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
304 #define FGP_STABLE 0x00000200
306 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
307 int fgp_flags, gfp_t gfp);
308 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
309 int fgp_flags, gfp_t gfp);
312 * filemap_get_folio - Find and get a folio.
313 * @mapping: The address_space to search.
314 * @index: The page index.
316 * Looks up the page cache entry at @mapping & @index. If a folio is
317 * present, it is returned with an increased refcount.
319 * Otherwise, %NULL is returned.
321 static inline struct folio *filemap_get_folio(struct address_space *mapping,
324 return __filemap_get_folio(mapping, index, 0, 0);
328 * find_get_page - find and get a page reference
329 * @mapping: the address_space to search
330 * @offset: the page index
332 * Looks up the page cache slot at @mapping & @offset. If there is a
333 * page cache page, it is returned with an increased refcount.
335 * Otherwise, %NULL is returned.
337 static inline struct page *find_get_page(struct address_space *mapping,
340 return pagecache_get_page(mapping, offset, 0, 0);
343 static inline struct page *find_get_page_flags(struct address_space *mapping,
344 pgoff_t offset, int fgp_flags)
346 return pagecache_get_page(mapping, offset, fgp_flags, 0);
350 * find_lock_page - locate, pin and lock a pagecache page
351 * @mapping: the address_space to search
352 * @index: the page index
354 * Looks up the page cache entry at @mapping & @index. If there is a
355 * page cache page, it is returned locked and with an increased
358 * Context: May sleep.
359 * Return: A struct page or %NULL if there is no page in the cache for this
362 static inline struct page *find_lock_page(struct address_space *mapping,
365 return pagecache_get_page(mapping, index, FGP_LOCK, 0);
369 * find_or_create_page - locate or add a pagecache page
370 * @mapping: the page's address_space
371 * @index: the page's index into the mapping
372 * @gfp_mask: page allocation mode
374 * Looks up the page cache slot at @mapping & @offset. If there is a
375 * page cache page, it is returned locked and with an increased
378 * If the page is not present, a new page is allocated using @gfp_mask
379 * and added to the page cache and the VM's LRU list. The page is
380 * returned locked and with an increased refcount.
382 * On memory exhaustion, %NULL is returned.
384 * find_or_create_page() may sleep, even if @gfp_flags specifies an
387 static inline struct page *find_or_create_page(struct address_space *mapping,
388 pgoff_t index, gfp_t gfp_mask)
390 return pagecache_get_page(mapping, index,
391 FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
396 * grab_cache_page_nowait - returns locked page at given index in given cache
397 * @mapping: target address_space
398 * @index: the page index
400 * Same as grab_cache_page(), but do not wait if the page is unavailable.
401 * This is intended for speculative data generators, where the data can
402 * be regenerated if the page couldn't be grabbed. This routine should
403 * be safe to call while holding the lock for another page.
405 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
406 * and deadlock against the caller's locked page.
408 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
411 return pagecache_get_page(mapping, index,
412 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
413 mapping_gfp_mask(mapping));
416 /* Does this page contain this index? */
417 static inline bool thp_contains(struct page *head, pgoff_t index)
419 /* HugeTLBfs indexes the page cache in units of hpage_size */
421 return head->index == index;
422 return page_index(head) == (index & ~(thp_nr_pages(head) - 1UL));
425 #define swapcache_index(folio) __page_file_index(&(folio)->page)
428 * folio_index - File index of a folio.
431 * For a folio which is either in the page cache or the swap cache,
432 * return its index within the address_space it belongs to. If you know
433 * the page is definitely in the page cache, you can look at the folio's
436 * Return: The index (offset in units of pages) of a folio in its file.
438 static inline pgoff_t folio_index(struct folio *folio)
440 if (unlikely(folio_test_swapcache(folio)))
441 return swapcache_index(folio);
446 * folio_next_index - Get the index of the next folio.
447 * @folio: The current folio.
449 * Return: The index of the folio which follows this folio in the file.
451 static inline pgoff_t folio_next_index(struct folio *folio)
453 return folio->index + folio_nr_pages(folio);
457 * folio_file_page - The page for a particular index.
458 * @folio: The folio which contains this index.
459 * @index: The index we want to look up.
461 * Sometimes after looking up a folio in the page cache, we need to
462 * obtain the specific page for an index (eg a page fault).
464 * Return: The page containing the file data for this index.
466 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
468 /* HugeTLBfs indexes the page cache in units of hpage_size */
469 if (folio_test_hugetlb(folio))
471 return folio_page(folio, index & (folio_nr_pages(folio) - 1));
475 * folio_contains - Does this folio contain this index?
477 * @index: The page index within the file.
479 * Context: The caller should have the page locked in order to prevent
480 * (eg) shmem from moving the page between the page cache and swap cache
481 * and changing its index in the middle of the operation.
482 * Return: true or false.
484 static inline bool folio_contains(struct folio *folio, pgoff_t index)
486 /* HugeTLBfs indexes the page cache in units of hpage_size */
487 if (folio_test_hugetlb(folio))
488 return folio->index == index;
489 return index - folio_index(folio) < folio_nr_pages(folio);
493 * Given the page we found in the page cache, return the page corresponding
494 * to this index in the file
496 static inline struct page *find_subpage(struct page *head, pgoff_t index)
498 /* HugeTLBfs wants the head page regardless */
502 return head + (index & (thp_nr_pages(head) - 1));
505 unsigned find_get_entries(struct address_space *mapping, pgoff_t start,
506 pgoff_t end, struct pagevec *pvec, pgoff_t *indices);
507 unsigned find_get_pages_range(struct address_space *mapping, pgoff_t *start,
508 pgoff_t end, unsigned int nr_pages,
509 struct page **pages);
510 static inline unsigned find_get_pages(struct address_space *mapping,
511 pgoff_t *start, unsigned int nr_pages,
514 return find_get_pages_range(mapping, start, (pgoff_t)-1, nr_pages,
517 unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t start,
518 unsigned int nr_pages, struct page **pages);
519 unsigned find_get_pages_range_tag(struct address_space *mapping, pgoff_t *index,
520 pgoff_t end, xa_mark_t tag, unsigned int nr_pages,
521 struct page **pages);
522 static inline unsigned find_get_pages_tag(struct address_space *mapping,
523 pgoff_t *index, xa_mark_t tag, unsigned int nr_pages,
526 return find_get_pages_range_tag(mapping, index, (pgoff_t)-1, tag,
530 struct page *grab_cache_page_write_begin(struct address_space *mapping,
531 pgoff_t index, unsigned flags);
534 * Returns locked page at given index in given cache, creating it if needed.
536 static inline struct page *grab_cache_page(struct address_space *mapping,
539 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
542 extern struct page * read_cache_page(struct address_space *mapping,
543 pgoff_t index, filler_t *filler, void *data);
544 extern struct page * read_cache_page_gfp(struct address_space *mapping,
545 pgoff_t index, gfp_t gfp_mask);
546 extern int read_cache_pages(struct address_space *mapping,
547 struct list_head *pages, filler_t *filler, void *data);
549 static inline struct page *read_mapping_page(struct address_space *mapping,
550 pgoff_t index, void *data)
552 return read_cache_page(mapping, index, NULL, data);
556 * Get index of the page within radix-tree (but not for hugetlb pages).
557 * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
559 static inline pgoff_t page_to_index(struct page *page)
563 if (likely(!PageTransTail(page)))
566 head = compound_head(page);
568 * We don't initialize ->index for tail pages: calculate based on
571 return head->index + page - head;
574 extern pgoff_t hugetlb_basepage_index(struct page *page);
577 * Get the offset in PAGE_SIZE (even for hugetlb pages).
578 * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
580 static inline pgoff_t page_to_pgoff(struct page *page)
582 if (unlikely(PageHuge(page)))
583 return hugetlb_basepage_index(page);
584 return page_to_index(page);
588 * Return byte-offset into filesystem object for page.
590 static inline loff_t page_offset(struct page *page)
592 return ((loff_t)page->index) << PAGE_SHIFT;
595 static inline loff_t page_file_offset(struct page *page)
597 return ((loff_t)page_index(page)) << PAGE_SHIFT;
601 * folio_pos - Returns the byte position of this folio in its file.
604 static inline loff_t folio_pos(struct folio *folio)
606 return page_offset(&folio->page);
610 * folio_file_pos - Returns the byte position of this folio in its file.
613 * This differs from folio_pos() for folios which belong to a swap file.
614 * NFS is the only filesystem today which needs to use folio_file_pos().
616 static inline loff_t folio_file_pos(struct folio *folio)
618 return page_file_offset(&folio->page);
621 extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
622 unsigned long address);
624 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
625 unsigned long address)
628 if (unlikely(is_vm_hugetlb_page(vma)))
629 return linear_hugepage_index(vma, address);
630 pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
631 pgoff += vma->vm_pgoff;
635 struct wait_page_key {
641 struct wait_page_queue {
644 wait_queue_entry_t wait;
647 static inline bool wake_page_match(struct wait_page_queue *wait_page,
648 struct wait_page_key *key)
650 if (wait_page->folio != key->folio)
654 if (wait_page->bit_nr != key->bit_nr)
660 void __folio_lock(struct folio *folio);
661 int __folio_lock_killable(struct folio *folio);
662 bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
664 void unlock_page(struct page *page);
665 void folio_unlock(struct folio *folio);
667 static inline bool folio_trylock(struct folio *folio)
669 return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
673 * Return true if the page was successfully locked
675 static inline int trylock_page(struct page *page)
677 return folio_trylock(page_folio(page));
680 static inline void folio_lock(struct folio *folio)
683 if (!folio_trylock(folio))
688 * lock_page may only be called if we have the page's inode pinned.
690 static inline void lock_page(struct page *page)
695 folio = page_folio(page);
696 if (!folio_trylock(folio))
700 static inline int folio_lock_killable(struct folio *folio)
703 if (!folio_trylock(folio))
704 return __folio_lock_killable(folio);
709 * lock_page_killable is like lock_page but can be interrupted by fatal
710 * signals. It returns 0 if it locked the page and -EINTR if it was
711 * killed while waiting.
713 static inline int lock_page_killable(struct page *page)
715 return folio_lock_killable(page_folio(page));
719 * lock_page_or_retry - Lock the page, unless this would block and the
720 * caller indicated that it can handle a retry.
722 * Return value and mmap_lock implications depend on flags; see
723 * __folio_lock_or_retry().
725 static inline bool lock_page_or_retry(struct page *page, struct mm_struct *mm,
731 folio = page_folio(page);
732 return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
736 * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
737 * and should not be used directly.
739 void folio_wait_bit(struct folio *folio, int bit_nr);
740 int folio_wait_bit_killable(struct folio *folio, int bit_nr);
743 * Wait for a folio to be unlocked.
745 * This must be called with the caller "holding" the folio,
746 * ie with increased "page->count" so that the folio won't
747 * go away during the wait..
749 static inline void folio_wait_locked(struct folio *folio)
751 if (folio_test_locked(folio))
752 folio_wait_bit(folio, PG_locked);
755 static inline int folio_wait_locked_killable(struct folio *folio)
757 if (!folio_test_locked(folio))
759 return folio_wait_bit_killable(folio, PG_locked);
762 static inline void wait_on_page_locked(struct page *page)
764 folio_wait_locked(page_folio(page));
767 static inline int wait_on_page_locked_killable(struct page *page)
769 return folio_wait_locked_killable(page_folio(page));
772 int put_and_wait_on_page_locked(struct page *page, int state);
773 void wait_on_page_writeback(struct page *page);
774 void folio_wait_writeback(struct folio *folio);
775 int folio_wait_writeback_killable(struct folio *folio);
776 void end_page_writeback(struct page *page);
777 void folio_end_writeback(struct folio *folio);
778 void wait_for_stable_page(struct page *page);
779 void folio_wait_stable(struct folio *folio);
780 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
781 static inline void __set_page_dirty(struct page *page,
782 struct address_space *mapping, int warn)
784 __folio_mark_dirty(page_folio(page), mapping, warn);
786 void folio_account_cleaned(struct folio *folio, struct address_space *mapping,
787 struct bdi_writeback *wb);
788 static inline void account_page_cleaned(struct page *page,
789 struct address_space *mapping, struct bdi_writeback *wb)
791 return folio_account_cleaned(page_folio(page), mapping, wb);
793 void __folio_cancel_dirty(struct folio *folio);
794 static inline void folio_cancel_dirty(struct folio *folio)
796 /* Avoid atomic ops, locking, etc. when not actually needed. */
797 if (folio_test_dirty(folio))
798 __folio_cancel_dirty(folio);
800 static inline void cancel_dirty_page(struct page *page)
802 folio_cancel_dirty(page_folio(page));
804 bool folio_clear_dirty_for_io(struct folio *folio);
805 bool clear_page_dirty_for_io(struct page *page);
806 int __must_check folio_write_one(struct folio *folio);
807 static inline int __must_check write_one_page(struct page *page)
809 return folio_write_one(page_folio(page));
812 int __set_page_dirty_nobuffers(struct page *page);
813 int __set_page_dirty_no_writeback(struct page *page);
815 void page_endio(struct page *page, bool is_write, int err);
817 void folio_end_private_2(struct folio *folio);
818 void folio_wait_private_2(struct folio *folio);
819 int folio_wait_private_2_killable(struct folio *folio);
822 * Add an arbitrary waiter to a page's wait queue
824 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
827 * Fault in userspace address range.
829 size_t fault_in_writeable(char __user *uaddr, size_t size);
830 size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
831 size_t fault_in_readable(const char __user *uaddr, size_t size);
833 int add_to_page_cache_locked(struct page *page, struct address_space *mapping,
834 pgoff_t index, gfp_t gfp);
835 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
836 pgoff_t index, gfp_t gfp);
837 int filemap_add_folio(struct address_space *mapping, struct folio *folio,
838 pgoff_t index, gfp_t gfp);
839 extern void delete_from_page_cache(struct page *page);
840 extern void __delete_from_page_cache(struct page *page, void *shadow);
841 void replace_page_cache_page(struct page *old, struct page *new);
842 void delete_from_page_cache_batch(struct address_space *mapping,
843 struct pagevec *pvec);
844 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
848 * Like add_to_page_cache_locked, but used to add newly allocated pages:
849 * the page is new, so we can just run __SetPageLocked() against it.
851 static inline int add_to_page_cache(struct page *page,
852 struct address_space *mapping, pgoff_t offset, gfp_t gfp_mask)
856 __SetPageLocked(page);
857 error = add_to_page_cache_locked(page, mapping, offset, gfp_mask);
859 __ClearPageLocked(page);
863 /* Must be non-static for BPF error injection */
864 int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
865 pgoff_t index, gfp_t gfp, void **shadowp);
868 * struct readahead_control - Describes a readahead request.
870 * A readahead request is for consecutive pages. Filesystems which
871 * implement the ->readahead method should call readahead_page() or
872 * readahead_page_batch() in a loop and attempt to start I/O against
873 * each page in the request.
875 * Most of the fields in this struct are private and should be accessed
876 * by the functions below.
878 * @file: The file, used primarily by network filesystems for authentication.
879 * May be NULL if invoked internally by the filesystem.
880 * @mapping: Readahead this filesystem object.
881 * @ra: File readahead state. May be NULL.
883 struct readahead_control {
885 struct address_space *mapping;
886 struct file_ra_state *ra;
887 /* private: use the readahead_* accessors instead */
889 unsigned int _nr_pages;
890 unsigned int _batch_count;
893 #define DEFINE_READAHEAD(ractl, f, r, m, i) \
894 struct readahead_control ractl = { \
901 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
903 void page_cache_ra_unbounded(struct readahead_control *,
904 unsigned long nr_to_read, unsigned long lookahead_count);
905 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
906 void page_cache_async_ra(struct readahead_control *, struct page *,
907 unsigned long req_count);
908 void readahead_expand(struct readahead_control *ractl,
909 loff_t new_start, size_t new_len);
912 * page_cache_sync_readahead - generic file readahead
913 * @mapping: address_space which holds the pagecache and I/O vectors
914 * @ra: file_ra_state which holds the readahead state
915 * @file: Used by the filesystem for authentication.
916 * @index: Index of first page to be read.
917 * @req_count: Total number of pages being read by the caller.
919 * page_cache_sync_readahead() should be called when a cache miss happened:
920 * it will submit the read. The readahead logic may decide to piggyback more
921 * pages onto the read request if access patterns suggest it will improve
925 void page_cache_sync_readahead(struct address_space *mapping,
926 struct file_ra_state *ra, struct file *file, pgoff_t index,
927 unsigned long req_count)
929 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
930 page_cache_sync_ra(&ractl, req_count);
934 * page_cache_async_readahead - file readahead for marked pages
935 * @mapping: address_space which holds the pagecache and I/O vectors
936 * @ra: file_ra_state which holds the readahead state
937 * @file: Used by the filesystem for authentication.
938 * @page: The page at @index which triggered the readahead call.
939 * @index: Index of first page to be read.
940 * @req_count: Total number of pages being read by the caller.
942 * page_cache_async_readahead() should be called when a page is used which
943 * is marked as PageReadahead; this is a marker to suggest that the application
944 * has used up enough of the readahead window that we should start pulling in
948 void page_cache_async_readahead(struct address_space *mapping,
949 struct file_ra_state *ra, struct file *file,
950 struct page *page, pgoff_t index, unsigned long req_count)
952 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
953 page_cache_async_ra(&ractl, page, req_count);
956 static inline struct folio *__readahead_folio(struct readahead_control *ractl)
960 BUG_ON(ractl->_batch_count > ractl->_nr_pages);
961 ractl->_nr_pages -= ractl->_batch_count;
962 ractl->_index += ractl->_batch_count;
964 if (!ractl->_nr_pages) {
965 ractl->_batch_count = 0;
969 folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
970 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
971 ractl->_batch_count = folio_nr_pages(folio);
977 * readahead_page - Get the next page to read.
978 * @ractl: The current readahead request.
980 * Context: The page is locked and has an elevated refcount. The caller
981 * should decreases the refcount once the page has been submitted for I/O
982 * and unlock the page once all I/O to that page has completed.
983 * Return: A pointer to the next page, or %NULL if we are done.
985 static inline struct page *readahead_page(struct readahead_control *ractl)
987 struct folio *folio = __readahead_folio(ractl);
993 * readahead_folio - Get the next folio to read.
994 * @ractl: The current readahead request.
996 * Context: The folio is locked. The caller should unlock the folio once
997 * all I/O to that folio has completed.
998 * Return: A pointer to the next folio, or %NULL if we are done.
1000 static inline struct folio *readahead_folio(struct readahead_control *ractl)
1002 struct folio *folio = __readahead_folio(ractl);
1009 static inline unsigned int __readahead_batch(struct readahead_control *rac,
1010 struct page **array, unsigned int array_sz)
1013 XA_STATE(xas, &rac->mapping->i_pages, 0);
1016 BUG_ON(rac->_batch_count > rac->_nr_pages);
1017 rac->_nr_pages -= rac->_batch_count;
1018 rac->_index += rac->_batch_count;
1019 rac->_batch_count = 0;
1021 xas_set(&xas, rac->_index);
1023 xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1024 if (xas_retry(&xas, page))
1026 VM_BUG_ON_PAGE(!PageLocked(page), page);
1027 VM_BUG_ON_PAGE(PageTail(page), page);
1029 rac->_batch_count += thp_nr_pages(page);
1032 * The page cache isn't using multi-index entries yet,
1033 * so the xas cursor needs to be manually moved to the
1034 * next index. This can be removed once the page cache
1038 xas_set(&xas, rac->_index + rac->_batch_count);
1049 * readahead_page_batch - Get a batch of pages to read.
1050 * @rac: The current readahead request.
1051 * @array: An array of pointers to struct page.
1053 * Context: The pages are locked and have an elevated refcount. The caller
1054 * should decreases the refcount once the page has been submitted for I/O
1055 * and unlock the page once all I/O to that page has completed.
1056 * Return: The number of pages placed in the array. 0 indicates the request
1059 #define readahead_page_batch(rac, array) \
1060 __readahead_batch(rac, array, ARRAY_SIZE(array))
1063 * readahead_pos - The byte offset into the file of this readahead request.
1064 * @rac: The readahead request.
1066 static inline loff_t readahead_pos(struct readahead_control *rac)
1068 return (loff_t)rac->_index * PAGE_SIZE;
1072 * readahead_length - The number of bytes in this readahead request.
1073 * @rac: The readahead request.
1075 static inline size_t readahead_length(struct readahead_control *rac)
1077 return rac->_nr_pages * PAGE_SIZE;
1081 * readahead_index - The index of the first page in this readahead request.
1082 * @rac: The readahead request.
1084 static inline pgoff_t readahead_index(struct readahead_control *rac)
1090 * readahead_count - The number of pages in this readahead request.
1091 * @rac: The readahead request.
1093 static inline unsigned int readahead_count(struct readahead_control *rac)
1095 return rac->_nr_pages;
1099 * readahead_batch_length - The number of bytes in the current batch.
1100 * @rac: The readahead request.
1102 static inline size_t readahead_batch_length(struct readahead_control *rac)
1104 return rac->_batch_count * PAGE_SIZE;
1107 static inline unsigned long dir_pages(struct inode *inode)
1109 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1114 * folio_mkwrite_check_truncate - check if folio was truncated
1115 * @folio: the folio to check
1116 * @inode: the inode to check the folio against
1118 * Return: the number of bytes in the folio up to EOF,
1119 * or -EFAULT if the folio was truncated.
1121 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1122 struct inode *inode)
1124 loff_t size = i_size_read(inode);
1125 pgoff_t index = size >> PAGE_SHIFT;
1126 size_t offset = offset_in_folio(folio, size);
1128 if (!folio->mapping)
1131 /* folio is wholly inside EOF */
1132 if (folio_next_index(folio) - 1 < index)
1133 return folio_size(folio);
1134 /* folio is wholly past EOF */
1135 if (folio->index > index || !offset)
1137 /* folio is partially inside EOF */
1142 * page_mkwrite_check_truncate - check if page was truncated
1143 * @page: the page to check
1144 * @inode: the inode to check the page against
1146 * Returns the number of bytes in the page up to EOF,
1147 * or -EFAULT if the page was truncated.
1149 static inline int page_mkwrite_check_truncate(struct page *page,
1150 struct inode *inode)
1152 loff_t size = i_size_read(inode);
1153 pgoff_t index = size >> PAGE_SHIFT;
1154 int offset = offset_in_page(size);
1156 if (page->mapping != inode->i_mapping)
1159 /* page is wholly inside EOF */
1160 if (page->index < index)
1162 /* page is wholly past EOF */
1163 if (page->index > index || !offset)
1165 /* page is partially inside EOF */
1170 * i_blocks_per_folio - How many blocks fit in this folio.
1171 * @inode: The inode which contains the blocks.
1172 * @folio: The folio.
1174 * If the block size is larger than the size of this folio, return zero.
1176 * Context: The caller should hold a refcount on the folio to prevent it
1178 * Return: The number of filesystem blocks covered by this folio.
1181 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1183 return folio_size(folio) >> inode->i_blkbits;
1187 unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1189 return i_blocks_per_folio(inode, page_folio(page));
1191 #endif /* _LINUX_PAGEMAP_H */