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 write_inode_now(struct inode *, int sync);
34 int filemap_fdatawrite(struct address_space *);
35 int filemap_flush(struct address_space *);
36 int filemap_fdatawait_keep_errors(struct address_space *mapping);
37 int filemap_fdatawait_range(struct address_space *, loff_t lstart, loff_t lend);
38 int filemap_fdatawait_range_keep_errors(struct address_space *mapping,
39 loff_t start_byte, loff_t end_byte);
41 static inline int filemap_fdatawait(struct address_space *mapping)
43 return filemap_fdatawait_range(mapping, 0, LLONG_MAX);
46 bool filemap_range_has_page(struct address_space *, loff_t lstart, loff_t lend);
47 int filemap_write_and_wait_range(struct address_space *mapping,
48 loff_t lstart, loff_t lend);
49 int __filemap_fdatawrite_range(struct address_space *mapping,
50 loff_t start, loff_t end, int sync_mode);
51 int filemap_fdatawrite_range(struct address_space *mapping,
52 loff_t start, loff_t end);
53 int filemap_check_errors(struct address_space *mapping);
54 void __filemap_set_wb_err(struct address_space *mapping, int err);
55 int filemap_fdatawrite_wbc(struct address_space *mapping,
56 struct writeback_control *wbc);
58 static inline int filemap_write_and_wait(struct address_space *mapping)
60 return filemap_write_and_wait_range(mapping, 0, LLONG_MAX);
64 * filemap_set_wb_err - set a writeback error on an address_space
65 * @mapping: mapping in which to set writeback error
66 * @err: error to be set in mapping
68 * When writeback fails in some way, we must record that error so that
69 * userspace can be informed when fsync and the like are called. We endeavor
70 * to report errors on any file that was open at the time of the error. Some
71 * internal callers also need to know when writeback errors have occurred.
73 * When a writeback error occurs, most filesystems will want to call
74 * filemap_set_wb_err to record the error in the mapping so that it will be
75 * automatically reported whenever fsync is called on the file.
77 static inline void filemap_set_wb_err(struct address_space *mapping, int err)
79 /* Fastpath for common case of no error */
81 __filemap_set_wb_err(mapping, err);
85 * filemap_check_wb_err - has an error occurred since the mark was sampled?
86 * @mapping: mapping to check for writeback errors
87 * @since: previously-sampled errseq_t
89 * Grab the errseq_t value from the mapping, and see if it has changed "since"
90 * the given value was sampled.
92 * If it has then report the latest error set, otherwise return 0.
94 static inline int filemap_check_wb_err(struct address_space *mapping,
97 return errseq_check(&mapping->wb_err, since);
101 * filemap_sample_wb_err - sample the current errseq_t to test for later errors
102 * @mapping: mapping to be sampled
104 * Writeback errors are always reported relative to a particular sample point
105 * in the past. This function provides those sample points.
107 static inline errseq_t filemap_sample_wb_err(struct address_space *mapping)
109 return errseq_sample(&mapping->wb_err);
113 * file_sample_sb_err - sample the current errseq_t to test for later errors
114 * @file: file pointer to be sampled
116 * Grab the most current superblock-level errseq_t value for the given
119 static inline errseq_t file_sample_sb_err(struct file *file)
121 return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err);
125 * Flush file data before changing attributes. Caller must hold any locks
126 * required to prevent further writes to this file until we're done setting
129 static inline int inode_drain_writes(struct inode *inode)
131 inode_dio_wait(inode);
132 return filemap_write_and_wait(inode->i_mapping);
135 static inline bool mapping_empty(struct address_space *mapping)
137 return xa_empty(&mapping->i_pages);
141 * mapping_shrinkable - test if page cache state allows inode reclaim
142 * @mapping: the page cache mapping
144 * This checks the mapping's cache state for the pupose of inode
145 * reclaim and LRU management.
147 * The caller is expected to hold the i_lock, but is not required to
148 * hold the i_pages lock, which usually protects cache state. That's
149 * because the i_lock and the list_lru lock that protect the inode and
150 * its LRU state don't nest inside the irq-safe i_pages lock.
152 * Cache deletions are performed under the i_lock, which ensures that
153 * when an inode goes empty, it will reliably get queued on the LRU.
155 * Cache additions do not acquire the i_lock and may race with this
156 * check, in which case we'll report the inode as shrinkable when it
157 * has cache pages. This is okay: the shrinker also checks the
158 * refcount and the referenced bit, which will be elevated or set in
159 * the process of adding new cache pages to an inode.
161 static inline bool mapping_shrinkable(struct address_space *mapping)
166 * On highmem systems, there could be lowmem pressure from the
167 * inodes before there is highmem pressure from the page
168 * cache. Make inodes shrinkable regardless of cache state.
170 if (IS_ENABLED(CONFIG_HIGHMEM))
173 /* Cache completely empty? Shrink away. */
174 head = rcu_access_pointer(mapping->i_pages.xa_head);
179 * The xarray stores single offset-0 entries directly in the
180 * head pointer, which allows non-resident page cache entries
181 * to escape the shadow shrinker's list of xarray nodes. The
182 * inode shrinker needs to pick them up under memory pressure.
184 if (!xa_is_node(head) && xa_is_value(head))
191 * Bits in mapping->flags.
194 AS_EIO = 0, /* IO error on async write */
195 AS_ENOSPC = 1, /* ENOSPC on async write */
196 AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */
197 AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */
198 AS_EXITING = 4, /* final truncate in progress */
199 /* writeback related tags are not used */
200 AS_NO_WRITEBACK_TAGS = 5,
201 AS_LARGE_FOLIO_SUPPORT = 6,
205 * mapping_set_error - record a writeback error in the address_space
206 * @mapping: the mapping in which an error should be set
207 * @error: the error to set in the mapping
209 * When writeback fails in some way, we must record that error so that
210 * userspace can be informed when fsync and the like are called. We endeavor
211 * to report errors on any file that was open at the time of the error. Some
212 * internal callers also need to know when writeback errors have occurred.
214 * When a writeback error occurs, most filesystems will want to call
215 * mapping_set_error to record the error in the mapping so that it can be
216 * reported when the application calls fsync(2).
218 static inline void mapping_set_error(struct address_space *mapping, int error)
223 /* Record in wb_err for checkers using errseq_t based tracking */
224 __filemap_set_wb_err(mapping, error);
226 /* Record it in superblock */
228 errseq_set(&mapping->host->i_sb->s_wb_err, error);
230 /* Record it in flags for now, for legacy callers */
231 if (error == -ENOSPC)
232 set_bit(AS_ENOSPC, &mapping->flags);
234 set_bit(AS_EIO, &mapping->flags);
237 static inline void mapping_set_unevictable(struct address_space *mapping)
239 set_bit(AS_UNEVICTABLE, &mapping->flags);
242 static inline void mapping_clear_unevictable(struct address_space *mapping)
244 clear_bit(AS_UNEVICTABLE, &mapping->flags);
247 static inline bool mapping_unevictable(struct address_space *mapping)
249 return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
252 static inline void mapping_set_exiting(struct address_space *mapping)
254 set_bit(AS_EXITING, &mapping->flags);
257 static inline int mapping_exiting(struct address_space *mapping)
259 return test_bit(AS_EXITING, &mapping->flags);
262 static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
264 set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
267 static inline int mapping_use_writeback_tags(struct address_space *mapping)
269 return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
272 static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
274 return mapping->gfp_mask;
277 /* Restricts the given gfp_mask to what the mapping allows. */
278 static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
281 return mapping_gfp_mask(mapping) & gfp_mask;
285 * This is non-atomic. Only to be used before the mapping is activated.
286 * Probably needs a barrier...
288 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
294 * mapping_set_large_folios() - Indicate the file supports large folios.
295 * @mapping: The file.
297 * The filesystem should call this function in its inode constructor to
298 * indicate that the VFS can use large folios to cache the contents of
301 * Context: This should not be called while the inode is active as it
304 static inline void mapping_set_large_folios(struct address_space *mapping)
306 __set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
310 * Large folio support currently depends on THP. These dependencies are
311 * being worked on but are not yet fixed.
313 static inline bool mapping_large_folio_support(struct address_space *mapping)
315 return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) &&
316 test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
319 static inline int filemap_nr_thps(struct address_space *mapping)
321 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
322 return atomic_read(&mapping->nr_thps);
328 static inline void filemap_nr_thps_inc(struct address_space *mapping)
330 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
331 if (!mapping_large_folio_support(mapping))
332 atomic_inc(&mapping->nr_thps);
334 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
338 static inline void filemap_nr_thps_dec(struct address_space *mapping)
340 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
341 if (!mapping_large_folio_support(mapping))
342 atomic_dec(&mapping->nr_thps);
344 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
348 struct address_space *page_mapping(struct page *);
349 struct address_space *folio_mapping(struct folio *);
350 struct address_space *swapcache_mapping(struct folio *);
353 * folio_file_mapping - Find the mapping this folio belongs to.
356 * For folios which are in the page cache, return the mapping that this
357 * page belongs to. Folios in the swap cache return the mapping of the
358 * swap file or swap device where the data is stored. This is different
359 * from the mapping returned by folio_mapping(). The only reason to
360 * use it is if, like NFS, you return 0 from ->activate_swapfile.
362 * Do not call this for folios which aren't in the page cache or swap cache.
364 static inline struct address_space *folio_file_mapping(struct folio *folio)
366 if (unlikely(folio_test_swapcache(folio)))
367 return swapcache_mapping(folio);
369 return folio->mapping;
372 static inline struct address_space *page_file_mapping(struct page *page)
374 return folio_file_mapping(page_folio(page));
378 * For file cache pages, return the address_space, otherwise return NULL
380 static inline struct address_space *page_mapping_file(struct page *page)
382 struct folio *folio = page_folio(page);
384 if (unlikely(folio_test_swapcache(folio)))
386 return folio_mapping(folio);
390 * folio_inode - Get the host inode for this folio.
393 * For folios which are in the page cache, return the inode that this folio
396 * Do not call this for folios which aren't in the page cache.
398 static inline struct inode *folio_inode(struct folio *folio)
400 return folio->mapping->host;
404 * folio_attach_private - Attach private data to a folio.
405 * @folio: Folio to attach data to.
406 * @data: Data to attach to folio.
408 * Attaching private data to a folio increments the page's reference count.
409 * The data must be detached before the folio will be freed.
411 static inline void folio_attach_private(struct folio *folio, void *data)
414 folio->private = data;
415 folio_set_private(folio);
419 * folio_change_private - Change private data on a folio.
420 * @folio: Folio to change the data on.
421 * @data: Data to set on the folio.
423 * Change the private data attached to a folio and return the old
424 * data. The page must previously have had data attached and the data
425 * must be detached before the folio will be freed.
427 * Return: Data that was previously attached to the folio.
429 static inline void *folio_change_private(struct folio *folio, void *data)
431 void *old = folio_get_private(folio);
433 folio->private = data;
438 * folio_detach_private - Detach private data from a folio.
439 * @folio: Folio to detach data from.
441 * Removes the data that was previously attached to the folio and decrements
442 * the refcount on the page.
444 * Return: Data that was attached to the folio.
446 static inline void *folio_detach_private(struct folio *folio)
448 void *data = folio_get_private(folio);
450 if (!folio_test_private(folio))
452 folio_clear_private(folio);
453 folio->private = NULL;
459 static inline void attach_page_private(struct page *page, void *data)
461 folio_attach_private(page_folio(page), data);
464 static inline void *detach_page_private(struct page *page)
466 return folio_detach_private(page_folio(page));
470 struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order);
472 static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order)
474 return folio_alloc(gfp, order);
478 static inline struct page *__page_cache_alloc(gfp_t gfp)
480 return &filemap_alloc_folio(gfp, 0)->page;
483 static inline struct page *page_cache_alloc(struct address_space *x)
485 return __page_cache_alloc(mapping_gfp_mask(x));
488 static inline gfp_t readahead_gfp_mask(struct address_space *x)
490 return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
493 typedef int filler_t(struct file *, struct folio *);
495 pgoff_t page_cache_next_miss(struct address_space *mapping,
496 pgoff_t index, unsigned long max_scan);
497 pgoff_t page_cache_prev_miss(struct address_space *mapping,
498 pgoff_t index, unsigned long max_scan);
500 #define FGP_ACCESSED 0x00000001
501 #define FGP_LOCK 0x00000002
502 #define FGP_CREAT 0x00000004
503 #define FGP_WRITE 0x00000008
504 #define FGP_NOFS 0x00000010
505 #define FGP_NOWAIT 0x00000020
506 #define FGP_FOR_MMAP 0x00000040
507 #define FGP_ENTRY 0x00000080
508 #define FGP_STABLE 0x00000100
510 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
511 int fgp_flags, gfp_t gfp);
512 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
513 int fgp_flags, gfp_t gfp);
516 * filemap_get_folio - Find and get a folio.
517 * @mapping: The address_space to search.
518 * @index: The page index.
520 * Looks up the page cache entry at @mapping & @index. If a folio is
521 * present, it is returned with an increased refcount.
523 * Otherwise, %NULL is returned.
525 static inline struct folio *filemap_get_folio(struct address_space *mapping,
528 return __filemap_get_folio(mapping, index, 0, 0);
532 * filemap_lock_folio - Find and lock a folio.
533 * @mapping: The address_space to search.
534 * @index: The page index.
536 * Looks up the page cache entry at @mapping & @index. If a folio is
537 * present, it is returned locked with an increased refcount.
539 * Context: May sleep.
540 * Return: A folio or %NULL if there is no folio in the cache for this
541 * index. Will not return a shadow, swap or DAX entry.
543 static inline struct folio *filemap_lock_folio(struct address_space *mapping,
546 return __filemap_get_folio(mapping, index, FGP_LOCK, 0);
550 * find_get_page - find and get a page reference
551 * @mapping: the address_space to search
552 * @offset: the page index
554 * Looks up the page cache slot at @mapping & @offset. If there is a
555 * page cache page, it is returned with an increased refcount.
557 * Otherwise, %NULL is returned.
559 static inline struct page *find_get_page(struct address_space *mapping,
562 return pagecache_get_page(mapping, offset, 0, 0);
565 static inline struct page *find_get_page_flags(struct address_space *mapping,
566 pgoff_t offset, int fgp_flags)
568 return pagecache_get_page(mapping, offset, fgp_flags, 0);
572 * find_lock_page - locate, pin and lock a pagecache page
573 * @mapping: the address_space to search
574 * @index: the page index
576 * Looks up the page cache entry at @mapping & @index. If there is a
577 * page cache page, it is returned locked and with an increased
580 * Context: May sleep.
581 * Return: A struct page or %NULL if there is no page in the cache for this
584 static inline struct page *find_lock_page(struct address_space *mapping,
587 return pagecache_get_page(mapping, index, FGP_LOCK, 0);
591 * find_or_create_page - locate or add a pagecache page
592 * @mapping: the page's address_space
593 * @index: the page's index into the mapping
594 * @gfp_mask: page allocation mode
596 * Looks up the page cache slot at @mapping & @offset. If there is a
597 * page cache page, it is returned locked and with an increased
600 * If the page is not present, a new page is allocated using @gfp_mask
601 * and added to the page cache and the VM's LRU list. The page is
602 * returned locked and with an increased refcount.
604 * On memory exhaustion, %NULL is returned.
606 * find_or_create_page() may sleep, even if @gfp_flags specifies an
609 static inline struct page *find_or_create_page(struct address_space *mapping,
610 pgoff_t index, gfp_t gfp_mask)
612 return pagecache_get_page(mapping, index,
613 FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
618 * grab_cache_page_nowait - returns locked page at given index in given cache
619 * @mapping: target address_space
620 * @index: the page index
622 * Same as grab_cache_page(), but do not wait if the page is unavailable.
623 * This is intended for speculative data generators, where the data can
624 * be regenerated if the page couldn't be grabbed. This routine should
625 * be safe to call while holding the lock for another page.
627 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
628 * and deadlock against the caller's locked page.
630 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
633 return pagecache_get_page(mapping, index,
634 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
635 mapping_gfp_mask(mapping));
638 #define swapcache_index(folio) __page_file_index(&(folio)->page)
641 * folio_index - File index of a folio.
644 * For a folio which is either in the page cache or the swap cache,
645 * return its index within the address_space it belongs to. If you know
646 * the page is definitely in the page cache, you can look at the folio's
649 * Return: The index (offset in units of pages) of a folio in its file.
651 static inline pgoff_t folio_index(struct folio *folio)
653 if (unlikely(folio_test_swapcache(folio)))
654 return swapcache_index(folio);
659 * folio_next_index - Get the index of the next folio.
660 * @folio: The current folio.
662 * Return: The index of the folio which follows this folio in the file.
664 static inline pgoff_t folio_next_index(struct folio *folio)
666 return folio->index + folio_nr_pages(folio);
670 * folio_file_page - The page for a particular index.
671 * @folio: The folio which contains this index.
672 * @index: The index we want to look up.
674 * Sometimes after looking up a folio in the page cache, we need to
675 * obtain the specific page for an index (eg a page fault).
677 * Return: The page containing the file data for this index.
679 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
681 /* HugeTLBfs indexes the page cache in units of hpage_size */
682 if (folio_test_hugetlb(folio))
684 return folio_page(folio, index & (folio_nr_pages(folio) - 1));
688 * folio_contains - Does this folio contain this index?
690 * @index: The page index within the file.
692 * Context: The caller should have the page locked in order to prevent
693 * (eg) shmem from moving the page between the page cache and swap cache
694 * and changing its index in the middle of the operation.
695 * Return: true or false.
697 static inline bool folio_contains(struct folio *folio, pgoff_t index)
699 /* HugeTLBfs indexes the page cache in units of hpage_size */
700 if (folio_test_hugetlb(folio))
701 return folio->index == index;
702 return index - folio_index(folio) < folio_nr_pages(folio);
706 * Given the page we found in the page cache, return the page corresponding
707 * to this index in the file
709 static inline struct page *find_subpage(struct page *head, pgoff_t index)
711 /* HugeTLBfs wants the head page regardless */
715 return head + (index & (thp_nr_pages(head) - 1));
718 unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start,
719 pgoff_t end, struct folio_batch *fbatch);
720 unsigned filemap_get_folios_contig(struct address_space *mapping,
721 pgoff_t *start, pgoff_t end, struct folio_batch *fbatch);
722 unsigned find_get_pages_range_tag(struct address_space *mapping, pgoff_t *index,
723 pgoff_t end, xa_mark_t tag, unsigned int nr_pages,
724 struct page **pages);
725 static inline unsigned find_get_pages_tag(struct address_space *mapping,
726 pgoff_t *index, xa_mark_t tag, unsigned int nr_pages,
729 return find_get_pages_range_tag(mapping, index, (pgoff_t)-1, tag,
733 struct page *grab_cache_page_write_begin(struct address_space *mapping,
737 * Returns locked page at given index in given cache, creating it if needed.
739 static inline struct page *grab_cache_page(struct address_space *mapping,
742 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
745 struct folio *read_cache_folio(struct address_space *, pgoff_t index,
746 filler_t *filler, struct file *file);
747 struct page *read_cache_page(struct address_space *, pgoff_t index,
748 filler_t *filler, struct file *file);
749 extern struct page * read_cache_page_gfp(struct address_space *mapping,
750 pgoff_t index, gfp_t gfp_mask);
752 static inline struct page *read_mapping_page(struct address_space *mapping,
753 pgoff_t index, struct file *file)
755 return read_cache_page(mapping, index, NULL, file);
758 static inline struct folio *read_mapping_folio(struct address_space *mapping,
759 pgoff_t index, struct file *file)
761 return read_cache_folio(mapping, index, NULL, file);
765 * Get index of the page within radix-tree (but not for hugetlb pages).
766 * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
768 static inline pgoff_t page_to_index(struct page *page)
772 if (likely(!PageTransTail(page)))
775 head = compound_head(page);
777 * We don't initialize ->index for tail pages: calculate based on
780 return head->index + page - head;
783 extern pgoff_t hugetlb_basepage_index(struct page *page);
786 * Get the offset in PAGE_SIZE (even for hugetlb pages).
787 * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
789 static inline pgoff_t page_to_pgoff(struct page *page)
791 if (unlikely(PageHuge(page)))
792 return hugetlb_basepage_index(page);
793 return page_to_index(page);
797 * Return byte-offset into filesystem object for page.
799 static inline loff_t page_offset(struct page *page)
801 return ((loff_t)page->index) << PAGE_SHIFT;
804 static inline loff_t page_file_offset(struct page *page)
806 return ((loff_t)page_index(page)) << PAGE_SHIFT;
810 * folio_pos - Returns the byte position of this folio in its file.
813 static inline loff_t folio_pos(struct folio *folio)
815 return page_offset(&folio->page);
819 * folio_file_pos - Returns the byte position of this folio in its file.
822 * This differs from folio_pos() for folios which belong to a swap file.
823 * NFS is the only filesystem today which needs to use folio_file_pos().
825 static inline loff_t folio_file_pos(struct folio *folio)
827 return page_file_offset(&folio->page);
831 * Get the offset in PAGE_SIZE (even for hugetlb folios).
832 * (TODO: hugetlb folios should have ->index in PAGE_SIZE)
834 static inline pgoff_t folio_pgoff(struct folio *folio)
836 if (unlikely(folio_test_hugetlb(folio)))
837 return hugetlb_basepage_index(&folio->page);
841 extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
842 unsigned long address);
844 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
845 unsigned long address)
848 if (unlikely(is_vm_hugetlb_page(vma)))
849 return linear_hugepage_index(vma, address);
850 pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
851 pgoff += vma->vm_pgoff;
855 struct wait_page_key {
861 struct wait_page_queue {
864 wait_queue_entry_t wait;
867 static inline bool wake_page_match(struct wait_page_queue *wait_page,
868 struct wait_page_key *key)
870 if (wait_page->folio != key->folio)
874 if (wait_page->bit_nr != key->bit_nr)
880 void __folio_lock(struct folio *folio);
881 int __folio_lock_killable(struct folio *folio);
882 bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
884 void unlock_page(struct page *page);
885 void folio_unlock(struct folio *folio);
888 * folio_trylock() - Attempt to lock a folio.
889 * @folio: The folio to attempt to lock.
891 * Sometimes it is undesirable to wait for a folio to be unlocked (eg
892 * when the locks are being taken in the wrong order, or if making
893 * progress through a batch of folios is more important than processing
894 * them in order). Usually folio_lock() is the correct function to call.
896 * Context: Any context.
897 * Return: Whether the lock was successfully acquired.
899 static inline bool folio_trylock(struct folio *folio)
901 return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
905 * Return true if the page was successfully locked
907 static inline int trylock_page(struct page *page)
909 return folio_trylock(page_folio(page));
913 * folio_lock() - Lock this folio.
914 * @folio: The folio to lock.
916 * The folio lock protects against many things, probably more than it
917 * should. It is primarily held while a folio is being brought uptodate,
918 * either from its backing file or from swap. It is also held while a
919 * folio is being truncated from its address_space, so holding the lock
920 * is sufficient to keep folio->mapping stable.
922 * The folio lock is also held while write() is modifying the page to
923 * provide POSIX atomicity guarantees (as long as the write does not
924 * cross a page boundary). Other modifications to the data in the folio
925 * do not hold the folio lock and can race with writes, eg DMA and stores
928 * Context: May sleep. If you need to acquire the locks of two or
929 * more folios, they must be in order of ascending index, if they are
930 * in the same address_space. If they are in different address_spaces,
931 * acquire the lock of the folio which belongs to the address_space which
932 * has the lowest address in memory first.
934 static inline void folio_lock(struct folio *folio)
937 if (!folio_trylock(folio))
942 * lock_page() - Lock the folio containing this page.
943 * @page: The page to lock.
945 * See folio_lock() for a description of what the lock protects.
946 * This is a legacy function and new code should probably use folio_lock()
949 * Context: May sleep. Pages in the same folio share a lock, so do not
950 * attempt to lock two pages which share a folio.
952 static inline void lock_page(struct page *page)
957 folio = page_folio(page);
958 if (!folio_trylock(folio))
963 * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
964 * @folio: The folio to lock.
966 * Attempts to lock the folio, like folio_lock(), except that the sleep
967 * to acquire the lock is interruptible by a fatal signal.
969 * Context: May sleep; see folio_lock().
970 * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
972 static inline int folio_lock_killable(struct folio *folio)
975 if (!folio_trylock(folio))
976 return __folio_lock_killable(folio);
981 * lock_page_killable is like lock_page but can be interrupted by fatal
982 * signals. It returns 0 if it locked the page and -EINTR if it was
983 * killed while waiting.
985 static inline int lock_page_killable(struct page *page)
987 return folio_lock_killable(page_folio(page));
991 * folio_lock_or_retry - Lock the folio, unless this would block and the
992 * caller indicated that it can handle a retry.
994 * Return value and mmap_lock implications depend on flags; see
995 * __folio_lock_or_retry().
997 static inline bool folio_lock_or_retry(struct folio *folio,
998 struct mm_struct *mm, unsigned int flags)
1001 return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
1005 * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
1006 * and should not be used directly.
1008 void folio_wait_bit(struct folio *folio, int bit_nr);
1009 int folio_wait_bit_killable(struct folio *folio, int bit_nr);
1012 * Wait for a folio to be unlocked.
1014 * This must be called with the caller "holding" the folio,
1015 * ie with increased folio reference count so that the folio won't
1016 * go away during the wait.
1018 static inline void folio_wait_locked(struct folio *folio)
1020 if (folio_test_locked(folio))
1021 folio_wait_bit(folio, PG_locked);
1024 static inline int folio_wait_locked_killable(struct folio *folio)
1026 if (!folio_test_locked(folio))
1028 return folio_wait_bit_killable(folio, PG_locked);
1031 static inline void wait_on_page_locked(struct page *page)
1033 folio_wait_locked(page_folio(page));
1036 static inline int wait_on_page_locked_killable(struct page *page)
1038 return folio_wait_locked_killable(page_folio(page));
1041 void wait_on_page_writeback(struct page *page);
1042 void folio_wait_writeback(struct folio *folio);
1043 int folio_wait_writeback_killable(struct folio *folio);
1044 void end_page_writeback(struct page *page);
1045 void folio_end_writeback(struct folio *folio);
1046 void wait_for_stable_page(struct page *page);
1047 void folio_wait_stable(struct folio *folio);
1048 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
1049 static inline void __set_page_dirty(struct page *page,
1050 struct address_space *mapping, int warn)
1052 __folio_mark_dirty(page_folio(page), mapping, warn);
1054 void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
1055 void __folio_cancel_dirty(struct folio *folio);
1056 static inline void folio_cancel_dirty(struct folio *folio)
1058 /* Avoid atomic ops, locking, etc. when not actually needed. */
1059 if (folio_test_dirty(folio))
1060 __folio_cancel_dirty(folio);
1062 bool folio_clear_dirty_for_io(struct folio *folio);
1063 bool clear_page_dirty_for_io(struct page *page);
1064 void folio_invalidate(struct folio *folio, size_t offset, size_t length);
1065 int __must_check folio_write_one(struct folio *folio);
1066 static inline int __must_check write_one_page(struct page *page)
1068 return folio_write_one(page_folio(page));
1071 int __set_page_dirty_nobuffers(struct page *page);
1072 bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
1074 #ifdef CONFIG_MIGRATION
1075 int filemap_migrate_folio(struct address_space *mapping, struct folio *dst,
1076 struct folio *src, enum migrate_mode mode);
1078 #define filemap_migrate_folio NULL
1080 void page_endio(struct page *page, bool is_write, int err);
1082 void folio_end_private_2(struct folio *folio);
1083 void folio_wait_private_2(struct folio *folio);
1084 int folio_wait_private_2_killable(struct folio *folio);
1087 * Add an arbitrary waiter to a page's wait queue
1089 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
1092 * Fault in userspace address range.
1094 size_t fault_in_writeable(char __user *uaddr, size_t size);
1095 size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
1096 size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
1097 size_t fault_in_readable(const char __user *uaddr, size_t size);
1099 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
1100 pgoff_t index, gfp_t gfp);
1101 int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1102 pgoff_t index, gfp_t gfp);
1103 void filemap_remove_folio(struct folio *folio);
1104 void __filemap_remove_folio(struct folio *folio, void *shadow);
1105 void replace_page_cache_folio(struct folio *old, struct folio *new);
1106 void delete_from_page_cache_batch(struct address_space *mapping,
1107 struct folio_batch *fbatch);
1108 bool filemap_release_folio(struct folio *folio, gfp_t gfp);
1109 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1112 /* Must be non-static for BPF error injection */
1113 int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1114 pgoff_t index, gfp_t gfp, void **shadowp);
1116 bool filemap_range_has_writeback(struct address_space *mapping,
1117 loff_t start_byte, loff_t end_byte);
1120 * filemap_range_needs_writeback - check if range potentially needs writeback
1121 * @mapping: address space within which to check
1122 * @start_byte: offset in bytes where the range starts
1123 * @end_byte: offset in bytes where the range ends (inclusive)
1125 * Find at least one page in the range supplied, usually used to check if
1126 * direct writing in this range will trigger a writeback. Used by O_DIRECT
1127 * read/write with IOCB_NOWAIT, to see if the caller needs to do
1128 * filemap_write_and_wait_range() before proceeding.
1130 * Return: %true if the caller should do filemap_write_and_wait_range() before
1131 * doing O_DIRECT to a page in this range, %false otherwise.
1133 static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1137 if (!mapping->nrpages)
1139 if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1140 !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1142 return filemap_range_has_writeback(mapping, start_byte, end_byte);
1146 * struct readahead_control - Describes a readahead request.
1148 * A readahead request is for consecutive pages. Filesystems which
1149 * implement the ->readahead method should call readahead_page() or
1150 * readahead_page_batch() in a loop and attempt to start I/O against
1151 * each page in the request.
1153 * Most of the fields in this struct are private and should be accessed
1154 * by the functions below.
1156 * @file: The file, used primarily by network filesystems for authentication.
1157 * May be NULL if invoked internally by the filesystem.
1158 * @mapping: Readahead this filesystem object.
1159 * @ra: File readahead state. May be NULL.
1161 struct readahead_control {
1163 struct address_space *mapping;
1164 struct file_ra_state *ra;
1165 /* private: use the readahead_* accessors instead */
1167 unsigned int _nr_pages;
1168 unsigned int _batch_count;
1170 unsigned long _pflags;
1173 #define DEFINE_READAHEAD(ractl, f, r, m, i) \
1174 struct readahead_control ractl = { \
1181 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
1183 void page_cache_ra_unbounded(struct readahead_control *,
1184 unsigned long nr_to_read, unsigned long lookahead_count);
1185 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
1186 void page_cache_async_ra(struct readahead_control *, struct folio *,
1187 unsigned long req_count);
1188 void readahead_expand(struct readahead_control *ractl,
1189 loff_t new_start, size_t new_len);
1192 * page_cache_sync_readahead - generic file readahead
1193 * @mapping: address_space which holds the pagecache and I/O vectors
1194 * @ra: file_ra_state which holds the readahead state
1195 * @file: Used by the filesystem for authentication.
1196 * @index: Index of first page to be read.
1197 * @req_count: Total number of pages being read by the caller.
1199 * page_cache_sync_readahead() should be called when a cache miss happened:
1200 * it will submit the read. The readahead logic may decide to piggyback more
1201 * pages onto the read request if access patterns suggest it will improve
1205 void page_cache_sync_readahead(struct address_space *mapping,
1206 struct file_ra_state *ra, struct file *file, pgoff_t index,
1207 unsigned long req_count)
1209 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1210 page_cache_sync_ra(&ractl, req_count);
1214 * page_cache_async_readahead - file readahead for marked pages
1215 * @mapping: address_space which holds the pagecache and I/O vectors
1216 * @ra: file_ra_state which holds the readahead state
1217 * @file: Used by the filesystem for authentication.
1218 * @folio: The folio at @index which triggered the readahead call.
1219 * @index: Index of first page to be read.
1220 * @req_count: Total number of pages being read by the caller.
1222 * page_cache_async_readahead() should be called when a page is used which
1223 * is marked as PageReadahead; this is a marker to suggest that the application
1224 * has used up enough of the readahead window that we should start pulling in
1228 void page_cache_async_readahead(struct address_space *mapping,
1229 struct file_ra_state *ra, struct file *file,
1230 struct folio *folio, pgoff_t index, unsigned long req_count)
1232 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1233 page_cache_async_ra(&ractl, folio, req_count);
1236 static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1238 struct folio *folio;
1240 BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1241 ractl->_nr_pages -= ractl->_batch_count;
1242 ractl->_index += ractl->_batch_count;
1244 if (!ractl->_nr_pages) {
1245 ractl->_batch_count = 0;
1249 folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1250 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1251 ractl->_batch_count = folio_nr_pages(folio);
1257 * readahead_page - Get the next page to read.
1258 * @ractl: The current readahead request.
1260 * Context: The page is locked and has an elevated refcount. The caller
1261 * should decreases the refcount once the page has been submitted for I/O
1262 * and unlock the page once all I/O to that page has completed.
1263 * Return: A pointer to the next page, or %NULL if we are done.
1265 static inline struct page *readahead_page(struct readahead_control *ractl)
1267 struct folio *folio = __readahead_folio(ractl);
1269 return &folio->page;
1273 * readahead_folio - Get the next folio to read.
1274 * @ractl: The current readahead request.
1276 * Context: The folio is locked. The caller should unlock the folio once
1277 * all I/O to that folio has completed.
1278 * Return: A pointer to the next folio, or %NULL if we are done.
1280 static inline struct folio *readahead_folio(struct readahead_control *ractl)
1282 struct folio *folio = __readahead_folio(ractl);
1289 static inline unsigned int __readahead_batch(struct readahead_control *rac,
1290 struct page **array, unsigned int array_sz)
1293 XA_STATE(xas, &rac->mapping->i_pages, 0);
1296 BUG_ON(rac->_batch_count > rac->_nr_pages);
1297 rac->_nr_pages -= rac->_batch_count;
1298 rac->_index += rac->_batch_count;
1299 rac->_batch_count = 0;
1301 xas_set(&xas, rac->_index);
1303 xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1304 if (xas_retry(&xas, page))
1306 VM_BUG_ON_PAGE(!PageLocked(page), page);
1307 VM_BUG_ON_PAGE(PageTail(page), page);
1309 rac->_batch_count += thp_nr_pages(page);
1319 * readahead_page_batch - Get a batch of pages to read.
1320 * @rac: The current readahead request.
1321 * @array: An array of pointers to struct page.
1323 * Context: The pages are locked and have an elevated refcount. The caller
1324 * should decreases the refcount once the page has been submitted for I/O
1325 * and unlock the page once all I/O to that page has completed.
1326 * Return: The number of pages placed in the array. 0 indicates the request
1329 #define readahead_page_batch(rac, array) \
1330 __readahead_batch(rac, array, ARRAY_SIZE(array))
1333 * readahead_pos - The byte offset into the file of this readahead request.
1334 * @rac: The readahead request.
1336 static inline loff_t readahead_pos(struct readahead_control *rac)
1338 return (loff_t)rac->_index * PAGE_SIZE;
1342 * readahead_length - The number of bytes in this readahead request.
1343 * @rac: The readahead request.
1345 static inline size_t readahead_length(struct readahead_control *rac)
1347 return rac->_nr_pages * PAGE_SIZE;
1351 * readahead_index - The index of the first page in this readahead request.
1352 * @rac: The readahead request.
1354 static inline pgoff_t readahead_index(struct readahead_control *rac)
1360 * readahead_count - The number of pages in this readahead request.
1361 * @rac: The readahead request.
1363 static inline unsigned int readahead_count(struct readahead_control *rac)
1365 return rac->_nr_pages;
1369 * readahead_batch_length - The number of bytes in the current batch.
1370 * @rac: The readahead request.
1372 static inline size_t readahead_batch_length(struct readahead_control *rac)
1374 return rac->_batch_count * PAGE_SIZE;
1377 static inline unsigned long dir_pages(struct inode *inode)
1379 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1384 * folio_mkwrite_check_truncate - check if folio was truncated
1385 * @folio: the folio to check
1386 * @inode: the inode to check the folio against
1388 * Return: the number of bytes in the folio up to EOF,
1389 * or -EFAULT if the folio was truncated.
1391 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1392 struct inode *inode)
1394 loff_t size = i_size_read(inode);
1395 pgoff_t index = size >> PAGE_SHIFT;
1396 size_t offset = offset_in_folio(folio, size);
1398 if (!folio->mapping)
1401 /* folio is wholly inside EOF */
1402 if (folio_next_index(folio) - 1 < index)
1403 return folio_size(folio);
1404 /* folio is wholly past EOF */
1405 if (folio->index > index || !offset)
1407 /* folio is partially inside EOF */
1412 * page_mkwrite_check_truncate - check if page was truncated
1413 * @page: the page to check
1414 * @inode: the inode to check the page against
1416 * Returns the number of bytes in the page up to EOF,
1417 * or -EFAULT if the page was truncated.
1419 static inline int page_mkwrite_check_truncate(struct page *page,
1420 struct inode *inode)
1422 loff_t size = i_size_read(inode);
1423 pgoff_t index = size >> PAGE_SHIFT;
1424 int offset = offset_in_page(size);
1426 if (page->mapping != inode->i_mapping)
1429 /* page is wholly inside EOF */
1430 if (page->index < index)
1432 /* page is wholly past EOF */
1433 if (page->index > index || !offset)
1435 /* page is partially inside EOF */
1440 * i_blocks_per_folio - How many blocks fit in this folio.
1441 * @inode: The inode which contains the blocks.
1442 * @folio: The folio.
1444 * If the block size is larger than the size of this folio, return zero.
1446 * Context: The caller should hold a refcount on the folio to prevent it
1448 * Return: The number of filesystem blocks covered by this folio.
1451 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1453 return folio_size(folio) >> inode->i_blkbits;
1457 unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1459 return i_blocks_per_folio(inode, page_folio(page));
1461 #endif /* _LINUX_PAGEMAP_H */