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);
44 static inline int filemap_fdatawait(struct address_space *mapping)
46 return filemap_fdatawait_range(mapping, 0, LLONG_MAX);
49 bool filemap_range_has_page(struct address_space *, loff_t lstart, loff_t lend);
50 int filemap_write_and_wait_range(struct address_space *mapping,
51 loff_t lstart, loff_t lend);
52 int __filemap_fdatawrite_range(struct address_space *mapping,
53 loff_t start, loff_t end, int sync_mode);
54 int filemap_fdatawrite_range(struct address_space *mapping,
55 loff_t start, loff_t end);
56 int filemap_check_errors(struct address_space *mapping);
57 void __filemap_set_wb_err(struct address_space *mapping, int err);
58 int filemap_fdatawrite_wbc(struct address_space *mapping,
59 struct writeback_control *wbc);
60 int kiocb_write_and_wait(struct kiocb *iocb, size_t count);
62 static inline int filemap_write_and_wait(struct address_space *mapping)
64 return filemap_write_and_wait_range(mapping, 0, LLONG_MAX);
68 * filemap_set_wb_err - set a writeback error on an address_space
69 * @mapping: mapping in which to set writeback error
70 * @err: error to be set in mapping
72 * When writeback fails in some way, we must record that error so that
73 * userspace can be informed when fsync and the like are called. We endeavor
74 * to report errors on any file that was open at the time of the error. Some
75 * internal callers also need to know when writeback errors have occurred.
77 * When a writeback error occurs, most filesystems will want to call
78 * filemap_set_wb_err to record the error in the mapping so that it will be
79 * automatically reported whenever fsync is called on the file.
81 static inline void filemap_set_wb_err(struct address_space *mapping, int err)
83 /* Fastpath for common case of no error */
85 __filemap_set_wb_err(mapping, err);
89 * filemap_check_wb_err - has an error occurred since the mark was sampled?
90 * @mapping: mapping to check for writeback errors
91 * @since: previously-sampled errseq_t
93 * Grab the errseq_t value from the mapping, and see if it has changed "since"
94 * the given value was sampled.
96 * If it has then report the latest error set, otherwise return 0.
98 static inline int filemap_check_wb_err(struct address_space *mapping,
101 return errseq_check(&mapping->wb_err, since);
105 * filemap_sample_wb_err - sample the current errseq_t to test for later errors
106 * @mapping: mapping to be sampled
108 * Writeback errors are always reported relative to a particular sample point
109 * in the past. This function provides those sample points.
111 static inline errseq_t filemap_sample_wb_err(struct address_space *mapping)
113 return errseq_sample(&mapping->wb_err);
117 * file_sample_sb_err - sample the current errseq_t to test for later errors
118 * @file: file pointer to be sampled
120 * Grab the most current superblock-level errseq_t value for the given
123 static inline errseq_t file_sample_sb_err(struct file *file)
125 return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err);
129 * Flush file data before changing attributes. Caller must hold any locks
130 * required to prevent further writes to this file until we're done setting
133 static inline int inode_drain_writes(struct inode *inode)
135 inode_dio_wait(inode);
136 return filemap_write_and_wait(inode->i_mapping);
139 static inline bool mapping_empty(struct address_space *mapping)
141 return xa_empty(&mapping->i_pages);
145 * mapping_shrinkable - test if page cache state allows inode reclaim
146 * @mapping: the page cache mapping
148 * This checks the mapping's cache state for the pupose of inode
149 * reclaim and LRU management.
151 * The caller is expected to hold the i_lock, but is not required to
152 * hold the i_pages lock, which usually protects cache state. That's
153 * because the i_lock and the list_lru lock that protect the inode and
154 * its LRU state don't nest inside the irq-safe i_pages lock.
156 * Cache deletions are performed under the i_lock, which ensures that
157 * when an inode goes empty, it will reliably get queued on the LRU.
159 * Cache additions do not acquire the i_lock and may race with this
160 * check, in which case we'll report the inode as shrinkable when it
161 * has cache pages. This is okay: the shrinker also checks the
162 * refcount and the referenced bit, which will be elevated or set in
163 * the process of adding new cache pages to an inode.
165 static inline bool mapping_shrinkable(struct address_space *mapping)
170 * On highmem systems, there could be lowmem pressure from the
171 * inodes before there is highmem pressure from the page
172 * cache. Make inodes shrinkable regardless of cache state.
174 if (IS_ENABLED(CONFIG_HIGHMEM))
177 /* Cache completely empty? Shrink away. */
178 head = rcu_access_pointer(mapping->i_pages.xa_head);
183 * The xarray stores single offset-0 entries directly in the
184 * head pointer, which allows non-resident page cache entries
185 * to escape the shadow shrinker's list of xarray nodes. The
186 * inode shrinker needs to pick them up under memory pressure.
188 if (!xa_is_node(head) && xa_is_value(head))
195 * Bits in mapping->flags.
198 AS_EIO = 0, /* IO error on async write */
199 AS_ENOSPC = 1, /* ENOSPC on async write */
200 AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */
201 AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */
202 AS_EXITING = 4, /* final truncate in progress */
203 /* writeback related tags are not used */
204 AS_NO_WRITEBACK_TAGS = 5,
205 AS_LARGE_FOLIO_SUPPORT = 6,
209 * mapping_set_error - record a writeback error in the address_space
210 * @mapping: the mapping in which an error should be set
211 * @error: the error to set in the mapping
213 * When writeback fails in some way, we must record that error so that
214 * userspace can be informed when fsync and the like are called. We endeavor
215 * to report errors on any file that was open at the time of the error. Some
216 * internal callers also need to know when writeback errors have occurred.
218 * When a writeback error occurs, most filesystems will want to call
219 * mapping_set_error to record the error in the mapping so that it can be
220 * reported when the application calls fsync(2).
222 static inline void mapping_set_error(struct address_space *mapping, int error)
227 /* Record in wb_err for checkers using errseq_t based tracking */
228 __filemap_set_wb_err(mapping, error);
230 /* Record it in superblock */
232 errseq_set(&mapping->host->i_sb->s_wb_err, error);
234 /* Record it in flags for now, for legacy callers */
235 if (error == -ENOSPC)
236 set_bit(AS_ENOSPC, &mapping->flags);
238 set_bit(AS_EIO, &mapping->flags);
241 static inline void mapping_set_unevictable(struct address_space *mapping)
243 set_bit(AS_UNEVICTABLE, &mapping->flags);
246 static inline void mapping_clear_unevictable(struct address_space *mapping)
248 clear_bit(AS_UNEVICTABLE, &mapping->flags);
251 static inline bool mapping_unevictable(struct address_space *mapping)
253 return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
256 static inline void mapping_set_exiting(struct address_space *mapping)
258 set_bit(AS_EXITING, &mapping->flags);
261 static inline int mapping_exiting(struct address_space *mapping)
263 return test_bit(AS_EXITING, &mapping->flags);
266 static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
268 set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
271 static inline int mapping_use_writeback_tags(struct address_space *mapping)
273 return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
276 static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
278 return mapping->gfp_mask;
281 /* Restricts the given gfp_mask to what the mapping allows. */
282 static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
285 return mapping_gfp_mask(mapping) & gfp_mask;
289 * This is non-atomic. Only to be used before the mapping is activated.
290 * Probably needs a barrier...
292 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
298 * mapping_set_large_folios() - Indicate the file supports large folios.
299 * @mapping: The file.
301 * The filesystem should call this function in its inode constructor to
302 * indicate that the VFS can use large folios to cache the contents of
305 * Context: This should not be called while the inode is active as it
308 static inline void mapping_set_large_folios(struct address_space *mapping)
310 __set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
314 * Large folio support currently depends on THP. These dependencies are
315 * being worked on but are not yet fixed.
317 static inline bool mapping_large_folio_support(struct address_space *mapping)
319 return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) &&
320 test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
323 static inline int filemap_nr_thps(struct address_space *mapping)
325 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
326 return atomic_read(&mapping->nr_thps);
332 static inline void filemap_nr_thps_inc(struct address_space *mapping)
334 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
335 if (!mapping_large_folio_support(mapping))
336 atomic_inc(&mapping->nr_thps);
338 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
342 static inline void filemap_nr_thps_dec(struct address_space *mapping)
344 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
345 if (!mapping_large_folio_support(mapping))
346 atomic_dec(&mapping->nr_thps);
348 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
352 struct address_space *page_mapping(struct page *);
353 struct address_space *folio_mapping(struct folio *);
354 struct address_space *swapcache_mapping(struct folio *);
357 * folio_file_mapping - Find the mapping this folio belongs to.
360 * For folios which are in the page cache, return the mapping that this
361 * page belongs to. Folios in the swap cache return the mapping of the
362 * swap file or swap device where the data is stored. This is different
363 * from the mapping returned by folio_mapping(). The only reason to
364 * use it is if, like NFS, you return 0 from ->activate_swapfile.
366 * Do not call this for folios which aren't in the page cache or swap cache.
368 static inline struct address_space *folio_file_mapping(struct folio *folio)
370 if (unlikely(folio_test_swapcache(folio)))
371 return swapcache_mapping(folio);
373 return folio->mapping;
376 static inline struct address_space *page_file_mapping(struct page *page)
378 return folio_file_mapping(page_folio(page));
382 * For file cache pages, return the address_space, otherwise return NULL
384 static inline struct address_space *page_mapping_file(struct page *page)
386 struct folio *folio = page_folio(page);
388 if (unlikely(folio_test_swapcache(folio)))
390 return folio_mapping(folio);
394 * folio_inode - Get the host inode for this folio.
397 * For folios which are in the page cache, return the inode that this folio
400 * Do not call this for folios which aren't in the page cache.
402 static inline struct inode *folio_inode(struct folio *folio)
404 return folio->mapping->host;
408 * folio_attach_private - Attach private data to a folio.
409 * @folio: Folio to attach data to.
410 * @data: Data to attach to folio.
412 * Attaching private data to a folio increments the page's reference count.
413 * The data must be detached before the folio will be freed.
415 static inline void folio_attach_private(struct folio *folio, void *data)
418 folio->private = data;
419 folio_set_private(folio);
423 * folio_change_private - Change private data on a folio.
424 * @folio: Folio to change the data on.
425 * @data: Data to set on the folio.
427 * Change the private data attached to a folio and return the old
428 * data. The page must previously have had data attached and the data
429 * must be detached before the folio will be freed.
431 * Return: Data that was previously attached to the folio.
433 static inline void *folio_change_private(struct folio *folio, void *data)
435 void *old = folio_get_private(folio);
437 folio->private = data;
442 * folio_detach_private - Detach private data from a folio.
443 * @folio: Folio to detach data from.
445 * Removes the data that was previously attached to the folio and decrements
446 * the refcount on the page.
448 * Return: Data that was attached to the folio.
450 static inline void *folio_detach_private(struct folio *folio)
452 void *data = folio_get_private(folio);
454 if (!folio_test_private(folio))
456 folio_clear_private(folio);
457 folio->private = NULL;
463 static inline void attach_page_private(struct page *page, void *data)
465 folio_attach_private(page_folio(page), data);
468 static inline void *detach_page_private(struct page *page)
470 return folio_detach_private(page_folio(page));
474 struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order);
476 static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order)
478 return folio_alloc(gfp, order);
482 static inline struct page *__page_cache_alloc(gfp_t gfp)
484 return &filemap_alloc_folio(gfp, 0)->page;
487 static inline struct page *page_cache_alloc(struct address_space *x)
489 return __page_cache_alloc(mapping_gfp_mask(x));
492 static inline gfp_t readahead_gfp_mask(struct address_space *x)
494 return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
497 typedef int filler_t(struct file *, struct folio *);
499 pgoff_t page_cache_next_miss(struct address_space *mapping,
500 pgoff_t index, unsigned long max_scan);
501 pgoff_t page_cache_prev_miss(struct address_space *mapping,
502 pgoff_t index, unsigned long max_scan);
504 #define FGP_ACCESSED 0x00000001
505 #define FGP_LOCK 0x00000002
506 #define FGP_CREAT 0x00000004
507 #define FGP_WRITE 0x00000008
508 #define FGP_NOFS 0x00000010
509 #define FGP_NOWAIT 0x00000020
510 #define FGP_FOR_MMAP 0x00000040
511 #define FGP_STABLE 0x00000080
513 #define FGP_WRITEBEGIN (FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE)
515 void *filemap_get_entry(struct address_space *mapping, pgoff_t index);
516 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
517 int fgp_flags, gfp_t gfp);
518 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
519 int fgp_flags, gfp_t gfp);
522 * filemap_get_folio - Find and get a folio.
523 * @mapping: The address_space to search.
524 * @index: The page index.
526 * Looks up the page cache entry at @mapping & @index. If a folio is
527 * present, it is returned with an increased refcount.
529 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
530 * this index. Will not return a shadow, swap or DAX entry.
532 static inline struct folio *filemap_get_folio(struct address_space *mapping,
535 return __filemap_get_folio(mapping, index, 0, 0);
539 * filemap_lock_folio - Find and lock a folio.
540 * @mapping: The address_space to search.
541 * @index: The page index.
543 * Looks up the page cache entry at @mapping & @index. If a folio is
544 * present, it is returned locked with an increased refcount.
546 * Context: May sleep.
547 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
548 * this index. Will not return a shadow, swap or DAX entry.
550 static inline struct folio *filemap_lock_folio(struct address_space *mapping,
553 return __filemap_get_folio(mapping, index, FGP_LOCK, 0);
557 * filemap_grab_folio - grab a folio from the page cache
558 * @mapping: The address space to search
559 * @index: The page index
561 * Looks up the page cache entry at @mapping & @index. If no folio is found,
562 * a new folio is created. The folio is locked, marked as accessed, and
565 * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found
566 * and failed to create a folio.
568 static inline struct folio *filemap_grab_folio(struct address_space *mapping,
571 return __filemap_get_folio(mapping, index,
572 FGP_LOCK | FGP_ACCESSED | FGP_CREAT,
573 mapping_gfp_mask(mapping));
577 * find_get_page - find and get a page reference
578 * @mapping: the address_space to search
579 * @offset: the page index
581 * Looks up the page cache slot at @mapping & @offset. If there is a
582 * page cache page, it is returned with an increased refcount.
584 * Otherwise, %NULL is returned.
586 static inline struct page *find_get_page(struct address_space *mapping,
589 return pagecache_get_page(mapping, offset, 0, 0);
592 static inline struct page *find_get_page_flags(struct address_space *mapping,
593 pgoff_t offset, int fgp_flags)
595 return pagecache_get_page(mapping, offset, fgp_flags, 0);
599 * find_lock_page - locate, pin and lock a pagecache page
600 * @mapping: the address_space to search
601 * @index: the page index
603 * Looks up the page cache entry at @mapping & @index. If there is a
604 * page cache page, it is returned locked and with an increased
607 * Context: May sleep.
608 * Return: A struct page or %NULL if there is no page in the cache for this
611 static inline struct page *find_lock_page(struct address_space *mapping,
614 return pagecache_get_page(mapping, index, FGP_LOCK, 0);
618 * find_or_create_page - locate or add a pagecache page
619 * @mapping: the page's address_space
620 * @index: the page's index into the mapping
621 * @gfp_mask: page allocation mode
623 * Looks up the page cache slot at @mapping & @offset. If there is a
624 * page cache page, it is returned locked and with an increased
627 * If the page is not present, a new page is allocated using @gfp_mask
628 * and added to the page cache and the VM's LRU list. The page is
629 * returned locked and with an increased refcount.
631 * On memory exhaustion, %NULL is returned.
633 * find_or_create_page() may sleep, even if @gfp_flags specifies an
636 static inline struct page *find_or_create_page(struct address_space *mapping,
637 pgoff_t index, gfp_t gfp_mask)
639 return pagecache_get_page(mapping, index,
640 FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
645 * grab_cache_page_nowait - returns locked page at given index in given cache
646 * @mapping: target address_space
647 * @index: the page index
649 * Same as grab_cache_page(), but do not wait if the page is unavailable.
650 * This is intended for speculative data generators, where the data can
651 * be regenerated if the page couldn't be grabbed. This routine should
652 * be safe to call while holding the lock for another page.
654 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
655 * and deadlock against the caller's locked page.
657 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
660 return pagecache_get_page(mapping, index,
661 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
662 mapping_gfp_mask(mapping));
665 #define swapcache_index(folio) __page_file_index(&(folio)->page)
668 * folio_index - File index of a folio.
671 * For a folio which is either in the page cache or the swap cache,
672 * return its index within the address_space it belongs to. If you know
673 * the page is definitely in the page cache, you can look at the folio's
676 * Return: The index (offset in units of pages) of a folio in its file.
678 static inline pgoff_t folio_index(struct folio *folio)
680 if (unlikely(folio_test_swapcache(folio)))
681 return swapcache_index(folio);
686 * folio_next_index - Get the index of the next folio.
687 * @folio: The current folio.
689 * Return: The index of the folio which follows this folio in the file.
691 static inline pgoff_t folio_next_index(struct folio *folio)
693 return folio->index + folio_nr_pages(folio);
697 * folio_file_page - The page for a particular index.
698 * @folio: The folio which contains this index.
699 * @index: The index we want to look up.
701 * Sometimes after looking up a folio in the page cache, we need to
702 * obtain the specific page for an index (eg a page fault).
704 * Return: The page containing the file data for this index.
706 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
708 /* HugeTLBfs indexes the page cache in units of hpage_size */
709 if (folio_test_hugetlb(folio))
711 return folio_page(folio, index & (folio_nr_pages(folio) - 1));
715 * folio_contains - Does this folio contain this index?
717 * @index: The page index within the file.
719 * Context: The caller should have the page locked in order to prevent
720 * (eg) shmem from moving the page between the page cache and swap cache
721 * and changing its index in the middle of the operation.
722 * Return: true or false.
724 static inline bool folio_contains(struct folio *folio, pgoff_t index)
726 /* HugeTLBfs indexes the page cache in units of hpage_size */
727 if (folio_test_hugetlb(folio))
728 return folio->index == index;
729 return index - folio_index(folio) < folio_nr_pages(folio);
733 * Given the page we found in the page cache, return the page corresponding
734 * to this index in the file
736 static inline struct page *find_subpage(struct page *head, pgoff_t index)
738 /* HugeTLBfs wants the head page regardless */
742 return head + (index & (thp_nr_pages(head) - 1));
745 unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start,
746 pgoff_t end, struct folio_batch *fbatch);
747 unsigned filemap_get_folios_contig(struct address_space *mapping,
748 pgoff_t *start, pgoff_t end, struct folio_batch *fbatch);
749 unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start,
750 pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch);
752 struct page *grab_cache_page_write_begin(struct address_space *mapping,
756 * Returns locked page at given index in given cache, creating it if needed.
758 static inline struct page *grab_cache_page(struct address_space *mapping,
761 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
764 struct folio *read_cache_folio(struct address_space *, pgoff_t index,
765 filler_t *filler, struct file *file);
766 struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index,
768 struct page *read_cache_page(struct address_space *, pgoff_t index,
769 filler_t *filler, struct file *file);
770 extern struct page * read_cache_page_gfp(struct address_space *mapping,
771 pgoff_t index, gfp_t gfp_mask);
773 static inline struct page *read_mapping_page(struct address_space *mapping,
774 pgoff_t index, struct file *file)
776 return read_cache_page(mapping, index, NULL, file);
779 static inline struct folio *read_mapping_folio(struct address_space *mapping,
780 pgoff_t index, struct file *file)
782 return read_cache_folio(mapping, index, NULL, file);
786 * Get index of the page within radix-tree (but not for hugetlb pages).
787 * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
789 static inline pgoff_t page_to_index(struct page *page)
793 if (likely(!PageTransTail(page)))
796 head = compound_head(page);
798 * We don't initialize ->index for tail pages: calculate based on
801 return head->index + page - head;
804 extern pgoff_t hugetlb_basepage_index(struct page *page);
807 * Get the offset in PAGE_SIZE (even for hugetlb pages).
808 * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
810 static inline pgoff_t page_to_pgoff(struct page *page)
812 if (unlikely(PageHuge(page)))
813 return hugetlb_basepage_index(page);
814 return page_to_index(page);
818 * Return byte-offset into filesystem object for page.
820 static inline loff_t page_offset(struct page *page)
822 return ((loff_t)page->index) << PAGE_SHIFT;
825 static inline loff_t page_file_offset(struct page *page)
827 return ((loff_t)page_index(page)) << PAGE_SHIFT;
831 * folio_pos - Returns the byte position of this folio in its file.
834 static inline loff_t folio_pos(struct folio *folio)
836 return page_offset(&folio->page);
840 * folio_file_pos - Returns the byte position of this folio in its file.
843 * This differs from folio_pos() for folios which belong to a swap file.
844 * NFS is the only filesystem today which needs to use folio_file_pos().
846 static inline loff_t folio_file_pos(struct folio *folio)
848 return page_file_offset(&folio->page);
852 * Get the offset in PAGE_SIZE (even for hugetlb folios).
853 * (TODO: hugetlb folios should have ->index in PAGE_SIZE)
855 static inline pgoff_t folio_pgoff(struct folio *folio)
857 if (unlikely(folio_test_hugetlb(folio)))
858 return hugetlb_basepage_index(&folio->page);
862 extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
863 unsigned long address);
865 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
866 unsigned long address)
869 if (unlikely(is_vm_hugetlb_page(vma)))
870 return linear_hugepage_index(vma, address);
871 pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
872 pgoff += vma->vm_pgoff;
876 struct wait_page_key {
882 struct wait_page_queue {
885 wait_queue_entry_t wait;
888 static inline bool wake_page_match(struct wait_page_queue *wait_page,
889 struct wait_page_key *key)
891 if (wait_page->folio != key->folio)
895 if (wait_page->bit_nr != key->bit_nr)
901 void __folio_lock(struct folio *folio);
902 int __folio_lock_killable(struct folio *folio);
903 bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
905 void unlock_page(struct page *page);
906 void folio_unlock(struct folio *folio);
909 * folio_trylock() - Attempt to lock a folio.
910 * @folio: The folio to attempt to lock.
912 * Sometimes it is undesirable to wait for a folio to be unlocked (eg
913 * when the locks are being taken in the wrong order, or if making
914 * progress through a batch of folios is more important than processing
915 * them in order). Usually folio_lock() is the correct function to call.
917 * Context: Any context.
918 * Return: Whether the lock was successfully acquired.
920 static inline bool folio_trylock(struct folio *folio)
922 return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
926 * Return true if the page was successfully locked
928 static inline int trylock_page(struct page *page)
930 return folio_trylock(page_folio(page));
934 * folio_lock() - Lock this folio.
935 * @folio: The folio to lock.
937 * The folio lock protects against many things, probably more than it
938 * should. It is primarily held while a folio is being brought uptodate,
939 * either from its backing file or from swap. It is also held while a
940 * folio is being truncated from its address_space, so holding the lock
941 * is sufficient to keep folio->mapping stable.
943 * The folio lock is also held while write() is modifying the page to
944 * provide POSIX atomicity guarantees (as long as the write does not
945 * cross a page boundary). Other modifications to the data in the folio
946 * do not hold the folio lock and can race with writes, eg DMA and stores
949 * Context: May sleep. If you need to acquire the locks of two or
950 * more folios, they must be in order of ascending index, if they are
951 * in the same address_space. If they are in different address_spaces,
952 * acquire the lock of the folio which belongs to the address_space which
953 * has the lowest address in memory first.
955 static inline void folio_lock(struct folio *folio)
958 if (!folio_trylock(folio))
963 * lock_page() - Lock the folio containing this page.
964 * @page: The page to lock.
966 * See folio_lock() for a description of what the lock protects.
967 * This is a legacy function and new code should probably use folio_lock()
970 * Context: May sleep. Pages in the same folio share a lock, so do not
971 * attempt to lock two pages which share a folio.
973 static inline void lock_page(struct page *page)
978 folio = page_folio(page);
979 if (!folio_trylock(folio))
984 * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
985 * @folio: The folio to lock.
987 * Attempts to lock the folio, like folio_lock(), except that the sleep
988 * to acquire the lock is interruptible by a fatal signal.
990 * Context: May sleep; see folio_lock().
991 * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
993 static inline int folio_lock_killable(struct folio *folio)
996 if (!folio_trylock(folio))
997 return __folio_lock_killable(folio);
1002 * folio_lock_or_retry - Lock the folio, unless this would block and the
1003 * caller indicated that it can handle a retry.
1005 * Return value and mmap_lock implications depend on flags; see
1006 * __folio_lock_or_retry().
1008 static inline bool folio_lock_or_retry(struct folio *folio,
1009 struct mm_struct *mm, unsigned int flags)
1012 return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
1016 * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
1017 * and should not be used directly.
1019 void folio_wait_bit(struct folio *folio, int bit_nr);
1020 int folio_wait_bit_killable(struct folio *folio, int bit_nr);
1023 * Wait for a folio to be unlocked.
1025 * This must be called with the caller "holding" the folio,
1026 * ie with increased folio reference count so that the folio won't
1027 * go away during the wait.
1029 static inline void folio_wait_locked(struct folio *folio)
1031 if (folio_test_locked(folio))
1032 folio_wait_bit(folio, PG_locked);
1035 static inline int folio_wait_locked_killable(struct folio *folio)
1037 if (!folio_test_locked(folio))
1039 return folio_wait_bit_killable(folio, PG_locked);
1042 static inline void wait_on_page_locked(struct page *page)
1044 folio_wait_locked(page_folio(page));
1047 static inline int wait_on_page_locked_killable(struct page *page)
1049 return folio_wait_locked_killable(page_folio(page));
1052 void wait_on_page_writeback(struct page *page);
1053 void folio_wait_writeback(struct folio *folio);
1054 int folio_wait_writeback_killable(struct folio *folio);
1055 void end_page_writeback(struct page *page);
1056 void folio_end_writeback(struct folio *folio);
1057 void wait_for_stable_page(struct page *page);
1058 void folio_wait_stable(struct folio *folio);
1059 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
1060 static inline void __set_page_dirty(struct page *page,
1061 struct address_space *mapping, int warn)
1063 __folio_mark_dirty(page_folio(page), mapping, warn);
1065 void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
1066 void __folio_cancel_dirty(struct folio *folio);
1067 static inline void folio_cancel_dirty(struct folio *folio)
1069 /* Avoid atomic ops, locking, etc. when not actually needed. */
1070 if (folio_test_dirty(folio))
1071 __folio_cancel_dirty(folio);
1073 bool folio_clear_dirty_for_io(struct folio *folio);
1074 bool clear_page_dirty_for_io(struct page *page);
1075 void folio_invalidate(struct folio *folio, size_t offset, size_t length);
1076 int __set_page_dirty_nobuffers(struct page *page);
1077 bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
1079 #ifdef CONFIG_MIGRATION
1080 int filemap_migrate_folio(struct address_space *mapping, struct folio *dst,
1081 struct folio *src, enum migrate_mode mode);
1083 #define filemap_migrate_folio NULL
1085 void folio_end_private_2(struct folio *folio);
1086 void folio_wait_private_2(struct folio *folio);
1087 int folio_wait_private_2_killable(struct folio *folio);
1090 * Add an arbitrary waiter to a page's wait queue
1092 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
1095 * Fault in userspace address range.
1097 size_t fault_in_writeable(char __user *uaddr, size_t size);
1098 size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
1099 size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
1100 size_t fault_in_readable(const char __user *uaddr, size_t size);
1102 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
1103 pgoff_t index, gfp_t gfp);
1104 int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1105 pgoff_t index, gfp_t gfp);
1106 void filemap_remove_folio(struct folio *folio);
1107 void __filemap_remove_folio(struct folio *folio, void *shadow);
1108 void replace_page_cache_folio(struct folio *old, struct folio *new);
1109 void delete_from_page_cache_batch(struct address_space *mapping,
1110 struct folio_batch *fbatch);
1111 bool filemap_release_folio(struct folio *folio, gfp_t gfp);
1112 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1115 /* Must be non-static for BPF error injection */
1116 int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1117 pgoff_t index, gfp_t gfp, void **shadowp);
1119 bool filemap_range_has_writeback(struct address_space *mapping,
1120 loff_t start_byte, loff_t end_byte);
1123 * filemap_range_needs_writeback - check if range potentially needs writeback
1124 * @mapping: address space within which to check
1125 * @start_byte: offset in bytes where the range starts
1126 * @end_byte: offset in bytes where the range ends (inclusive)
1128 * Find at least one page in the range supplied, usually used to check if
1129 * direct writing in this range will trigger a writeback. Used by O_DIRECT
1130 * read/write with IOCB_NOWAIT, to see if the caller needs to do
1131 * filemap_write_and_wait_range() before proceeding.
1133 * Return: %true if the caller should do filemap_write_and_wait_range() before
1134 * doing O_DIRECT to a page in this range, %false otherwise.
1136 static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1140 if (!mapping->nrpages)
1142 if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1143 !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1145 return filemap_range_has_writeback(mapping, start_byte, end_byte);
1149 * struct readahead_control - Describes a readahead request.
1151 * A readahead request is for consecutive pages. Filesystems which
1152 * implement the ->readahead method should call readahead_page() or
1153 * readahead_page_batch() in a loop and attempt to start I/O against
1154 * each page in the request.
1156 * Most of the fields in this struct are private and should be accessed
1157 * by the functions below.
1159 * @file: The file, used primarily by network filesystems for authentication.
1160 * May be NULL if invoked internally by the filesystem.
1161 * @mapping: Readahead this filesystem object.
1162 * @ra: File readahead state. May be NULL.
1164 struct readahead_control {
1166 struct address_space *mapping;
1167 struct file_ra_state *ra;
1168 /* private: use the readahead_* accessors instead */
1170 unsigned int _nr_pages;
1171 unsigned int _batch_count;
1173 unsigned long _pflags;
1176 #define DEFINE_READAHEAD(ractl, f, r, m, i) \
1177 struct readahead_control ractl = { \
1184 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
1186 void page_cache_ra_unbounded(struct readahead_control *,
1187 unsigned long nr_to_read, unsigned long lookahead_count);
1188 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
1189 void page_cache_async_ra(struct readahead_control *, struct folio *,
1190 unsigned long req_count);
1191 void readahead_expand(struct readahead_control *ractl,
1192 loff_t new_start, size_t new_len);
1195 * page_cache_sync_readahead - generic file readahead
1196 * @mapping: address_space which holds the pagecache and I/O vectors
1197 * @ra: file_ra_state which holds the readahead state
1198 * @file: Used by the filesystem for authentication.
1199 * @index: Index of first page to be read.
1200 * @req_count: Total number of pages being read by the caller.
1202 * page_cache_sync_readahead() should be called when a cache miss happened:
1203 * it will submit the read. The readahead logic may decide to piggyback more
1204 * pages onto the read request if access patterns suggest it will improve
1208 void page_cache_sync_readahead(struct address_space *mapping,
1209 struct file_ra_state *ra, struct file *file, pgoff_t index,
1210 unsigned long req_count)
1212 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1213 page_cache_sync_ra(&ractl, req_count);
1217 * page_cache_async_readahead - file readahead for marked pages
1218 * @mapping: address_space which holds the pagecache and I/O vectors
1219 * @ra: file_ra_state which holds the readahead state
1220 * @file: Used by the filesystem for authentication.
1221 * @folio: The folio at @index which triggered the readahead call.
1222 * @index: Index of first page to be read.
1223 * @req_count: Total number of pages being read by the caller.
1225 * page_cache_async_readahead() should be called when a page is used which
1226 * is marked as PageReadahead; this is a marker to suggest that the application
1227 * has used up enough of the readahead window that we should start pulling in
1231 void page_cache_async_readahead(struct address_space *mapping,
1232 struct file_ra_state *ra, struct file *file,
1233 struct folio *folio, pgoff_t index, unsigned long req_count)
1235 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1236 page_cache_async_ra(&ractl, folio, req_count);
1239 static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1241 struct folio *folio;
1243 BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1244 ractl->_nr_pages -= ractl->_batch_count;
1245 ractl->_index += ractl->_batch_count;
1247 if (!ractl->_nr_pages) {
1248 ractl->_batch_count = 0;
1252 folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1253 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1254 ractl->_batch_count = folio_nr_pages(folio);
1260 * readahead_page - Get the next page to read.
1261 * @ractl: The current readahead request.
1263 * Context: The page is locked and has an elevated refcount. The caller
1264 * should decreases the refcount once the page has been submitted for I/O
1265 * and unlock the page once all I/O to that page has completed.
1266 * Return: A pointer to the next page, or %NULL if we are done.
1268 static inline struct page *readahead_page(struct readahead_control *ractl)
1270 struct folio *folio = __readahead_folio(ractl);
1272 return &folio->page;
1276 * readahead_folio - Get the next folio to read.
1277 * @ractl: The current readahead request.
1279 * Context: The folio is locked. The caller should unlock the folio once
1280 * all I/O to that folio has completed.
1281 * Return: A pointer to the next folio, or %NULL if we are done.
1283 static inline struct folio *readahead_folio(struct readahead_control *ractl)
1285 struct folio *folio = __readahead_folio(ractl);
1292 static inline unsigned int __readahead_batch(struct readahead_control *rac,
1293 struct page **array, unsigned int array_sz)
1296 XA_STATE(xas, &rac->mapping->i_pages, 0);
1299 BUG_ON(rac->_batch_count > rac->_nr_pages);
1300 rac->_nr_pages -= rac->_batch_count;
1301 rac->_index += rac->_batch_count;
1302 rac->_batch_count = 0;
1304 xas_set(&xas, rac->_index);
1306 xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1307 if (xas_retry(&xas, page))
1309 VM_BUG_ON_PAGE(!PageLocked(page), page);
1310 VM_BUG_ON_PAGE(PageTail(page), page);
1312 rac->_batch_count += thp_nr_pages(page);
1322 * readahead_page_batch - Get a batch of pages to read.
1323 * @rac: The current readahead request.
1324 * @array: An array of pointers to struct page.
1326 * Context: The pages are locked and have an elevated refcount. The caller
1327 * should decreases the refcount once the page has been submitted for I/O
1328 * and unlock the page once all I/O to that page has completed.
1329 * Return: The number of pages placed in the array. 0 indicates the request
1332 #define readahead_page_batch(rac, array) \
1333 __readahead_batch(rac, array, ARRAY_SIZE(array))
1336 * readahead_pos - The byte offset into the file of this readahead request.
1337 * @rac: The readahead request.
1339 static inline loff_t readahead_pos(struct readahead_control *rac)
1341 return (loff_t)rac->_index * PAGE_SIZE;
1345 * readahead_length - The number of bytes in this readahead request.
1346 * @rac: The readahead request.
1348 static inline size_t readahead_length(struct readahead_control *rac)
1350 return rac->_nr_pages * PAGE_SIZE;
1354 * readahead_index - The index of the first page in this readahead request.
1355 * @rac: The readahead request.
1357 static inline pgoff_t readahead_index(struct readahead_control *rac)
1363 * readahead_count - The number of pages in this readahead request.
1364 * @rac: The readahead request.
1366 static inline unsigned int readahead_count(struct readahead_control *rac)
1368 return rac->_nr_pages;
1372 * readahead_batch_length - The number of bytes in the current batch.
1373 * @rac: The readahead request.
1375 static inline size_t readahead_batch_length(struct readahead_control *rac)
1377 return rac->_batch_count * PAGE_SIZE;
1380 static inline unsigned long dir_pages(struct inode *inode)
1382 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1387 * folio_mkwrite_check_truncate - check if folio was truncated
1388 * @folio: the folio to check
1389 * @inode: the inode to check the folio against
1391 * Return: the number of bytes in the folio up to EOF,
1392 * or -EFAULT if the folio was truncated.
1394 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1395 struct inode *inode)
1397 loff_t size = i_size_read(inode);
1398 pgoff_t index = size >> PAGE_SHIFT;
1399 size_t offset = offset_in_folio(folio, size);
1401 if (!folio->mapping)
1404 /* folio is wholly inside EOF */
1405 if (folio_next_index(folio) - 1 < index)
1406 return folio_size(folio);
1407 /* folio is wholly past EOF */
1408 if (folio->index > index || !offset)
1410 /* folio is partially inside EOF */
1415 * page_mkwrite_check_truncate - check if page was truncated
1416 * @page: the page to check
1417 * @inode: the inode to check the page against
1419 * Returns the number of bytes in the page up to EOF,
1420 * or -EFAULT if the page was truncated.
1422 static inline int page_mkwrite_check_truncate(struct page *page,
1423 struct inode *inode)
1425 loff_t size = i_size_read(inode);
1426 pgoff_t index = size >> PAGE_SHIFT;
1427 int offset = offset_in_page(size);
1429 if (page->mapping != inode->i_mapping)
1432 /* page is wholly inside EOF */
1433 if (page->index < index)
1435 /* page is wholly past EOF */
1436 if (page->index > index || !offset)
1438 /* page is partially inside EOF */
1443 * i_blocks_per_folio - How many blocks fit in this folio.
1444 * @inode: The inode which contains the blocks.
1445 * @folio: The folio.
1447 * If the block size is larger than the size of this folio, return zero.
1449 * Context: The caller should hold a refcount on the folio to prevent it
1451 * Return: The number of filesystem blocks covered by this folio.
1454 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1456 return folio_size(folio) >> inode->i_blkbits;
1460 unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1462 return i_blocks_per_folio(inode, page_folio(page));
1464 #endif /* _LINUX_PAGEMAP_H */