

3064
static struct kmem_cache *shmem_inode_cachep;
L
Linus Torvalds 已提交
3065 3066 3067

static struct inode *shmem_alloc_inode(struct super_block *sb)
{
3068 3069 3070
	struct shmem_inode_info *info;
	info = kmem_cache_alloc(shmem_inode_cachep, GFP_KERNEL);
	if (!info)
L
Linus Torvalds 已提交
3071
		return NULL;
3072
	return &info->vfs_inode;
L
Linus Torvalds 已提交
3073 3074
}

3075
static void shmem_destroy_callback(struct rcu_head *head)
N
Nick Piggin 已提交
3076 3077 3078 3079 3080
{
	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
3081 3082
static void shmem_destroy_inode(struct inode *inode)
{
3083
	if (S_ISREG(inode->i_mode))
L
Linus Torvalds 已提交
3084
		mpol_free_shared_policy(&SHMEM_I(inode)->policy);
3085
	call_rcu(&inode->i_rcu, shmem_destroy_callback);
L
Linus Torvalds 已提交
3086 3087
}

3088
static void shmem_init_inode(void *foo)
L
Linus Torvalds 已提交
3089
{
3090 3091
	struct shmem_inode_info *info = foo;
	inode_init_once(&info->vfs_inode);
L
Linus Torvalds 已提交
3092 3093
}

3094
static int shmem_init_inodecache(void)
L
Linus Torvalds 已提交
3095 3096 3097
{
	shmem_inode_cachep = kmem_cache_create("shmem_inode_cache",
				sizeof(struct shmem_inode_info),
3098
				0, SLAB_PANIC, shmem_init_inode);
L
Linus Torvalds 已提交
3099 3100 3101
	return 0;
}

3102
static void shmem_destroy_inodecache(void)
L
Linus Torvalds 已提交
3103
{
3104
	kmem_cache_destroy(shmem_inode_cachep);
L
Linus Torvalds 已提交
3105 3106
}

3107
static const struct address_space_operations shmem_aops = {
L
Linus Torvalds 已提交
3108
	.writepage	= shmem_writepage,
3109
	.set_page_dirty	= __set_page_dirty_no_writeback,
L
Linus Torvalds 已提交
3110
#ifdef CONFIG_TMPFS
N
Nick Piggin 已提交
3111 3112
	.write_begin	= shmem_write_begin,
	.write_end	= shmem_write_end,
L
Linus Torvalds 已提交
3113
#endif
3114
#ifdef CONFIG_MIGRATION
3115
	.migratepage	= migrate_page,
3116
#endif
3117
	.error_remove_page = generic_error_remove_page,
L
Linus Torvalds 已提交
3118 3119
};

3120
static const struct file_operations shmem_file_operations = {
L
Linus Torvalds 已提交
3121 3122
	.mmap		= shmem_mmap,
#ifdef CONFIG_TMPFS
3123
	.llseek		= shmem_file_llseek,
A
Al Viro 已提交
3124
	.read		= new_sync_read,
3125
	.write		= new_sync_write,
A
Al Viro 已提交
3126
	.read_iter	= shmem_file_read_iter,
3127
	.write_iter	= generic_file_write_iter,
3128
	.fsync		= noop_fsync,
3129
	.splice_read	= shmem_file_splice_read,
3130
	.splice_write	= iter_file_splice_write,
3131
	.fallocate	= shmem_fallocate,
L
Linus Torvalds 已提交
3132 3133 3134
#endif
};

3135
static const struct inode_operations shmem_inode_operations = {
3136
	.setattr	= shmem_setattr,
3137 3138 3139 3140 3141
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
C
Christoph Hellwig 已提交
3142
	.set_acl	= simple_set_acl,
3143
#endif
L
Linus Torvalds 已提交
3144 3145
};

3146
static const struct inode_operations shmem_dir_inode_operations = {
L
Linus Torvalds 已提交
3147 3148 3149 3150 3151 3152 3153 3154 3155
#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,
M
Miklos Szeredi 已提交
3156
	.rename2	= shmem_rename2,
3157
	.tmpfile	= shmem_tmpfile,
L
Linus Torvalds 已提交
3158
#endif
3159 3160 3161 3162 3163 3164
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
#endif
3165
#ifdef CONFIG_TMPFS_POSIX_ACL
3166
	.setattr	= shmem_setattr,
C
Christoph Hellwig 已提交
3167
	.set_acl	= simple_set_acl,
3168 3169 3170
#endif
};

3171
static const struct inode_operations shmem_special_inode_operations = {
3172 3173 3174 3175 3176 3177
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
#endif
3178
#ifdef CONFIG_TMPFS_POSIX_ACL
3179
	.setattr	= shmem_setattr,
C
Christoph Hellwig 已提交
3180
	.set_acl	= simple_set_acl,
3181
#endif
L
Linus Torvalds 已提交
3182 3183
};

3184
static const struct super_operations shmem_ops = {
L
Linus Torvalds 已提交
3185 3186 3187 3188 3189
	.alloc_inode	= shmem_alloc_inode,
	.destroy_inode	= shmem_destroy_inode,
#ifdef CONFIG_TMPFS
	.statfs		= shmem_statfs,
	.remount_fs	= shmem_remount_fs,
3190
	.show_options	= shmem_show_options,
L
Linus Torvalds 已提交
3191
#endif
A
Al Viro 已提交
3192
	.evict_inode	= shmem_evict_inode,
L
Linus Torvalds 已提交
3193 3194 3195 3196
	.drop_inode	= generic_delete_inode,
	.put_super	= shmem_put_super,
};

3197
static const struct vm_operations_struct shmem_vm_ops = {
3198
	.fault		= shmem_fault,
3199
	.map_pages	= filemap_map_pages,
L
Linus Torvalds 已提交
3200 3201 3202 3203 3204 3205
#ifdef CONFIG_NUMA
	.set_policy     = shmem_set_policy,
	.get_policy     = shmem_get_policy,
#endif
};

A
Al Viro 已提交
3206 3207
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
3208
{
A
Al Viro 已提交
3209
	return mount_nodev(fs_type, flags, data, shmem_fill_super);
L
Linus Torvalds 已提交
3210 3211
}

3212
static struct file_system_type shmem_fs_type = {
L
Linus Torvalds 已提交
3213 3214
	.owner		= THIS_MODULE,
	.name		= "tmpfs",
A
Al Viro 已提交
3215
	.mount		= shmem_mount,
L
Linus Torvalds 已提交
3216
	.kill_sb	= kill_litter_super,
3217
	.fs_flags	= FS_USERNS_MOUNT,
L
Linus Torvalds 已提交
3218 3219
};

3220
int __init shmem_init(void)
L
Linus Torvalds 已提交
3221 3222 3223
{
	int error;

3224 3225 3226 3227
	/* If rootfs called this, don't re-init */
	if (shmem_inode_cachep)
		return 0;

P
Peter Zijlstra 已提交
3228 3229 3230 3231
	error = bdi_init(&shmem_backing_dev_info);
	if (error)
		goto out4;

3232
	error = shmem_init_inodecache();
L
Linus Torvalds 已提交
3233 3234 3235
	if (error)
		goto out3;

3236
	error = register_filesystem(&shmem_fs_type);
L
Linus Torvalds 已提交
3237 3238 3239 3240
	if (error) {
		printk(KERN_ERR "Could not register tmpfs\n");
		goto out2;
	}
3241

3242
	shm_mnt = kern_mount(&shmem_fs_type);
L
Linus Torvalds 已提交
3243 3244 3245 3246 3247 3248 3249 3250
	if (IS_ERR(shm_mnt)) {
		error = PTR_ERR(shm_mnt);
		printk(KERN_ERR "Could not kern_mount tmpfs\n");
		goto out1;
	}
	return 0;

out1:
3251
	unregister_filesystem(&shmem_fs_type);
L
Linus Torvalds 已提交
3252
out2:
3253
	shmem_destroy_inodecache();
L
Linus Torvalds 已提交
3254
out3:
P
Peter Zijlstra 已提交
3255 3256
	bdi_destroy(&shmem_backing_dev_info);
out4:
L
Linus Torvalds 已提交
3257 3258 3259
	shm_mnt = ERR_PTR(error);
	return error;
}
3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271

#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.
 */

3272
static struct file_system_type shmem_fs_type = {
3273
	.name		= "tmpfs",
A
Al Viro 已提交
3274
	.mount		= ramfs_mount,
3275
	.kill_sb	= kill_litter_super,
3276
	.fs_flags	= FS_USERNS_MOUNT,
3277 3278
};

3279
int __init shmem_init(void)
3280
{
3281
	BUG_ON(register_filesystem(&shmem_fs_type) != 0);
3282

3283
	shm_mnt = kern_mount(&shmem_fs_type);
3284 3285 3286 3287 3288
	BUG_ON(IS_ERR(shm_mnt));

	return 0;
}

3289
int shmem_unuse(swp_entry_t swap, struct page *page)
3290 3291 3292 3293
{
	return 0;
}

H
Hugh Dickins 已提交
3294 3295 3296 3297 3298
int shmem_lock(struct file *file, int lock, struct user_struct *user)
{
	return 0;
}

3299 3300 3301 3302
void shmem_unlock_mapping(struct address_space *mapping)
{
}

3303
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
3304
{
3305
	truncate_inode_pages_range(inode->i_mapping, lstart, lend);
3306 3307 3308
}
EXPORT_SYMBOL_GPL(shmem_truncate_range);

3309 3310
#define shmem_vm_ops				generic_file_vm_ops
#define shmem_file_operations			ramfs_file_operations
3311
#define shmem_get_inode(sb, dir, mode, dev, flags)	ramfs_get_inode(sb, dir, mode, dev)
3312 3313
#define shmem_acct_size(flags, size)		0
#define shmem_unacct_size(flags, size)		do {} while (0)
3314 3315 3316 3317

#endif /* CONFIG_SHMEM */

/* common code */
L
Linus Torvalds 已提交
3318

3319
static struct dentry_operations anon_ops = {
3320
	.d_dname = simple_dname
3321 3322
};

3323 3324
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
3325
{
A
Al Viro 已提交
3326
	struct file *res;
L
Linus Torvalds 已提交
3327
	struct inode *inode;
3328
	struct path path;
3329
	struct super_block *sb;
L
Linus Torvalds 已提交
3330 3331 3332
	struct qstr this;

	if (IS_ERR(shm_mnt))
A
Al Viro 已提交
3333
		return ERR_CAST(shm_mnt);
L
Linus Torvalds 已提交
3334

3335
	if (size < 0 || size > MAX_LFS_FILESIZE)
L
Linus Torvalds 已提交
3336 3337 3338 3339 3340
		return ERR_PTR(-EINVAL);

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

A
Al Viro 已提交
3341
	res = ERR_PTR(-ENOMEM);
L
Linus Torvalds 已提交
3342 3343 3344
	this.name = name;
	this.len = strlen(name);
	this.hash = 0; /* will go */
3345
	sb = shm_mnt->mnt_sb;
3346
	path.mnt = mntget(shm_mnt);
3347
	path.dentry = d_alloc_pseudo(sb, &this);
3348
	if (!path.dentry)
L
Linus Torvalds 已提交
3349
		goto put_memory;
3350
	d_set_d_op(path.dentry, &anon_ops);
L
Linus Torvalds 已提交
3351

A
Al Viro 已提交
3352
	res = ERR_PTR(-ENOSPC);
3353
	inode = shmem_get_inode(sb, NULL, S_IFREG | S_IRWXUGO, 0, flags);
L
Linus Torvalds 已提交
3354
	if (!inode)
3355
		goto put_memory;
L
Linus Torvalds 已提交
3356

3357
	inode->i_flags |= i_flags;
3358
	d_instantiate(path.dentry, inode);
L
Linus Torvalds 已提交
3359
	inode->i_size = size;
3360
	clear_nlink(inode);	/* It is unlinked */
A
Al Viro 已提交
3361 3362
	res = ERR_PTR(ramfs_nommu_expand_for_mapping(inode, size));
	if (IS_ERR(res))
3363
		goto put_path;
3364

A
Al Viro 已提交
3365
	res = alloc_file(&path, FMODE_WRITE | FMODE_READ,
3366
		  &shmem_file_operations);
A
Al Viro 已提交
3367
	if (IS_ERR(res))
3368
		goto put_path;
3369

A
Al Viro 已提交
3370
	return res;
L
Linus Torvalds 已提交
3371 3372 3373

put_memory:
	shmem_unacct_size(flags, size);
3374 3375
put_path:
	path_put(&path);
A
Al Viro 已提交
3376
	return res;
L
Linus Torvalds 已提交
3377
}
3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403

/**
 * 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);
}
3404
EXPORT_SYMBOL_GPL(shmem_file_setup);
L
Linus Torvalds 已提交
3405

3406
/**
L
Linus Torvalds 已提交
3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424
 * 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;
}
3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437

/**
 * 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.
 *
3438 3439
 * i915_gem_object_get_pages_gtt() mixes __GFP_NORETRY | __GFP_NOWARN in
 * with the mapping_gfp_mask(), to avoid OOMing the machine unnecessarily.
3440 3441 3442 3443
 */
struct page *shmem_read_mapping_page_gfp(struct address_space *mapping,
					 pgoff_t index, gfp_t gfp)
{
3444 3445
#ifdef CONFIG_SHMEM
	struct inode *inode = mapping->host;
H
Hugh Dickins 已提交
3446
	struct page *page;
3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459
	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
	 */
3460
	return read_cache_page_gfp(mapping, index, gfp);
3461
#endif
3462 3463
}
EXPORT_SYMBOL_GPL(shmem_read_mapping_page_gfp);