shmem.c 89.8 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/uio.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 LIST_HEAD(shmem_swaplist);
195
static DEFINE_MUTEX(shmem_swaplist_mutex);
L
Linus Torvalds 已提交
196

197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221
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);
	}
}

222
/**
223
 * shmem_recalc_inode - recalculate the block usage of an inode
L
Linus Torvalds 已提交
224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240
 * @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) {
241 242 243
		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
244
		info->alloced -= freed;
245
		inode->i_blocks -= freed * BLOCKS_PER_PAGE;
L
Linus Torvalds 已提交
246 247 248 249
		shmem_unacct_blocks(info->flags, freed);
	}
}

250 251 252 253 254 255 256
/*
 * 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;
257
	void *item;
258 259

	VM_BUG_ON(!expected);
260
	VM_BUG_ON(!replacement);
261
	pslot = radix_tree_lookup_slot(&mapping->page_tree, index);
262 263 264
	if (!pslot)
		return -ENOENT;
	item = radix_tree_deref_slot_protected(pslot, &mapping->tree_lock);
265 266
	if (item != expected)
		return -ENOENT;
267
	radix_tree_replace_slot(pslot, replacement);
268 269 270
	return 0;
}

271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288
/*
 * 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);
}

289 290 291 292 293
/*
 * 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,
294
				   pgoff_t index, void *expected)
295
{
296
	int error;
297

298 299
	VM_BUG_ON_PAGE(!PageLocked(page), page);
	VM_BUG_ON_PAGE(!PageSwapBacked(page), page);
300

301 302 303 304 305
	page_cache_get(page);
	page->mapping = mapping;
	page->index = index;

	spin_lock_irq(&mapping->tree_lock);
306
	if (!expected)
307 308 309 310
		error = radix_tree_insert(&mapping->page_tree, index, page);
	else
		error = shmem_radix_tree_replace(mapping, index, expected,
								 page);
311
	if (!error) {
312 313 314 315 316 317 318 319
		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);
320 321 322 323
	}
	return error;
}

324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342
/*
 * 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);
}

343 344 345 346 347 348
/*
 * 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)
{
349
	void *old;
350 351

	spin_lock_irq(&mapping->tree_lock);
352
	old = radix_tree_delete_item(&mapping->page_tree, index, radswap);
353
	spin_unlock_irq(&mapping->tree_lock);
354 355 356 357
	if (old != radswap)
		return -ENOENT;
	free_swap_and_cache(radix_to_swp_entry(radswap));
	return 0;
358 359
}

360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
/*
 * 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.
		 */
378 379
		pvec.nr = find_get_entries(mapping, index,
					   PAGEVEC_SIZE, pvec.pages, indices);
380 381 382
		if (!pvec.nr)
			break;
		index = indices[pvec.nr - 1] + 1;
383
		pagevec_remove_exceptionals(&pvec);
384 385 386 387
		check_move_unevictable_pages(pvec.pages, pvec.nr);
		pagevec_release(&pvec);
		cond_resched();
	}
388 389 390 391
}

/*
 * Remove range of pages and swap entries from radix tree, and free them.
392
 * If !unfalloc, truncate or punch hole; if unfalloc, undo failed fallocate.
393
 */
394 395
static void shmem_undo_range(struct inode *inode, loff_t lstart, loff_t lend,
								 bool unfalloc)
L
Linus Torvalds 已提交
396
{
397
	struct address_space *mapping = inode->i_mapping;
L
Linus Torvalds 已提交
398
	struct shmem_inode_info *info = SHMEM_I(inode);
399
	pgoff_t start = (lstart + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
400 401 402
	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);
403
	struct pagevec pvec;
404 405
	pgoff_t indices[PAGEVEC_SIZE];
	long nr_swaps_freed = 0;
406
	pgoff_t index;
407 408
	int i;

409 410
	if (lend == -1)
		end = -1;	/* unsigned, so actually very big */
411 412 413

	pagevec_init(&pvec, 0);
	index = start;
414
	while (index < end) {
415 416 417
		pvec.nr = find_get_entries(mapping, index,
			min(end - index, (pgoff_t)PAGEVEC_SIZE),
			pvec.pages, indices);
418 419
		if (!pvec.nr)
			break;
420 421 422
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

423
			index = indices[i];
424
			if (index >= end)
425 426
				break;

427
			if (radix_tree_exceptional_entry(page)) {
428 429
				if (unfalloc)
					continue;
430 431
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
432
				continue;
433 434 435
			}

			if (!trylock_page(page))
436
				continue;
437 438
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
439
					VM_BUG_ON_PAGE(PageWriteback(page), page);
440 441
					truncate_inode_page(mapping, page);
				}
442 443 444
			}
			unlock_page(page);
		}
445
		pagevec_remove_exceptionals(&pvec);
446
		pagevec_release(&pvec);
447 448 449
		cond_resched();
		index++;
	}
L
Linus Torvalds 已提交
450

451
	if (partial_start) {
452 453 454
		struct page *page = NULL;
		shmem_getpage(inode, start - 1, &page, SGP_READ, NULL);
		if (page) {
455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470
			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);
471 472 473 474 475
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}
476 477
	if (start >= end)
		return;
478 479

	index = start;
480
	while (index < end) {
481
		cond_resched();
482 483

		pvec.nr = find_get_entries(mapping, index,
484
				min(end - index, (pgoff_t)PAGEVEC_SIZE),
485
				pvec.pages, indices);
486
		if (!pvec.nr) {
487 488
			/* If all gone or hole-punch or unfalloc, we're done */
			if (index == start || end != -1)
489
				break;
490
			/* But if truncating, restart to make sure all gone */
491 492 493 494 495 496
			index = start;
			continue;
		}
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

497
			index = indices[i];
498
			if (index >= end)
499 500
				break;

501
			if (radix_tree_exceptional_entry(page)) {
502 503
				if (unfalloc)
					continue;
504 505 506 507 508 509
				if (shmem_free_swap(mapping, index, page)) {
					/* Swap was replaced by page: retry */
					index--;
					break;
				}
				nr_swaps_freed++;
510 511 512
				continue;
			}

513
			lock_page(page);
514 515
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
516
					VM_BUG_ON_PAGE(PageWriteback(page), page);
517
					truncate_inode_page(mapping, page);
518 519 520 521 522
				} else {
					/* Page was replaced by swap: retry */
					unlock_page(page);
					index--;
					break;
523
				}
524
			}
525 526
			unlock_page(page);
		}
527
		pagevec_remove_exceptionals(&pvec);
528
		pagevec_release(&pvec);
529 530
		index++;
	}
531

L
Linus Torvalds 已提交
532
	spin_lock(&info->lock);
533
	info->swapped -= nr_swaps_freed;
L
Linus Torvalds 已提交
534 535
	shmem_recalc_inode(inode);
	spin_unlock(&info->lock);
536
}
L
Linus Torvalds 已提交
537

538 539 540
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
{
	shmem_undo_range(inode, lstart, lend, false);
541
	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
L
Linus Torvalds 已提交
542
}
543
EXPORT_SYMBOL_GPL(shmem_truncate_range);
L
Linus Torvalds 已提交
544

545 546 547 548 549 550
static int shmem_getattr(struct vfsmount *mnt, struct dentry *dentry,
			 struct kstat *stat)
{
	struct inode *inode = dentry->d_inode;
	struct shmem_inode_info *info = SHMEM_I(inode);

551 552 553 554 555
	if (info->alloced - info->swapped != inode->i_mapping->nrpages) {
		spin_lock(&info->lock);
		shmem_recalc_inode(inode);
		spin_unlock(&info->lock);
	}
556 557 558 559
	generic_fillattr(inode, stat);
	return 0;
}

560
static int shmem_setattr(struct dentry *dentry, struct iattr *attr)
L
Linus Torvalds 已提交
561
{
562
	struct inode *inode = d_inode(dentry);
D
David Herrmann 已提交
563
	struct shmem_inode_info *info = SHMEM_I(inode);
L
Linus Torvalds 已提交
564 565
	int error;

566 567 568 569
	error = inode_change_ok(inode, attr);
	if (error)
		return error;

570 571 572
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
		loff_t oldsize = inode->i_size;
		loff_t newsize = attr->ia_size;
573

D
David Herrmann 已提交
574 575 576 577 578
		/* protected by i_mutex */
		if ((newsize < oldsize && (info->seals & F_SEAL_SHRINK)) ||
		    (newsize > oldsize && (info->seals & F_SEAL_GROW)))
			return -EPERM;

579
		if (newsize != oldsize) {
580 581 582 583
			error = shmem_reacct_size(SHMEM_I(inode)->flags,
					oldsize, newsize);
			if (error)
				return error;
584 585 586
			i_size_write(inode, newsize);
			inode->i_ctime = inode->i_mtime = CURRENT_TIME;
		}
587
		if (newsize <= oldsize) {
588
			loff_t holebegin = round_up(newsize, PAGE_SIZE);
589 590 591 592 593 594
			if (oldsize > holebegin)
				unmap_mapping_range(inode->i_mapping,
							holebegin, 0, 1);
			if (info->alloced)
				shmem_truncate_range(inode,
							newsize, (loff_t)-1);
595
			/* unmap again to remove racily COWed private pages */
596 597 598
			if (oldsize > holebegin)
				unmap_mapping_range(inode->i_mapping,
							holebegin, 0, 1);
599
		}
L
Linus Torvalds 已提交
600 601
	}

602 603
	setattr_copy(inode, attr);
	if (attr->ia_valid & ATTR_MODE)
C
Christoph Hellwig 已提交
604
		error = posix_acl_chmod(inode, inode->i_mode);
L
Linus Torvalds 已提交
605 606 607
	return error;
}

A
Al Viro 已提交
608
static void shmem_evict_inode(struct inode *inode)
L
Linus Torvalds 已提交
609 610 611
{
	struct shmem_inode_info *info = SHMEM_I(inode);

612
	if (inode->i_mapping->a_ops == &shmem_aops) {
L
Linus Torvalds 已提交
613 614
		shmem_unacct_size(info->flags, inode->i_size);
		inode->i_size = 0;
615
		shmem_truncate_range(inode, 0, (loff_t)-1);
L
Linus Torvalds 已提交
616
		if (!list_empty(&info->swaplist)) {
617
			mutex_lock(&shmem_swaplist_mutex);
L
Linus Torvalds 已提交
618
			list_del_init(&info->swaplist);
619
			mutex_unlock(&shmem_swaplist_mutex);
L
Linus Torvalds 已提交
620
		}
621 622
	} else
		kfree(info->symlink);
623

624
	simple_xattrs_free(&info->xattrs);
625
	WARN_ON(inode->i_blocks);
626
	shmem_free_inode(inode->i_sb);
627
	clear_inode(inode);
L
Linus Torvalds 已提交
628 629
}

630 631 632
/*
 * If swap found in inode, free it and move page from swapcache to filecache.
 */
633
static int shmem_unuse_inode(struct shmem_inode_info *info,
634
			     swp_entry_t swap, struct page **pagep)
L
Linus Torvalds 已提交
635
{
636
	struct address_space *mapping = info->vfs_inode.i_mapping;
637
	void *radswap;
638
	pgoff_t index;
639 640
	gfp_t gfp;
	int error = 0;
L
Linus Torvalds 已提交
641

642
	radswap = swp_to_radix_entry(swap);
643
	index = radix_tree_locate_item(&mapping->page_tree, radswap);
644
	if (index == -1)
645
		return -EAGAIN;	/* tell shmem_unuse we found nothing */
646

H
Hugh Dickins 已提交
647 648
	/*
	 * Move _head_ to start search for next from here.
A
Al Viro 已提交
649
	 * But be careful: shmem_evict_inode checks list_empty without taking
H
Hugh Dickins 已提交
650
	 * mutex, and there's an instant in list_move_tail when info->swaplist
651
	 * would appear empty, if it were the only one on shmem_swaplist.
H
Hugh Dickins 已提交
652 653 654
	 */
	if (shmem_swaplist.next != &info->swaplist)
		list_move_tail(&shmem_swaplist, &info->swaplist);
655

656 657 658 659 660 661 662
	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
663 664 665 666 667 668 669 670 671 672 673 674 675 676 677
		 * 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).
678 679 680 681 682
		 */
		if (!page_swapcount(*pagep))
			error = -ENOENT;
	}

K
KAMEZAWA Hiroyuki 已提交
683
	/*
684 685 686
	 * 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
687
	 */
688 689
	if (!error)
		error = shmem_add_to_page_cache(*pagep, mapping, index,
690
						radswap);
691
	if (error != -ENOMEM) {
692 693 694 695
		/*
		 * Truncation and eviction use free_swap_and_cache(), which
		 * only does trylock page: if we raced, best clean up here.
		 */
696 697
		delete_from_swap_cache(*pagep);
		set_page_dirty(*pagep);
698 699 700 701 702 703
		if (!error) {
			spin_lock(&info->lock);
			info->swapped--;
			spin_unlock(&info->lock);
			swap_free(swap);
		}
L
Linus Torvalds 已提交
704
	}
705
	return error;
L
Linus Torvalds 已提交
706 707 708
}

/*
709
 * Search through swapped inodes to find and replace swap by page.
L
Linus Torvalds 已提交
710
 */
711
int shmem_unuse(swp_entry_t swap, struct page *page)
L
Linus Torvalds 已提交
712
{
713
	struct list_head *this, *next;
L
Linus Torvalds 已提交
714
	struct shmem_inode_info *info;
715
	struct mem_cgroup *memcg;
716 717 718 719
	int error = 0;

	/*
	 * There's a faint possibility that swap page was replaced before
720
	 * caller locked it: caller will come back later with the right page.
721
	 */
722
	if (unlikely(!PageSwapCache(page) || page_private(page) != swap.val))
723
		goto out;
724 725 726 727 728 729

	/*
	 * 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.
	 */
730
	error = mem_cgroup_try_charge(page, current->mm, GFP_KERNEL, &memcg);
731 732
	if (error)
		goto out;
733
	/* No radix_tree_preload: swap entry keeps a place for page in tree */
734
	error = -EAGAIN;
L
Linus Torvalds 已提交
735

736
	mutex_lock(&shmem_swaplist_mutex);
737 738
	list_for_each_safe(this, next, &shmem_swaplist) {
		info = list_entry(this, struct shmem_inode_info, swaplist);
739
		if (info->swapped)
740
			error = shmem_unuse_inode(info, swap, &page);
741 742
		else
			list_del_init(&info->swaplist);
743
		cond_resched();
744
		if (error != -EAGAIN)
745
			break;
746
		/* found nothing in this: move on to search the next */
L
Linus Torvalds 已提交
747
	}
748
	mutex_unlock(&shmem_swaplist_mutex);
749

750 751 752 753 754 755
	if (error) {
		if (error != -ENOMEM)
			error = 0;
		mem_cgroup_cancel_charge(page, memcg);
	} else
		mem_cgroup_commit_charge(page, memcg, true);
756
out:
H
Hugh Dickins 已提交
757 758
	unlock_page(page);
	page_cache_release(page);
759
	return error;
L
Linus Torvalds 已提交
760 761 762 763 764 765 766 767 768 769
}

/*
 * 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;
770 771
	swp_entry_t swap;
	pgoff_t index;
L
Linus Torvalds 已提交
772 773 774 775 776 777 778 779

	BUG_ON(!PageLocked(page));
	mapping = page->mapping;
	index = page->index;
	inode = mapping->host;
	info = SHMEM_I(inode);
	if (info->flags & VM_LOCKED)
		goto redirty;
780
	if (!total_swap_pages)
L
Linus Torvalds 已提交
781 782
		goto redirty;

783
	/*
C
Christoph Hellwig 已提交
784 785 786 787 788
	 * Our capabilities prevent regular writeback or sync from ever calling
	 * shmem_writepage; but a stacking filesystem might use ->writepage of
	 * its underlying filesystem, in which case tmpfs should write out to
	 * swap only in response to memory pressure, and not for the writeback
	 * threads or sync.
789
	 */
790 791 792 793
	if (!wbc->for_reclaim) {
		WARN_ON_ONCE(1);	/* Still happens? Tell us about it! */
		goto redirty;
	}
794 795 796 797 798

	/*
	 * 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.
799 800 801 802 803 804
	 *
	 * 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.
805 806
	 */
	if (!PageUptodate(page)) {
807 808 809 810 811
		if (inode->i_private) {
			struct shmem_falloc *shmem_falloc;
			spin_lock(&inode->i_lock);
			shmem_falloc = inode->i_private;
			if (shmem_falloc &&
812
			    !shmem_falloc->waitq &&
813 814 815 816 817 818 819 820 821
			    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;
		}
822 823 824 825 826
		clear_highpage(page);
		flush_dcache_page(page);
		SetPageUptodate(page);
	}

827 828 829
	swap = get_swap_page();
	if (!swap.val)
		goto redirty;
830

831 832
	/*
	 * Add inode to shmem_unuse()'s list of swapped-out inodes,
833 834
	 * if it's not already there.  Do it now before the page is
	 * moved to swap cache, when its pagelock no longer protects
835
	 * the inode from eviction.  But don't unlock the mutex until
836 837
	 * we've incremented swapped, because shmem_unuse_inode() will
	 * prune a !swapped inode from the swaplist under this mutex.
838
	 */
839 840 841
	mutex_lock(&shmem_swaplist_mutex);
	if (list_empty(&info->swaplist))
		list_add_tail(&info->swaplist, &shmem_swaplist);
842

843
	if (add_to_swap_cache(page, swap, GFP_ATOMIC) == 0) {
H
Hugh Dickins 已提交
844
		swap_shmem_alloc(swap);
845 846 847 848 849
		shmem_delete_from_page_cache(page, swp_to_radix_entry(swap));

		spin_lock(&info->lock);
		info->swapped++;
		shmem_recalc_inode(inode);
850
		spin_unlock(&info->lock);
851 852

		mutex_unlock(&shmem_swaplist_mutex);
853
		BUG_ON(page_mapped(page));
H
Hugh Dickins 已提交
854
		swap_writepage(page, wbc);
L
Linus Torvalds 已提交
855 856 857
		return 0;
	}

858
	mutex_unlock(&shmem_swaplist_mutex);
859
	swapcache_free(swap);
L
Linus Torvalds 已提交
860 861
redirty:
	set_page_dirty(page);
862 863 864 865
	if (wbc->for_reclaim)
		return AOP_WRITEPAGE_ACTIVATE;	/* Return with page locked */
	unlock_page(page);
	return 0;
L
Linus Torvalds 已提交
866 867 868
}

#ifdef CONFIG_NUMA
869
#ifdef CONFIG_TMPFS
870
static void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
871
{
872
	char buffer[64];
873

874
	if (!mpol || mpol->mode == MPOL_DEFAULT)
875
		return;		/* show nothing */
876

877
	mpol_to_str(buffer, sizeof(buffer), mpol);
878 879

	seq_printf(seq, ",mpol=%s", buffer);
880
}
881 882 883 884 885 886 887 888 889 890 891 892

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;
}
893 894
#endif /* CONFIG_TMPFS */

895 896
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
897 898
{
	struct vm_area_struct pvma;
M
Mel Gorman 已提交
899
	struct page *page;
900

L
Linus Torvalds 已提交
901
	/* Create a pseudo vma that just contains the policy */
902
	pvma.vm_start = 0;
903 904
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
905
	pvma.vm_ops = NULL;
M
Mel Gorman 已提交
906 907 908 909 910 911 912 913
	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
914 915
}

916
static struct page *shmem_alloc_page(gfp_t gfp,
917
			struct shmem_inode_info *info, pgoff_t index)
L
Linus Torvalds 已提交
918 919
{
	struct vm_area_struct pvma;
M
Mel Gorman 已提交
920
	struct page *page;
L
Linus Torvalds 已提交
921

922 923
	/* Create a pseudo vma that just contains the policy */
	pvma.vm_start = 0;
924 925
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
926
	pvma.vm_ops = NULL;
927
	pvma.vm_policy = mpol_shared_policy_lookup(&info->policy, index);
928

M
Mel Gorman 已提交
929 930 931 932 933 934
	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
935
}
936 937
#else /* !CONFIG_NUMA */
#ifdef CONFIG_TMPFS
938
static inline void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
939 940 941 942
{
}
#endif /* CONFIG_TMPFS */

943 944
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
945
{
946
	return swapin_readahead(swap, gfp, NULL, 0);
L
Linus Torvalds 已提交
947 948
}

949
static inline struct page *shmem_alloc_page(gfp_t gfp,
950
			struct shmem_inode_info *info, pgoff_t index)
L
Linus Torvalds 已提交
951
{
952
	return alloc_page(gfp);
L
Linus Torvalds 已提交
953
}
954
#endif /* CONFIG_NUMA */
L
Linus Torvalds 已提交
955

956 957 958 959 960 961 962
#if !defined(CONFIG_NUMA) || !defined(CONFIG_TMPFS)
static inline struct mempolicy *shmem_get_sbmpol(struct shmem_sb_info *sbinfo)
{
	return NULL;
}
#endif

963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002
/*
 * 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);
1003
	flush_dcache_page(newpage);
1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017

	__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);
1018 1019 1020 1021
	if (!error) {
		__inc_zone_page_state(newpage, NR_FILE_PAGES);
		__dec_zone_page_state(oldpage, NR_FILE_PAGES);
	}
1022 1023
	spin_unlock_irq(&swap_mapping->tree_lock);

1024 1025 1026 1027 1028 1029 1030 1031
	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 {
1032
		mem_cgroup_replace_page(oldpage, newpage);
1033 1034 1035
		lru_cache_add_anon(newpage);
		*pagep = newpage;
	}
1036 1037 1038 1039 1040 1041 1042

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

	unlock_page(oldpage);
	page_cache_release(oldpage);
	page_cache_release(oldpage);
1043
	return error;
1044 1045
}

L
Linus Torvalds 已提交
1046
/*
1047
 * shmem_getpage_gfp - find page in cache, or get from swap, or allocate
L
Linus Torvalds 已提交
1048 1049 1050 1051 1052
 *
 * 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
 */
1053
static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
1054
	struct page **pagep, enum sgp_type sgp, gfp_t gfp, int *fault_type)
L
Linus Torvalds 已提交
1055 1056
{
	struct address_space *mapping = inode->i_mapping;
1057
	struct shmem_inode_info *info;
L
Linus Torvalds 已提交
1058
	struct shmem_sb_info *sbinfo;
1059
	struct mem_cgroup *memcg;
H
Hugh Dickins 已提交
1060
	struct page *page;
L
Linus Torvalds 已提交
1061 1062
	swp_entry_t swap;
	int error;
1063
	int once = 0;
1064
	int alloced = 0;
L
Linus Torvalds 已提交
1065

1066
	if (index > (MAX_LFS_FILESIZE >> PAGE_CACHE_SHIFT))
L
Linus Torvalds 已提交
1067 1068
		return -EFBIG;
repeat:
1069
	swap.val = 0;
1070
	page = find_lock_entry(mapping, index);
1071 1072 1073 1074 1075
	if (radix_tree_exceptional_entry(page)) {
		swap = radix_to_swp_entry(page);
		page = NULL;
	}

1076
	if (sgp != SGP_WRITE && sgp != SGP_FALLOC &&
1077 1078 1079 1080 1081
	    ((loff_t)index << PAGE_CACHE_SHIFT) >= i_size_read(inode)) {
		error = -EINVAL;
		goto failed;
	}

1082 1083 1084
	if (page && sgp == SGP_WRITE)
		mark_page_accessed(page);

1085 1086 1087 1088 1089 1090 1091 1092
	/* fallocated page? */
	if (page && !PageUptodate(page)) {
		if (sgp != SGP_READ)
			goto clear;
		unlock_page(page);
		page_cache_release(page);
		page = NULL;
	}
1093 1094 1095
	if (page || (sgp == SGP_READ && !swap.val)) {
		*pagep = page;
		return 0;
H
Hugh Dickins 已提交
1096 1097 1098
	}

	/*
1099 1100
	 * Fast cache lookup did not find it:
	 * bring it back from swap or allocate.
H
Hugh Dickins 已提交
1101
	 */
1102 1103
	info = SHMEM_I(inode);
	sbinfo = SHMEM_SB(inode->i_sb);
L
Linus Torvalds 已提交
1104 1105 1106

	if (swap.val) {
		/* Look it up and read it in.. */
H
Hugh Dickins 已提交
1107 1108
		page = lookup_swap_cache(swap);
		if (!page) {
L
Linus Torvalds 已提交
1109
			/* here we actually do the io */
1110 1111
			if (fault_type)
				*fault_type |= VM_FAULT_MAJOR;
1112
			page = shmem_swapin(swap, gfp, info, index);
H
Hugh Dickins 已提交
1113
			if (!page) {
1114 1115
				error = -ENOMEM;
				goto failed;
L
Linus Torvalds 已提交
1116 1117 1118 1119
			}
		}

		/* We have to do this with page locked to prevent races */
1120
		lock_page(page);
1121
		if (!PageSwapCache(page) || page_private(page) != swap.val ||
1122
		    !shmem_confirm_swap(mapping, index, swap)) {
1123
			error = -EEXIST;	/* try again */
1124
			goto unlock;
1125
		}
H
Hugh Dickins 已提交
1126
		if (!PageUptodate(page)) {
L
Linus Torvalds 已提交
1127
			error = -EIO;
1128
			goto failed;
L
Linus Torvalds 已提交
1129
		}
1130 1131
		wait_on_page_writeback(page);

1132 1133 1134 1135
		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
1136
		}
H
Hugh Dickins 已提交
1137

1138
		error = mem_cgroup_try_charge(page, current->mm, gfp, &memcg);
1139
		if (!error) {
1140
			error = shmem_add_to_page_cache(page, mapping, index,
1141
						swp_to_radix_entry(swap));
1142 1143 1144 1145 1146 1147 1148 1149
			/*
			 * 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
1150
			 * the rest.
1151 1152 1153
			 * Reset swap.val? No, leave it so "failed" goes back to
			 * "repeat": reading a hole and writing should succeed.
			 */
1154 1155
			if (error) {
				mem_cgroup_cancel_charge(page, memcg);
1156
				delete_from_swap_cache(page);
1157
			}
1158
		}
1159 1160 1161
		if (error)
			goto failed;

1162 1163
		mem_cgroup_commit_charge(page, memcg, true);

1164
		spin_lock(&info->lock);
1165
		info->swapped--;
1166
		shmem_recalc_inode(inode);
H
Hugh Dickins 已提交
1167
		spin_unlock(&info->lock);
1168

1169 1170 1171
		if (sgp == SGP_WRITE)
			mark_page_accessed(page);

1172
		delete_from_swap_cache(page);
H
Hugh Dickins 已提交
1173 1174 1175
		set_page_dirty(page);
		swap_free(swap);

1176 1177 1178 1179
	} else {
		if (shmem_acct_block(info->flags)) {
			error = -ENOSPC;
			goto failed;
L
Linus Torvalds 已提交
1180
		}
1181
		if (sbinfo->max_blocks) {
1182
			if (percpu_counter_compare(&sbinfo->used_blocks,
1183 1184 1185 1186
						sbinfo->max_blocks) >= 0) {
				error = -ENOSPC;
				goto unacct;
			}
1187
			percpu_counter_inc(&sbinfo->used_blocks);
1188
		}
L
Linus Torvalds 已提交
1189

1190 1191 1192 1193
		page = shmem_alloc_page(gfp, info, index);
		if (!page) {
			error = -ENOMEM;
			goto decused;
L
Linus Torvalds 已提交
1194 1195
		}

1196
		__SetPageSwapBacked(page);
1197
		__set_page_locked(page);
1198
		if (sgp == SGP_WRITE)
1199
			__SetPageReferenced(page);
1200

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

		spin_lock(&info->lock);
L
Linus Torvalds 已提交
1218
		info->alloced++;
1219 1220
		inode->i_blocks += BLOCKS_PER_PAGE;
		shmem_recalc_inode(inode);
L
Linus Torvalds 已提交
1221
		spin_unlock(&info->lock);
1222
		alloced = true;
1223

1224
		/*
1225 1226 1227 1228 1229 1230 1231 1232 1233
		 * 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.
1234 1235 1236 1237 1238 1239
		 */
		if (sgp != SGP_WRITE) {
			clear_highpage(page);
			flush_dcache_page(page);
			SetPageUptodate(page);
		}
1240
		if (sgp == SGP_DIRTY)
H
Hugh Dickins 已提交
1241
			set_page_dirty(page);
L
Linus Torvalds 已提交
1242
	}
1243

1244
	/* Perhaps the file has been truncated since we checked */
1245
	if (sgp != SGP_WRITE && sgp != SGP_FALLOC &&
1246 1247
	    ((loff_t)index << PAGE_CACHE_SHIFT) >= i_size_read(inode)) {
		error = -EINVAL;
1248 1249 1250 1251
		if (alloced)
			goto trunc;
		else
			goto failed;
H
Hugh Dickins 已提交
1252
	}
1253 1254
	*pagep = page;
	return 0;
L
Linus Torvalds 已提交
1255

1256
	/*
1257
	 * Error recovery.
1258
	 */
1259
trunc:
1260
	info = SHMEM_I(inode);
1261 1262 1263 1264 1265
	ClearPageDirty(page);
	delete_from_page_cache(page);
	spin_lock(&info->lock);
	info->alloced--;
	inode->i_blocks -= BLOCKS_PER_PAGE;
1266
	spin_unlock(&info->lock);
1267
decused:
1268
	sbinfo = SHMEM_SB(inode->i_sb);
1269 1270 1271 1272 1273
	if (sbinfo->max_blocks)
		percpu_counter_add(&sbinfo->used_blocks, -1);
unacct:
	shmem_unacct_blocks(info->flags, 1);
failed:
1274 1275 1276 1277
	if (swap.val && error != -EINVAL &&
	    !shmem_confirm_swap(mapping, index, swap))
		error = -EEXIST;
unlock:
H
Hugh Dickins 已提交
1278
	if (page) {
1279
		unlock_page(page);
H
Hugh Dickins 已提交
1280
		page_cache_release(page);
1281 1282 1283 1284 1285 1286
	}
	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
1287
		goto repeat;
1288
	}
1289
	if (error == -EEXIST)	/* from above or from radix_tree_insert */
1290 1291
		goto repeat;
	return error;
L
Linus Torvalds 已提交
1292 1293
}

N
Nick Piggin 已提交
1294
static int shmem_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
L
Linus Torvalds 已提交
1295
{
A
Al Viro 已提交
1296
	struct inode *inode = file_inode(vma->vm_file);
L
Linus Torvalds 已提交
1297
	int error;
1298
	int ret = VM_FAULT_LOCKED;
L
Linus Torvalds 已提交
1299

1300 1301 1302 1303
	/*
	 * 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
1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315
	 * 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.
1316 1317 1318 1319 1320 1321
	 */
	if (unlikely(inode->i_private)) {
		struct shmem_falloc *shmem_falloc;

		spin_lock(&inode->i_lock);
		shmem_falloc = inode->i_private;
1322 1323 1324 1325 1326 1327 1328 1329
		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;
1330 1331
			if ((vmf->flags & FAULT_FLAG_ALLOW_RETRY) &&
			   !(vmf->flags & FAULT_FLAG_RETRY_NOWAIT)) {
1332
				/* It's polite to up mmap_sem if we can */
1333
				up_read(&vma->vm_mm->mmap_sem);
1334
				ret = VM_FAULT_RETRY;
1335
			}
1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353

			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;
1354
		}
1355
		spin_unlock(&inode->i_lock);
1356 1357
	}

1358
	error = shmem_getpage(inode, vmf->pgoff, &vmf->page, SGP_CACHE, &ret);
N
Nick Piggin 已提交
1359 1360
	if (error)
		return ((error == -ENOMEM) ? VM_FAULT_OOM : VM_FAULT_SIGBUS);
1361

1362 1363 1364 1365
	if (ret & VM_FAULT_MAJOR) {
		count_vm_event(PGMAJFAULT);
		mem_cgroup_count_vm_event(vma->vm_mm, PGMAJFAULT);
	}
1366
	return ret;
L
Linus Torvalds 已提交
1367 1368 1369
}

#ifdef CONFIG_NUMA
1370
static int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *mpol)
L
Linus Torvalds 已提交
1371
{
A
Al Viro 已提交
1372
	struct inode *inode = file_inode(vma->vm_file);
1373
	return mpol_set_shared_policy(&SHMEM_I(inode)->policy, vma, mpol);
L
Linus Torvalds 已提交
1374 1375
}

A
Adrian Bunk 已提交
1376 1377
static struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
					  unsigned long addr)
L
Linus Torvalds 已提交
1378
{
A
Al Viro 已提交
1379
	struct inode *inode = file_inode(vma->vm_file);
1380
	pgoff_t index;
L
Linus Torvalds 已提交
1381

1382 1383
	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
1384 1385 1386 1387 1388
}
#endif

int shmem_lock(struct file *file, int lock, struct user_struct *user)
{
A
Al Viro 已提交
1389
	struct inode *inode = file_inode(file);
L
Linus Torvalds 已提交
1390 1391 1392 1393 1394 1395 1396 1397
	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;
1398
		mapping_set_unevictable(file->f_mapping);
L
Linus Torvalds 已提交
1399 1400 1401 1402
	}
	if (!lock && (info->flags & VM_LOCKED) && user) {
		user_shm_unlock(inode->i_size, user);
		info->flags &= ~VM_LOCKED;
1403
		mapping_clear_unevictable(file->f_mapping);
L
Linus Torvalds 已提交
1404 1405
	}
	retval = 0;
1406

L
Linus Torvalds 已提交
1407 1408 1409 1410 1411
out_nomem:
	spin_unlock(&info->lock);
	return retval;
}

1412
static int shmem_mmap(struct file *file, struct vm_area_struct *vma)
L
Linus Torvalds 已提交
1413 1414 1415 1416 1417 1418
{
	file_accessed(file);
	vma->vm_ops = &shmem_vm_ops;
	return 0;
}

1419
static struct inode *shmem_get_inode(struct super_block *sb, const struct inode *dir,
1420
				     umode_t mode, dev_t dev, unsigned long flags)
L
Linus Torvalds 已提交
1421 1422 1423 1424 1425
{
	struct inode *inode;
	struct shmem_inode_info *info;
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);

1426 1427
	if (shmem_reserve_inode(sb))
		return NULL;
L
Linus Torvalds 已提交
1428 1429 1430

	inode = new_inode(sb);
	if (inode) {
1431
		inode->i_ino = get_next_ino();
1432
		inode_init_owner(inode, dir, mode);
L
Linus Torvalds 已提交
1433 1434
		inode->i_blocks = 0;
		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1435
		inode->i_generation = get_seconds();
L
Linus Torvalds 已提交
1436 1437 1438
		info = SHMEM_I(inode);
		memset(info, 0, (char *)inode - (char *)info);
		spin_lock_init(&info->lock);
D
David Herrmann 已提交
1439
		info->seals = F_SEAL_SEAL;
1440
		info->flags = flags & VM_NORESERVE;
L
Linus Torvalds 已提交
1441
		INIT_LIST_HEAD(&info->swaplist);
1442
		simple_xattrs_init(&info->xattrs);
1443
		cache_no_acl(inode);
L
Linus Torvalds 已提交
1444 1445 1446

		switch (mode & S_IFMT) {
		default:
1447
			inode->i_op = &shmem_special_inode_operations;
L
Linus Torvalds 已提交
1448 1449 1450
			init_special_inode(inode, mode, dev);
			break;
		case S_IFREG:
1451
			inode->i_mapping->a_ops = &shmem_aops;
L
Linus Torvalds 已提交
1452 1453
			inode->i_op = &shmem_inode_operations;
			inode->i_fop = &shmem_file_operations;
1454 1455
			mpol_shared_policy_init(&info->policy,
						 shmem_get_sbmpol(sbinfo));
L
Linus Torvalds 已提交
1456 1457
			break;
		case S_IFDIR:
1458
			inc_nlink(inode);
L
Linus Torvalds 已提交
1459 1460 1461 1462 1463 1464 1465 1466 1467 1468
			/* 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.
			 */
1469
			mpol_shared_policy_init(&info->policy, NULL);
L
Linus Torvalds 已提交
1470 1471
			break;
		}
1472 1473
	} else
		shmem_free_inode(sb);
L
Linus Torvalds 已提交
1474 1475 1476
	return inode;
}

1477 1478
bool shmem_mapping(struct address_space *mapping)
{
1479 1480 1481
	if (!mapping->host)
		return false;

C
Christoph Hellwig 已提交
1482
	return mapping->host->i_sb->s_op == &shmem_ops;
1483 1484
}

L
Linus Torvalds 已提交
1485
#ifdef CONFIG_TMPFS
1486
static const struct inode_operations shmem_symlink_inode_operations;
1487
static const struct inode_operations shmem_short_symlink_operations;
L
Linus Torvalds 已提交
1488

1489 1490 1491 1492 1493 1494
#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
1495
static int
N
Nick Piggin 已提交
1496 1497 1498
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
1499
{
N
Nick Piggin 已提交
1500
	struct inode *inode = mapping->host;
D
David Herrmann 已提交
1501
	struct shmem_inode_info *info = SHMEM_I(inode);
N
Nick Piggin 已提交
1502
	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
D
David Herrmann 已提交
1503 1504 1505 1506 1507 1508 1509 1510 1511

	/* 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;
	}

1512
	return shmem_getpage(inode, index, pagep, SGP_WRITE, NULL);
N
Nick Piggin 已提交
1513 1514 1515 1516 1517 1518 1519 1520 1521
}

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
1522 1523 1524
	if (pos + copied > inode->i_size)
		i_size_write(inode, pos + copied);

1525 1526 1527 1528 1529 1530 1531 1532
	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
1533
	set_page_dirty(page);
1534
	unlock_page(page);
N
Nick Piggin 已提交
1535 1536 1537
	page_cache_release(page);

	return copied;
L
Linus Torvalds 已提交
1538 1539
}

A
Al Viro 已提交
1540
static ssize_t shmem_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
L
Linus Torvalds 已提交
1541
{
1542 1543
	struct file *file = iocb->ki_filp;
	struct inode *inode = file_inode(file);
L
Linus Torvalds 已提交
1544
	struct address_space *mapping = inode->i_mapping;
1545 1546
	pgoff_t index;
	unsigned long offset;
1547
	enum sgp_type sgp = SGP_READ;
1548
	int error = 0;
A
Al Viro 已提交
1549
	ssize_t retval = 0;
1550
	loff_t *ppos = &iocb->ki_pos;
1551 1552 1553 1554 1555 1556

	/*
	 * 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
1557
	if (!iter_is_iovec(to))
1558
		sgp = SGP_DIRTY;
L
Linus Torvalds 已提交
1559 1560 1561 1562 1563 1564

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

	for (;;) {
		struct page *page = NULL;
1565 1566
		pgoff_t end_index;
		unsigned long nr, ret;
L
Linus Torvalds 已提交
1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577
		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;
		}

1578 1579 1580 1581
		error = shmem_getpage(inode, index, &page, sgp, NULL);
		if (error) {
			if (error == -EINVAL)
				error = 0;
L
Linus Torvalds 已提交
1582 1583
			break;
		}
H
Hugh Dickins 已提交
1584 1585
		if (page)
			unlock_page(page);
L
Linus Torvalds 已提交
1586 1587 1588

		/*
		 * We must evaluate after, since reads (unlike writes)
1589
		 * are called without i_mutex protection against truncate
L
Linus Torvalds 已提交
1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616
		 */
		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
1617
		} else {
L
Linus Torvalds 已提交
1618
			page = ZERO_PAGE(0);
N
Nick Piggin 已提交
1619 1620
			page_cache_get(page);
		}
L
Linus Torvalds 已提交
1621 1622 1623 1624 1625

		/*
		 * 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
1626
		ret = copy_page_to_iter(page, offset, nr, to);
1627
		retval += ret;
L
Linus Torvalds 已提交
1628 1629 1630 1631 1632
		offset += ret;
		index += offset >> PAGE_CACHE_SHIFT;
		offset &= ~PAGE_CACHE_MASK;

		page_cache_release(page);
A
Al Viro 已提交
1633
		if (!iov_iter_count(to))
L
Linus Torvalds 已提交
1634
			break;
1635 1636 1637 1638
		if (ret < nr) {
			error = -EFAULT;
			break;
		}
L
Linus Torvalds 已提交
1639 1640 1641 1642
		cond_resched();
	}

	*ppos = ((loff_t) index << PAGE_CACHE_SHIFT) + offset;
1643 1644
	file_accessed(file);
	return retval ? retval : error;
L
Linus Torvalds 已提交
1645 1646
}

1647 1648 1649 1650 1651
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;
1652
	struct inode *inode = mapping->host;
1653 1654 1655 1656 1657 1658 1659 1660 1661 1662
	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,
1663
		.nr_pages_max = PIPE_DEF_BUFFERS,
1664 1665 1666 1667 1668
		.flags = flags,
		.ops = &page_cache_pipe_buf_ops,
		.spd_release = spd_release_page,
	};

1669
	isize = i_size_read(inode);
1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682
	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;
1683
	nr_pages = min(req_pages, spd.nr_pages_max);
1684 1685 1686 1687 1688 1689

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

1690 1691 1692 1693 1694
	while (spd.nr_pages < nr_pages) {
		error = shmem_getpage(inode, index, &page, SGP_CACHE, NULL);
		if (error)
			break;
		unlock_page(page);
1695 1696 1697 1698 1699 1700 1701
		spd.pages[spd.nr_pages++] = page;
		index++;
	}

	index = *ppos >> PAGE_CACHE_SHIFT;
	nr_pages = spd.nr_pages;
	spd.nr_pages = 0;
1702

1703 1704 1705 1706 1707 1708 1709 1710 1711
	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];

1712 1713 1714 1715
		if (!PageUptodate(page) || page->mapping != mapping) {
			error = shmem_getpage(inode, index, &page,
							SGP_CACHE, NULL);
			if (error)
1716
				break;
1717 1718 1719
			unlock_page(page);
			page_cache_release(spd.pages[page_nr]);
			spd.pages[page_nr] = page;
1720
		}
1721 1722

		isize = i_size_read(inode);
1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751
		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);

1752
	splice_shrink_spd(&spd);
1753 1754 1755 1756 1757 1758 1759 1760

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

1761 1762 1763 1764
/*
 * llseek SEEK_DATA or SEEK_HOLE through the radix_tree.
 */
static pgoff_t shmem_seek_hole_data(struct address_space *mapping,
1765
				    pgoff_t index, pgoff_t end, int whence)
1766 1767 1768 1769 1770 1771 1772 1773 1774 1775
{
	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) {
1776
		pvec.nr = find_get_entries(mapping, index,
1777 1778
					pvec.nr, pvec.pages, indices);
		if (!pvec.nr) {
1779
			if (whence == SEEK_DATA)
1780 1781 1782 1783 1784
				index = end;
			break;
		}
		for (i = 0; i < pvec.nr; i++, index++) {
			if (index < indices[i]) {
1785
				if (whence == SEEK_HOLE) {
1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796
					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 ||
1797 1798
			    (page && whence == SEEK_DATA) ||
			    (!page && whence == SEEK_HOLE)) {
1799 1800 1801 1802
				done = true;
				break;
			}
		}
1803
		pagevec_remove_exceptionals(&pvec);
1804 1805 1806 1807 1808 1809 1810
		pagevec_release(&pvec);
		pvec.nr = PAGEVEC_SIZE;
		cond_resched();
	}
	return index;
}

1811
static loff_t shmem_file_llseek(struct file *file, loff_t offset, int whence)
1812 1813 1814 1815 1816 1817
{
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
	pgoff_t start, end;
	loff_t new_offset;

1818 1819
	if (whence != SEEK_DATA && whence != SEEK_HOLE)
		return generic_file_llseek_size(file, offset, whence,
1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830
					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;
1831
		new_offset = shmem_seek_hole_data(mapping, start, end, whence);
1832 1833 1834 1835
		new_offset <<= PAGE_CACHE_SHIFT;
		if (new_offset > offset) {
			if (new_offset < inode->i_size)
				offset = new_offset;
1836
			else if (whence == SEEK_DATA)
1837 1838 1839 1840 1841 1842
				offset = -ENXIO;
			else
				offset = inode->i_size;
		}
	}

1843 1844
	if (offset >= 0)
		offset = vfs_setpos(file, offset, MAX_LFS_FILESIZE);
1845 1846 1847 1848
	mutex_unlock(&inode->i_mutex);
	return offset;
}

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 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897
/*
 * 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
1898 1899
static int shmem_wait_for_pins(struct address_space *mapping)
{
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 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959
	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
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 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069
}

#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;
}

2070 2071 2072
static long shmem_fallocate(struct file *file, int mode, loff_t offset,
							 loff_t len)
{
A
Al Viro 已提交
2073
	struct inode *inode = file_inode(file);
2074
	struct shmem_sb_info *sbinfo = SHMEM_SB(inode->i_sb);
D
David Herrmann 已提交
2075
	struct shmem_inode_info *info = SHMEM_I(inode);
2076
	struct shmem_falloc shmem_falloc;
2077 2078
	pgoff_t start, index, end;
	int error;
2079

2080 2081 2082
	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
		return -EOPNOTSUPP;

2083 2084 2085 2086 2087 2088
	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;
2089
		DECLARE_WAIT_QUEUE_HEAD_ONSTACK(shmem_falloc_waitq);
2090

D
David Herrmann 已提交
2091 2092 2093 2094 2095 2096
		/* protected by i_mutex */
		if (info->seals & F_SEAL_WRITE) {
			error = -EPERM;
			goto out;
		}

2097
		shmem_falloc.waitq = &shmem_falloc_waitq;
2098 2099 2100 2101 2102 2103
		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);

2104 2105 2106 2107 2108
		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 */
2109 2110 2111 2112 2113

		spin_lock(&inode->i_lock);
		inode->i_private = NULL;
		wake_up_all(&shmem_falloc_waitq);
		spin_unlock(&inode->i_lock);
2114
		error = 0;
2115
		goto out;
2116 2117 2118 2119 2120 2121 2122
	}

	/* 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
2123 2124 2125 2126 2127
	if ((info->seals & F_SEAL_GROW) && offset + len > inode->i_size) {
		error = -EPERM;
		goto out;
	}

2128 2129 2130 2131 2132 2133
	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;
2134 2135
	}

2136
	shmem_falloc.waitq = NULL;
2137 2138 2139 2140 2141 2142 2143 2144
	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);

2145 2146 2147 2148 2149 2150 2151 2152 2153
	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;
2154 2155
		else if (shmem_falloc.nr_unswapped > shmem_falloc.nr_falloced)
			error = -ENOMEM;
2156
		else
2157
			error = shmem_getpage(inode, index, &page, SGP_FALLOC,
2158 2159
									NULL);
		if (error) {
2160 2161 2162 2163
			/* Remove the !PageUptodate pages we added */
			shmem_undo_range(inode,
				(loff_t)start << PAGE_CACHE_SHIFT,
				(loff_t)index << PAGE_CACHE_SHIFT, true);
2164
			goto undone;
2165 2166
		}

2167 2168 2169 2170 2171 2172 2173 2174
		/*
		 * 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++;

2175
		/*
2176 2177 2178
		 * 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
2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190
		 * 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;
2191 2192 2193 2194
undone:
	spin_lock(&inode->i_lock);
	inode->i_private = NULL;
	spin_unlock(&inode->i_lock);
2195
out:
2196 2197 2198 2199
	mutex_unlock(&inode->i_mutex);
	return error;
}

2200
static int shmem_statfs(struct dentry *dentry, struct kstatfs *buf)
L
Linus Torvalds 已提交
2201
{
2202
	struct shmem_sb_info *sbinfo = SHMEM_SB(dentry->d_sb);
L
Linus Torvalds 已提交
2203 2204 2205 2206

	buf->f_type = TMPFS_MAGIC;
	buf->f_bsize = PAGE_CACHE_SIZE;
	buf->f_namelen = NAME_MAX;
2207
	if (sbinfo->max_blocks) {
L
Linus Torvalds 已提交
2208
		buf->f_blocks = sbinfo->max_blocks;
2209 2210 2211
		buf->f_bavail =
		buf->f_bfree  = sbinfo->max_blocks -
				percpu_counter_sum(&sbinfo->used_blocks);
2212 2213
	}
	if (sbinfo->max_inodes) {
L
Linus Torvalds 已提交
2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224
		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
2225
shmem_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev)
L
Linus Torvalds 已提交
2226
{
2227
	struct inode *inode;
L
Linus Torvalds 已提交
2228 2229
	int error = -ENOSPC;

2230
	inode = shmem_get_inode(dir->i_sb, dir, mode, dev, VM_NORESERVE);
L
Linus Torvalds 已提交
2231
	if (inode) {
C
Christoph Hellwig 已提交
2232 2233 2234
		error = simple_acl_create(dir, inode);
		if (error)
			goto out_iput;
2235
		error = security_inode_init_security(inode, dir,
2236
						     &dentry->d_name,
2237
						     shmem_initxattrs, NULL);
C
Christoph Hellwig 已提交
2238 2239
		if (error && error != -EOPNOTSUPP)
			goto out_iput;
2240

A
Al Viro 已提交
2241
		error = 0;
L
Linus Torvalds 已提交
2242 2243 2244 2245 2246 2247
		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
2248 2249 2250
out_iput:
	iput(inode);
	return error;
L
Linus Torvalds 已提交
2251 2252
}

2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263
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
2264 2265 2266 2267 2268
		if (error && error != -EOPNOTSUPP)
			goto out_iput;
		error = simple_acl_create(dir, inode);
		if (error)
			goto out_iput;
2269 2270 2271
		d_tmpfile(dentry, inode);
	}
	return error;
C
Christoph Hellwig 已提交
2272 2273 2274
out_iput:
	iput(inode);
	return error;
2275 2276
}

2277
static int shmem_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
2278 2279 2280 2281 2282
{
	int error;

	if ((error = shmem_mknod(dir, dentry, mode | S_IFDIR, 0)))
		return error;
2283
	inc_nlink(dir);
L
Linus Torvalds 已提交
2284 2285 2286
	return 0;
}

A
Al Viro 已提交
2287
static int shmem_create(struct inode *dir, struct dentry *dentry, umode_t mode,
A
Al Viro 已提交
2288
		bool excl)
L
Linus Torvalds 已提交
2289 2290 2291 2292 2293 2294 2295 2296 2297
{
	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)
{
2298
	struct inode *inode = d_inode(old_dentry);
2299
	int ret;
L
Linus Torvalds 已提交
2300 2301 2302 2303 2304 2305

	/*
	 * 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.
	 */
2306 2307 2308
	ret = shmem_reserve_inode(inode->i_sb);
	if (ret)
		goto out;
L
Linus Torvalds 已提交
2309 2310 2311

	dir->i_size += BOGO_DIRENT_SIZE;
	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
2312
	inc_nlink(inode);
A
Al Viro 已提交
2313
	ihold(inode);	/* New dentry reference */
L
Linus Torvalds 已提交
2314 2315
	dget(dentry);		/* Extra pinning count for the created dentry */
	d_instantiate(dentry, inode);
2316 2317
out:
	return ret;
L
Linus Torvalds 已提交
2318 2319 2320 2321
}

static int shmem_unlink(struct inode *dir, struct dentry *dentry)
{
2322
	struct inode *inode = d_inode(dentry);
L
Linus Torvalds 已提交
2323

2324 2325
	if (inode->i_nlink > 1 && !S_ISDIR(inode->i_mode))
		shmem_free_inode(inode->i_sb);
L
Linus Torvalds 已提交
2326 2327 2328

	dir->i_size -= BOGO_DIRENT_SIZE;
	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
2329
	drop_nlink(inode);
L
Linus Torvalds 已提交
2330 2331 2332 2333 2334 2335 2336 2337 2338
	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;

2339
	drop_nlink(d_inode(dentry));
2340
	drop_nlink(dir);
L
Linus Torvalds 已提交
2341 2342 2343
	return shmem_unlink(dir, dentry);
}

M
Miklos Szeredi 已提交
2344 2345
static int shmem_exchange(struct inode *old_dir, struct dentry *old_dentry, struct inode *new_dir, struct dentry *new_dentry)
{
2346 2347
	bool old_is_dir = d_is_dir(old_dentry);
	bool new_is_dir = d_is_dir(new_dentry);
M
Miklos Szeredi 已提交
2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359

	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 =
2360 2361
	d_inode(old_dentry)->i_ctime =
	d_inode(new_dentry)->i_ctime = CURRENT_TIME;
M
Miklos Szeredi 已提交
2362 2363 2364 2365

	return 0;
}

M
Miklos Szeredi 已提交
2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391
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
2392 2393 2394 2395 2396 2397
/*
 * 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
2398
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
2399
{
2400
	struct inode *inode = d_inode(old_dentry);
L
Linus Torvalds 已提交
2401 2402
	int they_are_dirs = S_ISDIR(inode->i_mode);

M
Miklos Szeredi 已提交
2403
	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT))
M
Miklos Szeredi 已提交
2404 2405
		return -EINVAL;

M
Miklos Szeredi 已提交
2406 2407 2408
	if (flags & RENAME_EXCHANGE)
		return shmem_exchange(old_dir, old_dentry, new_dir, new_dentry);

L
Linus Torvalds 已提交
2409 2410 2411
	if (!simple_empty(new_dentry))
		return -ENOTEMPTY;

M
Miklos Szeredi 已提交
2412 2413 2414 2415 2416 2417 2418 2419
	if (flags & RENAME_WHITEOUT) {
		int error;

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

2420
	if (d_really_is_positive(new_dentry)) {
L
Linus Torvalds 已提交
2421
		(void) shmem_unlink(new_dir, new_dentry);
2422
		if (they_are_dirs) {
2423
			drop_nlink(d_inode(new_dentry));
2424
			drop_nlink(old_dir);
2425
		}
L
Linus Torvalds 已提交
2426
	} else if (they_are_dirs) {
2427
		drop_nlink(old_dir);
2428
		inc_nlink(new_dir);
L
Linus Torvalds 已提交
2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443
	}

	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
2444
	struct page *page;
L
Linus Torvalds 已提交
2445 2446 2447 2448 2449 2450 2451
	char *kaddr;
	struct shmem_inode_info *info;

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

2452
	inode = shmem_get_inode(dir->i_sb, dir, S_IFLNK|S_IRWXUGO, 0, VM_NORESERVE);
L
Linus Torvalds 已提交
2453 2454 2455
	if (!inode)
		return -ENOSPC;

2456
	error = security_inode_init_security(inode, dir, &dentry->d_name,
2457
					     shmem_initxattrs, NULL);
2458 2459 2460 2461 2462 2463 2464 2465
	if (error) {
		if (error != -EOPNOTSUPP) {
			iput(inode);
			return error;
		}
		error = 0;
	}

L
Linus Torvalds 已提交
2466 2467
	info = SHMEM_I(inode);
	inode->i_size = len-1;
2468 2469 2470 2471 2472 2473 2474
	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;
A
Al Viro 已提交
2475
		inode->i_link = info->symlink;
L
Linus Torvalds 已提交
2476 2477 2478 2479 2480 2481
	} else {
		error = shmem_getpage(inode, 0, &page, SGP_WRITE, NULL);
		if (error) {
			iput(inode);
			return error;
		}
2482
		inode->i_mapping->a_ops = &shmem_aops;
L
Linus Torvalds 已提交
2483
		inode->i_op = &shmem_symlink_inode_operations;
2484
		kaddr = kmap_atomic(page);
L
Linus Torvalds 已提交
2485
		memcpy(kaddr, symname, len);
2486
		kunmap_atomic(kaddr);
2487
		SetPageUptodate(page);
L
Linus Torvalds 已提交
2488
		set_page_dirty(page);
2489
		unlock_page(page);
L
Linus Torvalds 已提交
2490 2491 2492 2493 2494 2495 2496 2497 2498
		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;
}

2499
static const char *shmem_follow_link(struct dentry *dentry, void **cookie)
L
Linus Torvalds 已提交
2500 2501
{
	struct page *page = NULL;
2502
	int error = shmem_getpage(d_inode(dentry), 0, &page, SGP_READ, NULL);
2503 2504 2505 2506 2507
	if (error)
		return ERR_PTR(error);
	unlock_page(page);
	*cookie = page;
	return kmap(page);
L
Linus Torvalds 已提交
2508 2509
}

2510
static void shmem_put_link(struct inode *unused, void *cookie)
L
Linus Torvalds 已提交
2511
{
2512 2513 2514 2515
	struct page *page = cookie;
	kunmap(page);
	mark_page_accessed(page);
	page_cache_release(page);
L
Linus Torvalds 已提交
2516 2517
}

2518
#ifdef CONFIG_TMPFS_XATTR
2519
/*
2520 2521
 * Superblocks without xattr inode operations may get some security.* xattr
 * support from the LSM "for free". As soon as we have any other xattrs
2522 2523 2524 2525
 * like ACLs, we also need to implement the security.* handlers at
 * filesystem level, though.
 */

2526 2527 2528 2529 2530 2531 2532 2533 2534
/*
 * 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;
2535
	struct simple_xattr *new_xattr;
2536 2537 2538
	size_t len;

	for (xattr = xattr_array; xattr->name != NULL; xattr++) {
2539
		new_xattr = simple_xattr_alloc(xattr->value, xattr->value_len);
2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555
		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);

2556
		simple_xattr_list_add(&info->xattrs, new_xattr);
2557 2558 2559 2560 2561
	}

	return 0;
}

2562
static const struct xattr_handler *shmem_xattr_handlers[] = {
2563
#ifdef CONFIG_TMPFS_POSIX_ACL
C
Christoph Hellwig 已提交
2564 2565
	&posix_acl_access_xattr_handler,
	&posix_acl_default_xattr_handler,
2566
#endif
2567 2568
	NULL
};
2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591

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)
{
2592
	struct shmem_inode_info *info = SHMEM_I(d_inode(dentry));
2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606
	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;

2607
	return simple_xattr_get(&info->xattrs, name, buffer, size);
2608 2609 2610 2611 2612
}

static int shmem_setxattr(struct dentry *dentry, const char *name,
			  const void *value, size_t size, int flags)
{
2613
	struct shmem_inode_info *info = SHMEM_I(d_inode(dentry));
2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627
	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;

2628
	return simple_xattr_set(&info->xattrs, name, value, size, flags);
2629 2630 2631 2632
}

static int shmem_removexattr(struct dentry *dentry, const char *name)
{
2633
	struct shmem_inode_info *info = SHMEM_I(d_inode(dentry));
2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647
	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;

2648
	return simple_xattr_remove(&info->xattrs, name);
2649 2650 2651 2652
}

static ssize_t shmem_listxattr(struct dentry *dentry, char *buffer, size_t size)
{
2653
	struct shmem_inode_info *info = SHMEM_I(d_inode(dentry));
2654
	return simple_xattr_list(&info->xattrs, buffer, size);
2655 2656 2657
}
#endif /* CONFIG_TMPFS_XATTR */

2658
static const struct inode_operations shmem_short_symlink_operations = {
2659
	.readlink	= generic_readlink,
A
Al Viro 已提交
2660
	.follow_link	= simple_follow_link,
2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677
#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,
2678
#endif
2679
};
2680

2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693
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
2694 2695
static struct dentry *shmem_fh_to_dentry(struct super_block *sb,
		struct fid *fid, int fh_len, int fh_type)
2696 2697
{
	struct inode *inode;
C
Christoph Hellwig 已提交
2698
	struct dentry *dentry = NULL;
2699
	u64 inum;
C
Christoph Hellwig 已提交
2700 2701 2702

	if (fh_len < 3)
		return NULL;
2703

2704 2705 2706
	inum = fid->raw[2];
	inum = (inum << 32) | fid->raw[1];

C
Christoph Hellwig 已提交
2707 2708
	inode = ilookup5(sb, (unsigned long)(inum + fid->raw[0]),
			shmem_match, fid->raw);
2709
	if (inode) {
C
Christoph Hellwig 已提交
2710
		dentry = d_find_alias(inode);
2711 2712 2713
		iput(inode);
	}

C
Christoph Hellwig 已提交
2714
	return dentry;
2715 2716
}

A
Al Viro 已提交
2717 2718
static int shmem_encode_fh(struct inode *inode, __u32 *fh, int *len,
				struct inode *parent)
2719
{
2720 2721
	if (*len < 3) {
		*len = 3;
2722
		return FILEID_INVALID;
2723
	}
2724

A
Al Viro 已提交
2725
	if (inode_unhashed(inode)) {
2726 2727 2728 2729 2730 2731 2732
		/* 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
2733
		if (inode_unhashed(inode))
2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746
			__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;
}

2747
static const struct export_operations shmem_export_ops = {
2748 2749
	.get_parent     = shmem_get_parent,
	.encode_fh      = shmem_encode_fh,
C
Christoph Hellwig 已提交
2750
	.fh_to_dentry	= shmem_fh_to_dentry,
2751 2752
};

2753 2754
static int shmem_parse_options(char *options, struct shmem_sb_info *sbinfo,
			       bool remount)
L
Linus Torvalds 已提交
2755 2756
{
	char *this_char, *value, *rest;
G
Greg Thelen 已提交
2757
	struct mempolicy *mpol = NULL;
2758 2759
	uid_t uid;
	gid_t gid;
L
Linus Torvalds 已提交
2760

2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777
	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
2778 2779 2780 2781 2782 2783 2784 2785
		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
2786
			goto error;
L
Linus Torvalds 已提交
2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799
		}

		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;
2800 2801
			sbinfo->max_blocks =
				DIV_ROUND_UP(size, PAGE_CACHE_SIZE);
L
Linus Torvalds 已提交
2802
		} else if (!strcmp(this_char,"nr_blocks")) {
2803
			sbinfo->max_blocks = memparse(value, &rest);
L
Linus Torvalds 已提交
2804 2805 2806
			if (*rest)
				goto bad_val;
		} else if (!strcmp(this_char,"nr_inodes")) {
2807
			sbinfo->max_inodes = memparse(value, &rest);
L
Linus Torvalds 已提交
2808 2809 2810
			if (*rest)
				goto bad_val;
		} else if (!strcmp(this_char,"mode")) {
2811
			if (remount)
L
Linus Torvalds 已提交
2812
				continue;
2813
			sbinfo->mode = simple_strtoul(value, &rest, 8) & 07777;
L
Linus Torvalds 已提交
2814 2815 2816
			if (*rest)
				goto bad_val;
		} else if (!strcmp(this_char,"uid")) {
2817
			if (remount)
L
Linus Torvalds 已提交
2818
				continue;
2819
			uid = simple_strtoul(value, &rest, 0);
L
Linus Torvalds 已提交
2820 2821
			if (*rest)
				goto bad_val;
2822 2823 2824
			sbinfo->uid = make_kuid(current_user_ns(), uid);
			if (!uid_valid(sbinfo->uid))
				goto bad_val;
L
Linus Torvalds 已提交
2825
		} else if (!strcmp(this_char,"gid")) {
2826
			if (remount)
L
Linus Torvalds 已提交
2827
				continue;
2828
			gid = simple_strtoul(value, &rest, 0);
L
Linus Torvalds 已提交
2829 2830
			if (*rest)
				goto bad_val;
2831 2832 2833
			sbinfo->gid = make_kgid(current_user_ns(), gid);
			if (!gid_valid(sbinfo->gid))
				goto bad_val;
2834
		} else if (!strcmp(this_char,"mpol")) {
G
Greg Thelen 已提交
2835 2836 2837
			mpol_put(mpol);
			mpol = NULL;
			if (mpol_parse_str(value, &mpol))
2838
				goto bad_val;
L
Linus Torvalds 已提交
2839 2840 2841
		} else {
			printk(KERN_ERR "tmpfs: Bad mount option %s\n",
			       this_char);
G
Greg Thelen 已提交
2842
			goto error;
L
Linus Torvalds 已提交
2843 2844
		}
	}
G
Greg Thelen 已提交
2845
	sbinfo->mpol = mpol;
L
Linus Torvalds 已提交
2846 2847 2848 2849 2850
	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
2851 2852
error:
	mpol_put(mpol);
L
Linus Torvalds 已提交
2853 2854 2855 2856 2857 2858 2859
	return 1;

}

static int shmem_remount_fs(struct super_block *sb, int *flags, char *data)
{
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);
2860
	struct shmem_sb_info config = *sbinfo;
2861 2862 2863
	unsigned long inodes;
	int error = -EINVAL;

2864
	config.mpol = NULL;
2865
	if (shmem_parse_options(data, &config, true))
2866
		return error;
L
Linus Torvalds 已提交
2867

2868 2869
	spin_lock(&sbinfo->stat_lock);
	inodes = sbinfo->max_inodes - sbinfo->free_inodes;
2870
	if (percpu_counter_compare(&sbinfo->used_blocks, config.max_blocks) > 0)
2871
		goto out;
2872
	if (config.max_inodes < inodes)
2873 2874
		goto out;
	/*
2875
	 * Those tests disallow limited->unlimited while any are in use;
2876 2877 2878
	 * but we must separately disallow unlimited->limited, because
	 * in that case we have no record of how much is already in use.
	 */
2879
	if (config.max_blocks && !sbinfo->max_blocks)
2880
		goto out;
2881
	if (config.max_inodes && !sbinfo->max_inodes)
2882 2883 2884
		goto out;

	error = 0;
2885 2886 2887
	sbinfo->max_blocks  = config.max_blocks;
	sbinfo->max_inodes  = config.max_inodes;
	sbinfo->free_inodes = config.max_inodes - inodes;
2888

2889 2890 2891 2892 2893 2894 2895
	/*
	 * Preserve previous mempolicy unless mpol remount option was specified.
	 */
	if (config.mpol) {
		mpol_put(sbinfo->mpol);
		sbinfo->mpol = config.mpol;	/* transfers initial ref */
	}
2896 2897 2898
out:
	spin_unlock(&sbinfo->stat_lock);
	return error;
L
Linus Torvalds 已提交
2899
}
2900

2901
static int shmem_show_options(struct seq_file *seq, struct dentry *root)
2902
{
2903
	struct shmem_sb_info *sbinfo = SHMEM_SB(root->d_sb);
2904 2905 2906 2907 2908 2909 2910

	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))
2911
		seq_printf(seq, ",mode=%03ho", sbinfo->mode);
2912 2913 2914 2915 2916 2917
	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));
2918
	shmem_show_mpol(seq, sbinfo->mpol);
2919 2920
	return 0;
}
D
David Herrmann 已提交
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 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991

#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;
}

2992
#endif /* CONFIG_TMPFS */
L
Linus Torvalds 已提交
2993 2994 2995

static void shmem_put_super(struct super_block *sb)
{
2996 2997 2998
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);

	percpu_counter_destroy(&sbinfo->used_blocks);
G
Greg Thelen 已提交
2999
	mpol_put(sbinfo->mpol);
3000
	kfree(sbinfo);
L
Linus Torvalds 已提交
3001 3002 3003
	sb->s_fs_info = NULL;
}

3004
int shmem_fill_super(struct super_block *sb, void *data, int silent)
L
Linus Torvalds 已提交
3005 3006
{
	struct inode *inode;
3007
	struct shmem_sb_info *sbinfo;
3008 3009 3010
	int err = -ENOMEM;

	/* Round up to L1_CACHE_BYTES to resist false sharing */
3011
	sbinfo = kzalloc(max((int)sizeof(struct shmem_sb_info),
3012 3013 3014 3015 3016
				L1_CACHE_BYTES), GFP_KERNEL);
	if (!sbinfo)
		return -ENOMEM;

	sbinfo->mode = S_IRWXUGO | S_ISVTX;
3017 3018
	sbinfo->uid = current_fsuid();
	sbinfo->gid = current_fsgid();
3019
	sb->s_fs_info = sbinfo;
L
Linus Torvalds 已提交
3020

3021
#ifdef CONFIG_TMPFS
L
Linus Torvalds 已提交
3022 3023 3024 3025 3026
	/*
	 * 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.
	 */
3027
	if (!(sb->s_flags & MS_KERNMOUNT)) {
3028 3029 3030 3031 3032 3033
		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;
		}
3034 3035
	} else {
		sb->s_flags |= MS_NOUSER;
L
Linus Torvalds 已提交
3036
	}
3037
	sb->s_export_op = &shmem_export_ops;
H
Hugh Dickins 已提交
3038
	sb->s_flags |= MS_NOSEC;
L
Linus Torvalds 已提交
3039 3040 3041 3042
#else
	sb->s_flags |= MS_NOUSER;
#endif

3043
	spin_lock_init(&sbinfo->stat_lock);
3044
	if (percpu_counter_init(&sbinfo->used_blocks, 0, GFP_KERNEL))
3045
		goto failed;
3046
	sbinfo->free_inodes = sbinfo->max_inodes;
3047

3048
	sb->s_maxbytes = MAX_LFS_FILESIZE;
L
Linus Torvalds 已提交
3049 3050 3051 3052
	sb->s_blocksize = PAGE_CACHE_SIZE;
	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
	sb->s_magic = TMPFS_MAGIC;
	sb->s_op = &shmem_ops;
3053
	sb->s_time_gran = 1;
3054
#ifdef CONFIG_TMPFS_XATTR
3055
	sb->s_xattr = shmem_xattr_handlers;
3056 3057
#endif
#ifdef CONFIG_TMPFS_POSIX_ACL
3058 3059
	sb->s_flags |= MS_POSIXACL;
#endif
3060

3061
	inode = shmem_get_inode(sb, NULL, S_IFDIR | sbinfo->mode, 0, VM_NORESERVE);
L
Linus Torvalds 已提交
3062 3063
	if (!inode)
		goto failed;
3064 3065
	inode->i_uid = sbinfo->uid;
	inode->i_gid = sbinfo->gid;
A
Al Viro 已提交
3066 3067
	sb->s_root = d_make_root(inode);
	if (!sb->s_root)
3068
		goto failed;
L
Linus Torvalds 已提交
3069 3070 3071 3072 3073 3074 3075
	return 0;

failed:
	shmem_put_super(sb);
	return err;
}

3076
static struct kmem_cache *shmem_inode_cachep;
L
Linus Torvalds 已提交
3077 3078 3079

static struct inode *shmem_alloc_inode(struct super_block *sb)
{
3080 3081 3082
	struct shmem_inode_info *info;
	info = kmem_cache_alloc(shmem_inode_cachep, GFP_KERNEL);
	if (!info)
L
Linus Torvalds 已提交
3083
		return NULL;
3084
	return &info->vfs_inode;
L
Linus Torvalds 已提交
3085 3086
}

3087
static void shmem_destroy_callback(struct rcu_head *head)
N
Nick Piggin 已提交
3088 3089 3090 3091 3092
{
	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
3093 3094
static void shmem_destroy_inode(struct inode *inode)
{
3095
	if (S_ISREG(inode->i_mode))
L
Linus Torvalds 已提交
3096
		mpol_free_shared_policy(&SHMEM_I(inode)->policy);
3097
	call_rcu(&inode->i_rcu, shmem_destroy_callback);
L
Linus Torvalds 已提交
3098 3099
}

3100
static void shmem_init_inode(void *foo)
L
Linus Torvalds 已提交
3101
{
3102 3103
	struct shmem_inode_info *info = foo;
	inode_init_once(&info->vfs_inode);
L
Linus Torvalds 已提交
3104 3105
}

3106
static int shmem_init_inodecache(void)
L
Linus Torvalds 已提交
3107 3108 3109
{
	shmem_inode_cachep = kmem_cache_create("shmem_inode_cache",
				sizeof(struct shmem_inode_info),
3110
				0, SLAB_PANIC, shmem_init_inode);
L
Linus Torvalds 已提交
3111 3112 3113
	return 0;
}

3114
static void shmem_destroy_inodecache(void)
L
Linus Torvalds 已提交
3115
{
3116
	kmem_cache_destroy(shmem_inode_cachep);
L
Linus Torvalds 已提交
3117 3118
}

3119
static const struct address_space_operations shmem_aops = {
L
Linus Torvalds 已提交
3120
	.writepage	= shmem_writepage,
3121
	.set_page_dirty	= __set_page_dirty_no_writeback,
L
Linus Torvalds 已提交
3122
#ifdef CONFIG_TMPFS
N
Nick Piggin 已提交
3123 3124
	.write_begin	= shmem_write_begin,
	.write_end	= shmem_write_end,
L
Linus Torvalds 已提交
3125
#endif
3126
#ifdef CONFIG_MIGRATION
3127
	.migratepage	= migrate_page,
3128
#endif
3129
	.error_remove_page = generic_error_remove_page,
L
Linus Torvalds 已提交
3130 3131
};

3132
static const struct file_operations shmem_file_operations = {
L
Linus Torvalds 已提交
3133 3134
	.mmap		= shmem_mmap,
#ifdef CONFIG_TMPFS
3135
	.llseek		= shmem_file_llseek,
A
Al Viro 已提交
3136
	.read_iter	= shmem_file_read_iter,
3137
	.write_iter	= generic_file_write_iter,
3138
	.fsync		= noop_fsync,
3139
	.splice_read	= shmem_file_splice_read,
3140
	.splice_write	= iter_file_splice_write,
3141
	.fallocate	= shmem_fallocate,
L
Linus Torvalds 已提交
3142 3143 3144
#endif
};

3145
static const struct inode_operations shmem_inode_operations = {
3146
	.getattr	= shmem_getattr,
3147
	.setattr	= shmem_setattr,
3148 3149 3150 3151 3152
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
C
Christoph Hellwig 已提交
3153
	.set_acl	= simple_set_acl,
3154
#endif
L
Linus Torvalds 已提交
3155 3156
};

3157
static const struct inode_operations shmem_dir_inode_operations = {
L
Linus Torvalds 已提交
3158 3159 3160 3161 3162 3163 3164 3165 3166
#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
3167
	.rename2	= shmem_rename2,
3168
	.tmpfile	= shmem_tmpfile,
L
Linus Torvalds 已提交
3169
#endif
3170 3171 3172 3173 3174 3175
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
#endif
3176
#ifdef CONFIG_TMPFS_POSIX_ACL
3177
	.setattr	= shmem_setattr,
C
Christoph Hellwig 已提交
3178
	.set_acl	= simple_set_acl,
3179 3180 3181
#endif
};

3182
static const struct inode_operations shmem_special_inode_operations = {
3183 3184 3185 3186 3187 3188
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
#endif
3189
#ifdef CONFIG_TMPFS_POSIX_ACL
3190
	.setattr	= shmem_setattr,
C
Christoph Hellwig 已提交
3191
	.set_acl	= simple_set_acl,
3192
#endif
L
Linus Torvalds 已提交
3193 3194
};

3195
static const struct super_operations shmem_ops = {
L
Linus Torvalds 已提交
3196 3197 3198 3199 3200
	.alloc_inode	= shmem_alloc_inode,
	.destroy_inode	= shmem_destroy_inode,
#ifdef CONFIG_TMPFS
	.statfs		= shmem_statfs,
	.remount_fs	= shmem_remount_fs,
3201
	.show_options	= shmem_show_options,
L
Linus Torvalds 已提交
3202
#endif
A
Al Viro 已提交
3203
	.evict_inode	= shmem_evict_inode,
L
Linus Torvalds 已提交
3204 3205 3206 3207
	.drop_inode	= generic_delete_inode,
	.put_super	= shmem_put_super,
};

3208
static const struct vm_operations_struct shmem_vm_ops = {
3209
	.fault		= shmem_fault,
3210
	.map_pages	= filemap_map_pages,
L
Linus Torvalds 已提交
3211 3212 3213 3214 3215 3216
#ifdef CONFIG_NUMA
	.set_policy     = shmem_set_policy,
	.get_policy     = shmem_get_policy,
#endif
};

A
Al Viro 已提交
3217 3218
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
3219
{
A
Al Viro 已提交
3220
	return mount_nodev(fs_type, flags, data, shmem_fill_super);
L
Linus Torvalds 已提交
3221 3222
}

3223
static struct file_system_type shmem_fs_type = {
L
Linus Torvalds 已提交
3224 3225
	.owner		= THIS_MODULE,
	.name		= "tmpfs",
A
Al Viro 已提交
3226
	.mount		= shmem_mount,
L
Linus Torvalds 已提交
3227
	.kill_sb	= kill_litter_super,
3228
	.fs_flags	= FS_USERNS_MOUNT,
L
Linus Torvalds 已提交
3229 3230
};

3231
int __init shmem_init(void)
L
Linus Torvalds 已提交
3232 3233 3234
{
	int error;

3235 3236 3237 3238
	/* If rootfs called this, don't re-init */
	if (shmem_inode_cachep)
		return 0;

3239
	error = shmem_init_inodecache();
L
Linus Torvalds 已提交
3240 3241 3242
	if (error)
		goto out3;

3243
	error = register_filesystem(&shmem_fs_type);
L
Linus Torvalds 已提交
3244 3245 3246 3247
	if (error) {
		printk(KERN_ERR "Could not register tmpfs\n");
		goto out2;
	}
3248

3249
	shm_mnt = kern_mount(&shmem_fs_type);
L
Linus Torvalds 已提交
3250 3251 3252 3253 3254 3255 3256 3257
	if (IS_ERR(shm_mnt)) {
		error = PTR_ERR(shm_mnt);
		printk(KERN_ERR "Could not kern_mount tmpfs\n");
		goto out1;
	}
	return 0;

out1:
3258
	unregister_filesystem(&shmem_fs_type);
L
Linus Torvalds 已提交
3259
out2:
3260
	shmem_destroy_inodecache();
L
Linus Torvalds 已提交
3261 3262 3263 3264
out3:
	shm_mnt = ERR_PTR(error);
	return error;
}
3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276

#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.
 */

3277
static struct file_system_type shmem_fs_type = {
3278
	.name		= "tmpfs",
A
Al Viro 已提交
3279
	.mount		= ramfs_mount,
3280
	.kill_sb	= kill_litter_super,
3281
	.fs_flags	= FS_USERNS_MOUNT,
3282 3283
};

3284
int __init shmem_init(void)
3285
{
3286
	BUG_ON(register_filesystem(&shmem_fs_type) != 0);
3287

3288
	shm_mnt = kern_mount(&shmem_fs_type);
3289 3290 3291 3292 3293
	BUG_ON(IS_ERR(shm_mnt));

	return 0;
}

3294
int shmem_unuse(swp_entry_t swap, struct page *page)
3295 3296 3297 3298
{
	return 0;
}

H
Hugh Dickins 已提交
3299 3300 3301 3302 3303
int shmem_lock(struct file *file, int lock, struct user_struct *user)
{
	return 0;
}

3304 3305 3306 3307
void shmem_unlock_mapping(struct address_space *mapping)
{
}

3308
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
3309
{
3310
	truncate_inode_pages_range(inode->i_mapping, lstart, lend);
3311 3312 3313
}
EXPORT_SYMBOL_GPL(shmem_truncate_range);

3314 3315
#define shmem_vm_ops				generic_file_vm_ops
#define shmem_file_operations			ramfs_file_operations
3316
#define shmem_get_inode(sb, dir, mode, dev, flags)	ramfs_get_inode(sb, dir, mode, dev)
3317 3318
#define shmem_acct_size(flags, size)		0
#define shmem_unacct_size(flags, size)		do {} while (0)
3319 3320 3321 3322

#endif /* CONFIG_SHMEM */

/* common code */
L
Linus Torvalds 已提交
3323

3324
static struct dentry_operations anon_ops = {
3325
	.d_dname = simple_dname
3326 3327
};

3328 3329
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
3330
{
A
Al Viro 已提交
3331
	struct file *res;
L
Linus Torvalds 已提交
3332
	struct inode *inode;
3333
	struct path path;
3334
	struct super_block *sb;
L
Linus Torvalds 已提交
3335 3336 3337
	struct qstr this;

	if (IS_ERR(shm_mnt))
A
Al Viro 已提交
3338
		return ERR_CAST(shm_mnt);
L
Linus Torvalds 已提交
3339

3340
	if (size < 0 || size > MAX_LFS_FILESIZE)
L
Linus Torvalds 已提交
3341 3342 3343 3344 3345
		return ERR_PTR(-EINVAL);

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

A
Al Viro 已提交
3346
	res = ERR_PTR(-ENOMEM);
L
Linus Torvalds 已提交
3347 3348 3349
	this.name = name;
	this.len = strlen(name);
	this.hash = 0; /* will go */
3350
	sb = shm_mnt->mnt_sb;
3351
	path.mnt = mntget(shm_mnt);
3352
	path.dentry = d_alloc_pseudo(sb, &this);
3353
	if (!path.dentry)
L
Linus Torvalds 已提交
3354
		goto put_memory;
3355
	d_set_d_op(path.dentry, &anon_ops);
L
Linus Torvalds 已提交
3356

A
Al Viro 已提交
3357
	res = ERR_PTR(-ENOSPC);
3358
	inode = shmem_get_inode(sb, NULL, S_IFREG | S_IRWXUGO, 0, flags);
L
Linus Torvalds 已提交
3359
	if (!inode)
3360
		goto put_memory;
L
Linus Torvalds 已提交
3361

3362
	inode->i_flags |= i_flags;
3363
	d_instantiate(path.dentry, inode);
L
Linus Torvalds 已提交
3364
	inode->i_size = size;
3365
	clear_nlink(inode);	/* It is unlinked */
A
Al Viro 已提交
3366 3367
	res = ERR_PTR(ramfs_nommu_expand_for_mapping(inode, size));
	if (IS_ERR(res))
3368
		goto put_path;
3369

A
Al Viro 已提交
3370
	res = alloc_file(&path, FMODE_WRITE | FMODE_READ,
3371
		  &shmem_file_operations);
A
Al Viro 已提交
3372
	if (IS_ERR(res))
3373
		goto put_path;
3374

A
Al Viro 已提交
3375
	return res;
L
Linus Torvalds 已提交
3376 3377 3378

put_memory:
	shmem_unacct_size(flags, size);
3379 3380
put_path:
	path_put(&path);
A
Al Viro 已提交
3381
	return res;
L
Linus Torvalds 已提交
3382
}
3383 3384 3385 3386 3387

/**
 * 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
3388 3389
 *	higher layer.  The users are the big_key and shm implementations.  LSM
 *	checks are provided at the key or shm level rather than the inode.
3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408
 * @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);
}
3409
EXPORT_SYMBOL_GPL(shmem_file_setup);
L
Linus Torvalds 已提交
3410

3411
/**
L
Linus Torvalds 已提交
3412 3413 3414 3415 3416 3417 3418 3419
 * 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;

3420 3421 3422 3423 3424 3425 3426
	/*
	 * Cloning a new file under mmap_sem leads to a lock ordering conflict
	 * between XFS directory reading and selinux: since this file is only
	 * accessible to the user through its mapping, use S_PRIVATE flag to
	 * bypass file security, in the same way as shmem_kernel_file_setup().
	 */
	file = __shmem_file_setup("dev/zero", size, vma->vm_flags, S_PRIVATE);
L
Linus Torvalds 已提交
3427 3428 3429 3430 3431 3432 3433 3434 3435
	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;
}
3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448

/**
 * 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.
 *
3449 3450
 * i915_gem_object_get_pages_gtt() mixes __GFP_NORETRY | __GFP_NOWARN in
 * with the mapping_gfp_mask(), to avoid OOMing the machine unnecessarily.
3451 3452 3453 3454
 */
struct page *shmem_read_mapping_page_gfp(struct address_space *mapping,
					 pgoff_t index, gfp_t gfp)
{
3455 3456
#ifdef CONFIG_SHMEM
	struct inode *inode = mapping->host;
H
Hugh Dickins 已提交
3457
	struct page *page;
3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470
	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
	 */
3471
	return read_cache_page_gfp(mapping, index, gfp);
3472
#endif
3473 3474
}
EXPORT_SYMBOL_GPL(shmem_read_mapping_page_gfp);