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>

76 77
#include "internal.h"

H
Hugh Dickins 已提交
78
#define BLOCKS_PER_PAGE  (PAGE_CACHE_SIZE/512)
L
Linus Torvalds 已提交
79 80 81 82 83
#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

84 85 86
/* Symlink up to this size is kmalloc'ed instead of using a swappable page */
#define SHORT_SYMLINK_LEN 128

87
/*
88 89 90
 * 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.
91 92
 */
struct shmem_falloc {
93
	wait_queue_head_t *waitq; /* faults into hole wait for punch to end */
94 95 96 97 98 99
	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 */
};

100
/* Flag allocation requirements to shmem_getpage */
L
Linus Torvalds 已提交
101 102 103
enum sgp_type {
	SGP_READ,	/* don't exceed i_size, don't allocate page */
	SGP_CACHE,	/* don't exceed i_size, may allocate page */
104
	SGP_DIRTY,	/* like SGP_CACHE, but set new page dirty */
105 106
	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
107 108
};

A
Andrew Morton 已提交
109
#ifdef CONFIG_TMPFS
110 111 112 113 114 115 116 117 118
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
119
#endif
120

121 122 123
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);
124 125 126 127 128 129 130 131 132
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
133 134 135 136 137 138 139 140 141 142 143 144 145 146

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)
{
147
	return (flags & VM_NORESERVE) ?
148
		0 : security_vm_enough_memory_mm(current->mm, VM_ACCT(size));
L
Linus Torvalds 已提交
149 150 151 152
}

static inline void shmem_unacct_size(unsigned long flags, loff_t size)
{
153
	if (!(flags & VM_NORESERVE))
L
Linus Torvalds 已提交
154 155 156
		vm_unacct_memory(VM_ACCT(size));
}

157 158 159 160 161 162 163 164 165 166 167 168 169
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
170 171 172 173 174 175 176 177
/*
 * ... 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)
{
178
	return (flags & VM_NORESERVE) ?
179
		security_vm_enough_memory_mm(current->mm, VM_ACCT(PAGE_CACHE_SIZE)) : 0;
L
Linus Torvalds 已提交
180 181 182 183
}

static inline void shmem_unacct_blocks(unsigned long flags, long pages)
{
184
	if (flags & VM_NORESERVE)
L
Linus Torvalds 已提交
185 186 187
		vm_unacct_memory(pages * VM_ACCT(PAGE_CACHE_SIZE));
}

188
static const struct super_operations shmem_ops;
189
static const struct address_space_operations shmem_aops;
190
static const struct file_operations shmem_file_operations;
191 192 193
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;
194
static const struct vm_operations_struct shmem_vm_ops;
L
Linus Torvalds 已提交
195 196

static LIST_HEAD(shmem_swaplist);
197
static DEFINE_MUTEX(shmem_swaplist_mutex);
L
Linus Torvalds 已提交
198

199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
static int shmem_reserve_inode(struct super_block *sb)
{
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);
	if (sbinfo->max_inodes) {
		spin_lock(&sbinfo->stat_lock);
		if (!sbinfo->free_inodes) {
			spin_unlock(&sbinfo->stat_lock);
			return -ENOSPC;
		}
		sbinfo->free_inodes--;
		spin_unlock(&sbinfo->stat_lock);
	}
	return 0;
}

static void shmem_free_inode(struct super_block *sb)
{
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);
	if (sbinfo->max_inodes) {
		spin_lock(&sbinfo->stat_lock);
		sbinfo->free_inodes++;
		spin_unlock(&sbinfo->stat_lock);
	}
}

224
/**
225
 * shmem_recalc_inode - recalculate the block usage of an inode
L
Linus Torvalds 已提交
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242
 * @inode: inode to recalc
 *
 * We have to calculate the free blocks since the mm can drop
 * undirtied hole pages behind our back.
 *
 * But normally   info->alloced == inode->i_mapping->nrpages + info->swapped
 * So mm freed is info->alloced - (inode->i_mapping->nrpages + info->swapped)
 *
 * It has to be called with the spinlock held.
 */
static void shmem_recalc_inode(struct inode *inode)
{
	struct shmem_inode_info *info = SHMEM_I(inode);
	long freed;

	freed = info->alloced - info->swapped - inode->i_mapping->nrpages;
	if (freed > 0) {
243 244 245
		struct shmem_sb_info *sbinfo = SHMEM_SB(inode->i_sb);
		if (sbinfo->max_blocks)
			percpu_counter_add(&sbinfo->used_blocks, -freed);
L
Linus Torvalds 已提交
246
		info->alloced -= freed;
247
		inode->i_blocks -= freed * BLOCKS_PER_PAGE;
L
Linus Torvalds 已提交
248 249 250 251
		shmem_unacct_blocks(info->flags, freed);
	}
}

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

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

273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290
/*
 * Sometimes, before we decide whether to proceed or to fail, we must check
 * that an entry was not already brought back from swap by a racing thread.
 *
 * Checking page is not enough: by the time a SwapCache page is locked, it
 * might be reused, and again be SwapCache, using the same swap as before.
 */
static bool shmem_confirm_swap(struct address_space *mapping,
			       pgoff_t index, swp_entry_t swap)
{
	void *item;

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

291 292 293 294 295
/*
 * 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,
296
				   pgoff_t index, void *expected)
297
{
298
	int error;
299

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

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

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

326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344
/*
 * Like delete_from_page_cache, but substitutes swap for page.
 */
static void shmem_delete_from_page_cache(struct page *page, void *radswap)
{
	struct address_space *mapping = page->mapping;
	int error;

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

345 346 347 348 349 350
/*
 * Remove swap entry from radix tree, free the swap and its page cache.
 */
static int shmem_free_swap(struct address_space *mapping,
			   pgoff_t index, void *radswap)
{
351
	void *old;
352 353

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

362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379
/*
 * SysV IPC SHM_UNLOCK restore Unevictable pages to their evictable lists.
 */
void shmem_unlock_mapping(struct address_space *mapping)
{
	struct pagevec pvec;
	pgoff_t indices[PAGEVEC_SIZE];
	pgoff_t index = 0;

	pagevec_init(&pvec, 0);
	/*
	 * Minor point, but we might as well stop if someone else SHM_LOCKs it.
	 */
	while (!mapping_unevictable(mapping)) {
		/*
		 * Avoid pagevec_lookup(): find_get_pages() returns 0 as if it
		 * has finished, if it hits a row of PAGEVEC_SIZE swap entries.
		 */
380 381
		pvec.nr = find_get_entries(mapping, index,
					   PAGEVEC_SIZE, pvec.pages, indices);
382 383 384
		if (!pvec.nr)
			break;
		index = indices[pvec.nr - 1] + 1;
385
		pagevec_remove_exceptionals(&pvec);
386 387 388 389
		check_move_unevictable_pages(pvec.pages, pvec.nr);
		pagevec_release(&pvec);
		cond_resched();
	}
390 391 392 393
}

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

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

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

425
			index = indices[i];
426
			if (index >= end)
427 428
				break;

429
			if (radix_tree_exceptional_entry(page)) {
430 431
				if (unfalloc)
					continue;
432 433
				nr_swaps_freed += !shmem_free_swap(mapping,
								index, page);
434
				continue;
435 436 437
			}

			if (!trylock_page(page))
438
				continue;
439 440
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
441
					VM_BUG_ON_PAGE(PageWriteback(page), page);
442 443
					truncate_inode_page(mapping, page);
				}
444 445 446
			}
			unlock_page(page);
		}
447
		pagevec_remove_exceptionals(&pvec);
448
		pagevec_release(&pvec);
449 450 451
		cond_resched();
		index++;
	}
L
Linus Torvalds 已提交
452

453
	if (partial_start) {
454 455 456
		struct page *page = NULL;
		shmem_getpage(inode, start - 1, &page, SGP_READ, NULL);
		if (page) {
457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472
			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);
473 474 475 476 477
			set_page_dirty(page);
			unlock_page(page);
			page_cache_release(page);
		}
	}
478 479
	if (start >= end)
		return;
480 481

	index = start;
482
	while (index < end) {
483
		cond_resched();
484 485

		pvec.nr = find_get_entries(mapping, index,
486
				min(end - index, (pgoff_t)PAGEVEC_SIZE),
487
				pvec.pages, indices);
488
		if (!pvec.nr) {
489 490
			/* If all gone or hole-punch or unfalloc, we're done */
			if (index == start || end != -1)
491
				break;
492
			/* But if truncating, restart to make sure all gone */
493 494 495 496 497 498
			index = start;
			continue;
		}
		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];

499
			index = indices[i];
500
			if (index >= end)
501 502
				break;

503
			if (radix_tree_exceptional_entry(page)) {
504 505
				if (unfalloc)
					continue;
506 507 508 509 510 511
				if (shmem_free_swap(mapping, index, page)) {
					/* Swap was replaced by page: retry */
					index--;
					break;
				}
				nr_swaps_freed++;
512 513 514
				continue;
			}

515
			lock_page(page);
516 517
			if (!unfalloc || !PageUptodate(page)) {
				if (page->mapping == mapping) {
518
					VM_BUG_ON_PAGE(PageWriteback(page), page);
519
					truncate_inode_page(mapping, page);
520 521 522 523 524
				} else {
					/* Page was replaced by swap: retry */
					unlock_page(page);
					index--;
					break;
525
				}
526
			}
527 528
			unlock_page(page);
		}
529
		pagevec_remove_exceptionals(&pvec);
530
		pagevec_release(&pvec);
531 532
		index++;
	}
533

L
Linus Torvalds 已提交
534
	spin_lock(&info->lock);
535
	info->swapped -= nr_swaps_freed;
L
Linus Torvalds 已提交
536 537
	shmem_recalc_inode(inode);
	spin_unlock(&info->lock);
538
}
L
Linus Torvalds 已提交
539

540 541 542
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
{
	shmem_undo_range(inode, lstart, lend, false);
543
	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
L
Linus Torvalds 已提交
544
}
545
EXPORT_SYMBOL_GPL(shmem_truncate_range);
L
Linus Torvalds 已提交
546

547 548 549 550 551 552
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);

553 554 555 556 557
	if (info->alloced - info->swapped != inode->i_mapping->nrpages) {
		spin_lock(&info->lock);
		shmem_recalc_inode(inode);
		spin_unlock(&info->lock);
	}
558 559 560 561
	generic_fillattr(inode, stat);
	return 0;
}

562
static int shmem_setattr(struct dentry *dentry, struct iattr *attr)
L
Linus Torvalds 已提交
563
{
564
	struct inode *inode = d_inode(dentry);
D
David Herrmann 已提交
565
	struct shmem_inode_info *info = SHMEM_I(inode);
L
Linus Torvalds 已提交
566 567
	int error;

568 569 570 571
	error = inode_change_ok(inode, attr);
	if (error)
		return error;

572 573 574
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
		loff_t oldsize = inode->i_size;
		loff_t newsize = attr->ia_size;
575

D
David Herrmann 已提交
576 577 578 579 580
		/* protected by i_mutex */
		if ((newsize < oldsize && (info->seals & F_SEAL_SHRINK)) ||
		    (newsize > oldsize && (info->seals & F_SEAL_GROW)))
			return -EPERM;

581
		if (newsize != oldsize) {
582 583 584 585
			error = shmem_reacct_size(SHMEM_I(inode)->flags,
					oldsize, newsize);
			if (error)
				return error;
586 587 588
			i_size_write(inode, newsize);
			inode->i_ctime = inode->i_mtime = CURRENT_TIME;
		}
589
		if (newsize <= oldsize) {
590
			loff_t holebegin = round_up(newsize, PAGE_SIZE);
591 592 593 594 595 596
			if (oldsize > holebegin)
				unmap_mapping_range(inode->i_mapping,
							holebegin, 0, 1);
			if (info->alloced)
				shmem_truncate_range(inode,
							newsize, (loff_t)-1);
597
			/* unmap again to remove racily COWed private pages */
598 599 600
			if (oldsize > holebegin)
				unmap_mapping_range(inode->i_mapping,
							holebegin, 0, 1);
601
		}
L
Linus Torvalds 已提交
602 603
	}

604 605
	setattr_copy(inode, attr);
	if (attr->ia_valid & ATTR_MODE)
C
Christoph Hellwig 已提交
606
		error = posix_acl_chmod(inode, inode->i_mode);
L
Linus Torvalds 已提交
607 608 609
	return error;
}

A
Al Viro 已提交
610
static void shmem_evict_inode(struct inode *inode)
L
Linus Torvalds 已提交
611 612 613
{
	struct shmem_inode_info *info = SHMEM_I(inode);

614
	if (inode->i_mapping->a_ops == &shmem_aops) {
L
Linus Torvalds 已提交
615 616
		shmem_unacct_size(info->flags, inode->i_size);
		inode->i_size = 0;
617
		shmem_truncate_range(inode, 0, (loff_t)-1);
L
Linus Torvalds 已提交
618
		if (!list_empty(&info->swaplist)) {
619
			mutex_lock(&shmem_swaplist_mutex);
L
Linus Torvalds 已提交
620
			list_del_init(&info->swaplist);
621
			mutex_unlock(&shmem_swaplist_mutex);
L
Linus Torvalds 已提交
622
		}
623 624
	} else
		kfree(info->symlink);
625

626
	simple_xattrs_free(&info->xattrs);
627
	WARN_ON(inode->i_blocks);
628
	shmem_free_inode(inode->i_sb);
629
	clear_inode(inode);
L
Linus Torvalds 已提交
630 631
}

632 633 634
/*
 * If swap found in inode, free it and move page from swapcache to filecache.
 */
635
static int shmem_unuse_inode(struct shmem_inode_info *info,
636
			     swp_entry_t swap, struct page **pagep)
L
Linus Torvalds 已提交
637
{
638
	struct address_space *mapping = info->vfs_inode.i_mapping;
639
	void *radswap;
640
	pgoff_t index;
641 642
	gfp_t gfp;
	int error = 0;
L
Linus Torvalds 已提交
643

644
	radswap = swp_to_radix_entry(swap);
645
	index = radix_tree_locate_item(&mapping->page_tree, radswap);
646
	if (index == -1)
647
		return -EAGAIN;	/* tell shmem_unuse we found nothing */
648

H
Hugh Dickins 已提交
649 650
	/*
	 * Move _head_ to start search for next from here.
A
Al Viro 已提交
651
	 * But be careful: shmem_evict_inode checks list_empty without taking
H
Hugh Dickins 已提交
652
	 * mutex, and there's an instant in list_move_tail when info->swaplist
653
	 * would appear empty, if it were the only one on shmem_swaplist.
H
Hugh Dickins 已提交
654 655 656
	 */
	if (shmem_swaplist.next != &info->swaplist)
		list_move_tail(&shmem_swaplist, &info->swaplist);
657

658 659 660 661 662 663 664
	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
665 666 667 668 669 670 671 672 673 674 675 676 677 678 679
		 * 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).
680 681 682 683 684
		 */
		if (!page_swapcount(*pagep))
			error = -ENOENT;
	}

K
KAMEZAWA Hiroyuki 已提交
685
	/*
686 687 688
	 * 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
689
	 */
690 691
	if (!error)
		error = shmem_add_to_page_cache(*pagep, mapping, index,
692
						radswap);
693
	if (error != -ENOMEM) {
694 695 696 697
		/*
		 * Truncation and eviction use free_swap_and_cache(), which
		 * only does trylock page: if we raced, best clean up here.
		 */
698 699
		delete_from_swap_cache(*pagep);
		set_page_dirty(*pagep);
700 701 702 703 704 705
		if (!error) {
			spin_lock(&info->lock);
			info->swapped--;
			spin_unlock(&info->lock);
			swap_free(swap);
		}
L
Linus Torvalds 已提交
706
	}
707
	return error;
L
Linus Torvalds 已提交
708 709 710
}

/*
711
 * Search through swapped inodes to find and replace swap by page.
L
Linus Torvalds 已提交
712
 */
713
int shmem_unuse(swp_entry_t swap, struct page *page)
L
Linus Torvalds 已提交
714
{
715
	struct list_head *this, *next;
L
Linus Torvalds 已提交
716
	struct shmem_inode_info *info;
717
	struct mem_cgroup *memcg;
718 719 720 721
	int error = 0;

	/*
	 * There's a faint possibility that swap page was replaced before
722
	 * caller locked it: caller will come back later with the right page.
723
	 */
724
	if (unlikely(!PageSwapCache(page) || page_private(page) != swap.val))
725
		goto out;
726 727 728 729 730 731

	/*
	 * 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.
	 */
732
	error = mem_cgroup_try_charge(page, current->mm, GFP_KERNEL, &memcg);
733 734
	if (error)
		goto out;
735
	/* No radix_tree_preload: swap entry keeps a place for page in tree */
736
	error = -EAGAIN;
L
Linus Torvalds 已提交
737

738
	mutex_lock(&shmem_swaplist_mutex);
739 740
	list_for_each_safe(this, next, &shmem_swaplist) {
		info = list_entry(this, struct shmem_inode_info, swaplist);
741
		if (info->swapped)
742
			error = shmem_unuse_inode(info, swap, &page);
743 744
		else
			list_del_init(&info->swaplist);
745
		cond_resched();
746
		if (error != -EAGAIN)
747
			break;
748
		/* found nothing in this: move on to search the next */
L
Linus Torvalds 已提交
749
	}
750
	mutex_unlock(&shmem_swaplist_mutex);
751

752 753 754 755 756 757
	if (error) {
		if (error != -ENOMEM)
			error = 0;
		mem_cgroup_cancel_charge(page, memcg);
	} else
		mem_cgroup_commit_charge(page, memcg, true);
758
out:
H
Hugh Dickins 已提交
759 760
	unlock_page(page);
	page_cache_release(page);
761
	return error;
L
Linus Torvalds 已提交
762 763 764 765 766 767 768 769 770 771
}

/*
 * 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;
772 773
	swp_entry_t swap;
	pgoff_t index;
L
Linus Torvalds 已提交
774 775 776 777 778 779 780 781

	BUG_ON(!PageLocked(page));
	mapping = page->mapping;
	index = page->index;
	inode = mapping->host;
	info = SHMEM_I(inode);
	if (info->flags & VM_LOCKED)
		goto redirty;
782
	if (!total_swap_pages)
L
Linus Torvalds 已提交
783 784
		goto redirty;

785
	/*
C
Christoph Hellwig 已提交
786 787 788 789 790
	 * 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.
791
	 */
792 793 794 795
	if (!wbc->for_reclaim) {
		WARN_ON_ONCE(1);	/* Still happens? Tell us about it! */
		goto redirty;
	}
796 797 798 799 800

	/*
	 * 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.
801 802 803 804 805 806
	 *
	 * 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.
807 808
	 */
	if (!PageUptodate(page)) {
809 810 811 812 813
		if (inode->i_private) {
			struct shmem_falloc *shmem_falloc;
			spin_lock(&inode->i_lock);
			shmem_falloc = inode->i_private;
			if (shmem_falloc &&
814
			    !shmem_falloc->waitq &&
815 816 817 818 819 820 821 822 823
			    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;
		}
824 825 826 827 828
		clear_highpage(page);
		flush_dcache_page(page);
		SetPageUptodate(page);
	}

829 830 831
	swap = get_swap_page();
	if (!swap.val)
		goto redirty;
832

833 834
	/*
	 * Add inode to shmem_unuse()'s list of swapped-out inodes,
835 836
	 * if it's not already there.  Do it now before the page is
	 * moved to swap cache, when its pagelock no longer protects
837
	 * the inode from eviction.  But don't unlock the mutex until
838 839
	 * we've incremented swapped, because shmem_unuse_inode() will
	 * prune a !swapped inode from the swaplist under this mutex.
840
	 */
841 842 843
	mutex_lock(&shmem_swaplist_mutex);
	if (list_empty(&info->swaplist))
		list_add_tail(&info->swaplist, &shmem_swaplist);
844

845
	if (add_to_swap_cache(page, swap, GFP_ATOMIC) == 0) {
H
Hugh Dickins 已提交
846
		swap_shmem_alloc(swap);
847 848 849 850 851
		shmem_delete_from_page_cache(page, swp_to_radix_entry(swap));

		spin_lock(&info->lock);
		info->swapped++;
		shmem_recalc_inode(inode);
852
		spin_unlock(&info->lock);
853 854

		mutex_unlock(&shmem_swaplist_mutex);
855
		BUG_ON(page_mapped(page));
H
Hugh Dickins 已提交
856
		swap_writepage(page, wbc);
L
Linus Torvalds 已提交
857 858 859
		return 0;
	}

860
	mutex_unlock(&shmem_swaplist_mutex);
861
	swapcache_free(swap);
L
Linus Torvalds 已提交
862 863
redirty:
	set_page_dirty(page);
864 865 866 867
	if (wbc->for_reclaim)
		return AOP_WRITEPAGE_ACTIVATE;	/* Return with page locked */
	unlock_page(page);
	return 0;
L
Linus Torvalds 已提交
868 869 870
}

#ifdef CONFIG_NUMA
871
#ifdef CONFIG_TMPFS
872
static void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
873
{
874
	char buffer[64];
875

876
	if (!mpol || mpol->mode == MPOL_DEFAULT)
877
		return;		/* show nothing */
878

879
	mpol_to_str(buffer, sizeof(buffer), mpol);
880 881

	seq_printf(seq, ",mpol=%s", buffer);
882
}
883 884 885 886 887 888 889 890 891 892 893 894

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;
}
895 896
#endif /* CONFIG_TMPFS */

897 898
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
899 900
{
	struct vm_area_struct pvma;
M
Mel Gorman 已提交
901
	struct page *page;
902

L
Linus Torvalds 已提交
903
	/* Create a pseudo vma that just contains the policy */
904
	pvma.vm_start = 0;
905 906
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
907
	pvma.vm_ops = NULL;
M
Mel Gorman 已提交
908 909 910 911 912 913 914 915
	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
916 917
}

918
static struct page *shmem_alloc_page(gfp_t gfp,
919
			struct shmem_inode_info *info, pgoff_t index)
L
Linus Torvalds 已提交
920 921
{
	struct vm_area_struct pvma;
M
Mel Gorman 已提交
922
	struct page *page;
L
Linus Torvalds 已提交
923

924 925
	/* Create a pseudo vma that just contains the policy */
	pvma.vm_start = 0;
926 927
	/* Bias interleave by inode number to distribute better across nodes */
	pvma.vm_pgoff = index + info->vfs_inode.i_ino;
928
	pvma.vm_ops = NULL;
929
	pvma.vm_policy = mpol_shared_policy_lookup(&info->policy, index);
930

M
Mel Gorman 已提交
931 932 933 934 935 936
	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
937
}
938 939
#else /* !CONFIG_NUMA */
#ifdef CONFIG_TMPFS
940
static inline void shmem_show_mpol(struct seq_file *seq, struct mempolicy *mpol)
941 942 943 944
{
}
#endif /* CONFIG_TMPFS */

945 946
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
947
{
948
	return swapin_readahead(swap, gfp, NULL, 0);
L
Linus Torvalds 已提交
949 950
}

951
static inline struct page *shmem_alloc_page(gfp_t gfp,
952
			struct shmem_inode_info *info, pgoff_t index)
L
Linus Torvalds 已提交
953
{
954
	return alloc_page(gfp);
L
Linus Torvalds 已提交
955
}
956
#endif /* CONFIG_NUMA */
L
Linus Torvalds 已提交
957

958 959 960 961 962 963 964
#if !defined(CONFIG_NUMA) || !defined(CONFIG_TMPFS)
static inline struct mempolicy *shmem_get_sbmpol(struct shmem_sb_info *sbinfo)
{
	return NULL;
}
#endif

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 1003 1004
/*
 * 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);
1005
	flush_dcache_page(newpage);
1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019

	__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);
1020 1021 1022 1023
	if (!error) {
		__inc_zone_page_state(newpage, NR_FILE_PAGES);
		__dec_zone_page_state(oldpage, NR_FILE_PAGES);
	}
1024 1025
	spin_unlock_irq(&swap_mapping->tree_lock);

1026 1027 1028 1029 1030 1031 1032 1033
	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 {
1034
		mem_cgroup_replace_page(oldpage, newpage);
1035 1036 1037
		lru_cache_add_anon(newpage);
		*pagep = newpage;
	}
1038 1039 1040 1041 1042 1043 1044

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

	unlock_page(oldpage);
	page_cache_release(oldpage);
	page_cache_release(oldpage);
1045
	return error;
1046 1047
}

L
Linus Torvalds 已提交
1048
/*
1049
 * shmem_getpage_gfp - find page in cache, or get from swap, or allocate
L
Linus Torvalds 已提交
1050 1051 1052 1053 1054
 *
 * 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
 */
1055
static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
1056
	struct page **pagep, enum sgp_type sgp, gfp_t gfp, int *fault_type)
L
Linus Torvalds 已提交
1057 1058
{
	struct address_space *mapping = inode->i_mapping;
1059
	struct shmem_inode_info *info;
L
Linus Torvalds 已提交
1060
	struct shmem_sb_info *sbinfo;
1061
	struct mem_cgroup *memcg;
H
Hugh Dickins 已提交
1062
	struct page *page;
L
Linus Torvalds 已提交
1063 1064
	swp_entry_t swap;
	int error;
1065
	int once = 0;
1066
	int alloced = 0;
L
Linus Torvalds 已提交
1067

1068
	if (index > (MAX_LFS_FILESIZE >> PAGE_CACHE_SHIFT))
L
Linus Torvalds 已提交
1069 1070
		return -EFBIG;
repeat:
1071
	swap.val = 0;
1072
	page = find_lock_entry(mapping, index);
1073 1074 1075 1076 1077
	if (radix_tree_exceptional_entry(page)) {
		swap = radix_to_swp_entry(page);
		page = NULL;
	}

1078
	if (sgp != SGP_WRITE && sgp != SGP_FALLOC &&
1079 1080 1081 1082 1083
	    ((loff_t)index << PAGE_CACHE_SHIFT) >= i_size_read(inode)) {
		error = -EINVAL;
		goto failed;
	}

1084 1085 1086
	if (page && sgp == SGP_WRITE)
		mark_page_accessed(page);

1087 1088 1089 1090 1091 1092 1093 1094
	/* fallocated page? */
	if (page && !PageUptodate(page)) {
		if (sgp != SGP_READ)
			goto clear;
		unlock_page(page);
		page_cache_release(page);
		page = NULL;
	}
1095 1096 1097
	if (page || (sgp == SGP_READ && !swap.val)) {
		*pagep = page;
		return 0;
H
Hugh Dickins 已提交
1098 1099 1100
	}

	/*
1101 1102
	 * Fast cache lookup did not find it:
	 * bring it back from swap or allocate.
H
Hugh Dickins 已提交
1103
	 */
1104 1105
	info = SHMEM_I(inode);
	sbinfo = SHMEM_SB(inode->i_sb);
L
Linus Torvalds 已提交
1106 1107 1108

	if (swap.val) {
		/* Look it up and read it in.. */
H
Hugh Dickins 已提交
1109 1110
		page = lookup_swap_cache(swap);
		if (!page) {
L
Linus Torvalds 已提交
1111
			/* here we actually do the io */
1112 1113
			if (fault_type)
				*fault_type |= VM_FAULT_MAJOR;
1114
			page = shmem_swapin(swap, gfp, info, index);
H
Hugh Dickins 已提交
1115
			if (!page) {
1116 1117
				error = -ENOMEM;
				goto failed;
L
Linus Torvalds 已提交
1118 1119 1120 1121
			}
		}

		/* We have to do this with page locked to prevent races */
1122
		lock_page(page);
1123
		if (!PageSwapCache(page) || page_private(page) != swap.val ||
1124
		    !shmem_confirm_swap(mapping, index, swap)) {
1125
			error = -EEXIST;	/* try again */
1126
			goto unlock;
1127
		}
H
Hugh Dickins 已提交
1128
		if (!PageUptodate(page)) {
L
Linus Torvalds 已提交
1129
			error = -EIO;
1130
			goto failed;
L
Linus Torvalds 已提交
1131
		}
1132 1133
		wait_on_page_writeback(page);

1134 1135 1136 1137
		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
1138
		}
H
Hugh Dickins 已提交
1139

1140
		error = mem_cgroup_try_charge(page, current->mm, gfp, &memcg);
1141
		if (!error) {
1142
			error = shmem_add_to_page_cache(page, mapping, index,
1143
						swp_to_radix_entry(swap));
1144 1145 1146 1147 1148 1149 1150 1151
			/*
			 * 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
1152
			 * the rest.
1153 1154 1155
			 * Reset swap.val? No, leave it so "failed" goes back to
			 * "repeat": reading a hole and writing should succeed.
			 */
1156 1157
			if (error) {
				mem_cgroup_cancel_charge(page, memcg);
1158
				delete_from_swap_cache(page);
1159
			}
1160
		}
1161 1162 1163
		if (error)
			goto failed;

1164 1165
		mem_cgroup_commit_charge(page, memcg, true);

1166
		spin_lock(&info->lock);
1167
		info->swapped--;
1168
		shmem_recalc_inode(inode);
H
Hugh Dickins 已提交
1169
		spin_unlock(&info->lock);
1170

1171 1172 1173
		if (sgp == SGP_WRITE)
			mark_page_accessed(page);

1174
		delete_from_swap_cache(page);
H
Hugh Dickins 已提交
1175 1176 1177
		set_page_dirty(page);
		swap_free(swap);

1178 1179 1180 1181
	} else {
		if (shmem_acct_block(info->flags)) {
			error = -ENOSPC;
			goto failed;
L
Linus Torvalds 已提交
1182
		}
1183
		if (sbinfo->max_blocks) {
1184
			if (percpu_counter_compare(&sbinfo->used_blocks,
1185 1186 1187 1188
						sbinfo->max_blocks) >= 0) {
				error = -ENOSPC;
				goto unacct;
			}
1189
			percpu_counter_inc(&sbinfo->used_blocks);
1190
		}
L
Linus Torvalds 已提交
1191

1192 1193 1194 1195
		page = shmem_alloc_page(gfp, info, index);
		if (!page) {
			error = -ENOMEM;
			goto decused;
L
Linus Torvalds 已提交
1196 1197
		}

1198
		__SetPageSwapBacked(page);
1199
		__set_page_locked(page);
1200
		if (sgp == SGP_WRITE)
1201
			__SetPageReferenced(page);
1202

1203
		error = mem_cgroup_try_charge(page, current->mm, gfp, &memcg);
1204 1205
		if (error)
			goto decused;
1206
		error = radix_tree_maybe_preload(gfp & GFP_RECLAIM_MASK);
1207 1208
		if (!error) {
			error = shmem_add_to_page_cache(page, mapping, index,
1209
							NULL);
1210 1211 1212
			radix_tree_preload_end();
		}
		if (error) {
1213
			mem_cgroup_cancel_charge(page, memcg);
1214 1215
			goto decused;
		}
1216
		mem_cgroup_commit_charge(page, memcg, false);
1217 1218 1219
		lru_cache_add_anon(page);

		spin_lock(&info->lock);
L
Linus Torvalds 已提交
1220
		info->alloced++;
1221 1222
		inode->i_blocks += BLOCKS_PER_PAGE;
		shmem_recalc_inode(inode);
L
Linus Torvalds 已提交
1223
		spin_unlock(&info->lock);
1224
		alloced = true;
1225

1226
		/*
1227 1228 1229 1230 1231 1232 1233 1234 1235
		 * 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.
1236 1237 1238 1239 1240 1241
		 */
		if (sgp != SGP_WRITE) {
			clear_highpage(page);
			flush_dcache_page(page);
			SetPageUptodate(page);
		}
1242
		if (sgp == SGP_DIRTY)
H
Hugh Dickins 已提交
1243
			set_page_dirty(page);
L
Linus Torvalds 已提交
1244
	}
1245

1246
	/* Perhaps the file has been truncated since we checked */
1247
	if (sgp != SGP_WRITE && sgp != SGP_FALLOC &&
1248 1249
	    ((loff_t)index << PAGE_CACHE_SHIFT) >= i_size_read(inode)) {
		error = -EINVAL;
1250 1251 1252 1253
		if (alloced)
			goto trunc;
		else
			goto failed;
H
Hugh Dickins 已提交
1254
	}
1255 1256
	*pagep = page;
	return 0;
L
Linus Torvalds 已提交
1257

1258
	/*
1259
	 * Error recovery.
1260
	 */
1261
trunc:
1262
	info = SHMEM_I(inode);
1263 1264 1265 1266 1267
	ClearPageDirty(page);
	delete_from_page_cache(page);
	spin_lock(&info->lock);
	info->alloced--;
	inode->i_blocks -= BLOCKS_PER_PAGE;
1268
	spin_unlock(&info->lock);
1269
decused:
1270
	sbinfo = SHMEM_SB(inode->i_sb);
1271 1272 1273 1274 1275
	if (sbinfo->max_blocks)
		percpu_counter_add(&sbinfo->used_blocks, -1);
unacct:
	shmem_unacct_blocks(info->flags, 1);
failed:
1276 1277 1278 1279
	if (swap.val && error != -EINVAL &&
	    !shmem_confirm_swap(mapping, index, swap))
		error = -EEXIST;
unlock:
H
Hugh Dickins 已提交
1280
	if (page) {
1281
		unlock_page(page);
H
Hugh Dickins 已提交
1282
		page_cache_release(page);
1283 1284 1285 1286 1287 1288
	}
	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
1289
		goto repeat;
1290
	}
1291
	if (error == -EEXIST)	/* from above or from radix_tree_insert */
1292 1293
		goto repeat;
	return error;
L
Linus Torvalds 已提交
1294 1295
}

N
Nick Piggin 已提交
1296
static int shmem_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
L
Linus Torvalds 已提交
1297
{
A
Al Viro 已提交
1298
	struct inode *inode = file_inode(vma->vm_file);
L
Linus Torvalds 已提交
1299
	int error;
1300
	int ret = VM_FAULT_LOCKED;
L
Linus Torvalds 已提交
1301

1302 1303 1304 1305
	/*
	 * 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
1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317
	 * 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.
1318 1319 1320 1321 1322 1323
	 */
	if (unlikely(inode->i_private)) {
		struct shmem_falloc *shmem_falloc;

		spin_lock(&inode->i_lock);
		shmem_falloc = inode->i_private;
1324 1325 1326 1327 1328 1329 1330 1331
		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;
1332 1333
			if ((vmf->flags & FAULT_FLAG_ALLOW_RETRY) &&
			   !(vmf->flags & FAULT_FLAG_RETRY_NOWAIT)) {
1334
				/* It's polite to up mmap_sem if we can */
1335
				up_read(&vma->vm_mm->mmap_sem);
1336
				ret = VM_FAULT_RETRY;
1337
			}
1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355

			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;
1356
		}
1357
		spin_unlock(&inode->i_lock);
1358 1359
	}

1360
	error = shmem_getpage(inode, vmf->pgoff, &vmf->page, SGP_CACHE, &ret);
N
Nick Piggin 已提交
1361 1362
	if (error)
		return ((error == -ENOMEM) ? VM_FAULT_OOM : VM_FAULT_SIGBUS);
1363

1364 1365 1366 1367
	if (ret & VM_FAULT_MAJOR) {
		count_vm_event(PGMAJFAULT);
		mem_cgroup_count_vm_event(vma->vm_mm, PGMAJFAULT);
	}
1368
	return ret;
L
Linus Torvalds 已提交
1369 1370 1371
}

#ifdef CONFIG_NUMA
1372
static int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *mpol)
L
Linus Torvalds 已提交
1373
{
A
Al Viro 已提交
1374
	struct inode *inode = file_inode(vma->vm_file);
1375
	return mpol_set_shared_policy(&SHMEM_I(inode)->policy, vma, mpol);
L
Linus Torvalds 已提交
1376 1377
}

A
Adrian Bunk 已提交
1378 1379
static struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
					  unsigned long addr)
L
Linus Torvalds 已提交
1380
{
A
Al Viro 已提交
1381
	struct inode *inode = file_inode(vma->vm_file);
1382
	pgoff_t index;
L
Linus Torvalds 已提交
1383

1384 1385
	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
1386 1387 1388 1389 1390
}
#endif

int shmem_lock(struct file *file, int lock, struct user_struct *user)
{
A
Al Viro 已提交
1391
	struct inode *inode = file_inode(file);
L
Linus Torvalds 已提交
1392 1393 1394 1395 1396 1397 1398 1399
	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;
1400
		mapping_set_unevictable(file->f_mapping);
L
Linus Torvalds 已提交
1401 1402 1403 1404
	}
	if (!lock && (info->flags & VM_LOCKED) && user) {
		user_shm_unlock(inode->i_size, user);
		info->flags &= ~VM_LOCKED;
1405
		mapping_clear_unevictable(file->f_mapping);
L
Linus Torvalds 已提交
1406 1407
	}
	retval = 0;
1408

L
Linus Torvalds 已提交
1409 1410 1411 1412 1413
out_nomem:
	spin_unlock(&info->lock);
	return retval;
}

1414
static int shmem_mmap(struct file *file, struct vm_area_struct *vma)
L
Linus Torvalds 已提交
1415 1416 1417 1418 1419 1420
{
	file_accessed(file);
	vma->vm_ops = &shmem_vm_ops;
	return 0;
}

1421
static struct inode *shmem_get_inode(struct super_block *sb, const struct inode *dir,
1422
				     umode_t mode, dev_t dev, unsigned long flags)
L
Linus Torvalds 已提交
1423 1424 1425 1426 1427
{
	struct inode *inode;
	struct shmem_inode_info *info;
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);

1428 1429
	if (shmem_reserve_inode(sb))
		return NULL;
L
Linus Torvalds 已提交
1430 1431 1432

	inode = new_inode(sb);
	if (inode) {
1433
		inode->i_ino = get_next_ino();
1434
		inode_init_owner(inode, dir, mode);
L
Linus Torvalds 已提交
1435 1436
		inode->i_blocks = 0;
		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1437
		inode->i_generation = get_seconds();
L
Linus Torvalds 已提交
1438 1439 1440
		info = SHMEM_I(inode);
		memset(info, 0, (char *)inode - (char *)info);
		spin_lock_init(&info->lock);
D
David Herrmann 已提交
1441
		info->seals = F_SEAL_SEAL;
1442
		info->flags = flags & VM_NORESERVE;
L
Linus Torvalds 已提交
1443
		INIT_LIST_HEAD(&info->swaplist);
1444
		simple_xattrs_init(&info->xattrs);
1445
		cache_no_acl(inode);
L
Linus Torvalds 已提交
1446 1447 1448

		switch (mode & S_IFMT) {
		default:
1449
			inode->i_op = &shmem_special_inode_operations;
L
Linus Torvalds 已提交
1450 1451 1452
			init_special_inode(inode, mode, dev);
			break;
		case S_IFREG:
1453
			inode->i_mapping->a_ops = &shmem_aops;
L
Linus Torvalds 已提交
1454 1455
			inode->i_op = &shmem_inode_operations;
			inode->i_fop = &shmem_file_operations;
1456 1457
			mpol_shared_policy_init(&info->policy,
						 shmem_get_sbmpol(sbinfo));
L
Linus Torvalds 已提交
1458 1459
			break;
		case S_IFDIR:
1460
			inc_nlink(inode);
L
Linus Torvalds 已提交
1461 1462 1463 1464 1465 1466 1467 1468 1469 1470
			/* 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.
			 */
1471
			mpol_shared_policy_init(&info->policy, NULL);
L
Linus Torvalds 已提交
1472 1473
			break;
		}
1474 1475
	} else
		shmem_free_inode(sb);
L
Linus Torvalds 已提交
1476 1477 1478
	return inode;
}

1479 1480
bool shmem_mapping(struct address_space *mapping)
{
1481 1482 1483
	if (!mapping->host)
		return false;

C
Christoph Hellwig 已提交
1484
	return mapping->host->i_sb->s_op == &shmem_ops;
1485 1486
}

L
Linus Torvalds 已提交
1487
#ifdef CONFIG_TMPFS
1488
static const struct inode_operations shmem_symlink_inode_operations;
1489
static const struct inode_operations shmem_short_symlink_operations;
L
Linus Torvalds 已提交
1490

1491 1492 1493 1494 1495 1496
#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
1497
static int
N
Nick Piggin 已提交
1498 1499 1500
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
1501
{
N
Nick Piggin 已提交
1502
	struct inode *inode = mapping->host;
D
David Herrmann 已提交
1503
	struct shmem_inode_info *info = SHMEM_I(inode);
N
Nick Piggin 已提交
1504
	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
D
David Herrmann 已提交
1505 1506 1507 1508 1509 1510 1511 1512 1513

	/* 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;
	}

1514
	return shmem_getpage(inode, index, pagep, SGP_WRITE, NULL);
N
Nick Piggin 已提交
1515 1516 1517 1518 1519 1520 1521 1522 1523
}

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
1524 1525 1526
	if (pos + copied > inode->i_size)
		i_size_write(inode, pos + copied);

1527 1528 1529 1530 1531 1532 1533 1534
	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
1535
	set_page_dirty(page);
1536
	unlock_page(page);
N
Nick Piggin 已提交
1537 1538 1539
	page_cache_release(page);

	return copied;
L
Linus Torvalds 已提交
1540 1541
}

A
Al Viro 已提交
1542
static ssize_t shmem_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
L
Linus Torvalds 已提交
1543
{
1544 1545
	struct file *file = iocb->ki_filp;
	struct inode *inode = file_inode(file);
L
Linus Torvalds 已提交
1546
	struct address_space *mapping = inode->i_mapping;
1547 1548
	pgoff_t index;
	unsigned long offset;
1549
	enum sgp_type sgp = SGP_READ;
1550
	int error = 0;
A
Al Viro 已提交
1551
	ssize_t retval = 0;
1552
	loff_t *ppos = &iocb->ki_pos;
1553 1554 1555 1556 1557 1558

	/*
	 * 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
1559
	if (!iter_is_iovec(to))
1560
		sgp = SGP_DIRTY;
L
Linus Torvalds 已提交
1561 1562 1563 1564 1565 1566

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

	for (;;) {
		struct page *page = NULL;
1567 1568
		pgoff_t end_index;
		unsigned long nr, ret;
L
Linus Torvalds 已提交
1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579
		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;
		}

1580 1581 1582 1583
		error = shmem_getpage(inode, index, &page, sgp, NULL);
		if (error) {
			if (error == -EINVAL)
				error = 0;
L
Linus Torvalds 已提交
1584 1585
			break;
		}
H
Hugh Dickins 已提交
1586 1587
		if (page)
			unlock_page(page);
L
Linus Torvalds 已提交
1588 1589 1590

		/*
		 * We must evaluate after, since reads (unlike writes)
1591
		 * are called without i_mutex protection against truncate
L
Linus Torvalds 已提交
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 1617 1618
		 */
		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
1619
		} else {
L
Linus Torvalds 已提交
1620
			page = ZERO_PAGE(0);
N
Nick Piggin 已提交
1621 1622
			page_cache_get(page);
		}
L
Linus Torvalds 已提交
1623 1624 1625 1626 1627

		/*
		 * 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
1628
		ret = copy_page_to_iter(page, offset, nr, to);
1629
		retval += ret;
L
Linus Torvalds 已提交
1630 1631 1632 1633 1634
		offset += ret;
		index += offset >> PAGE_CACHE_SHIFT;
		offset &= ~PAGE_CACHE_MASK;

		page_cache_release(page);
A
Al Viro 已提交
1635
		if (!iov_iter_count(to))
L
Linus Torvalds 已提交
1636
			break;
1637 1638 1639 1640
		if (ret < nr) {
			error = -EFAULT;
			break;
		}
L
Linus Torvalds 已提交
1641 1642 1643 1644
		cond_resched();
	}

	*ppos = ((loff_t) index << PAGE_CACHE_SHIFT) + offset;
1645 1646
	file_accessed(file);
	return retval ? retval : error;
L
Linus Torvalds 已提交
1647 1648
}

1649 1650 1651 1652 1653
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;
1654
	struct inode *inode = mapping->host;
1655 1656 1657 1658 1659 1660 1661 1662 1663 1664
	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,
1665
		.nr_pages_max = PIPE_DEF_BUFFERS,
1666 1667 1668 1669 1670
		.flags = flags,
		.ops = &page_cache_pipe_buf_ops,
		.spd_release = spd_release_page,
	};

1671
	isize = i_size_read(inode);
1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684
	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;
1685
	nr_pages = min(req_pages, spd.nr_pages_max);
1686 1687 1688 1689 1690 1691

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

1692 1693 1694 1695 1696
	while (spd.nr_pages < nr_pages) {
		error = shmem_getpage(inode, index, &page, SGP_CACHE, NULL);
		if (error)
			break;
		unlock_page(page);
1697 1698 1699 1700 1701 1702 1703
		spd.pages[spd.nr_pages++] = page;
		index++;
	}

	index = *ppos >> PAGE_CACHE_SHIFT;
	nr_pages = spd.nr_pages;
	spd.nr_pages = 0;
1704

1705 1706 1707 1708 1709 1710 1711 1712 1713
	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];

1714 1715 1716 1717
		if (!PageUptodate(page) || page->mapping != mapping) {
			error = shmem_getpage(inode, index, &page,
							SGP_CACHE, NULL);
			if (error)
1718
				break;
1719 1720 1721
			unlock_page(page);
			page_cache_release(spd.pages[page_nr]);
			spd.pages[page_nr] = page;
1722
		}
1723 1724

		isize = i_size_read(inode);
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 1752 1753
		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);

1754
	splice_shrink_spd(&spd);
1755 1756 1757 1758 1759 1760 1761 1762

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

1763 1764 1765 1766
/*
 * llseek SEEK_DATA or SEEK_HOLE through the radix_tree.
 */
static pgoff_t shmem_seek_hole_data(struct address_space *mapping,
1767
				    pgoff_t index, pgoff_t end, int whence)
1768 1769 1770 1771 1772 1773 1774 1775 1776 1777
{
	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) {
1778
		pvec.nr = find_get_entries(mapping, index,
1779 1780
					pvec.nr, pvec.pages, indices);
		if (!pvec.nr) {
1781
			if (whence == SEEK_DATA)
1782 1783 1784 1785 1786
				index = end;
			break;
		}
		for (i = 0; i < pvec.nr; i++, index++) {
			if (index < indices[i]) {
1787
				if (whence == SEEK_HOLE) {
1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798
					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 ||
1799 1800
			    (page && whence == SEEK_DATA) ||
			    (!page && whence == SEEK_HOLE)) {
1801 1802 1803 1804
				done = true;
				break;
			}
		}
1805
		pagevec_remove_exceptionals(&pvec);
1806 1807 1808 1809 1810 1811 1812
		pagevec_release(&pvec);
		pvec.nr = PAGEVEC_SIZE;
		cond_resched();
	}
	return index;
}

1813
static loff_t shmem_file_llseek(struct file *file, loff_t offset, int whence)
1814 1815 1816 1817 1818 1819
{
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
	pgoff_t start, end;
	loff_t new_offset;

1820 1821
	if (whence != SEEK_DATA && whence != SEEK_HOLE)
		return generic_file_llseek_size(file, offset, whence,
1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832
					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;
1833
		new_offset = shmem_seek_hole_data(mapping, start, end, whence);
1834 1835 1836 1837
		new_offset <<= PAGE_CACHE_SHIFT;
		if (new_offset > offset) {
			if (new_offset < inode->i_size)
				offset = new_offset;
1838
			else if (whence == SEEK_DATA)
1839 1840 1841 1842 1843 1844
				offset = -ENXIO;
			else
				offset = inode->i_size;
		}
	}

1845 1846
	if (offset >= 0)
		offset = vfs_setpos(file, offset, MAX_LFS_FILESIZE);
1847 1848 1849 1850
	mutex_unlock(&inode->i_mutex);
	return offset;
}

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 1898 1899
/*
 * 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
1900 1901
static int shmem_wait_for_pins(struct address_space *mapping)
{
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 1960 1961
	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
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 2070 2071
}

#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;
}

2072 2073 2074
static long shmem_fallocate(struct file *file, int mode, loff_t offset,
							 loff_t len)
{
A
Al Viro 已提交
2075
	struct inode *inode = file_inode(file);
2076
	struct shmem_sb_info *sbinfo = SHMEM_SB(inode->i_sb);
D
David Herrmann 已提交
2077
	struct shmem_inode_info *info = SHMEM_I(inode);
2078
	struct shmem_falloc shmem_falloc;
2079 2080
	pgoff_t start, index, end;
	int error;
2081

2082 2083 2084
	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
		return -EOPNOTSUPP;

2085 2086 2087 2088 2089 2090
	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;
2091
		DECLARE_WAIT_QUEUE_HEAD_ONSTACK(shmem_falloc_waitq);
2092

D
David Herrmann 已提交
2093 2094 2095 2096 2097 2098
		/* protected by i_mutex */
		if (info->seals & F_SEAL_WRITE) {
			error = -EPERM;
			goto out;
		}

2099
		shmem_falloc.waitq = &shmem_falloc_waitq;
2100 2101 2102 2103 2104 2105
		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);

2106 2107 2108 2109 2110
		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 */
2111 2112 2113 2114 2115

		spin_lock(&inode->i_lock);
		inode->i_private = NULL;
		wake_up_all(&shmem_falloc_waitq);
		spin_unlock(&inode->i_lock);
2116
		error = 0;
2117
		goto out;
2118 2119 2120 2121 2122 2123 2124
	}

	/* 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
2125 2126 2127 2128 2129
	if ((info->seals & F_SEAL_GROW) && offset + len > inode->i_size) {
		error = -EPERM;
		goto out;
	}

2130 2131 2132 2133 2134 2135
	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;
2136 2137
	}

2138
	shmem_falloc.waitq = NULL;
2139 2140 2141 2142 2143 2144 2145 2146
	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);

2147 2148 2149 2150 2151 2152 2153 2154 2155
	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;
2156 2157
		else if (shmem_falloc.nr_unswapped > shmem_falloc.nr_falloced)
			error = -ENOMEM;
2158
		else
2159
			error = shmem_getpage(inode, index, &page, SGP_FALLOC,
2160 2161
									NULL);
		if (error) {
2162 2163 2164 2165
			/* Remove the !PageUptodate pages we added */
			shmem_undo_range(inode,
				(loff_t)start << PAGE_CACHE_SHIFT,
				(loff_t)index << PAGE_CACHE_SHIFT, true);
2166
			goto undone;
2167 2168
		}

2169 2170 2171 2172 2173 2174 2175 2176
		/*
		 * 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++;

2177
		/*
2178 2179 2180
		 * 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
2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192
		 * 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;
2193 2194 2195 2196
undone:
	spin_lock(&inode->i_lock);
	inode->i_private = NULL;
	spin_unlock(&inode->i_lock);
2197
out:
2198 2199 2200 2201
	mutex_unlock(&inode->i_mutex);
	return error;
}

2202
static int shmem_statfs(struct dentry *dentry, struct kstatfs *buf)
L
Linus Torvalds 已提交
2203
{
2204
	struct shmem_sb_info *sbinfo = SHMEM_SB(dentry->d_sb);
L
Linus Torvalds 已提交
2205 2206 2207 2208

	buf->f_type = TMPFS_MAGIC;
	buf->f_bsize = PAGE_CACHE_SIZE;
	buf->f_namelen = NAME_MAX;
2209
	if (sbinfo->max_blocks) {
L
Linus Torvalds 已提交
2210
		buf->f_blocks = sbinfo->max_blocks;
2211 2212 2213
		buf->f_bavail =
		buf->f_bfree  = sbinfo->max_blocks -
				percpu_counter_sum(&sbinfo->used_blocks);
2214 2215
	}
	if (sbinfo->max_inodes) {
L
Linus Torvalds 已提交
2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226
		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
2227
shmem_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev)
L
Linus Torvalds 已提交
2228
{
2229
	struct inode *inode;
L
Linus Torvalds 已提交
2230 2231
	int error = -ENOSPC;

2232
	inode = shmem_get_inode(dir->i_sb, dir, mode, dev, VM_NORESERVE);
L
Linus Torvalds 已提交
2233
	if (inode) {
C
Christoph Hellwig 已提交
2234 2235 2236
		error = simple_acl_create(dir, inode);
		if (error)
			goto out_iput;
2237
		error = security_inode_init_security(inode, dir,
2238
						     &dentry->d_name,
2239
						     shmem_initxattrs, NULL);
C
Christoph Hellwig 已提交
2240 2241
		if (error && error != -EOPNOTSUPP)
			goto out_iput;
2242

A
Al Viro 已提交
2243
		error = 0;
L
Linus Torvalds 已提交
2244 2245 2246 2247 2248 2249
		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
2250 2251 2252
out_iput:
	iput(inode);
	return error;
L
Linus Torvalds 已提交
2253 2254
}

2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265
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
2266 2267 2268 2269 2270
		if (error && error != -EOPNOTSUPP)
			goto out_iput;
		error = simple_acl_create(dir, inode);
		if (error)
			goto out_iput;
2271 2272 2273
		d_tmpfile(dentry, inode);
	}
	return error;
C
Christoph Hellwig 已提交
2274 2275 2276
out_iput:
	iput(inode);
	return error;
2277 2278
}

2279
static int shmem_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
2280 2281 2282 2283 2284
{
	int error;

	if ((error = shmem_mknod(dir, dentry, mode | S_IFDIR, 0)))
		return error;
2285
	inc_nlink(dir);
L
Linus Torvalds 已提交
2286 2287 2288
	return 0;
}

A
Al Viro 已提交
2289
static int shmem_create(struct inode *dir, struct dentry *dentry, umode_t mode,
A
Al Viro 已提交
2290
		bool excl)
L
Linus Torvalds 已提交
2291 2292 2293 2294 2295 2296 2297 2298 2299
{
	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)
{
2300
	struct inode *inode = d_inode(old_dentry);
2301
	int ret;
L
Linus Torvalds 已提交
2302 2303 2304 2305 2306 2307

	/*
	 * 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.
	 */
2308 2309 2310
	ret = shmem_reserve_inode(inode->i_sb);
	if (ret)
		goto out;
L
Linus Torvalds 已提交
2311 2312 2313

	dir->i_size += BOGO_DIRENT_SIZE;
	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
2314
	inc_nlink(inode);
A
Al Viro 已提交
2315
	ihold(inode);	/* New dentry reference */
L
Linus Torvalds 已提交
2316 2317
	dget(dentry);		/* Extra pinning count for the created dentry */
	d_instantiate(dentry, inode);
2318 2319
out:
	return ret;
L
Linus Torvalds 已提交
2320 2321 2322 2323
}

static int shmem_unlink(struct inode *dir, struct dentry *dentry)
{
2324
	struct inode *inode = d_inode(dentry);
L
Linus Torvalds 已提交
2325

2326 2327
	if (inode->i_nlink > 1 && !S_ISDIR(inode->i_mode))
		shmem_free_inode(inode->i_sb);
L
Linus Torvalds 已提交
2328 2329 2330

	dir->i_size -= BOGO_DIRENT_SIZE;
	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
2331
	drop_nlink(inode);
L
Linus Torvalds 已提交
2332 2333 2334 2335 2336 2337 2338 2339 2340
	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;

2341
	drop_nlink(d_inode(dentry));
2342
	drop_nlink(dir);
L
Linus Torvalds 已提交
2343 2344 2345
	return shmem_unlink(dir, dentry);
}

M
Miklos Szeredi 已提交
2346 2347
static int shmem_exchange(struct inode *old_dir, struct dentry *old_dentry, struct inode *new_dir, struct dentry *new_dentry)
{
2348 2349
	bool old_is_dir = d_is_dir(old_dentry);
	bool new_is_dir = d_is_dir(new_dentry);
M
Miklos Szeredi 已提交
2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361

	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 =
2362 2363
	d_inode(old_dentry)->i_ctime =
	d_inode(new_dentry)->i_ctime = CURRENT_TIME;
M
Miklos Szeredi 已提交
2364 2365 2366 2367

	return 0;
}

M
Miklos Szeredi 已提交
2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393
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
2394 2395 2396 2397 2398 2399
/*
 * 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
2400
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
2401
{
2402
	struct inode *inode = d_inode(old_dentry);
L
Linus Torvalds 已提交
2403 2404
	int they_are_dirs = S_ISDIR(inode->i_mode);

M
Miklos Szeredi 已提交
2405
	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT))
M
Miklos Szeredi 已提交
2406 2407
		return -EINVAL;

M
Miklos Szeredi 已提交
2408 2409 2410
	if (flags & RENAME_EXCHANGE)
		return shmem_exchange(old_dir, old_dentry, new_dir, new_dentry);

L
Linus Torvalds 已提交
2411 2412 2413
	if (!simple_empty(new_dentry))
		return -ENOTEMPTY;

M
Miklos Szeredi 已提交
2414 2415 2416 2417 2418 2419 2420 2421
	if (flags & RENAME_WHITEOUT) {
		int error;

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

2422
	if (d_really_is_positive(new_dentry)) {
L
Linus Torvalds 已提交
2423
		(void) shmem_unlink(new_dir, new_dentry);
2424
		if (they_are_dirs) {
2425
			drop_nlink(d_inode(new_dentry));
2426
			drop_nlink(old_dir);
2427
		}
L
Linus Torvalds 已提交
2428
	} else if (they_are_dirs) {
2429
		drop_nlink(old_dir);
2430
		inc_nlink(new_dir);
L
Linus Torvalds 已提交
2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445
	}

	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
2446
	struct page *page;
L
Linus Torvalds 已提交
2447 2448 2449 2450 2451 2452 2453
	char *kaddr;
	struct shmem_inode_info *info;

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

2454
	inode = shmem_get_inode(dir->i_sb, dir, S_IFLNK|S_IRWXUGO, 0, VM_NORESERVE);
L
Linus Torvalds 已提交
2455 2456 2457
	if (!inode)
		return -ENOSPC;

2458
	error = security_inode_init_security(inode, dir, &dentry->d_name,
2459
					     shmem_initxattrs, NULL);
2460 2461 2462 2463 2464 2465 2466 2467
	if (error) {
		if (error != -EOPNOTSUPP) {
			iput(inode);
			return error;
		}
		error = 0;
	}

L
Linus Torvalds 已提交
2468 2469
	info = SHMEM_I(inode);
	inode->i_size = len-1;
2470 2471 2472 2473 2474 2475 2476
	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
2477
		inode->i_link = info->symlink;
L
Linus Torvalds 已提交
2478 2479 2480 2481 2482 2483
	} else {
		error = shmem_getpage(inode, 0, &page, SGP_WRITE, NULL);
		if (error) {
			iput(inode);
			return error;
		}
2484
		inode->i_mapping->a_ops = &shmem_aops;
L
Linus Torvalds 已提交
2485
		inode->i_op = &shmem_symlink_inode_operations;
2486
		kaddr = kmap_atomic(page);
L
Linus Torvalds 已提交
2487
		memcpy(kaddr, symname, len);
2488
		kunmap_atomic(kaddr);
2489
		SetPageUptodate(page);
L
Linus Torvalds 已提交
2490
		set_page_dirty(page);
2491
		unlock_page(page);
L
Linus Torvalds 已提交
2492 2493 2494 2495 2496 2497 2498 2499 2500
		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;
}

2501
static const char *shmem_follow_link(struct dentry *dentry, void **cookie)
L
Linus Torvalds 已提交
2502 2503
{
	struct page *page = NULL;
2504
	int error = shmem_getpage(d_inode(dentry), 0, &page, SGP_READ, NULL);
2505 2506 2507 2508 2509
	if (error)
		return ERR_PTR(error);
	unlock_page(page);
	*cookie = page;
	return kmap(page);
L
Linus Torvalds 已提交
2510 2511
}

2512
static void shmem_put_link(struct inode *unused, void *cookie)
L
Linus Torvalds 已提交
2513
{
2514 2515 2516 2517
	struct page *page = cookie;
	kunmap(page);
	mark_page_accessed(page);
	page_cache_release(page);
L
Linus Torvalds 已提交
2518 2519
}

2520
#ifdef CONFIG_TMPFS_XATTR
2521
/*
2522 2523
 * Superblocks without xattr inode operations may get some security.* xattr
 * support from the LSM "for free". As soon as we have any other xattrs
2524 2525 2526 2527
 * like ACLs, we also need to implement the security.* handlers at
 * filesystem level, though.
 */

2528 2529 2530 2531 2532 2533 2534 2535 2536
/*
 * 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;
2537
	struct simple_xattr *new_xattr;
2538 2539 2540
	size_t len;

	for (xattr = xattr_array; xattr->name != NULL; xattr++) {
2541
		new_xattr = simple_xattr_alloc(xattr->value, xattr->value_len);
2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557
		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);

2558
		simple_xattr_list_add(&info->xattrs, new_xattr);
2559 2560 2561 2562 2563
	}

	return 0;
}

2564
static const struct xattr_handler *shmem_xattr_handlers[] = {
2565
#ifdef CONFIG_TMPFS_POSIX_ACL
C
Christoph Hellwig 已提交
2566 2567
	&posix_acl_access_xattr_handler,
	&posix_acl_default_xattr_handler,
2568
#endif
2569 2570
	NULL
};
2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593

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)
{
2594
	struct shmem_inode_info *info = SHMEM_I(d_inode(dentry));
2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608
	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;

2609
	return simple_xattr_get(&info->xattrs, name, buffer, size);
2610 2611 2612 2613 2614
}

static int shmem_setxattr(struct dentry *dentry, const char *name,
			  const void *value, size_t size, int flags)
{
2615
	struct shmem_inode_info *info = SHMEM_I(d_inode(dentry));
2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629
	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;

2630
	return simple_xattr_set(&info->xattrs, name, value, size, flags);
2631 2632 2633 2634
}

static int shmem_removexattr(struct dentry *dentry, const char *name)
{
2635
	struct shmem_inode_info *info = SHMEM_I(d_inode(dentry));
2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649
	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;

2650
	return simple_xattr_remove(&info->xattrs, name);
2651 2652 2653 2654
}

static ssize_t shmem_listxattr(struct dentry *dentry, char *buffer, size_t size)
{
2655
	struct shmem_inode_info *info = SHMEM_I(d_inode(dentry));
2656
	return simple_xattr_list(&info->xattrs, buffer, size);
2657 2658 2659
}
#endif /* CONFIG_TMPFS_XATTR */

2660
static const struct inode_operations shmem_short_symlink_operations = {
2661
	.readlink	= generic_readlink,
A
Al Viro 已提交
2662
	.follow_link	= simple_follow_link,
2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679
#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,
2680
#endif
2681
};
2682

2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695
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
2696 2697
static struct dentry *shmem_fh_to_dentry(struct super_block *sb,
		struct fid *fid, int fh_len, int fh_type)
2698 2699
{
	struct inode *inode;
C
Christoph Hellwig 已提交
2700
	struct dentry *dentry = NULL;
2701
	u64 inum;
C
Christoph Hellwig 已提交
2702 2703 2704

	if (fh_len < 3)
		return NULL;
2705

2706 2707 2708
	inum = fid->raw[2];
	inum = (inum << 32) | fid->raw[1];

C
Christoph Hellwig 已提交
2709 2710
	inode = ilookup5(sb, (unsigned long)(inum + fid->raw[0]),
			shmem_match, fid->raw);
2711
	if (inode) {
C
Christoph Hellwig 已提交
2712
		dentry = d_find_alias(inode);
2713 2714 2715
		iput(inode);
	}

C
Christoph Hellwig 已提交
2716
	return dentry;
2717 2718
}

A
Al Viro 已提交
2719 2720
static int shmem_encode_fh(struct inode *inode, __u32 *fh, int *len,
				struct inode *parent)
2721
{
2722 2723
	if (*len < 3) {
		*len = 3;
2724
		return FILEID_INVALID;
2725
	}
2726

A
Al Viro 已提交
2727
	if (inode_unhashed(inode)) {
2728 2729 2730 2731 2732 2733 2734
		/* 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
2735
		if (inode_unhashed(inode))
2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748
			__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;
}

2749
static const struct export_operations shmem_export_ops = {
2750 2751
	.get_parent     = shmem_get_parent,
	.encode_fh      = shmem_encode_fh,
C
Christoph Hellwig 已提交
2752
	.fh_to_dentry	= shmem_fh_to_dentry,
2753 2754
};

2755 2756
static int shmem_parse_options(char *options, struct shmem_sb_info *sbinfo,
			       bool remount)
L
Linus Torvalds 已提交
2757 2758
{
	char *this_char, *value, *rest;
G
Greg Thelen 已提交
2759
	struct mempolicy *mpol = NULL;
2760 2761
	uid_t uid;
	gid_t gid;
L
Linus Torvalds 已提交
2762

2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779
	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
2780 2781 2782 2783 2784 2785 2786 2787
		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
2788
			goto error;
L
Linus Torvalds 已提交
2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801
		}

		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;
2802 2803
			sbinfo->max_blocks =
				DIV_ROUND_UP(size, PAGE_CACHE_SIZE);
L
Linus Torvalds 已提交
2804
		} else if (!strcmp(this_char,"nr_blocks")) {
2805
			sbinfo->max_blocks = memparse(value, &rest);
L
Linus Torvalds 已提交
2806 2807 2808
			if (*rest)
				goto bad_val;
		} else if (!strcmp(this_char,"nr_inodes")) {
2809
			sbinfo->max_inodes = memparse(value, &rest);
L
Linus Torvalds 已提交
2810 2811 2812
			if (*rest)
				goto bad_val;
		} else if (!strcmp(this_char,"mode")) {
2813
			if (remount)
L
Linus Torvalds 已提交
2814
				continue;
2815
			sbinfo->mode = simple_strtoul(value, &rest, 8) & 07777;
L
Linus Torvalds 已提交
2816 2817 2818
			if (*rest)
				goto bad_val;
		} else if (!strcmp(this_char,"uid")) {
2819
			if (remount)
L
Linus Torvalds 已提交
2820
				continue;
2821
			uid = simple_strtoul(value, &rest, 0);
L
Linus Torvalds 已提交
2822 2823
			if (*rest)
				goto bad_val;
2824 2825 2826
			sbinfo->uid = make_kuid(current_user_ns(), uid);
			if (!uid_valid(sbinfo->uid))
				goto bad_val;
L
Linus Torvalds 已提交
2827
		} else if (!strcmp(this_char,"gid")) {
2828
			if (remount)
L
Linus Torvalds 已提交
2829
				continue;
2830
			gid = simple_strtoul(value, &rest, 0);
L
Linus Torvalds 已提交
2831 2832
			if (*rest)
				goto bad_val;
2833 2834 2835
			sbinfo->gid = make_kgid(current_user_ns(), gid);
			if (!gid_valid(sbinfo->gid))
				goto bad_val;
2836
		} else if (!strcmp(this_char,"mpol")) {
G
Greg Thelen 已提交
2837 2838 2839
			mpol_put(mpol);
			mpol = NULL;
			if (mpol_parse_str(value, &mpol))
2840
				goto bad_val;
L
Linus Torvalds 已提交
2841 2842 2843
		} else {
			printk(KERN_ERR "tmpfs: Bad mount option %s\n",
			       this_char);
G
Greg Thelen 已提交
2844
			goto error;
L
Linus Torvalds 已提交
2845 2846
		}
	}
G
Greg Thelen 已提交
2847
	sbinfo->mpol = mpol;
L
Linus Torvalds 已提交
2848 2849 2850 2851 2852
	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
2853 2854
error:
	mpol_put(mpol);
L
Linus Torvalds 已提交
2855 2856 2857 2858 2859 2860 2861
	return 1;

}

static int shmem_remount_fs(struct super_block *sb, int *flags, char *data)
{
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);
2862
	struct shmem_sb_info config = *sbinfo;
2863 2864 2865
	unsigned long inodes;
	int error = -EINVAL;

2866
	config.mpol = NULL;
2867
	if (shmem_parse_options(data, &config, true))
2868
		return error;
L
Linus Torvalds 已提交
2869

2870 2871
	spin_lock(&sbinfo->stat_lock);
	inodes = sbinfo->max_inodes - sbinfo->free_inodes;
2872
	if (percpu_counter_compare(&sbinfo->used_blocks, config.max_blocks) > 0)
2873
		goto out;
2874
	if (config.max_inodes < inodes)
2875 2876
		goto out;
	/*
2877
	 * Those tests disallow limited->unlimited while any are in use;
2878 2879 2880
	 * but we must separately disallow unlimited->limited, because
	 * in that case we have no record of how much is already in use.
	 */
2881
	if (config.max_blocks && !sbinfo->max_blocks)
2882
		goto out;
2883
	if (config.max_inodes && !sbinfo->max_inodes)
2884 2885 2886
		goto out;

	error = 0;
2887 2888 2889
	sbinfo->max_blocks  = config.max_blocks;
	sbinfo->max_inodes  = config.max_inodes;
	sbinfo->free_inodes = config.max_inodes - inodes;
2890

2891 2892 2893 2894 2895 2896 2897
	/*
	 * Preserve previous mempolicy unless mpol remount option was specified.
	 */
	if (config.mpol) {
		mpol_put(sbinfo->mpol);
		sbinfo->mpol = config.mpol;	/* transfers initial ref */
	}
2898 2899 2900
out:
	spin_unlock(&sbinfo->stat_lock);
	return error;
L
Linus Torvalds 已提交
2901
}
2902

2903
static int shmem_show_options(struct seq_file *seq, struct dentry *root)
2904
{
2905
	struct shmem_sb_info *sbinfo = SHMEM_SB(root->d_sb);
2906 2907 2908 2909 2910 2911 2912

	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))
2913
		seq_printf(seq, ",mode=%03ho", sbinfo->mode);
2914 2915 2916 2917 2918 2919
	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));
2920
	shmem_show_mpol(seq, sbinfo->mpol);
2921 2922
	return 0;
}
D
David Herrmann 已提交
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 2992 2993

#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;
}

2994
#endif /* CONFIG_TMPFS */
L
Linus Torvalds 已提交
2995 2996 2997

static void shmem_put_super(struct super_block *sb)
{
2998 2999 3000
	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);

	percpu_counter_destroy(&sbinfo->used_blocks);
G
Greg Thelen 已提交
3001
	mpol_put(sbinfo->mpol);
3002
	kfree(sbinfo);
L
Linus Torvalds 已提交
3003 3004 3005
	sb->s_fs_info = NULL;
}

3006
int shmem_fill_super(struct super_block *sb, void *data, int silent)
L
Linus Torvalds 已提交
3007 3008
{
	struct inode *inode;
3009
	struct shmem_sb_info *sbinfo;
3010 3011 3012
	int err = -ENOMEM;

	/* Round up to L1_CACHE_BYTES to resist false sharing */
3013
	sbinfo = kzalloc(max((int)sizeof(struct shmem_sb_info),
3014 3015 3016 3017 3018
				L1_CACHE_BYTES), GFP_KERNEL);
	if (!sbinfo)
		return -ENOMEM;

	sbinfo->mode = S_IRWXUGO | S_ISVTX;
3019 3020
	sbinfo->uid = current_fsuid();
	sbinfo->gid = current_fsgid();
3021
	sb->s_fs_info = sbinfo;
L
Linus Torvalds 已提交
3022

3023
#ifdef CONFIG_TMPFS
L
Linus Torvalds 已提交
3024 3025 3026 3027 3028
	/*
	 * 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.
	 */
3029
	if (!(sb->s_flags & MS_KERNMOUNT)) {
3030 3031 3032 3033 3034 3035
		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;
		}
3036 3037
	} else {
		sb->s_flags |= MS_NOUSER;
L
Linus Torvalds 已提交
3038
	}
3039
	sb->s_export_op = &shmem_export_ops;
H
Hugh Dickins 已提交
3040
	sb->s_flags |= MS_NOSEC;
L
Linus Torvalds 已提交
3041 3042 3043 3044
#else
	sb->s_flags |= MS_NOUSER;
#endif

3045
	spin_lock_init(&sbinfo->stat_lock);
3046
	if (percpu_counter_init(&sbinfo->used_blocks, 0, GFP_KERNEL))
3047
		goto failed;
3048
	sbinfo->free_inodes = sbinfo->max_inodes;
3049

3050
	sb->s_maxbytes = MAX_LFS_FILESIZE;
L
Linus Torvalds 已提交
3051 3052 3053 3054
	sb->s_blocksize = PAGE_CACHE_SIZE;
	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
	sb->s_magic = TMPFS_MAGIC;
	sb->s_op = &shmem_ops;
3055
	sb->s_time_gran = 1;
3056
#ifdef CONFIG_TMPFS_XATTR
3057
	sb->s_xattr = shmem_xattr_handlers;
3058 3059
#endif
#ifdef CONFIG_TMPFS_POSIX_ACL
3060 3061
	sb->s_flags |= MS_POSIXACL;
#endif
3062

3063
	inode = shmem_get_inode(sb, NULL, S_IFDIR | sbinfo->mode, 0, VM_NORESERVE);
L
Linus Torvalds 已提交
3064 3065
	if (!inode)
		goto failed;
3066 3067
	inode->i_uid = sbinfo->uid;
	inode->i_gid = sbinfo->gid;
A
Al Viro 已提交
3068 3069
	sb->s_root = d_make_root(inode);
	if (!sb->s_root)
3070
		goto failed;
L
Linus Torvalds 已提交
3071 3072 3073 3074 3075 3076 3077
	return 0;

failed:
	shmem_put_super(sb);
	return err;
}

3078
static struct kmem_cache *shmem_inode_cachep;
L
Linus Torvalds 已提交
3079 3080 3081

static struct inode *shmem_alloc_inode(struct super_block *sb)
{
3082 3083 3084
	struct shmem_inode_info *info;
	info = kmem_cache_alloc(shmem_inode_cachep, GFP_KERNEL);
	if (!info)
L
Linus Torvalds 已提交
3085
		return NULL;
3086
	return &info->vfs_inode;
L
Linus Torvalds 已提交
3087 3088
}

3089
static void shmem_destroy_callback(struct rcu_head *head)
N
Nick Piggin 已提交
3090 3091 3092 3093 3094
{
	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
3095 3096
static void shmem_destroy_inode(struct inode *inode)
{
3097
	if (S_ISREG(inode->i_mode))
L
Linus Torvalds 已提交
3098
		mpol_free_shared_policy(&SHMEM_I(inode)->policy);
3099
	call_rcu(&inode->i_rcu, shmem_destroy_callback);
L
Linus Torvalds 已提交
3100 3101
}

3102
static void shmem_init_inode(void *foo)
L
Linus Torvalds 已提交
3103
{
3104 3105
	struct shmem_inode_info *info = foo;
	inode_init_once(&info->vfs_inode);
L
Linus Torvalds 已提交
3106 3107
}

3108
static int shmem_init_inodecache(void)
L
Linus Torvalds 已提交
3109 3110 3111
{
	shmem_inode_cachep = kmem_cache_create("shmem_inode_cache",
				sizeof(struct shmem_inode_info),
3112
				0, SLAB_PANIC, shmem_init_inode);
L
Linus Torvalds 已提交
3113 3114 3115
	return 0;
}

3116
static void shmem_destroy_inodecache(void)
L
Linus Torvalds 已提交
3117
{
3118
	kmem_cache_destroy(shmem_inode_cachep);
L
Linus Torvalds 已提交
3119 3120
}

3121
static const struct address_space_operations shmem_aops = {
L
Linus Torvalds 已提交
3122
	.writepage	= shmem_writepage,
3123
	.set_page_dirty	= __set_page_dirty_no_writeback,
L
Linus Torvalds 已提交
3124
#ifdef CONFIG_TMPFS
N
Nick Piggin 已提交
3125 3126
	.write_begin	= shmem_write_begin,
	.write_end	= shmem_write_end,
L
Linus Torvalds 已提交
3127
#endif
3128
#ifdef CONFIG_MIGRATION
3129
	.migratepage	= migrate_page,
3130
#endif
3131
	.error_remove_page = generic_error_remove_page,
L
Linus Torvalds 已提交
3132 3133
};

3134
static const struct file_operations shmem_file_operations = {
L
Linus Torvalds 已提交
3135 3136
	.mmap		= shmem_mmap,
#ifdef CONFIG_TMPFS
3137
	.llseek		= shmem_file_llseek,
A
Al Viro 已提交
3138
	.read_iter	= shmem_file_read_iter,
3139
	.write_iter	= generic_file_write_iter,
3140
	.fsync		= noop_fsync,
3141
	.splice_read	= shmem_file_splice_read,
3142
	.splice_write	= iter_file_splice_write,
3143
	.fallocate	= shmem_fallocate,
L
Linus Torvalds 已提交
3144 3145 3146
#endif
};

3147
static const struct inode_operations shmem_inode_operations = {
3148
	.getattr	= shmem_getattr,
3149
	.setattr	= shmem_setattr,
3150 3151 3152 3153 3154
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
C
Christoph Hellwig 已提交
3155
	.set_acl	= simple_set_acl,
3156
#endif
L
Linus Torvalds 已提交
3157 3158
};

3159
static const struct inode_operations shmem_dir_inode_operations = {
L
Linus Torvalds 已提交
3160 3161 3162 3163 3164 3165 3166 3167 3168
#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
3169
	.rename2	= shmem_rename2,
3170
	.tmpfile	= shmem_tmpfile,
L
Linus Torvalds 已提交
3171
#endif
3172 3173 3174 3175 3176 3177
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
#endif
3178
#ifdef CONFIG_TMPFS_POSIX_ACL
3179
	.setattr	= shmem_setattr,
C
Christoph Hellwig 已提交
3180
	.set_acl	= simple_set_acl,
3181 3182 3183
#endif
};

3184
static const struct inode_operations shmem_special_inode_operations = {
3185 3186 3187 3188 3189 3190
#ifdef CONFIG_TMPFS_XATTR
	.setxattr	= shmem_setxattr,
	.getxattr	= shmem_getxattr,
	.listxattr	= shmem_listxattr,
	.removexattr	= shmem_removexattr,
#endif
3191
#ifdef CONFIG_TMPFS_POSIX_ACL
3192
	.setattr	= shmem_setattr,
C
Christoph Hellwig 已提交
3193
	.set_acl	= simple_set_acl,
3194
#endif
L
Linus Torvalds 已提交
3195 3196
};

3197
static const struct super_operations shmem_ops = {
L
Linus Torvalds 已提交
3198 3199 3200 3201 3202
	.alloc_inode	= shmem_alloc_inode,
	.destroy_inode	= shmem_destroy_inode,
#ifdef CONFIG_TMPFS
	.statfs		= shmem_statfs,
	.remount_fs	= shmem_remount_fs,
3203
	.show_options	= shmem_show_options,
L
Linus Torvalds 已提交
3204
#endif
A
Al Viro 已提交
3205
	.evict_inode	= shmem_evict_inode,
L
Linus Torvalds 已提交
3206 3207 3208 3209
	.drop_inode	= generic_delete_inode,
	.put_super	= shmem_put_super,
};

3210
static const struct vm_operations_struct shmem_vm_ops = {
3211
	.fault		= shmem_fault,
3212
	.map_pages	= filemap_map_pages,
L
Linus Torvalds 已提交
3213 3214 3215 3216 3217 3218
#ifdef CONFIG_NUMA
	.set_policy     = shmem_set_policy,
	.get_policy     = shmem_get_policy,
#endif
};

A
Al Viro 已提交
3219 3220
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
3221
{
A
Al Viro 已提交
3222
	return mount_nodev(fs_type, flags, data, shmem_fill_super);
L
Linus Torvalds 已提交
3223 3224
}

3225
static struct file_system_type shmem_fs_type = {
L
Linus Torvalds 已提交
3226 3227
	.owner		= THIS_MODULE,
	.name		= "tmpfs",
A
Al Viro 已提交
3228
	.mount		= shmem_mount,
L
Linus Torvalds 已提交
3229
	.kill_sb	= kill_litter_super,
3230
	.fs_flags	= FS_USERNS_MOUNT,
L
Linus Torvalds 已提交
3231 3232
};

3233
int __init shmem_init(void)
L
Linus Torvalds 已提交
3234 3235 3236
{
	int error;

3237 3238 3239 3240
	/* If rootfs called this, don't re-init */
	if (shmem_inode_cachep)
		return 0;

3241
	error = shmem_init_inodecache();
L
Linus Torvalds 已提交
3242 3243 3244
	if (error)
		goto out3;

3245
	error = register_filesystem(&shmem_fs_type);
L
Linus Torvalds 已提交
3246 3247 3248 3249
	if (error) {
		printk(KERN_ERR "Could not register tmpfs\n");
		goto out2;
	}
3250

3251
	shm_mnt = kern_mount(&shmem_fs_type);
L
Linus Torvalds 已提交
3252 3253 3254 3255 3256 3257 3258 3259
	if (IS_ERR(shm_mnt)) {
		error = PTR_ERR(shm_mnt);
		printk(KERN_ERR "Could not kern_mount tmpfs\n");
		goto out1;
	}
	return 0;

out1:
3260
	unregister_filesystem(&shmem_fs_type);
L
Linus Torvalds 已提交
3261
out2:
3262
	shmem_destroy_inodecache();
L
Linus Torvalds 已提交
3263 3264 3265 3266
out3:
	shm_mnt = ERR_PTR(error);
	return error;
}
3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278

#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.
 */

3279
static struct file_system_type shmem_fs_type = {
3280
	.name		= "tmpfs",
A
Al Viro 已提交
3281
	.mount		= ramfs_mount,
3282
	.kill_sb	= kill_litter_super,
3283
	.fs_flags	= FS_USERNS_MOUNT,
3284 3285
};

3286
int __init shmem_init(void)
3287
{
3288
	BUG_ON(register_filesystem(&shmem_fs_type) != 0);
3289

3290
	shm_mnt = kern_mount(&shmem_fs_type);
3291 3292 3293 3294 3295
	BUG_ON(IS_ERR(shm_mnt));

	return 0;
}

3296
int shmem_unuse(swp_entry_t swap, struct page *page)
3297 3298 3299 3300
{
	return 0;
}

H
Hugh Dickins 已提交
3301 3302 3303 3304 3305
int shmem_lock(struct file *file, int lock, struct user_struct *user)
{
	return 0;
}

3306 3307 3308 3309
void shmem_unlock_mapping(struct address_space *mapping)
{
}

3310
void shmem_truncate_range(struct inode *inode, loff_t lstart, loff_t lend)
3311
{
3312
	truncate_inode_pages_range(inode->i_mapping, lstart, lend);
3313 3314 3315
}
EXPORT_SYMBOL_GPL(shmem_truncate_range);

3316 3317
#define shmem_vm_ops				generic_file_vm_ops
#define shmem_file_operations			ramfs_file_operations
3318
#define shmem_get_inode(sb, dir, mode, dev, flags)	ramfs_get_inode(sb, dir, mode, dev)
3319 3320
#define shmem_acct_size(flags, size)		0
#define shmem_unacct_size(flags, size)		do {} while (0)
3321 3322 3323 3324

#endif /* CONFIG_SHMEM */

/* common code */
L
Linus Torvalds 已提交
3325

3326
static struct dentry_operations anon_ops = {
3327
	.d_dname = simple_dname
3328 3329
};

3330 3331
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
3332
{
A
Al Viro 已提交
3333
	struct file *res;
L
Linus Torvalds 已提交
3334
	struct inode *inode;
3335
	struct path path;
3336
	struct super_block *sb;
L
Linus Torvalds 已提交
3337 3338 3339
	struct qstr this;

	if (IS_ERR(shm_mnt))
A
Al Viro 已提交
3340
		return ERR_CAST(shm_mnt);
L
Linus Torvalds 已提交
3341

3342
	if (size < 0 || size > MAX_LFS_FILESIZE)
L
Linus Torvalds 已提交
3343 3344 3345 3346 3347
		return ERR_PTR(-EINVAL);

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

A
Al Viro 已提交
3348
	res = ERR_PTR(-ENOMEM);
L
Linus Torvalds 已提交
3349 3350 3351
	this.name = name;
	this.len = strlen(name);
	this.hash = 0; /* will go */
3352
	sb = shm_mnt->mnt_sb;
3353
	path.mnt = mntget(shm_mnt);
3354
	path.dentry = d_alloc_pseudo(sb, &this);
3355
	if (!path.dentry)
L
Linus Torvalds 已提交
3356
		goto put_memory;
3357
	d_set_d_op(path.dentry, &anon_ops);
L
Linus Torvalds 已提交
3358

A
Al Viro 已提交
3359
	res = ERR_PTR(-ENOSPC);
3360
	inode = shmem_get_inode(sb, NULL, S_IFREG | S_IRWXUGO, 0, flags);
L
Linus Torvalds 已提交
3361
	if (!inode)
3362
		goto put_memory;
L
Linus Torvalds 已提交
3363

3364
	inode->i_flags |= i_flags;
3365
	d_instantiate(path.dentry, inode);
L
Linus Torvalds 已提交
3366
	inode->i_size = size;
3367
	clear_nlink(inode);	/* It is unlinked */
A
Al Viro 已提交
3368 3369
	res = ERR_PTR(ramfs_nommu_expand_for_mapping(inode, size));
	if (IS_ERR(res))
3370
		goto put_path;
3371

A
Al Viro 已提交
3372
	res = alloc_file(&path, FMODE_WRITE | FMODE_READ,
3373
		  &shmem_file_operations);
A
Al Viro 已提交
3374
	if (IS_ERR(res))
3375
		goto put_path;
3376

A
Al Viro 已提交
3377
	return res;
L
Linus Torvalds 已提交
3378 3379 3380

put_memory:
	shmem_unacct_size(flags, size);
3381 3382
put_path:
	path_put(&path);
A
Al Viro 已提交
3383
	return res;
L
Linus Torvalds 已提交
3384
}
3385 3386 3387 3388 3389

/**
 * 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
3390 3391
 *	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.
3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410
 * @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);
}
3411
EXPORT_SYMBOL_GPL(shmem_file_setup);
L
Linus Torvalds 已提交
3412

3413
/**
L
Linus Torvalds 已提交
3414 3415 3416 3417 3418 3419 3420 3421
 * 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;

3422 3423 3424 3425 3426 3427 3428
	/*
	 * 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
3429 3430 3431 3432 3433 3434 3435 3436 3437
	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;
}
3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450

/**
 * 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.
 *
3451 3452
 * i915_gem_object_get_pages_gtt() mixes __GFP_NORETRY | __GFP_NOWARN in
 * with the mapping_gfp_mask(), to avoid OOMing the machine unnecessarily.
3453 3454 3455 3456
 */
struct page *shmem_read_mapping_page_gfp(struct address_space *mapping,
					 pgoff_t index, gfp_t gfp)
{
3457 3458
#ifdef CONFIG_SHMEM
	struct inode *inode = mapping->host;
H
Hugh Dickins 已提交
3459
	struct page *page;
3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472
	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
	 */
3473
	return read_cache_page_gfp(mapping, index, gfp);
3474
#endif
3475 3476
}
EXPORT_SYMBOL_GPL(shmem_read_mapping_page_gfp);