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 static inline bool mapping_empty(struct address_space *mapping)
23 return xa_empty(&mapping->i_pages);
27 * mapping_shrinkable - test if page cache state allows inode reclaim
28 * @mapping: the page cache mapping
30 * This checks the mapping's cache state for the pupose of inode
31 * reclaim and LRU management.
33 * The caller is expected to hold the i_lock, but is not required to
34 * hold the i_pages lock, which usually protects cache state. That's
35 * because the i_lock and the list_lru lock that protect the inode and
36 * its LRU state don't nest inside the irq-safe i_pages lock.
38 * Cache deletions are performed under the i_lock, which ensures that
39 * when an inode goes empty, it will reliably get queued on the LRU.
41 * Cache additions do not acquire the i_lock and may race with this
42 * check, in which case we'll report the inode as shrinkable when it
43 * has cache pages. This is okay: the shrinker also checks the
44 * refcount and the referenced bit, which will be elevated or set in
45 * the process of adding new cache pages to an inode.
47 static inline bool mapping_shrinkable(struct address_space *mapping)
52 * On highmem systems, there could be lowmem pressure from the
53 * inodes before there is highmem pressure from the page
54 * cache. Make inodes shrinkable regardless of cache state.
56 if (IS_ENABLED(CONFIG_HIGHMEM))
59 /* Cache completely empty? Shrink away. */
60 head = rcu_access_pointer(mapping->i_pages.xa_head);
65 * The xarray stores single offset-0 entries directly in the
66 * head pointer, which allows non-resident page cache entries
67 * to escape the shadow shrinker's list of xarray nodes. The
68 * inode shrinker needs to pick them up under memory pressure.
70 if (!xa_is_node(head) && xa_is_value(head))
77 * Bits in mapping->flags.
80 AS_EIO = 0, /* IO error on async write */
81 AS_ENOSPC = 1, /* ENOSPC on async write */
82 AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */
83 AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */
84 AS_EXITING = 4, /* final truncate in progress */
85 /* writeback related tags are not used */
86 AS_NO_WRITEBACK_TAGS = 5,
87 AS_THP_SUPPORT = 6, /* THPs supported */
91 * mapping_set_error - record a writeback error in the address_space
92 * @mapping: the mapping in which an error should be set
93 * @error: the error to set in the mapping
95 * When writeback fails in some way, we must record that error so that
96 * userspace can be informed when fsync and the like are called. We endeavor
97 * to report errors on any file that was open at the time of the error. Some
98 * internal callers also need to know when writeback errors have occurred.
100 * When a writeback error occurs, most filesystems will want to call
101 * mapping_set_error to record the error in the mapping so that it can be
102 * reported when the application calls fsync(2).
104 static inline void mapping_set_error(struct address_space *mapping, int error)
109 /* Record in wb_err for checkers using errseq_t based tracking */
110 __filemap_set_wb_err(mapping, error);
112 /* Record it in superblock */
114 errseq_set(&mapping->host->i_sb->s_wb_err, error);
116 /* Record it in flags for now, for legacy callers */
117 if (error == -ENOSPC)
118 set_bit(AS_ENOSPC, &mapping->flags);
120 set_bit(AS_EIO, &mapping->flags);
123 static inline void mapping_set_unevictable(struct address_space *mapping)
125 set_bit(AS_UNEVICTABLE, &mapping->flags);
128 static inline void mapping_clear_unevictable(struct address_space *mapping)
130 clear_bit(AS_UNEVICTABLE, &mapping->flags);
133 static inline bool mapping_unevictable(struct address_space *mapping)
135 return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
138 static inline void mapping_set_exiting(struct address_space *mapping)
140 set_bit(AS_EXITING, &mapping->flags);
143 static inline int mapping_exiting(struct address_space *mapping)
145 return test_bit(AS_EXITING, &mapping->flags);
148 static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
150 set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
153 static inline int mapping_use_writeback_tags(struct address_space *mapping)
155 return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
158 static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
160 return mapping->gfp_mask;
163 /* Restricts the given gfp_mask to what the mapping allows. */
164 static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
167 return mapping_gfp_mask(mapping) & gfp_mask;
171 * This is non-atomic. Only to be used before the mapping is activated.
172 * Probably needs a barrier...
174 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
179 static inline bool mapping_thp_support(struct address_space *mapping)
181 return test_bit(AS_THP_SUPPORT, &mapping->flags);
184 static inline int filemap_nr_thps(struct address_space *mapping)
186 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
187 return atomic_read(&mapping->nr_thps);
193 static inline void filemap_nr_thps_inc(struct address_space *mapping)
195 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
196 if (!mapping_thp_support(mapping))
197 atomic_inc(&mapping->nr_thps);
203 static inline void filemap_nr_thps_dec(struct address_space *mapping)
205 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
206 if (!mapping_thp_support(mapping))
207 atomic_dec(&mapping->nr_thps);
213 void release_pages(struct page **pages, int nr);
215 struct address_space *page_mapping(struct page *);
216 struct address_space *folio_mapping(struct folio *);
217 struct address_space *swapcache_mapping(struct folio *);
220 * folio_file_mapping - Find the mapping this folio belongs to.
223 * For folios which are in the page cache, return the mapping that this
224 * page belongs to. Folios in the swap cache return the mapping of the
225 * swap file or swap device where the data is stored. This is different
226 * from the mapping returned by folio_mapping(). The only reason to
227 * use it is if, like NFS, you return 0 from ->activate_swapfile.
229 * Do not call this for folios which aren't in the page cache or swap cache.
231 static inline struct address_space *folio_file_mapping(struct folio *folio)
233 if (unlikely(folio_test_swapcache(folio)))
234 return swapcache_mapping(folio);
236 return folio->mapping;
239 static inline struct address_space *page_file_mapping(struct page *page)
241 return folio_file_mapping(page_folio(page));
245 * For file cache pages, return the address_space, otherwise return NULL
247 static inline struct address_space *page_mapping_file(struct page *page)
249 struct folio *folio = page_folio(page);
251 if (unlikely(folio_test_swapcache(folio)))
253 return folio_mapping(folio);
257 * folio_inode - Get the host inode for this folio.
260 * For folios which are in the page cache, return the inode that this folio
263 * Do not call this for folios which aren't in the page cache.
265 static inline struct inode *folio_inode(struct folio *folio)
267 return folio->mapping->host;
270 static inline bool page_cache_add_speculative(struct page *page, int count)
272 VM_BUG_ON_PAGE(PageTail(page), page);
273 return folio_ref_try_add_rcu((struct folio *)page, count);
276 static inline bool page_cache_get_speculative(struct page *page)
278 return page_cache_add_speculative(page, 1);
282 * folio_attach_private - Attach private data to a folio.
283 * @folio: Folio to attach data to.
284 * @data: Data to attach to folio.
286 * Attaching private data to a folio increments the page's reference count.
287 * The data must be detached before the folio will be freed.
289 static inline void folio_attach_private(struct folio *folio, void *data)
292 folio->private = data;
293 folio_set_private(folio);
297 * folio_change_private - Change private data on a folio.
298 * @folio: Folio to change the data on.
299 * @data: Data to set on the folio.
301 * Change the private data attached to a folio and return the old
302 * data. The page must previously have had data attached and the data
303 * must be detached before the folio will be freed.
305 * Return: Data that was previously attached to the folio.
307 static inline void *folio_change_private(struct folio *folio, void *data)
309 void *old = folio_get_private(folio);
311 folio->private = data;
316 * folio_detach_private - Detach private data from a folio.
317 * @folio: Folio to detach data from.
319 * Removes the data that was previously attached to the folio and decrements
320 * the refcount on the page.
322 * Return: Data that was attached to the folio.
324 static inline void *folio_detach_private(struct folio *folio)
326 void *data = folio_get_private(folio);
328 if (!folio_test_private(folio))
330 folio_clear_private(folio);
331 folio->private = NULL;
337 static inline void attach_page_private(struct page *page, void *data)
339 folio_attach_private(page_folio(page), data);
342 static inline void *detach_page_private(struct page *page)
344 return folio_detach_private(page_folio(page));
348 struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order);
350 static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order)
352 return folio_alloc(gfp, order);
356 static inline struct page *__page_cache_alloc(gfp_t gfp)
358 return &filemap_alloc_folio(gfp, 0)->page;
361 static inline struct page *page_cache_alloc(struct address_space *x)
363 return __page_cache_alloc(mapping_gfp_mask(x));
366 static inline gfp_t readahead_gfp_mask(struct address_space *x)
368 return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
371 typedef int filler_t(void *, struct page *);
373 pgoff_t page_cache_next_miss(struct address_space *mapping,
374 pgoff_t index, unsigned long max_scan);
375 pgoff_t page_cache_prev_miss(struct address_space *mapping,
376 pgoff_t index, unsigned long max_scan);
378 #define FGP_ACCESSED 0x00000001
379 #define FGP_LOCK 0x00000002
380 #define FGP_CREAT 0x00000004
381 #define FGP_WRITE 0x00000008
382 #define FGP_NOFS 0x00000010
383 #define FGP_NOWAIT 0x00000020
384 #define FGP_FOR_MMAP 0x00000040
385 #define FGP_HEAD 0x00000080
386 #define FGP_ENTRY 0x00000100
387 #define FGP_STABLE 0x00000200
389 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
390 int fgp_flags, gfp_t gfp);
391 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
392 int fgp_flags, gfp_t gfp);
395 * filemap_get_folio - Find and get a folio.
396 * @mapping: The address_space to search.
397 * @index: The page index.
399 * Looks up the page cache entry at @mapping & @index. If a folio is
400 * present, it is returned with an increased refcount.
402 * Otherwise, %NULL is returned.
404 static inline struct folio *filemap_get_folio(struct address_space *mapping,
407 return __filemap_get_folio(mapping, index, 0, 0);
411 * find_get_page - find and get a page reference
412 * @mapping: the address_space to search
413 * @offset: the page index
415 * Looks up the page cache slot at @mapping & @offset. If there is a
416 * page cache page, it is returned with an increased refcount.
418 * Otherwise, %NULL is returned.
420 static inline struct page *find_get_page(struct address_space *mapping,
423 return pagecache_get_page(mapping, offset, 0, 0);
426 static inline struct page *find_get_page_flags(struct address_space *mapping,
427 pgoff_t offset, int fgp_flags)
429 return pagecache_get_page(mapping, offset, fgp_flags, 0);
433 * find_lock_page - locate, pin and lock a pagecache page
434 * @mapping: the address_space to search
435 * @index: the page index
437 * Looks up the page cache entry at @mapping & @index. If there is a
438 * page cache page, it is returned locked and with an increased
441 * Context: May sleep.
442 * Return: A struct page or %NULL if there is no page in the cache for this
445 static inline struct page *find_lock_page(struct address_space *mapping,
448 return pagecache_get_page(mapping, index, FGP_LOCK, 0);
452 * find_or_create_page - locate or add a pagecache page
453 * @mapping: the page's address_space
454 * @index: the page's index into the mapping
455 * @gfp_mask: page allocation mode
457 * Looks up the page cache slot at @mapping & @offset. If there is a
458 * page cache page, it is returned locked and with an increased
461 * If the page is not present, a new page is allocated using @gfp_mask
462 * and added to the page cache and the VM's LRU list. The page is
463 * returned locked and with an increased refcount.
465 * On memory exhaustion, %NULL is returned.
467 * find_or_create_page() may sleep, even if @gfp_flags specifies an
470 static inline struct page *find_or_create_page(struct address_space *mapping,
471 pgoff_t index, gfp_t gfp_mask)
473 return pagecache_get_page(mapping, index,
474 FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
479 * grab_cache_page_nowait - returns locked page at given index in given cache
480 * @mapping: target address_space
481 * @index: the page index
483 * Same as grab_cache_page(), but do not wait if the page is unavailable.
484 * This is intended for speculative data generators, where the data can
485 * be regenerated if the page couldn't be grabbed. This routine should
486 * be safe to call while holding the lock for another page.
488 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
489 * and deadlock against the caller's locked page.
491 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
494 return pagecache_get_page(mapping, index,
495 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
496 mapping_gfp_mask(mapping));
499 /* Does this page contain this index? */
500 static inline bool thp_contains(struct page *head, pgoff_t index)
502 /* HugeTLBfs indexes the page cache in units of hpage_size */
504 return head->index == index;
505 return page_index(head) == (index & ~(thp_nr_pages(head) - 1UL));
508 #define swapcache_index(folio) __page_file_index(&(folio)->page)
511 * folio_index - File index of a folio.
514 * For a folio which is either in the page cache or the swap cache,
515 * return its index within the address_space it belongs to. If you know
516 * the page is definitely in the page cache, you can look at the folio's
519 * Return: The index (offset in units of pages) of a folio in its file.
521 static inline pgoff_t folio_index(struct folio *folio)
523 if (unlikely(folio_test_swapcache(folio)))
524 return swapcache_index(folio);
529 * folio_next_index - Get the index of the next folio.
530 * @folio: The current folio.
532 * Return: The index of the folio which follows this folio in the file.
534 static inline pgoff_t folio_next_index(struct folio *folio)
536 return folio->index + folio_nr_pages(folio);
540 * folio_file_page - The page for a particular index.
541 * @folio: The folio which contains this index.
542 * @index: The index we want to look up.
544 * Sometimes after looking up a folio in the page cache, we need to
545 * obtain the specific page for an index (eg a page fault).
547 * Return: The page containing the file data for this index.
549 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
551 /* HugeTLBfs indexes the page cache in units of hpage_size */
552 if (folio_test_hugetlb(folio))
554 return folio_page(folio, index & (folio_nr_pages(folio) - 1));
558 * folio_contains - Does this folio contain this index?
560 * @index: The page index within the file.
562 * Context: The caller should have the page locked in order to prevent
563 * (eg) shmem from moving the page between the page cache and swap cache
564 * and changing its index in the middle of the operation.
565 * Return: true or false.
567 static inline bool folio_contains(struct folio *folio, pgoff_t index)
569 /* HugeTLBfs indexes the page cache in units of hpage_size */
570 if (folio_test_hugetlb(folio))
571 return folio->index == index;
572 return index - folio_index(folio) < folio_nr_pages(folio);
576 * Given the page we found in the page cache, return the page corresponding
577 * to this index in the file
579 static inline struct page *find_subpage(struct page *head, pgoff_t index)
581 /* HugeTLBfs wants the head page regardless */
585 return head + (index & (thp_nr_pages(head) - 1));
588 unsigned find_get_entries(struct address_space *mapping, pgoff_t start,
589 pgoff_t end, struct pagevec *pvec, pgoff_t *indices);
590 unsigned find_get_pages_range(struct address_space *mapping, pgoff_t *start,
591 pgoff_t end, unsigned int nr_pages,
592 struct page **pages);
593 static inline unsigned find_get_pages(struct address_space *mapping,
594 pgoff_t *start, unsigned int nr_pages,
597 return find_get_pages_range(mapping, start, (pgoff_t)-1, nr_pages,
600 unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t start,
601 unsigned int nr_pages, struct page **pages);
602 unsigned find_get_pages_range_tag(struct address_space *mapping, pgoff_t *index,
603 pgoff_t end, xa_mark_t tag, unsigned int nr_pages,
604 struct page **pages);
605 static inline unsigned find_get_pages_tag(struct address_space *mapping,
606 pgoff_t *index, xa_mark_t tag, unsigned int nr_pages,
609 return find_get_pages_range_tag(mapping, index, (pgoff_t)-1, tag,
613 struct page *grab_cache_page_write_begin(struct address_space *mapping,
614 pgoff_t index, unsigned flags);
617 * Returns locked page at given index in given cache, creating it if needed.
619 static inline struct page *grab_cache_page(struct address_space *mapping,
622 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
625 extern struct page * read_cache_page(struct address_space *mapping,
626 pgoff_t index, filler_t *filler, void *data);
627 extern struct page * read_cache_page_gfp(struct address_space *mapping,
628 pgoff_t index, gfp_t gfp_mask);
629 extern int read_cache_pages(struct address_space *mapping,
630 struct list_head *pages, filler_t *filler, void *data);
632 static inline struct page *read_mapping_page(struct address_space *mapping,
633 pgoff_t index, void *data)
635 return read_cache_page(mapping, index, NULL, data);
639 * Get index of the page within radix-tree (but not for hugetlb pages).
640 * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
642 static inline pgoff_t page_to_index(struct page *page)
646 if (likely(!PageTransTail(page)))
649 head = compound_head(page);
651 * We don't initialize ->index for tail pages: calculate based on
654 return head->index + page - head;
657 extern pgoff_t hugetlb_basepage_index(struct page *page);
660 * Get the offset in PAGE_SIZE (even for hugetlb pages).
661 * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
663 static inline pgoff_t page_to_pgoff(struct page *page)
665 if (unlikely(PageHuge(page)))
666 return hugetlb_basepage_index(page);
667 return page_to_index(page);
671 * Return byte-offset into filesystem object for page.
673 static inline loff_t page_offset(struct page *page)
675 return ((loff_t)page->index) << PAGE_SHIFT;
678 static inline loff_t page_file_offset(struct page *page)
680 return ((loff_t)page_index(page)) << PAGE_SHIFT;
684 * folio_pos - Returns the byte position of this folio in its file.
687 static inline loff_t folio_pos(struct folio *folio)
689 return page_offset(&folio->page);
693 * folio_file_pos - Returns the byte position of this folio in its file.
696 * This differs from folio_pos() for folios which belong to a swap file.
697 * NFS is the only filesystem today which needs to use folio_file_pos().
699 static inline loff_t folio_file_pos(struct folio *folio)
701 return page_file_offset(&folio->page);
704 extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
705 unsigned long address);
707 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
708 unsigned long address)
711 if (unlikely(is_vm_hugetlb_page(vma)))
712 return linear_hugepage_index(vma, address);
713 pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
714 pgoff += vma->vm_pgoff;
718 struct wait_page_key {
724 struct wait_page_queue {
727 wait_queue_entry_t wait;
730 static inline bool wake_page_match(struct wait_page_queue *wait_page,
731 struct wait_page_key *key)
733 if (wait_page->folio != key->folio)
737 if (wait_page->bit_nr != key->bit_nr)
743 void __folio_lock(struct folio *folio);
744 int __folio_lock_killable(struct folio *folio);
745 bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
747 void unlock_page(struct page *page);
748 void folio_unlock(struct folio *folio);
750 static inline bool folio_trylock(struct folio *folio)
752 return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
756 * Return true if the page was successfully locked
758 static inline int trylock_page(struct page *page)
760 return folio_trylock(page_folio(page));
763 static inline void folio_lock(struct folio *folio)
766 if (!folio_trylock(folio))
771 * lock_page may only be called if we have the page's inode pinned.
773 static inline void lock_page(struct page *page)
778 folio = page_folio(page);
779 if (!folio_trylock(folio))
783 static inline int folio_lock_killable(struct folio *folio)
786 if (!folio_trylock(folio))
787 return __folio_lock_killable(folio);
792 * lock_page_killable is like lock_page but can be interrupted by fatal
793 * signals. It returns 0 if it locked the page and -EINTR if it was
794 * killed while waiting.
796 static inline int lock_page_killable(struct page *page)
798 return folio_lock_killable(page_folio(page));
802 * lock_page_or_retry - Lock the page, unless this would block and the
803 * caller indicated that it can handle a retry.
805 * Return value and mmap_lock implications depend on flags; see
806 * __folio_lock_or_retry().
808 static inline bool lock_page_or_retry(struct page *page, struct mm_struct *mm,
814 folio = page_folio(page);
815 return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
819 * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
820 * and should not be used directly.
822 void folio_wait_bit(struct folio *folio, int bit_nr);
823 int folio_wait_bit_killable(struct folio *folio, int bit_nr);
826 * Wait for a folio to be unlocked.
828 * This must be called with the caller "holding" the folio,
829 * ie with increased "page->count" so that the folio won't
830 * go away during the wait..
832 static inline void folio_wait_locked(struct folio *folio)
834 if (folio_test_locked(folio))
835 folio_wait_bit(folio, PG_locked);
838 static inline int folio_wait_locked_killable(struct folio *folio)
840 if (!folio_test_locked(folio))
842 return folio_wait_bit_killable(folio, PG_locked);
845 static inline void wait_on_page_locked(struct page *page)
847 folio_wait_locked(page_folio(page));
850 static inline int wait_on_page_locked_killable(struct page *page)
852 return folio_wait_locked_killable(page_folio(page));
855 int put_and_wait_on_page_locked(struct page *page, int state);
856 void wait_on_page_writeback(struct page *page);
857 void folio_wait_writeback(struct folio *folio);
858 int folio_wait_writeback_killable(struct folio *folio);
859 void end_page_writeback(struct page *page);
860 void folio_end_writeback(struct folio *folio);
861 void wait_for_stable_page(struct page *page);
862 void folio_wait_stable(struct folio *folio);
863 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
864 static inline void __set_page_dirty(struct page *page,
865 struct address_space *mapping, int warn)
867 __folio_mark_dirty(page_folio(page), mapping, warn);
869 void folio_account_cleaned(struct folio *folio, struct address_space *mapping,
870 struct bdi_writeback *wb);
871 static inline void account_page_cleaned(struct page *page,
872 struct address_space *mapping, struct bdi_writeback *wb)
874 return folio_account_cleaned(page_folio(page), mapping, wb);
876 void __folio_cancel_dirty(struct folio *folio);
877 static inline void folio_cancel_dirty(struct folio *folio)
879 /* Avoid atomic ops, locking, etc. when not actually needed. */
880 if (folio_test_dirty(folio))
881 __folio_cancel_dirty(folio);
883 static inline void cancel_dirty_page(struct page *page)
885 folio_cancel_dirty(page_folio(page));
887 bool folio_clear_dirty_for_io(struct folio *folio);
888 bool clear_page_dirty_for_io(struct page *page);
889 int __must_check folio_write_one(struct folio *folio);
890 static inline int __must_check write_one_page(struct page *page)
892 return folio_write_one(page_folio(page));
895 int __set_page_dirty_nobuffers(struct page *page);
896 int __set_page_dirty_no_writeback(struct page *page);
898 void page_endio(struct page *page, bool is_write, int err);
900 void folio_end_private_2(struct folio *folio);
901 void folio_wait_private_2(struct folio *folio);
902 int folio_wait_private_2_killable(struct folio *folio);
905 * Add an arbitrary waiter to a page's wait queue
907 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
910 * Fault in userspace address range.
912 size_t fault_in_writeable(char __user *uaddr, size_t size);
913 size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
914 size_t fault_in_readable(const char __user *uaddr, size_t size);
916 int add_to_page_cache_locked(struct page *page, struct address_space *mapping,
917 pgoff_t index, gfp_t gfp);
918 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
919 pgoff_t index, gfp_t gfp);
920 int filemap_add_folio(struct address_space *mapping, struct folio *folio,
921 pgoff_t index, gfp_t gfp);
922 extern void delete_from_page_cache(struct page *page);
923 extern void __delete_from_page_cache(struct page *page, void *shadow);
924 void replace_page_cache_page(struct page *old, struct page *new);
925 void delete_from_page_cache_batch(struct address_space *mapping,
926 struct pagevec *pvec);
927 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
931 * Like add_to_page_cache_locked, but used to add newly allocated pages:
932 * the page is new, so we can just run __SetPageLocked() against it.
934 static inline int add_to_page_cache(struct page *page,
935 struct address_space *mapping, pgoff_t offset, gfp_t gfp_mask)
939 __SetPageLocked(page);
940 error = add_to_page_cache_locked(page, mapping, offset, gfp_mask);
942 __ClearPageLocked(page);
946 /* Must be non-static for BPF error injection */
947 int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
948 pgoff_t index, gfp_t gfp, void **shadowp);
951 * struct readahead_control - Describes a readahead request.
953 * A readahead request is for consecutive pages. Filesystems which
954 * implement the ->readahead method should call readahead_page() or
955 * readahead_page_batch() in a loop and attempt to start I/O against
956 * each page in the request.
958 * Most of the fields in this struct are private and should be accessed
959 * by the functions below.
961 * @file: The file, used primarily by network filesystems for authentication.
962 * May be NULL if invoked internally by the filesystem.
963 * @mapping: Readahead this filesystem object.
964 * @ra: File readahead state. May be NULL.
966 struct readahead_control {
968 struct address_space *mapping;
969 struct file_ra_state *ra;
970 /* private: use the readahead_* accessors instead */
972 unsigned int _nr_pages;
973 unsigned int _batch_count;
976 #define DEFINE_READAHEAD(ractl, f, r, m, i) \
977 struct readahead_control ractl = { \
984 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
986 void page_cache_ra_unbounded(struct readahead_control *,
987 unsigned long nr_to_read, unsigned long lookahead_count);
988 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
989 void page_cache_async_ra(struct readahead_control *, struct page *,
990 unsigned long req_count);
991 void readahead_expand(struct readahead_control *ractl,
992 loff_t new_start, size_t new_len);
995 * page_cache_sync_readahead - generic file readahead
996 * @mapping: address_space which holds the pagecache and I/O vectors
997 * @ra: file_ra_state which holds the readahead state
998 * @file: Used by the filesystem for authentication.
999 * @index: Index of first page to be read.
1000 * @req_count: Total number of pages being read by the caller.
1002 * page_cache_sync_readahead() should be called when a cache miss happened:
1003 * it will submit the read. The readahead logic may decide to piggyback more
1004 * pages onto the read request if access patterns suggest it will improve
1008 void page_cache_sync_readahead(struct address_space *mapping,
1009 struct file_ra_state *ra, struct file *file, pgoff_t index,
1010 unsigned long req_count)
1012 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1013 page_cache_sync_ra(&ractl, req_count);
1017 * page_cache_async_readahead - file readahead for marked pages
1018 * @mapping: address_space which holds the pagecache and I/O vectors
1019 * @ra: file_ra_state which holds the readahead state
1020 * @file: Used by the filesystem for authentication.
1021 * @page: The page at @index which triggered the readahead call.
1022 * @index: Index of first page to be read.
1023 * @req_count: Total number of pages being read by the caller.
1025 * page_cache_async_readahead() should be called when a page is used which
1026 * is marked as PageReadahead; this is a marker to suggest that the application
1027 * has used up enough of the readahead window that we should start pulling in
1031 void page_cache_async_readahead(struct address_space *mapping,
1032 struct file_ra_state *ra, struct file *file,
1033 struct page *page, pgoff_t index, unsigned long req_count)
1035 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1036 page_cache_async_ra(&ractl, page, req_count);
1039 static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1041 struct folio *folio;
1043 BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1044 ractl->_nr_pages -= ractl->_batch_count;
1045 ractl->_index += ractl->_batch_count;
1047 if (!ractl->_nr_pages) {
1048 ractl->_batch_count = 0;
1052 folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1053 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1054 ractl->_batch_count = folio_nr_pages(folio);
1060 * readahead_page - Get the next page to read.
1061 * @ractl: The current readahead request.
1063 * Context: The page is locked and has an elevated refcount. The caller
1064 * should decreases the refcount once the page has been submitted for I/O
1065 * and unlock the page once all I/O to that page has completed.
1066 * Return: A pointer to the next page, or %NULL if we are done.
1068 static inline struct page *readahead_page(struct readahead_control *ractl)
1070 struct folio *folio = __readahead_folio(ractl);
1072 return &folio->page;
1076 * readahead_folio - Get the next folio to read.
1077 * @ractl: The current readahead request.
1079 * Context: The folio is locked. The caller should unlock the folio once
1080 * all I/O to that folio has completed.
1081 * Return: A pointer to the next folio, or %NULL if we are done.
1083 static inline struct folio *readahead_folio(struct readahead_control *ractl)
1085 struct folio *folio = __readahead_folio(ractl);
1092 static inline unsigned int __readahead_batch(struct readahead_control *rac,
1093 struct page **array, unsigned int array_sz)
1096 XA_STATE(xas, &rac->mapping->i_pages, 0);
1099 BUG_ON(rac->_batch_count > rac->_nr_pages);
1100 rac->_nr_pages -= rac->_batch_count;
1101 rac->_index += rac->_batch_count;
1102 rac->_batch_count = 0;
1104 xas_set(&xas, rac->_index);
1106 xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1107 if (xas_retry(&xas, page))
1109 VM_BUG_ON_PAGE(!PageLocked(page), page);
1110 VM_BUG_ON_PAGE(PageTail(page), page);
1112 rac->_batch_count += thp_nr_pages(page);
1115 * The page cache isn't using multi-index entries yet,
1116 * so the xas cursor needs to be manually moved to the
1117 * next index. This can be removed once the page cache
1121 xas_set(&xas, rac->_index + rac->_batch_count);
1132 * readahead_page_batch - Get a batch of pages to read.
1133 * @rac: The current readahead request.
1134 * @array: An array of pointers to struct page.
1136 * Context: The pages are locked and have an elevated refcount. The caller
1137 * should decreases the refcount once the page has been submitted for I/O
1138 * and unlock the page once all I/O to that page has completed.
1139 * Return: The number of pages placed in the array. 0 indicates the request
1142 #define readahead_page_batch(rac, array) \
1143 __readahead_batch(rac, array, ARRAY_SIZE(array))
1146 * readahead_pos - The byte offset into the file of this readahead request.
1147 * @rac: The readahead request.
1149 static inline loff_t readahead_pos(struct readahead_control *rac)
1151 return (loff_t)rac->_index * PAGE_SIZE;
1155 * readahead_length - The number of bytes in this readahead request.
1156 * @rac: The readahead request.
1158 static inline size_t readahead_length(struct readahead_control *rac)
1160 return rac->_nr_pages * PAGE_SIZE;
1164 * readahead_index - The index of the first page in this readahead request.
1165 * @rac: The readahead request.
1167 static inline pgoff_t readahead_index(struct readahead_control *rac)
1173 * readahead_count - The number of pages in this readahead request.
1174 * @rac: The readahead request.
1176 static inline unsigned int readahead_count(struct readahead_control *rac)
1178 return rac->_nr_pages;
1182 * readahead_batch_length - The number of bytes in the current batch.
1183 * @rac: The readahead request.
1185 static inline size_t readahead_batch_length(struct readahead_control *rac)
1187 return rac->_batch_count * PAGE_SIZE;
1190 static inline unsigned long dir_pages(struct inode *inode)
1192 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1197 * folio_mkwrite_check_truncate - check if folio was truncated
1198 * @folio: the folio to check
1199 * @inode: the inode to check the folio against
1201 * Return: the number of bytes in the folio up to EOF,
1202 * or -EFAULT if the folio was truncated.
1204 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1205 struct inode *inode)
1207 loff_t size = i_size_read(inode);
1208 pgoff_t index = size >> PAGE_SHIFT;
1209 size_t offset = offset_in_folio(folio, size);
1211 if (!folio->mapping)
1214 /* folio is wholly inside EOF */
1215 if (folio_next_index(folio) - 1 < index)
1216 return folio_size(folio);
1217 /* folio is wholly past EOF */
1218 if (folio->index > index || !offset)
1220 /* folio is partially inside EOF */
1225 * page_mkwrite_check_truncate - check if page was truncated
1226 * @page: the page to check
1227 * @inode: the inode to check the page against
1229 * Returns the number of bytes in the page up to EOF,
1230 * or -EFAULT if the page was truncated.
1232 static inline int page_mkwrite_check_truncate(struct page *page,
1233 struct inode *inode)
1235 loff_t size = i_size_read(inode);
1236 pgoff_t index = size >> PAGE_SHIFT;
1237 int offset = offset_in_page(size);
1239 if (page->mapping != inode->i_mapping)
1242 /* page is wholly inside EOF */
1243 if (page->index < index)
1245 /* page is wholly past EOF */
1246 if (page->index > index || !offset)
1248 /* page is partially inside EOF */
1253 * i_blocks_per_folio - How many blocks fit in this folio.
1254 * @inode: The inode which contains the blocks.
1255 * @folio: The folio.
1257 * If the block size is larger than the size of this folio, return zero.
1259 * Context: The caller should hold a refcount on the folio to prevent it
1261 * Return: The number of filesystem blocks covered by this folio.
1264 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1266 return folio_size(folio) >> inode->i_blkbits;
1270 unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1272 return i_blocks_per_folio(inode, page_folio(page));
1274 #endif /* _LINUX_PAGEMAP_H */