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 * Bits in mapping->flags.
30 AS_EIO = 0, /* IO error on async write */
31 AS_ENOSPC = 1, /* ENOSPC on async write */
32 AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */
33 AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */
34 AS_EXITING = 4, /* final truncate in progress */
35 /* writeback related tags are not used */
36 AS_NO_WRITEBACK_TAGS = 5,
37 AS_THP_SUPPORT = 6, /* THPs supported */
41 * mapping_set_error - record a writeback error in the address_space
42 * @mapping: the mapping in which an error should be set
43 * @error: the error to set in the mapping
45 * When writeback fails in some way, we must record that error so that
46 * userspace can be informed when fsync and the like are called. We endeavor
47 * to report errors on any file that was open at the time of the error. Some
48 * internal callers also need to know when writeback errors have occurred.
50 * When a writeback error occurs, most filesystems will want to call
51 * mapping_set_error to record the error in the mapping so that it can be
52 * reported when the application calls fsync(2).
54 static inline void mapping_set_error(struct address_space *mapping, int error)
59 /* Record in wb_err for checkers using errseq_t based tracking */
60 __filemap_set_wb_err(mapping, error);
62 /* Record it in superblock */
64 errseq_set(&mapping->host->i_sb->s_wb_err, error);
66 /* Record it in flags for now, for legacy callers */
68 set_bit(AS_ENOSPC, &mapping->flags);
70 set_bit(AS_EIO, &mapping->flags);
73 static inline void mapping_set_unevictable(struct address_space *mapping)
75 set_bit(AS_UNEVICTABLE, &mapping->flags);
78 static inline void mapping_clear_unevictable(struct address_space *mapping)
80 clear_bit(AS_UNEVICTABLE, &mapping->flags);
83 static inline bool mapping_unevictable(struct address_space *mapping)
85 return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
88 static inline void mapping_set_exiting(struct address_space *mapping)
90 set_bit(AS_EXITING, &mapping->flags);
93 static inline int mapping_exiting(struct address_space *mapping)
95 return test_bit(AS_EXITING, &mapping->flags);
98 static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
100 set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
103 static inline int mapping_use_writeback_tags(struct address_space *mapping)
105 return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
108 static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
110 return mapping->gfp_mask;
113 /* Restricts the given gfp_mask to what the mapping allows. */
114 static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
117 return mapping_gfp_mask(mapping) & gfp_mask;
121 * This is non-atomic. Only to be used before the mapping is activated.
122 * Probably needs a barrier...
124 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
129 static inline bool mapping_thp_support(struct address_space *mapping)
131 return test_bit(AS_THP_SUPPORT, &mapping->flags);
134 static inline int filemap_nr_thps(struct address_space *mapping)
136 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
137 return atomic_read(&mapping->nr_thps);
143 static inline void filemap_nr_thps_inc(struct address_space *mapping)
145 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
146 if (!mapping_thp_support(mapping))
147 atomic_inc(&mapping->nr_thps);
153 static inline void filemap_nr_thps_dec(struct address_space *mapping)
155 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
156 if (!mapping_thp_support(mapping))
157 atomic_dec(&mapping->nr_thps);
163 void release_pages(struct page **pages, int nr);
165 struct address_space *page_mapping(struct page *);
166 struct address_space *folio_mapping(struct folio *);
167 struct address_space *swapcache_mapping(struct folio *);
170 * folio_file_mapping - Find the mapping this folio belongs to.
173 * For folios which are in the page cache, return the mapping that this
174 * page belongs to. Folios in the swap cache return the mapping of the
175 * swap file or swap device where the data is stored. This is different
176 * from the mapping returned by folio_mapping(). The only reason to
177 * use it is if, like NFS, you return 0 from ->activate_swapfile.
179 * Do not call this for folios which aren't in the page cache or swap cache.
181 static inline struct address_space *folio_file_mapping(struct folio *folio)
183 if (unlikely(folio_test_swapcache(folio)))
184 return swapcache_mapping(folio);
186 return folio->mapping;
189 static inline struct address_space *page_file_mapping(struct page *page)
191 return folio_file_mapping(page_folio(page));
195 * For file cache pages, return the address_space, otherwise return NULL
197 static inline struct address_space *page_mapping_file(struct page *page)
199 struct folio *folio = page_folio(page);
201 if (unlikely(folio_test_swapcache(folio)))
203 return folio_mapping(folio);
206 static inline bool page_cache_add_speculative(struct page *page, int count)
208 VM_BUG_ON_PAGE(PageTail(page), page);
209 return folio_ref_try_add_rcu((struct folio *)page, count);
212 static inline bool page_cache_get_speculative(struct page *page)
214 return page_cache_add_speculative(page, 1);
218 * folio_attach_private - Attach private data to a folio.
219 * @folio: Folio to attach data to.
220 * @data: Data to attach to folio.
222 * Attaching private data to a folio increments the page's reference count.
223 * The data must be detached before the folio will be freed.
225 static inline void folio_attach_private(struct folio *folio, void *data)
228 folio->private = data;
229 folio_set_private(folio);
233 * folio_detach_private - Detach private data from a folio.
234 * @folio: Folio to detach data from.
236 * Removes the data that was previously attached to the folio and decrements
237 * the refcount on the page.
239 * Return: Data that was attached to the folio.
241 static inline void *folio_detach_private(struct folio *folio)
243 void *data = folio_get_private(folio);
245 if (!folio_test_private(folio))
247 folio_clear_private(folio);
248 folio->private = NULL;
254 static inline void attach_page_private(struct page *page, void *data)
256 folio_attach_private(page_folio(page), data);
259 static inline void *detach_page_private(struct page *page)
261 return folio_detach_private(page_folio(page));
265 struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order);
267 static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order)
269 return folio_alloc(gfp, order);
273 static inline struct page *__page_cache_alloc(gfp_t gfp)
275 return &filemap_alloc_folio(gfp, 0)->page;
278 static inline struct page *page_cache_alloc(struct address_space *x)
280 return __page_cache_alloc(mapping_gfp_mask(x));
283 static inline gfp_t readahead_gfp_mask(struct address_space *x)
285 return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
288 typedef int filler_t(void *, struct page *);
290 pgoff_t page_cache_next_miss(struct address_space *mapping,
291 pgoff_t index, unsigned long max_scan);
292 pgoff_t page_cache_prev_miss(struct address_space *mapping,
293 pgoff_t index, unsigned long max_scan);
295 #define FGP_ACCESSED 0x00000001
296 #define FGP_LOCK 0x00000002
297 #define FGP_CREAT 0x00000004
298 #define FGP_WRITE 0x00000008
299 #define FGP_NOFS 0x00000010
300 #define FGP_NOWAIT 0x00000020
301 #define FGP_FOR_MMAP 0x00000040
302 #define FGP_HEAD 0x00000080
303 #define FGP_ENTRY 0x00000100
305 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
306 int fgp_flags, gfp_t gfp);
307 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
308 int fgp_flags, gfp_t gfp);
311 * filemap_get_folio - Find and get a folio.
312 * @mapping: The address_space to search.
313 * @index: The page index.
315 * Looks up the page cache entry at @mapping & @index. If a folio is
316 * present, it is returned with an increased refcount.
318 * Otherwise, %NULL is returned.
320 static inline struct folio *filemap_get_folio(struct address_space *mapping,
323 return __filemap_get_folio(mapping, index, 0, 0);
327 * find_get_page - find and get a page reference
328 * @mapping: the address_space to search
329 * @offset: the page index
331 * Looks up the page cache slot at @mapping & @offset. If there is a
332 * page cache page, it is returned with an increased refcount.
334 * Otherwise, %NULL is returned.
336 static inline struct page *find_get_page(struct address_space *mapping,
339 return pagecache_get_page(mapping, offset, 0, 0);
342 static inline struct page *find_get_page_flags(struct address_space *mapping,
343 pgoff_t offset, int fgp_flags)
345 return pagecache_get_page(mapping, offset, fgp_flags, 0);
349 * find_lock_page - locate, pin and lock a pagecache page
350 * @mapping: the address_space to search
351 * @index: the page index
353 * Looks up the page cache entry at @mapping & @index. If there is a
354 * page cache page, it is returned locked and with an increased
357 * Context: May sleep.
358 * Return: A struct page or %NULL if there is no page in the cache for this
361 static inline struct page *find_lock_page(struct address_space *mapping,
364 return pagecache_get_page(mapping, index, FGP_LOCK, 0);
368 * find_or_create_page - locate or add a pagecache page
369 * @mapping: the page's address_space
370 * @index: the page's index into the mapping
371 * @gfp_mask: page allocation mode
373 * Looks up the page cache slot at @mapping & @offset. If there is a
374 * page cache page, it is returned locked and with an increased
377 * If the page is not present, a new page is allocated using @gfp_mask
378 * and added to the page cache and the VM's LRU list. The page is
379 * returned locked and with an increased refcount.
381 * On memory exhaustion, %NULL is returned.
383 * find_or_create_page() may sleep, even if @gfp_flags specifies an
386 static inline struct page *find_or_create_page(struct address_space *mapping,
387 pgoff_t index, gfp_t gfp_mask)
389 return pagecache_get_page(mapping, index,
390 FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
395 * grab_cache_page_nowait - returns locked page at given index in given cache
396 * @mapping: target address_space
397 * @index: the page index
399 * Same as grab_cache_page(), but do not wait if the page is unavailable.
400 * This is intended for speculative data generators, where the data can
401 * be regenerated if the page couldn't be grabbed. This routine should
402 * be safe to call while holding the lock for another page.
404 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
405 * and deadlock against the caller's locked page.
407 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
410 return pagecache_get_page(mapping, index,
411 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
412 mapping_gfp_mask(mapping));
415 /* Does this page contain this index? */
416 static inline bool thp_contains(struct page *head, pgoff_t index)
418 /* HugeTLBfs indexes the page cache in units of hpage_size */
420 return head->index == index;
421 return page_index(head) == (index & ~(thp_nr_pages(head) - 1UL));
424 #define swapcache_index(folio) __page_file_index(&(folio)->page)
427 * folio_index - File index of a folio.
430 * For a folio which is either in the page cache or the swap cache,
431 * return its index within the address_space it belongs to. If you know
432 * the page is definitely in the page cache, you can look at the folio's
435 * Return: The index (offset in units of pages) of a folio in its file.
437 static inline pgoff_t folio_index(struct folio *folio)
439 if (unlikely(folio_test_swapcache(folio)))
440 return swapcache_index(folio);
445 * folio_next_index - Get the index of the next folio.
446 * @folio: The current folio.
448 * Return: The index of the folio which follows this folio in the file.
450 static inline pgoff_t folio_next_index(struct folio *folio)
452 return folio->index + folio_nr_pages(folio);
456 * folio_file_page - The page for a particular index.
457 * @folio: The folio which contains this index.
458 * @index: The index we want to look up.
460 * Sometimes after looking up a folio in the page cache, we need to
461 * obtain the specific page for an index (eg a page fault).
463 * Return: The page containing the file data for this index.
465 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
467 /* HugeTLBfs indexes the page cache in units of hpage_size */
468 if (folio_test_hugetlb(folio))
470 return folio_page(folio, index & (folio_nr_pages(folio) - 1));
474 * folio_contains - Does this folio contain this index?
476 * @index: The page index within the file.
478 * Context: The caller should have the page locked in order to prevent
479 * (eg) shmem from moving the page between the page cache and swap cache
480 * and changing its index in the middle of the operation.
481 * Return: true or false.
483 static inline bool folio_contains(struct folio *folio, pgoff_t index)
485 /* HugeTLBfs indexes the page cache in units of hpage_size */
486 if (folio_test_hugetlb(folio))
487 return folio->index == index;
488 return index - folio_index(folio) < folio_nr_pages(folio);
492 * Given the page we found in the page cache, return the page corresponding
493 * to this index in the file
495 static inline struct page *find_subpage(struct page *head, pgoff_t index)
497 /* HugeTLBfs wants the head page regardless */
501 return head + (index & (thp_nr_pages(head) - 1));
504 unsigned find_get_entries(struct address_space *mapping, pgoff_t start,
505 pgoff_t end, struct pagevec *pvec, pgoff_t *indices);
506 unsigned find_get_pages_range(struct address_space *mapping, pgoff_t *start,
507 pgoff_t end, unsigned int nr_pages,
508 struct page **pages);
509 static inline unsigned find_get_pages(struct address_space *mapping,
510 pgoff_t *start, unsigned int nr_pages,
513 return find_get_pages_range(mapping, start, (pgoff_t)-1, nr_pages,
516 unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t start,
517 unsigned int nr_pages, struct page **pages);
518 unsigned find_get_pages_range_tag(struct address_space *mapping, pgoff_t *index,
519 pgoff_t end, xa_mark_t tag, unsigned int nr_pages,
520 struct page **pages);
521 static inline unsigned find_get_pages_tag(struct address_space *mapping,
522 pgoff_t *index, xa_mark_t tag, unsigned int nr_pages,
525 return find_get_pages_range_tag(mapping, index, (pgoff_t)-1, tag,
529 struct page *grab_cache_page_write_begin(struct address_space *mapping,
530 pgoff_t index, unsigned flags);
533 * Returns locked page at given index in given cache, creating it if needed.
535 static inline struct page *grab_cache_page(struct address_space *mapping,
538 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
541 extern struct page * read_cache_page(struct address_space *mapping,
542 pgoff_t index, filler_t *filler, void *data);
543 extern struct page * read_cache_page_gfp(struct address_space *mapping,
544 pgoff_t index, gfp_t gfp_mask);
545 extern int read_cache_pages(struct address_space *mapping,
546 struct list_head *pages, filler_t *filler, void *data);
548 static inline struct page *read_mapping_page(struct address_space *mapping,
549 pgoff_t index, void *data)
551 return read_cache_page(mapping, index, NULL, data);
555 * Get index of the page within radix-tree (but not for hugetlb pages).
556 * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
558 static inline pgoff_t page_to_index(struct page *page)
562 if (likely(!PageTransTail(page)))
565 head = compound_head(page);
567 * We don't initialize ->index for tail pages: calculate based on
570 return head->index + page - head;
573 extern pgoff_t hugetlb_basepage_index(struct page *page);
576 * Get the offset in PAGE_SIZE (even for hugetlb pages).
577 * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
579 static inline pgoff_t page_to_pgoff(struct page *page)
581 if (unlikely(PageHuge(page)))
582 return hugetlb_basepage_index(page);
583 return page_to_index(page);
587 * Return byte-offset into filesystem object for page.
589 static inline loff_t page_offset(struct page *page)
591 return ((loff_t)page->index) << PAGE_SHIFT;
594 static inline loff_t page_file_offset(struct page *page)
596 return ((loff_t)page_index(page)) << PAGE_SHIFT;
600 * folio_pos - Returns the byte position of this folio in its file.
603 static inline loff_t folio_pos(struct folio *folio)
605 return page_offset(&folio->page);
609 * folio_file_pos - Returns the byte position of this folio in its file.
612 * This differs from folio_pos() for folios which belong to a swap file.
613 * NFS is the only filesystem today which needs to use folio_file_pos().
615 static inline loff_t folio_file_pos(struct folio *folio)
617 return page_file_offset(&folio->page);
620 extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
621 unsigned long address);
623 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
624 unsigned long address)
627 if (unlikely(is_vm_hugetlb_page(vma)))
628 return linear_hugepage_index(vma, address);
629 pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
630 pgoff += vma->vm_pgoff;
634 struct wait_page_key {
640 struct wait_page_queue {
643 wait_queue_entry_t wait;
646 static inline bool wake_page_match(struct wait_page_queue *wait_page,
647 struct wait_page_key *key)
649 if (wait_page->folio != key->folio)
653 if (wait_page->bit_nr != key->bit_nr)
659 void __folio_lock(struct folio *folio);
660 int __folio_lock_killable(struct folio *folio);
661 bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
663 void unlock_page(struct page *page);
664 void folio_unlock(struct folio *folio);
666 static inline bool folio_trylock(struct folio *folio)
668 return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
672 * Return true if the page was successfully locked
674 static inline int trylock_page(struct page *page)
676 return folio_trylock(page_folio(page));
679 static inline void folio_lock(struct folio *folio)
682 if (!folio_trylock(folio))
687 * lock_page may only be called if we have the page's inode pinned.
689 static inline void lock_page(struct page *page)
694 folio = page_folio(page);
695 if (!folio_trylock(folio))
699 static inline int folio_lock_killable(struct folio *folio)
702 if (!folio_trylock(folio))
703 return __folio_lock_killable(folio);
708 * lock_page_killable is like lock_page but can be interrupted by fatal
709 * signals. It returns 0 if it locked the page and -EINTR if it was
710 * killed while waiting.
712 static inline int lock_page_killable(struct page *page)
714 return folio_lock_killable(page_folio(page));
718 * lock_page_or_retry - Lock the page, unless this would block and the
719 * caller indicated that it can handle a retry.
721 * Return value and mmap_lock implications depend on flags; see
722 * __folio_lock_or_retry().
724 static inline bool lock_page_or_retry(struct page *page, struct mm_struct *mm,
730 folio = page_folio(page);
731 return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
735 * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
736 * and should not be used directly.
738 void folio_wait_bit(struct folio *folio, int bit_nr);
739 int folio_wait_bit_killable(struct folio *folio, int bit_nr);
742 * Wait for a folio to be unlocked.
744 * This must be called with the caller "holding" the folio,
745 * ie with increased "page->count" so that the folio won't
746 * go away during the wait..
748 static inline void folio_wait_locked(struct folio *folio)
750 if (folio_test_locked(folio))
751 folio_wait_bit(folio, PG_locked);
754 static inline int folio_wait_locked_killable(struct folio *folio)
756 if (!folio_test_locked(folio))
758 return folio_wait_bit_killable(folio, PG_locked);
761 static inline void wait_on_page_locked(struct page *page)
763 folio_wait_locked(page_folio(page));
766 static inline int wait_on_page_locked_killable(struct page *page)
768 return folio_wait_locked_killable(page_folio(page));
771 int put_and_wait_on_page_locked(struct page *page, int state);
772 void wait_on_page_writeback(struct page *page);
773 void folio_wait_writeback(struct folio *folio);
774 int folio_wait_writeback_killable(struct folio *folio);
775 void end_page_writeback(struct page *page);
776 void folio_end_writeback(struct folio *folio);
777 void wait_for_stable_page(struct page *page);
778 void folio_wait_stable(struct folio *folio);
779 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
780 static inline void __set_page_dirty(struct page *page,
781 struct address_space *mapping, int warn)
783 __folio_mark_dirty(page_folio(page), mapping, warn);
785 void folio_account_cleaned(struct folio *folio, struct address_space *mapping,
786 struct bdi_writeback *wb);
787 static inline void account_page_cleaned(struct page *page,
788 struct address_space *mapping, struct bdi_writeback *wb)
790 return folio_account_cleaned(page_folio(page), mapping, wb);
792 void __folio_cancel_dirty(struct folio *folio);
793 static inline void folio_cancel_dirty(struct folio *folio)
795 /* Avoid atomic ops, locking, etc. when not actually needed. */
796 if (folio_test_dirty(folio))
797 __folio_cancel_dirty(folio);
799 static inline void cancel_dirty_page(struct page *page)
801 folio_cancel_dirty(page_folio(page));
803 bool folio_clear_dirty_for_io(struct folio *folio);
804 bool clear_page_dirty_for_io(struct page *page);
806 int __set_page_dirty_nobuffers(struct page *page);
807 int __set_page_dirty_no_writeback(struct page *page);
809 void page_endio(struct page *page, bool is_write, int err);
811 void folio_end_private_2(struct folio *folio);
812 void folio_wait_private_2(struct folio *folio);
813 int folio_wait_private_2_killable(struct folio *folio);
816 * Add an arbitrary waiter to a page's wait queue
818 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
821 * Fault everything in given userspace address range in.
823 static inline int fault_in_pages_writeable(char __user *uaddr, size_t size)
825 char __user *end = uaddr + size - 1;
827 if (unlikely(size == 0))
830 if (unlikely(uaddr > end))
833 * Writing zeroes into userspace here is OK, because we know that if
834 * the zero gets there, we'll be overwriting it.
837 if (unlikely(__put_user(0, uaddr) != 0))
840 } while (uaddr <= end);
842 /* Check whether the range spilled into the next page. */
843 if (((unsigned long)uaddr & PAGE_MASK) ==
844 ((unsigned long)end & PAGE_MASK))
845 return __put_user(0, end);
850 static inline int fault_in_pages_readable(const char __user *uaddr, size_t size)
853 const char __user *end = uaddr + size - 1;
855 if (unlikely(size == 0))
858 if (unlikely(uaddr > end))
862 if (unlikely(__get_user(c, uaddr) != 0))
865 } while (uaddr <= end);
867 /* Check whether the range spilled into the next page. */
868 if (((unsigned long)uaddr & PAGE_MASK) ==
869 ((unsigned long)end & PAGE_MASK)) {
870 return __get_user(c, end);
877 int add_to_page_cache_locked(struct page *page, struct address_space *mapping,
878 pgoff_t index, gfp_t gfp);
879 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
880 pgoff_t index, gfp_t gfp);
881 int filemap_add_folio(struct address_space *mapping, struct folio *folio,
882 pgoff_t index, gfp_t gfp);
883 extern void delete_from_page_cache(struct page *page);
884 extern void __delete_from_page_cache(struct page *page, void *shadow);
885 void replace_page_cache_page(struct page *old, struct page *new);
886 void delete_from_page_cache_batch(struct address_space *mapping,
887 struct pagevec *pvec);
888 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
892 * Like add_to_page_cache_locked, but used to add newly allocated pages:
893 * the page is new, so we can just run __SetPageLocked() against it.
895 static inline int add_to_page_cache(struct page *page,
896 struct address_space *mapping, pgoff_t offset, gfp_t gfp_mask)
900 __SetPageLocked(page);
901 error = add_to_page_cache_locked(page, mapping, offset, gfp_mask);
903 __ClearPageLocked(page);
907 /* Must be non-static for BPF error injection */
908 int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
909 pgoff_t index, gfp_t gfp, void **shadowp);
912 * struct readahead_control - Describes a readahead request.
914 * A readahead request is for consecutive pages. Filesystems which
915 * implement the ->readahead method should call readahead_page() or
916 * readahead_page_batch() in a loop and attempt to start I/O against
917 * each page in the request.
919 * Most of the fields in this struct are private and should be accessed
920 * by the functions below.
922 * @file: The file, used primarily by network filesystems for authentication.
923 * May be NULL if invoked internally by the filesystem.
924 * @mapping: Readahead this filesystem object.
925 * @ra: File readahead state. May be NULL.
927 struct readahead_control {
929 struct address_space *mapping;
930 struct file_ra_state *ra;
931 /* private: use the readahead_* accessors instead */
933 unsigned int _nr_pages;
934 unsigned int _batch_count;
937 #define DEFINE_READAHEAD(ractl, f, r, m, i) \
938 struct readahead_control ractl = { \
945 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
947 void page_cache_ra_unbounded(struct readahead_control *,
948 unsigned long nr_to_read, unsigned long lookahead_count);
949 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
950 void page_cache_async_ra(struct readahead_control *, struct page *,
951 unsigned long req_count);
952 void readahead_expand(struct readahead_control *ractl,
953 loff_t new_start, size_t new_len);
956 * page_cache_sync_readahead - generic file readahead
957 * @mapping: address_space which holds the pagecache and I/O vectors
958 * @ra: file_ra_state which holds the readahead state
959 * @file: Used by the filesystem for authentication.
960 * @index: Index of first page to be read.
961 * @req_count: Total number of pages being read by the caller.
963 * page_cache_sync_readahead() should be called when a cache miss happened:
964 * it will submit the read. The readahead logic may decide to piggyback more
965 * pages onto the read request if access patterns suggest it will improve
969 void page_cache_sync_readahead(struct address_space *mapping,
970 struct file_ra_state *ra, struct file *file, pgoff_t index,
971 unsigned long req_count)
973 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
974 page_cache_sync_ra(&ractl, req_count);
978 * page_cache_async_readahead - file readahead for marked pages
979 * @mapping: address_space which holds the pagecache and I/O vectors
980 * @ra: file_ra_state which holds the readahead state
981 * @file: Used by the filesystem for authentication.
982 * @page: The page at @index which triggered the readahead call.
983 * @index: Index of first page to be read.
984 * @req_count: Total number of pages being read by the caller.
986 * page_cache_async_readahead() should be called when a page is used which
987 * is marked as PageReadahead; this is a marker to suggest that the application
988 * has used up enough of the readahead window that we should start pulling in
992 void page_cache_async_readahead(struct address_space *mapping,
993 struct file_ra_state *ra, struct file *file,
994 struct page *page, pgoff_t index, unsigned long req_count)
996 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
997 page_cache_async_ra(&ractl, page, req_count);
1000 static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1002 struct folio *folio;
1004 BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1005 ractl->_nr_pages -= ractl->_batch_count;
1006 ractl->_index += ractl->_batch_count;
1008 if (!ractl->_nr_pages) {
1009 ractl->_batch_count = 0;
1013 folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1014 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1015 ractl->_batch_count = folio_nr_pages(folio);
1021 * readahead_page - Get the next page to read.
1022 * @ractl: The current readahead request.
1024 * Context: The page is locked and has an elevated refcount. The caller
1025 * should decreases the refcount once the page has been submitted for I/O
1026 * and unlock the page once all I/O to that page has completed.
1027 * Return: A pointer to the next page, or %NULL if we are done.
1029 static inline struct page *readahead_page(struct readahead_control *ractl)
1031 struct folio *folio = __readahead_folio(ractl);
1033 return &folio->page;
1037 * readahead_folio - Get the next folio to read.
1038 * @ractl: The current readahead request.
1040 * Context: The folio is locked. The caller should unlock the folio once
1041 * all I/O to that folio has completed.
1042 * Return: A pointer to the next folio, or %NULL if we are done.
1044 static inline struct folio *readahead_folio(struct readahead_control *ractl)
1046 struct folio *folio = __readahead_folio(ractl);
1053 static inline unsigned int __readahead_batch(struct readahead_control *rac,
1054 struct page **array, unsigned int array_sz)
1057 XA_STATE(xas, &rac->mapping->i_pages, 0);
1060 BUG_ON(rac->_batch_count > rac->_nr_pages);
1061 rac->_nr_pages -= rac->_batch_count;
1062 rac->_index += rac->_batch_count;
1063 rac->_batch_count = 0;
1065 xas_set(&xas, rac->_index);
1067 xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1068 if (xas_retry(&xas, page))
1070 VM_BUG_ON_PAGE(!PageLocked(page), page);
1071 VM_BUG_ON_PAGE(PageTail(page), page);
1073 rac->_batch_count += thp_nr_pages(page);
1076 * The page cache isn't using multi-index entries yet,
1077 * so the xas cursor needs to be manually moved to the
1078 * next index. This can be removed once the page cache
1082 xas_set(&xas, rac->_index + rac->_batch_count);
1093 * readahead_page_batch - Get a batch of pages to read.
1094 * @rac: The current readahead request.
1095 * @array: An array of pointers to struct page.
1097 * Context: The pages are locked and have an elevated refcount. The caller
1098 * should decreases the refcount once the page has been submitted for I/O
1099 * and unlock the page once all I/O to that page has completed.
1100 * Return: The number of pages placed in the array. 0 indicates the request
1103 #define readahead_page_batch(rac, array) \
1104 __readahead_batch(rac, array, ARRAY_SIZE(array))
1107 * readahead_pos - The byte offset into the file of this readahead request.
1108 * @rac: The readahead request.
1110 static inline loff_t readahead_pos(struct readahead_control *rac)
1112 return (loff_t)rac->_index * PAGE_SIZE;
1116 * readahead_length - The number of bytes in this readahead request.
1117 * @rac: The readahead request.
1119 static inline size_t readahead_length(struct readahead_control *rac)
1121 return rac->_nr_pages * PAGE_SIZE;
1125 * readahead_index - The index of the first page in this readahead request.
1126 * @rac: The readahead request.
1128 static inline pgoff_t readahead_index(struct readahead_control *rac)
1134 * readahead_count - The number of pages in this readahead request.
1135 * @rac: The readahead request.
1137 static inline unsigned int readahead_count(struct readahead_control *rac)
1139 return rac->_nr_pages;
1143 * readahead_batch_length - The number of bytes in the current batch.
1144 * @rac: The readahead request.
1146 static inline size_t readahead_batch_length(struct readahead_control *rac)
1148 return rac->_batch_count * PAGE_SIZE;
1151 static inline unsigned long dir_pages(struct inode *inode)
1153 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1158 * folio_mkwrite_check_truncate - check if folio was truncated
1159 * @folio: the folio to check
1160 * @inode: the inode to check the folio against
1162 * Return: the number of bytes in the folio up to EOF,
1163 * or -EFAULT if the folio was truncated.
1165 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1166 struct inode *inode)
1168 loff_t size = i_size_read(inode);
1169 pgoff_t index = size >> PAGE_SHIFT;
1170 size_t offset = offset_in_folio(folio, size);
1172 if (!folio->mapping)
1175 /* folio is wholly inside EOF */
1176 if (folio_next_index(folio) - 1 < index)
1177 return folio_size(folio);
1178 /* folio is wholly past EOF */
1179 if (folio->index > index || !offset)
1181 /* folio is partially inside EOF */
1186 * page_mkwrite_check_truncate - check if page was truncated
1187 * @page: the page to check
1188 * @inode: the inode to check the page against
1190 * Returns the number of bytes in the page up to EOF,
1191 * or -EFAULT if the page was truncated.
1193 static inline int page_mkwrite_check_truncate(struct page *page,
1194 struct inode *inode)
1196 loff_t size = i_size_read(inode);
1197 pgoff_t index = size >> PAGE_SHIFT;
1198 int offset = offset_in_page(size);
1200 if (page->mapping != inode->i_mapping)
1203 /* page is wholly inside EOF */
1204 if (page->index < index)
1206 /* page is wholly past EOF */
1207 if (page->index > index || !offset)
1209 /* page is partially inside EOF */
1214 * i_blocks_per_folio - How many blocks fit in this folio.
1215 * @inode: The inode which contains the blocks.
1216 * @folio: The folio.
1218 * If the block size is larger than the size of this folio, return zero.
1220 * Context: The caller should hold a refcount on the folio to prevent it
1222 * Return: The number of filesystem blocks covered by this folio.
1225 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1227 return folio_size(folio) >> inode->i_blkbits;
1231 unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1233 return i_blocks_per_folio(inode, page_folio(page));
1235 #endif /* _LINUX_PAGEMAP_H */