1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_PAGEMAP_H
3 #define _LINUX_PAGEMAP_H
6 * Copyright 1995 Linus Torvalds
10 #include <linux/list.h>
11 #include <linux/highmem.h>
12 #include <linux/compiler.h>
13 #include <linux/uaccess.h>
14 #include <linux/gfp.h>
15 #include <linux/bitops.h>
16 #include <linux/hardirq.h> /* for in_interrupt() */
17 #include <linux/hugetlb_inline.h>
21 unsigned long invalidate_mapping_pages(struct address_space *mapping,
22 pgoff_t start, pgoff_t end);
24 static inline void invalidate_remote_inode(struct inode *inode)
26 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
27 S_ISLNK(inode->i_mode))
28 invalidate_mapping_pages(inode->i_mapping, 0, -1);
30 int invalidate_inode_pages2(struct address_space *mapping);
31 int invalidate_inode_pages2_range(struct address_space *mapping,
32 pgoff_t start, pgoff_t end);
33 int kiocb_invalidate_pages(struct kiocb *iocb, size_t count);
34 void kiocb_invalidate_post_direct_write(struct kiocb *iocb, size_t count);
36 int write_inode_now(struct inode *, int sync);
37 int filemap_fdatawrite(struct address_space *);
38 int filemap_flush(struct address_space *);
39 int filemap_fdatawait_keep_errors(struct address_space *mapping);
40 int filemap_fdatawait_range(struct address_space *, loff_t lstart, loff_t lend);
41 int filemap_fdatawait_range_keep_errors(struct address_space *mapping,
42 loff_t start_byte, loff_t end_byte);
43 int filemap_invalidate_inode(struct inode *inode, bool flush,
44 loff_t start, loff_t end);
46 static inline int filemap_fdatawait(struct address_space *mapping)
48 return filemap_fdatawait_range(mapping, 0, LLONG_MAX);
51 bool filemap_range_has_page(struct address_space *, loff_t lstart, loff_t lend);
52 int filemap_write_and_wait_range(struct address_space *mapping,
53 loff_t lstart, loff_t lend);
54 int __filemap_fdatawrite_range(struct address_space *mapping,
55 loff_t start, loff_t end, int sync_mode);
56 int filemap_fdatawrite_range(struct address_space *mapping,
57 loff_t start, loff_t end);
58 int filemap_check_errors(struct address_space *mapping);
59 void __filemap_set_wb_err(struct address_space *mapping, int err);
60 int filemap_fdatawrite_wbc(struct address_space *mapping,
61 struct writeback_control *wbc);
62 int kiocb_write_and_wait(struct kiocb *iocb, size_t count);
64 static inline int filemap_write_and_wait(struct address_space *mapping)
66 return filemap_write_and_wait_range(mapping, 0, LLONG_MAX);
70 * filemap_set_wb_err - set a writeback error on an address_space
71 * @mapping: mapping in which to set writeback error
72 * @err: error to be set in mapping
74 * When writeback fails in some way, we must record that error so that
75 * userspace can be informed when fsync and the like are called. We endeavor
76 * to report errors on any file that was open at the time of the error. Some
77 * internal callers also need to know when writeback errors have occurred.
79 * When a writeback error occurs, most filesystems will want to call
80 * filemap_set_wb_err to record the error in the mapping so that it will be
81 * automatically reported whenever fsync is called on the file.
83 static inline void filemap_set_wb_err(struct address_space *mapping, int err)
85 /* Fastpath for common case of no error */
87 __filemap_set_wb_err(mapping, err);
91 * filemap_check_wb_err - has an error occurred since the mark was sampled?
92 * @mapping: mapping to check for writeback errors
93 * @since: previously-sampled errseq_t
95 * Grab the errseq_t value from the mapping, and see if it has changed "since"
96 * the given value was sampled.
98 * If it has then report the latest error set, otherwise return 0.
100 static inline int filemap_check_wb_err(struct address_space *mapping,
103 return errseq_check(&mapping->wb_err, since);
107 * filemap_sample_wb_err - sample the current errseq_t to test for later errors
108 * @mapping: mapping to be sampled
110 * Writeback errors are always reported relative to a particular sample point
111 * in the past. This function provides those sample points.
113 static inline errseq_t filemap_sample_wb_err(struct address_space *mapping)
115 return errseq_sample(&mapping->wb_err);
119 * file_sample_sb_err - sample the current errseq_t to test for later errors
120 * @file: file pointer to be sampled
122 * Grab the most current superblock-level errseq_t value for the given
125 static inline errseq_t file_sample_sb_err(struct file *file)
127 return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err);
131 * Flush file data before changing attributes. Caller must hold any locks
132 * required to prevent further writes to this file until we're done setting
135 static inline int inode_drain_writes(struct inode *inode)
137 inode_dio_wait(inode);
138 return filemap_write_and_wait(inode->i_mapping);
141 static inline bool mapping_empty(struct address_space *mapping)
143 return xa_empty(&mapping->i_pages);
147 * mapping_shrinkable - test if page cache state allows inode reclaim
148 * @mapping: the page cache mapping
150 * This checks the mapping's cache state for the pupose of inode
151 * reclaim and LRU management.
153 * The caller is expected to hold the i_lock, but is not required to
154 * hold the i_pages lock, which usually protects cache state. That's
155 * because the i_lock and the list_lru lock that protect the inode and
156 * its LRU state don't nest inside the irq-safe i_pages lock.
158 * Cache deletions are performed under the i_lock, which ensures that
159 * when an inode goes empty, it will reliably get queued on the LRU.
161 * Cache additions do not acquire the i_lock and may race with this
162 * check, in which case we'll report the inode as shrinkable when it
163 * has cache pages. This is okay: the shrinker also checks the
164 * refcount and the referenced bit, which will be elevated or set in
165 * the process of adding new cache pages to an inode.
167 static inline bool mapping_shrinkable(struct address_space *mapping)
172 * On highmem systems, there could be lowmem pressure from the
173 * inodes before there is highmem pressure from the page
174 * cache. Make inodes shrinkable regardless of cache state.
176 if (IS_ENABLED(CONFIG_HIGHMEM))
179 /* Cache completely empty? Shrink away. */
180 head = rcu_access_pointer(mapping->i_pages.xa_head);
185 * The xarray stores single offset-0 entries directly in the
186 * head pointer, which allows non-resident page cache entries
187 * to escape the shadow shrinker's list of xarray nodes. The
188 * inode shrinker needs to pick them up under memory pressure.
190 if (!xa_is_node(head) && xa_is_value(head))
197 * Bits in mapping->flags.
200 AS_EIO = 0, /* IO error on async write */
201 AS_ENOSPC = 1, /* ENOSPC on async write */
202 AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */
203 AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */
204 AS_EXITING = 4, /* final truncate in progress */
205 /* writeback related tags are not used */
206 AS_NO_WRITEBACK_TAGS = 5,
207 AS_LARGE_FOLIO_SUPPORT = 6,
208 AS_RELEASE_ALWAYS, /* Call ->release_folio(), even if no private data */
209 AS_STABLE_WRITES, /* must wait for writeback before modifying
211 AS_INACCESSIBLE, /* Do not attempt direct R/W access to the mapping,
212 including to move the mapping */
216 * mapping_set_error - record a writeback error in the address_space
217 * @mapping: the mapping in which an error should be set
218 * @error: the error to set in the mapping
220 * When writeback fails in some way, we must record that error so that
221 * userspace can be informed when fsync and the like are called. We endeavor
222 * to report errors on any file that was open at the time of the error. Some
223 * internal callers also need to know when writeback errors have occurred.
225 * When a writeback error occurs, most filesystems will want to call
226 * mapping_set_error to record the error in the mapping so that it can be
227 * reported when the application calls fsync(2).
229 static inline void mapping_set_error(struct address_space *mapping, int error)
234 /* Record in wb_err for checkers using errseq_t based tracking */
235 __filemap_set_wb_err(mapping, error);
237 /* Record it in superblock */
239 errseq_set(&mapping->host->i_sb->s_wb_err, error);
241 /* Record it in flags for now, for legacy callers */
242 if (error == -ENOSPC)
243 set_bit(AS_ENOSPC, &mapping->flags);
245 set_bit(AS_EIO, &mapping->flags);
248 static inline void mapping_set_unevictable(struct address_space *mapping)
250 set_bit(AS_UNEVICTABLE, &mapping->flags);
253 static inline void mapping_clear_unevictable(struct address_space *mapping)
255 clear_bit(AS_UNEVICTABLE, &mapping->flags);
258 static inline bool mapping_unevictable(struct address_space *mapping)
260 return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
263 static inline void mapping_set_exiting(struct address_space *mapping)
265 set_bit(AS_EXITING, &mapping->flags);
268 static inline int mapping_exiting(struct address_space *mapping)
270 return test_bit(AS_EXITING, &mapping->flags);
273 static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
275 set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
278 static inline int mapping_use_writeback_tags(struct address_space *mapping)
280 return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
283 static inline bool mapping_release_always(const struct address_space *mapping)
285 return test_bit(AS_RELEASE_ALWAYS, &mapping->flags);
288 static inline void mapping_set_release_always(struct address_space *mapping)
290 set_bit(AS_RELEASE_ALWAYS, &mapping->flags);
293 static inline void mapping_clear_release_always(struct address_space *mapping)
295 clear_bit(AS_RELEASE_ALWAYS, &mapping->flags);
298 static inline bool mapping_stable_writes(const struct address_space *mapping)
300 return test_bit(AS_STABLE_WRITES, &mapping->flags);
303 static inline void mapping_set_stable_writes(struct address_space *mapping)
305 set_bit(AS_STABLE_WRITES, &mapping->flags);
308 static inline void mapping_clear_stable_writes(struct address_space *mapping)
310 clear_bit(AS_STABLE_WRITES, &mapping->flags);
313 static inline void mapping_set_inaccessible(struct address_space *mapping)
316 * It's expected inaccessible mappings are also unevictable. Compaction
317 * migrate scanner (isolate_migratepages_block()) relies on this to
318 * reduce page locking.
320 set_bit(AS_UNEVICTABLE, &mapping->flags);
321 set_bit(AS_INACCESSIBLE, &mapping->flags);
324 static inline bool mapping_inaccessible(struct address_space *mapping)
326 return test_bit(AS_INACCESSIBLE, &mapping->flags);
329 static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
331 return mapping->gfp_mask;
334 /* Restricts the given gfp_mask to what the mapping allows. */
335 static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
338 return mapping_gfp_mask(mapping) & gfp_mask;
342 * This is non-atomic. Only to be used before the mapping is activated.
343 * Probably needs a barrier...
345 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
351 * There are some parts of the kernel which assume that PMD entries
352 * are exactly HPAGE_PMD_ORDER. Those should be fixed, but until then,
353 * limit the maximum allocation order to PMD size. I'm not aware of any
354 * assumptions about maximum order if THP are disabled, but 8 seems like
355 * a good order (that's 1MB if you're using 4kB pages)
357 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
358 #define PREFERRED_MAX_PAGECACHE_ORDER HPAGE_PMD_ORDER
360 #define PREFERRED_MAX_PAGECACHE_ORDER 8
364 * xas_split_alloc() does not support arbitrary orders. This implies no
365 * 512MB THP on ARM64 with 64KB base page size.
367 #define MAX_XAS_ORDER (XA_CHUNK_SHIFT * 2 - 1)
368 #define MAX_PAGECACHE_ORDER min(MAX_XAS_ORDER, PREFERRED_MAX_PAGECACHE_ORDER)
371 * mapping_set_large_folios() - Indicate the file supports large folios.
372 * @mapping: The file.
374 * The filesystem should call this function in its inode constructor to
375 * indicate that the VFS can use large folios to cache the contents of
378 * Context: This should not be called while the inode is active as it
381 static inline void mapping_set_large_folios(struct address_space *mapping)
383 __set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
387 * Large folio support currently depends on THP. These dependencies are
388 * being worked on but are not yet fixed.
390 static inline bool mapping_large_folio_support(struct address_space *mapping)
392 /* AS_LARGE_FOLIO_SUPPORT is only reasonable for pagecache folios */
393 VM_WARN_ONCE((unsigned long)mapping & PAGE_MAPPING_ANON,
394 "Anonymous mapping always supports large folio");
396 return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) &&
397 test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
400 /* Return the maximum folio size for this pagecache mapping, in bytes. */
401 static inline size_t mapping_max_folio_size(struct address_space *mapping)
403 if (mapping_large_folio_support(mapping))
404 return PAGE_SIZE << MAX_PAGECACHE_ORDER;
408 static inline int filemap_nr_thps(struct address_space *mapping)
410 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
411 return atomic_read(&mapping->nr_thps);
417 static inline void filemap_nr_thps_inc(struct address_space *mapping)
419 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
420 if (!mapping_large_folio_support(mapping))
421 atomic_inc(&mapping->nr_thps);
423 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
427 static inline void filemap_nr_thps_dec(struct address_space *mapping)
429 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
430 if (!mapping_large_folio_support(mapping))
431 atomic_dec(&mapping->nr_thps);
433 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
437 struct address_space *folio_mapping(struct folio *);
438 struct address_space *swapcache_mapping(struct folio *);
441 * folio_file_mapping - Find the mapping this folio belongs to.
444 * For folios which are in the page cache, return the mapping that this
445 * page belongs to. Folios in the swap cache return the mapping of the
446 * swap file or swap device where the data is stored. This is different
447 * from the mapping returned by folio_mapping(). The only reason to
448 * use it is if, like NFS, you return 0 from ->activate_swapfile.
450 * Do not call this for folios which aren't in the page cache or swap cache.
452 static inline struct address_space *folio_file_mapping(struct folio *folio)
454 if (unlikely(folio_test_swapcache(folio)))
455 return swapcache_mapping(folio);
457 return folio->mapping;
461 * folio_flush_mapping - Find the file mapping this folio belongs to.
464 * For folios which are in the page cache, return the mapping that this
465 * page belongs to. Anonymous folios return NULL, even if they're in
466 * the swap cache. Other kinds of folio also return NULL.
468 * This is ONLY used by architecture cache flushing code. If you aren't
469 * writing cache flushing code, you want either folio_mapping() or
470 * folio_file_mapping().
472 static inline struct address_space *folio_flush_mapping(struct folio *folio)
474 if (unlikely(folio_test_swapcache(folio)))
477 return folio_mapping(folio);
480 static inline struct address_space *page_file_mapping(struct page *page)
482 return folio_file_mapping(page_folio(page));
486 * folio_inode - Get the host inode for this folio.
489 * For folios which are in the page cache, return the inode that this folio
492 * Do not call this for folios which aren't in the page cache.
494 static inline struct inode *folio_inode(struct folio *folio)
496 return folio->mapping->host;
500 * folio_attach_private - Attach private data to a folio.
501 * @folio: Folio to attach data to.
502 * @data: Data to attach to folio.
504 * Attaching private data to a folio increments the page's reference count.
505 * The data must be detached before the folio will be freed.
507 static inline void folio_attach_private(struct folio *folio, void *data)
510 folio->private = data;
511 folio_set_private(folio);
515 * folio_change_private - Change private data on a folio.
516 * @folio: Folio to change the data on.
517 * @data: Data to set on the folio.
519 * Change the private data attached to a folio and return the old
520 * data. The page must previously have had data attached and the data
521 * must be detached before the folio will be freed.
523 * Return: Data that was previously attached to the folio.
525 static inline void *folio_change_private(struct folio *folio, void *data)
527 void *old = folio_get_private(folio);
529 folio->private = data;
534 * folio_detach_private - Detach private data from a folio.
535 * @folio: Folio to detach data from.
537 * Removes the data that was previously attached to the folio and decrements
538 * the refcount on the page.
540 * Return: Data that was attached to the folio.
542 static inline void *folio_detach_private(struct folio *folio)
544 void *data = folio_get_private(folio);
546 if (!folio_test_private(folio))
548 folio_clear_private(folio);
549 folio->private = NULL;
555 static inline void attach_page_private(struct page *page, void *data)
557 folio_attach_private(page_folio(page), data);
560 static inline void *detach_page_private(struct page *page)
562 return folio_detach_private(page_folio(page));
566 struct folio *filemap_alloc_folio_noprof(gfp_t gfp, unsigned int order);
568 static inline struct folio *filemap_alloc_folio_noprof(gfp_t gfp, unsigned int order)
570 return folio_alloc_noprof(gfp, order);
574 #define filemap_alloc_folio(...) \
575 alloc_hooks(filemap_alloc_folio_noprof(__VA_ARGS__))
577 static inline struct page *__page_cache_alloc(gfp_t gfp)
579 return &filemap_alloc_folio(gfp, 0)->page;
582 static inline gfp_t readahead_gfp_mask(struct address_space *x)
584 return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
587 typedef int filler_t(struct file *, struct folio *);
589 pgoff_t page_cache_next_miss(struct address_space *mapping,
590 pgoff_t index, unsigned long max_scan);
591 pgoff_t page_cache_prev_miss(struct address_space *mapping,
592 pgoff_t index, unsigned long max_scan);
595 * typedef fgf_t - Flags for getting folios from the page cache.
597 * Most users of the page cache will not need to use these flags;
598 * there are convenience functions such as filemap_get_folio() and
599 * filemap_lock_folio(). For users which need more control over exactly
600 * what is done with the folios, these flags to __filemap_get_folio()
603 * * %FGP_ACCESSED - The folio will be marked accessed.
604 * * %FGP_LOCK - The folio is returned locked.
605 * * %FGP_CREAT - If no folio is present then a new folio is allocated,
606 * added to the page cache and the VM's LRU list. The folio is
608 * * %FGP_FOR_MMAP - The caller wants to do its own locking dance if the
609 * folio is already in cache. If the folio was allocated, unlock it
610 * before returning so the caller can do the same dance.
611 * * %FGP_WRITE - The folio will be written to by the caller.
612 * * %FGP_NOFS - __GFP_FS will get cleared in gfp.
613 * * %FGP_NOWAIT - Don't block on the folio lock.
614 * * %FGP_STABLE - Wait for the folio to be stable (finished writeback)
615 * * %FGP_WRITEBEGIN - The flags to use in a filesystem write_begin()
618 typedef unsigned int __bitwise fgf_t;
620 #define FGP_ACCESSED ((__force fgf_t)0x00000001)
621 #define FGP_LOCK ((__force fgf_t)0x00000002)
622 #define FGP_CREAT ((__force fgf_t)0x00000004)
623 #define FGP_WRITE ((__force fgf_t)0x00000008)
624 #define FGP_NOFS ((__force fgf_t)0x00000010)
625 #define FGP_NOWAIT ((__force fgf_t)0x00000020)
626 #define FGP_FOR_MMAP ((__force fgf_t)0x00000040)
627 #define FGP_STABLE ((__force fgf_t)0x00000080)
628 #define FGF_GET_ORDER(fgf) (((__force unsigned)fgf) >> 26) /* top 6 bits */
630 #define FGP_WRITEBEGIN (FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE)
633 * fgf_set_order - Encode a length in the fgf_t flags.
634 * @size: The suggested size of the folio to create.
636 * The caller of __filemap_get_folio() can use this to suggest a preferred
637 * size for the folio that is created. If there is already a folio at
638 * the index, it will be returned, no matter what its size. If a folio
639 * is freshly created, it may be of a different size than requested
640 * due to alignment constraints, memory pressure, or the presence of
641 * other folios at nearby indices.
643 static inline fgf_t fgf_set_order(size_t size)
645 unsigned int shift = ilog2(size);
647 if (shift <= PAGE_SHIFT)
649 return (__force fgf_t)((shift - PAGE_SHIFT) << 26);
652 void *filemap_get_entry(struct address_space *mapping, pgoff_t index);
653 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
654 fgf_t fgp_flags, gfp_t gfp);
655 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
656 fgf_t fgp_flags, gfp_t gfp);
659 * filemap_get_folio - Find and get a folio.
660 * @mapping: The address_space to search.
661 * @index: The page index.
663 * Looks up the page cache entry at @mapping & @index. If a folio is
664 * present, it is returned with an increased refcount.
666 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
667 * this index. Will not return a shadow, swap or DAX entry.
669 static inline struct folio *filemap_get_folio(struct address_space *mapping,
672 return __filemap_get_folio(mapping, index, 0, 0);
676 * filemap_lock_folio - Find and lock a folio.
677 * @mapping: The address_space to search.
678 * @index: The page index.
680 * Looks up the page cache entry at @mapping & @index. If a folio is
681 * present, it is returned locked with an increased refcount.
683 * Context: May sleep.
684 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
685 * this index. Will not return a shadow, swap or DAX entry.
687 static inline struct folio *filemap_lock_folio(struct address_space *mapping,
690 return __filemap_get_folio(mapping, index, FGP_LOCK, 0);
694 * filemap_grab_folio - grab a folio from the page cache
695 * @mapping: The address space to search
696 * @index: The page index
698 * Looks up the page cache entry at @mapping & @index. If no folio is found,
699 * a new folio is created. The folio is locked, marked as accessed, and
702 * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found
703 * and failed to create a folio.
705 static inline struct folio *filemap_grab_folio(struct address_space *mapping,
708 return __filemap_get_folio(mapping, index,
709 FGP_LOCK | FGP_ACCESSED | FGP_CREAT,
710 mapping_gfp_mask(mapping));
714 * find_get_page - find and get a page reference
715 * @mapping: the address_space to search
716 * @offset: the page index
718 * Looks up the page cache slot at @mapping & @offset. If there is a
719 * page cache page, it is returned with an increased refcount.
721 * Otherwise, %NULL is returned.
723 static inline struct page *find_get_page(struct address_space *mapping,
726 return pagecache_get_page(mapping, offset, 0, 0);
729 static inline struct page *find_get_page_flags(struct address_space *mapping,
730 pgoff_t offset, fgf_t fgp_flags)
732 return pagecache_get_page(mapping, offset, fgp_flags, 0);
736 * find_lock_page - locate, pin and lock a pagecache page
737 * @mapping: the address_space to search
738 * @index: the page index
740 * Looks up the page cache entry at @mapping & @index. If there is a
741 * page cache page, it is returned locked and with an increased
744 * Context: May sleep.
745 * Return: A struct page or %NULL if there is no page in the cache for this
748 static inline struct page *find_lock_page(struct address_space *mapping,
751 return pagecache_get_page(mapping, index, FGP_LOCK, 0);
755 * find_or_create_page - locate or add a pagecache page
756 * @mapping: the page's address_space
757 * @index: the page's index into the mapping
758 * @gfp_mask: page allocation mode
760 * Looks up the page cache slot at @mapping & @offset. If there is a
761 * page cache page, it is returned locked and with an increased
764 * If the page is not present, a new page is allocated using @gfp_mask
765 * and added to the page cache and the VM's LRU list. The page is
766 * returned locked and with an increased refcount.
768 * On memory exhaustion, %NULL is returned.
770 * find_or_create_page() may sleep, even if @gfp_flags specifies an
773 static inline struct page *find_or_create_page(struct address_space *mapping,
774 pgoff_t index, gfp_t gfp_mask)
776 return pagecache_get_page(mapping, index,
777 FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
782 * grab_cache_page_nowait - returns locked page at given index in given cache
783 * @mapping: target address_space
784 * @index: the page index
786 * Same as grab_cache_page(), but do not wait if the page is unavailable.
787 * This is intended for speculative data generators, where the data can
788 * be regenerated if the page couldn't be grabbed. This routine should
789 * be safe to call while holding the lock for another page.
791 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
792 * and deadlock against the caller's locked page.
794 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
797 return pagecache_get_page(mapping, index,
798 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
799 mapping_gfp_mask(mapping));
802 extern pgoff_t __folio_swap_cache_index(struct folio *folio);
805 * folio_index - File index of a folio.
808 * For a folio which is either in the page cache or the swap cache,
809 * return its index within the address_space it belongs to. If you know
810 * the page is definitely in the page cache, you can look at the folio's
813 * Return: The index (offset in units of pages) of a folio in its file.
815 static inline pgoff_t folio_index(struct folio *folio)
817 if (unlikely(folio_test_swapcache(folio)))
818 return __folio_swap_cache_index(folio);
823 * folio_next_index - Get the index of the next folio.
824 * @folio: The current folio.
826 * Return: The index of the folio which follows this folio in the file.
828 static inline pgoff_t folio_next_index(struct folio *folio)
830 return folio->index + folio_nr_pages(folio);
834 * folio_file_page - The page for a particular index.
835 * @folio: The folio which contains this index.
836 * @index: The index we want to look up.
838 * Sometimes after looking up a folio in the page cache, we need to
839 * obtain the specific page for an index (eg a page fault).
841 * Return: The page containing the file data for this index.
843 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
845 return folio_page(folio, index & (folio_nr_pages(folio) - 1));
849 * folio_contains - Does this folio contain this index?
851 * @index: The page index within the file.
853 * Context: The caller should have the page locked in order to prevent
854 * (eg) shmem from moving the page between the page cache and swap cache
855 * and changing its index in the middle of the operation.
856 * Return: true or false.
858 static inline bool folio_contains(struct folio *folio, pgoff_t index)
860 return index - folio_index(folio) < folio_nr_pages(folio);
864 * Given the page we found in the page cache, return the page corresponding
865 * to this index in the file
867 static inline struct page *find_subpage(struct page *head, pgoff_t index)
869 /* HugeTLBfs wants the head page regardless */
873 return head + (index & (thp_nr_pages(head) - 1));
876 unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start,
877 pgoff_t end, struct folio_batch *fbatch);
878 unsigned filemap_get_folios_contig(struct address_space *mapping,
879 pgoff_t *start, pgoff_t end, struct folio_batch *fbatch);
880 unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start,
881 pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch);
883 struct page *grab_cache_page_write_begin(struct address_space *mapping,
887 * Returns locked page at given index in given cache, creating it if needed.
889 static inline struct page *grab_cache_page(struct address_space *mapping,
892 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
895 struct folio *read_cache_folio(struct address_space *, pgoff_t index,
896 filler_t *filler, struct file *file);
897 struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index,
899 struct page *read_cache_page(struct address_space *, pgoff_t index,
900 filler_t *filler, struct file *file);
901 extern struct page * read_cache_page_gfp(struct address_space *mapping,
902 pgoff_t index, gfp_t gfp_mask);
904 static inline struct page *read_mapping_page(struct address_space *mapping,
905 pgoff_t index, struct file *file)
907 return read_cache_page(mapping, index, NULL, file);
910 static inline struct folio *read_mapping_folio(struct address_space *mapping,
911 pgoff_t index, struct file *file)
913 return read_cache_folio(mapping, index, NULL, file);
917 * Get the offset in PAGE_SIZE (even for hugetlb pages).
919 static inline pgoff_t page_to_pgoff(struct page *page)
923 if (likely(!PageTransTail(page)))
926 head = compound_head(page);
928 * We don't initialize ->index for tail pages: calculate based on
931 return head->index + page - head;
935 * Return byte-offset into filesystem object for page.
937 static inline loff_t page_offset(struct page *page)
939 return ((loff_t)page->index) << PAGE_SHIFT;
943 * folio_pos - Returns the byte position of this folio in its file.
946 static inline loff_t folio_pos(struct folio *folio)
948 return page_offset(&folio->page);
952 * Get the offset in PAGE_SIZE (even for hugetlb folios).
954 static inline pgoff_t folio_pgoff(struct folio *folio)
959 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
960 unsigned long address)
963 pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
964 pgoff += vma->vm_pgoff;
968 struct wait_page_key {
974 struct wait_page_queue {
977 wait_queue_entry_t wait;
980 static inline bool wake_page_match(struct wait_page_queue *wait_page,
981 struct wait_page_key *key)
983 if (wait_page->folio != key->folio)
987 if (wait_page->bit_nr != key->bit_nr)
993 void __folio_lock(struct folio *folio);
994 int __folio_lock_killable(struct folio *folio);
995 vm_fault_t __folio_lock_or_retry(struct folio *folio, struct vm_fault *vmf);
996 void unlock_page(struct page *page);
997 void folio_unlock(struct folio *folio);
1000 * folio_trylock() - Attempt to lock a folio.
1001 * @folio: The folio to attempt to lock.
1003 * Sometimes it is undesirable to wait for a folio to be unlocked (eg
1004 * when the locks are being taken in the wrong order, or if making
1005 * progress through a batch of folios is more important than processing
1006 * them in order). Usually folio_lock() is the correct function to call.
1008 * Context: Any context.
1009 * Return: Whether the lock was successfully acquired.
1011 static inline bool folio_trylock(struct folio *folio)
1013 return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
1017 * Return true if the page was successfully locked
1019 static inline bool trylock_page(struct page *page)
1021 return folio_trylock(page_folio(page));
1025 * folio_lock() - Lock this folio.
1026 * @folio: The folio to lock.
1028 * The folio lock protects against many things, probably more than it
1029 * should. It is primarily held while a folio is being brought uptodate,
1030 * either from its backing file or from swap. It is also held while a
1031 * folio is being truncated from its address_space, so holding the lock
1032 * is sufficient to keep folio->mapping stable.
1034 * The folio lock is also held while write() is modifying the page to
1035 * provide POSIX atomicity guarantees (as long as the write does not
1036 * cross a page boundary). Other modifications to the data in the folio
1037 * do not hold the folio lock and can race with writes, eg DMA and stores
1040 * Context: May sleep. If you need to acquire the locks of two or
1041 * more folios, they must be in order of ascending index, if they are
1042 * in the same address_space. If they are in different address_spaces,
1043 * acquire the lock of the folio which belongs to the address_space which
1044 * has the lowest address in memory first.
1046 static inline void folio_lock(struct folio *folio)
1049 if (!folio_trylock(folio))
1050 __folio_lock(folio);
1054 * lock_page() - Lock the folio containing this page.
1055 * @page: The page to lock.
1057 * See folio_lock() for a description of what the lock protects.
1058 * This is a legacy function and new code should probably use folio_lock()
1061 * Context: May sleep. Pages in the same folio share a lock, so do not
1062 * attempt to lock two pages which share a folio.
1064 static inline void lock_page(struct page *page)
1066 struct folio *folio;
1069 folio = page_folio(page);
1070 if (!folio_trylock(folio))
1071 __folio_lock(folio);
1075 * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
1076 * @folio: The folio to lock.
1078 * Attempts to lock the folio, like folio_lock(), except that the sleep
1079 * to acquire the lock is interruptible by a fatal signal.
1081 * Context: May sleep; see folio_lock().
1082 * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
1084 static inline int folio_lock_killable(struct folio *folio)
1087 if (!folio_trylock(folio))
1088 return __folio_lock_killable(folio);
1093 * folio_lock_or_retry - Lock the folio, unless this would block and the
1094 * caller indicated that it can handle a retry.
1096 * Return value and mmap_lock implications depend on flags; see
1097 * __folio_lock_or_retry().
1099 static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
1100 struct vm_fault *vmf)
1103 if (!folio_trylock(folio))
1104 return __folio_lock_or_retry(folio, vmf);
1109 * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
1110 * and should not be used directly.
1112 void folio_wait_bit(struct folio *folio, int bit_nr);
1113 int folio_wait_bit_killable(struct folio *folio, int bit_nr);
1116 * Wait for a folio to be unlocked.
1118 * This must be called with the caller "holding" the folio,
1119 * ie with increased folio reference count so that the folio won't
1120 * go away during the wait.
1122 static inline void folio_wait_locked(struct folio *folio)
1124 if (folio_test_locked(folio))
1125 folio_wait_bit(folio, PG_locked);
1128 static inline int folio_wait_locked_killable(struct folio *folio)
1130 if (!folio_test_locked(folio))
1132 return folio_wait_bit_killable(folio, PG_locked);
1135 static inline void wait_on_page_locked(struct page *page)
1137 folio_wait_locked(page_folio(page));
1140 void folio_end_read(struct folio *folio, bool success);
1141 void wait_on_page_writeback(struct page *page);
1142 void folio_wait_writeback(struct folio *folio);
1143 int folio_wait_writeback_killable(struct folio *folio);
1144 void end_page_writeback(struct page *page);
1145 void folio_end_writeback(struct folio *folio);
1146 void wait_for_stable_page(struct page *page);
1147 void folio_wait_stable(struct folio *folio);
1148 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
1149 void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
1150 void __folio_cancel_dirty(struct folio *folio);
1151 static inline void folio_cancel_dirty(struct folio *folio)
1153 /* Avoid atomic ops, locking, etc. when not actually needed. */
1154 if (folio_test_dirty(folio))
1155 __folio_cancel_dirty(folio);
1157 bool folio_clear_dirty_for_io(struct folio *folio);
1158 bool clear_page_dirty_for_io(struct page *page);
1159 void folio_invalidate(struct folio *folio, size_t offset, size_t length);
1160 bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
1162 #ifdef CONFIG_MIGRATION
1163 int filemap_migrate_folio(struct address_space *mapping, struct folio *dst,
1164 struct folio *src, enum migrate_mode mode);
1166 #define filemap_migrate_folio NULL
1168 void folio_end_private_2(struct folio *folio);
1169 void folio_wait_private_2(struct folio *folio);
1170 int folio_wait_private_2_killable(struct folio *folio);
1173 * Add an arbitrary waiter to a page's wait queue
1175 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
1178 * Fault in userspace address range.
1180 size_t fault_in_writeable(char __user *uaddr, size_t size);
1181 size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
1182 size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
1183 size_t fault_in_readable(const char __user *uaddr, size_t size);
1185 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
1186 pgoff_t index, gfp_t gfp);
1187 int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1188 pgoff_t index, gfp_t gfp);
1189 void filemap_remove_folio(struct folio *folio);
1190 void __filemap_remove_folio(struct folio *folio, void *shadow);
1191 void replace_page_cache_folio(struct folio *old, struct folio *new);
1192 void delete_from_page_cache_batch(struct address_space *mapping,
1193 struct folio_batch *fbatch);
1194 bool filemap_release_folio(struct folio *folio, gfp_t gfp);
1195 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1198 /* Must be non-static for BPF error injection */
1199 int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1200 pgoff_t index, gfp_t gfp, void **shadowp);
1202 bool filemap_range_has_writeback(struct address_space *mapping,
1203 loff_t start_byte, loff_t end_byte);
1206 * filemap_range_needs_writeback - check if range potentially needs writeback
1207 * @mapping: address space within which to check
1208 * @start_byte: offset in bytes where the range starts
1209 * @end_byte: offset in bytes where the range ends (inclusive)
1211 * Find at least one page in the range supplied, usually used to check if
1212 * direct writing in this range will trigger a writeback. Used by O_DIRECT
1213 * read/write with IOCB_NOWAIT, to see if the caller needs to do
1214 * filemap_write_and_wait_range() before proceeding.
1216 * Return: %true if the caller should do filemap_write_and_wait_range() before
1217 * doing O_DIRECT to a page in this range, %false otherwise.
1219 static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1223 if (!mapping->nrpages)
1225 if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1226 !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1228 return filemap_range_has_writeback(mapping, start_byte, end_byte);
1232 * struct readahead_control - Describes a readahead request.
1234 * A readahead request is for consecutive pages. Filesystems which
1235 * implement the ->readahead method should call readahead_page() or
1236 * readahead_page_batch() in a loop and attempt to start I/O against
1237 * each page in the request.
1239 * Most of the fields in this struct are private and should be accessed
1240 * by the functions below.
1242 * @file: The file, used primarily by network filesystems for authentication.
1243 * May be NULL if invoked internally by the filesystem.
1244 * @mapping: Readahead this filesystem object.
1245 * @ra: File readahead state. May be NULL.
1247 struct readahead_control {
1249 struct address_space *mapping;
1250 struct file_ra_state *ra;
1251 /* private: use the readahead_* accessors instead */
1253 unsigned int _nr_pages;
1254 unsigned int _batch_count;
1256 unsigned long _pflags;
1259 #define DEFINE_READAHEAD(ractl, f, r, m, i) \
1260 struct readahead_control ractl = { \
1267 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
1269 void page_cache_ra_unbounded(struct readahead_control *,
1270 unsigned long nr_to_read, unsigned long lookahead_count);
1271 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
1272 void page_cache_async_ra(struct readahead_control *, struct folio *,
1273 unsigned long req_count);
1274 void readahead_expand(struct readahead_control *ractl,
1275 loff_t new_start, size_t new_len);
1278 * page_cache_sync_readahead - generic file readahead
1279 * @mapping: address_space which holds the pagecache and I/O vectors
1280 * @ra: file_ra_state which holds the readahead state
1281 * @file: Used by the filesystem for authentication.
1282 * @index: Index of first page to be read.
1283 * @req_count: Total number of pages being read by the caller.
1285 * page_cache_sync_readahead() should be called when a cache miss happened:
1286 * it will submit the read. The readahead logic may decide to piggyback more
1287 * pages onto the read request if access patterns suggest it will improve
1291 void page_cache_sync_readahead(struct address_space *mapping,
1292 struct file_ra_state *ra, struct file *file, pgoff_t index,
1293 unsigned long req_count)
1295 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1296 page_cache_sync_ra(&ractl, req_count);
1300 * page_cache_async_readahead - file readahead for marked pages
1301 * @mapping: address_space which holds the pagecache and I/O vectors
1302 * @ra: file_ra_state which holds the readahead state
1303 * @file: Used by the filesystem for authentication.
1304 * @folio: The folio which triggered the readahead call.
1305 * @req_count: Total number of pages being read by the caller.
1307 * page_cache_async_readahead() should be called when a page is used which
1308 * is marked as PageReadahead; this is a marker to suggest that the application
1309 * has used up enough of the readahead window that we should start pulling in
1313 void page_cache_async_readahead(struct address_space *mapping,
1314 struct file_ra_state *ra, struct file *file,
1315 struct folio *folio, unsigned long req_count)
1317 DEFINE_READAHEAD(ractl, file, ra, mapping, folio->index);
1318 page_cache_async_ra(&ractl, folio, req_count);
1321 static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1323 struct folio *folio;
1325 BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1326 ractl->_nr_pages -= ractl->_batch_count;
1327 ractl->_index += ractl->_batch_count;
1329 if (!ractl->_nr_pages) {
1330 ractl->_batch_count = 0;
1334 folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1335 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1336 ractl->_batch_count = folio_nr_pages(folio);
1342 * readahead_page - Get the next page to read.
1343 * @ractl: The current readahead request.
1345 * Context: The page is locked and has an elevated refcount. The caller
1346 * should decreases the refcount once the page has been submitted for I/O
1347 * and unlock the page once all I/O to that page has completed.
1348 * Return: A pointer to the next page, or %NULL if we are done.
1350 static inline struct page *readahead_page(struct readahead_control *ractl)
1352 struct folio *folio = __readahead_folio(ractl);
1354 return &folio->page;
1358 * readahead_folio - Get the next folio to read.
1359 * @ractl: The current readahead request.
1361 * Context: The folio is locked. The caller should unlock the folio once
1362 * all I/O to that folio has completed.
1363 * Return: A pointer to the next folio, or %NULL if we are done.
1365 static inline struct folio *readahead_folio(struct readahead_control *ractl)
1367 struct folio *folio = __readahead_folio(ractl);
1374 static inline unsigned int __readahead_batch(struct readahead_control *rac,
1375 struct page **array, unsigned int array_sz)
1378 XA_STATE(xas, &rac->mapping->i_pages, 0);
1381 BUG_ON(rac->_batch_count > rac->_nr_pages);
1382 rac->_nr_pages -= rac->_batch_count;
1383 rac->_index += rac->_batch_count;
1384 rac->_batch_count = 0;
1386 xas_set(&xas, rac->_index);
1388 xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1389 if (xas_retry(&xas, page))
1391 VM_BUG_ON_PAGE(!PageLocked(page), page);
1392 VM_BUG_ON_PAGE(PageTail(page), page);
1394 rac->_batch_count += thp_nr_pages(page);
1404 * readahead_page_batch - Get a batch of pages to read.
1405 * @rac: The current readahead request.
1406 * @array: An array of pointers to struct page.
1408 * Context: The pages are locked and have an elevated refcount. The caller
1409 * should decreases the refcount once the page has been submitted for I/O
1410 * and unlock the page once all I/O to that page has completed.
1411 * Return: The number of pages placed in the array. 0 indicates the request
1414 #define readahead_page_batch(rac, array) \
1415 __readahead_batch(rac, array, ARRAY_SIZE(array))
1418 * readahead_pos - The byte offset into the file of this readahead request.
1419 * @rac: The readahead request.
1421 static inline loff_t readahead_pos(struct readahead_control *rac)
1423 return (loff_t)rac->_index * PAGE_SIZE;
1427 * readahead_length - The number of bytes in this readahead request.
1428 * @rac: The readahead request.
1430 static inline size_t readahead_length(struct readahead_control *rac)
1432 return rac->_nr_pages * PAGE_SIZE;
1436 * readahead_index - The index of the first page in this readahead request.
1437 * @rac: The readahead request.
1439 static inline pgoff_t readahead_index(struct readahead_control *rac)
1445 * readahead_count - The number of pages in this readahead request.
1446 * @rac: The readahead request.
1448 static inline unsigned int readahead_count(struct readahead_control *rac)
1450 return rac->_nr_pages;
1454 * readahead_batch_length - The number of bytes in the current batch.
1455 * @rac: The readahead request.
1457 static inline size_t readahead_batch_length(struct readahead_control *rac)
1459 return rac->_batch_count * PAGE_SIZE;
1462 static inline unsigned long dir_pages(struct inode *inode)
1464 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1469 * folio_mkwrite_check_truncate - check if folio was truncated
1470 * @folio: the folio to check
1471 * @inode: the inode to check the folio against
1473 * Return: the number of bytes in the folio up to EOF,
1474 * or -EFAULT if the folio was truncated.
1476 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1477 struct inode *inode)
1479 loff_t size = i_size_read(inode);
1480 pgoff_t index = size >> PAGE_SHIFT;
1481 size_t offset = offset_in_folio(folio, size);
1483 if (!folio->mapping)
1486 /* folio is wholly inside EOF */
1487 if (folio_next_index(folio) - 1 < index)
1488 return folio_size(folio);
1489 /* folio is wholly past EOF */
1490 if (folio->index > index || !offset)
1492 /* folio is partially inside EOF */
1497 * page_mkwrite_check_truncate - check if page was truncated
1498 * @page: the page to check
1499 * @inode: the inode to check the page against
1501 * Returns the number of bytes in the page up to EOF,
1502 * or -EFAULT if the page was truncated.
1504 static inline int page_mkwrite_check_truncate(struct page *page,
1505 struct inode *inode)
1507 loff_t size = i_size_read(inode);
1508 pgoff_t index = size >> PAGE_SHIFT;
1509 int offset = offset_in_page(size);
1511 if (page->mapping != inode->i_mapping)
1514 /* page is wholly inside EOF */
1515 if (page->index < index)
1517 /* page is wholly past EOF */
1518 if (page->index > index || !offset)
1520 /* page is partially inside EOF */
1525 * i_blocks_per_folio - How many blocks fit in this folio.
1526 * @inode: The inode which contains the blocks.
1527 * @folio: The folio.
1529 * If the block size is larger than the size of this folio, return zero.
1531 * Context: The caller should hold a refcount on the folio to prevent it
1533 * Return: The number of filesystem blocks covered by this folio.
1536 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1538 return folio_size(folio) >> inode->i_blkbits;
1540 #endif /* _LINUX_PAGEMAP_H */