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);
256 static inline bool page_cache_add_speculative(struct page *page, int count)
258 VM_BUG_ON_PAGE(PageTail(page), page);
259 return folio_ref_try_add_rcu((struct folio *)page, count);
262 static inline bool page_cache_get_speculative(struct page *page)
264 return page_cache_add_speculative(page, 1);
268 * folio_attach_private - Attach private data to a folio.
269 * @folio: Folio to attach data to.
270 * @data: Data to attach to folio.
272 * Attaching private data to a folio increments the page's reference count.
273 * The data must be detached before the folio will be freed.
275 static inline void folio_attach_private(struct folio *folio, void *data)
278 folio->private = data;
279 folio_set_private(folio);
283 * folio_detach_private - Detach private data from a folio.
284 * @folio: Folio to detach data from.
286 * Removes the data that was previously attached to the folio and decrements
287 * the refcount on the page.
289 * Return: Data that was attached to the folio.
291 static inline void *folio_detach_private(struct folio *folio)
293 void *data = folio_get_private(folio);
295 if (!folio_test_private(folio))
297 folio_clear_private(folio);
298 folio->private = NULL;
304 static inline void attach_page_private(struct page *page, void *data)
306 folio_attach_private(page_folio(page), data);
309 static inline void *detach_page_private(struct page *page)
311 return folio_detach_private(page_folio(page));
315 struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order);
317 static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order)
319 return folio_alloc(gfp, order);
323 static inline struct page *__page_cache_alloc(gfp_t gfp)
325 return &filemap_alloc_folio(gfp, 0)->page;
328 static inline struct page *page_cache_alloc(struct address_space *x)
330 return __page_cache_alloc(mapping_gfp_mask(x));
333 static inline gfp_t readahead_gfp_mask(struct address_space *x)
335 return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
338 typedef int filler_t(void *, struct page *);
340 pgoff_t page_cache_next_miss(struct address_space *mapping,
341 pgoff_t index, unsigned long max_scan);
342 pgoff_t page_cache_prev_miss(struct address_space *mapping,
343 pgoff_t index, unsigned long max_scan);
345 #define FGP_ACCESSED 0x00000001
346 #define FGP_LOCK 0x00000002
347 #define FGP_CREAT 0x00000004
348 #define FGP_WRITE 0x00000008
349 #define FGP_NOFS 0x00000010
350 #define FGP_NOWAIT 0x00000020
351 #define FGP_FOR_MMAP 0x00000040
352 #define FGP_HEAD 0x00000080
353 #define FGP_ENTRY 0x00000100
354 #define FGP_STABLE 0x00000200
356 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
357 int fgp_flags, gfp_t gfp);
358 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
359 int fgp_flags, gfp_t gfp);
362 * filemap_get_folio - Find and get a folio.
363 * @mapping: The address_space to search.
364 * @index: The page index.
366 * Looks up the page cache entry at @mapping & @index. If a folio is
367 * present, it is returned with an increased refcount.
369 * Otherwise, %NULL is returned.
371 static inline struct folio *filemap_get_folio(struct address_space *mapping,
374 return __filemap_get_folio(mapping, index, 0, 0);
378 * find_get_page - find and get a page reference
379 * @mapping: the address_space to search
380 * @offset: the page index
382 * Looks up the page cache slot at @mapping & @offset. If there is a
383 * page cache page, it is returned with an increased refcount.
385 * Otherwise, %NULL is returned.
387 static inline struct page *find_get_page(struct address_space *mapping,
390 return pagecache_get_page(mapping, offset, 0, 0);
393 static inline struct page *find_get_page_flags(struct address_space *mapping,
394 pgoff_t offset, int fgp_flags)
396 return pagecache_get_page(mapping, offset, fgp_flags, 0);
400 * find_lock_page - locate, pin and lock a pagecache page
401 * @mapping: the address_space to search
402 * @index: the page index
404 * Looks up the page cache entry at @mapping & @index. If there is a
405 * page cache page, it is returned locked and with an increased
408 * Context: May sleep.
409 * Return: A struct page or %NULL if there is no page in the cache for this
412 static inline struct page *find_lock_page(struct address_space *mapping,
415 return pagecache_get_page(mapping, index, FGP_LOCK, 0);
419 * find_or_create_page - locate or add a pagecache page
420 * @mapping: the page's address_space
421 * @index: the page's index into the mapping
422 * @gfp_mask: page allocation mode
424 * Looks up the page cache slot at @mapping & @offset. If there is a
425 * page cache page, it is returned locked and with an increased
428 * If the page is not present, a new page is allocated using @gfp_mask
429 * and added to the page cache and the VM's LRU list. The page is
430 * returned locked and with an increased refcount.
432 * On memory exhaustion, %NULL is returned.
434 * find_or_create_page() may sleep, even if @gfp_flags specifies an
437 static inline struct page *find_or_create_page(struct address_space *mapping,
438 pgoff_t index, gfp_t gfp_mask)
440 return pagecache_get_page(mapping, index,
441 FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
446 * grab_cache_page_nowait - returns locked page at given index in given cache
447 * @mapping: target address_space
448 * @index: the page index
450 * Same as grab_cache_page(), but do not wait if the page is unavailable.
451 * This is intended for speculative data generators, where the data can
452 * be regenerated if the page couldn't be grabbed. This routine should
453 * be safe to call while holding the lock for another page.
455 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
456 * and deadlock against the caller's locked page.
458 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
461 return pagecache_get_page(mapping, index,
462 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
463 mapping_gfp_mask(mapping));
466 /* Does this page contain this index? */
467 static inline bool thp_contains(struct page *head, pgoff_t index)
469 /* HugeTLBfs indexes the page cache in units of hpage_size */
471 return head->index == index;
472 return page_index(head) == (index & ~(thp_nr_pages(head) - 1UL));
475 #define swapcache_index(folio) __page_file_index(&(folio)->page)
478 * folio_index - File index of a folio.
481 * For a folio which is either in the page cache or the swap cache,
482 * return its index within the address_space it belongs to. If you know
483 * the page is definitely in the page cache, you can look at the folio's
486 * Return: The index (offset in units of pages) of a folio in its file.
488 static inline pgoff_t folio_index(struct folio *folio)
490 if (unlikely(folio_test_swapcache(folio)))
491 return swapcache_index(folio);
496 * folio_next_index - Get the index of the next folio.
497 * @folio: The current folio.
499 * Return: The index of the folio which follows this folio in the file.
501 static inline pgoff_t folio_next_index(struct folio *folio)
503 return folio->index + folio_nr_pages(folio);
507 * folio_file_page - The page for a particular index.
508 * @folio: The folio which contains this index.
509 * @index: The index we want to look up.
511 * Sometimes after looking up a folio in the page cache, we need to
512 * obtain the specific page for an index (eg a page fault).
514 * Return: The page containing the file data for this index.
516 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
518 /* HugeTLBfs indexes the page cache in units of hpage_size */
519 if (folio_test_hugetlb(folio))
521 return folio_page(folio, index & (folio_nr_pages(folio) - 1));
525 * folio_contains - Does this folio contain this index?
527 * @index: The page index within the file.
529 * Context: The caller should have the page locked in order to prevent
530 * (eg) shmem from moving the page between the page cache and swap cache
531 * and changing its index in the middle of the operation.
532 * Return: true or false.
534 static inline bool folio_contains(struct folio *folio, pgoff_t index)
536 /* HugeTLBfs indexes the page cache in units of hpage_size */
537 if (folio_test_hugetlb(folio))
538 return folio->index == index;
539 return index - folio_index(folio) < folio_nr_pages(folio);
543 * Given the page we found in the page cache, return the page corresponding
544 * to this index in the file
546 static inline struct page *find_subpage(struct page *head, pgoff_t index)
548 /* HugeTLBfs wants the head page regardless */
552 return head + (index & (thp_nr_pages(head) - 1));
555 unsigned find_get_entries(struct address_space *mapping, pgoff_t start,
556 pgoff_t end, struct pagevec *pvec, pgoff_t *indices);
557 unsigned find_get_pages_range(struct address_space *mapping, pgoff_t *start,
558 pgoff_t end, unsigned int nr_pages,
559 struct page **pages);
560 static inline unsigned find_get_pages(struct address_space *mapping,
561 pgoff_t *start, unsigned int nr_pages,
564 return find_get_pages_range(mapping, start, (pgoff_t)-1, nr_pages,
567 unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t start,
568 unsigned int nr_pages, struct page **pages);
569 unsigned find_get_pages_range_tag(struct address_space *mapping, pgoff_t *index,
570 pgoff_t end, xa_mark_t tag, unsigned int nr_pages,
571 struct page **pages);
572 static inline unsigned find_get_pages_tag(struct address_space *mapping,
573 pgoff_t *index, xa_mark_t tag, unsigned int nr_pages,
576 return find_get_pages_range_tag(mapping, index, (pgoff_t)-1, tag,
580 struct page *grab_cache_page_write_begin(struct address_space *mapping,
581 pgoff_t index, unsigned flags);
584 * Returns locked page at given index in given cache, creating it if needed.
586 static inline struct page *grab_cache_page(struct address_space *mapping,
589 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
592 extern struct page * read_cache_page(struct address_space *mapping,
593 pgoff_t index, filler_t *filler, void *data);
594 extern struct page * read_cache_page_gfp(struct address_space *mapping,
595 pgoff_t index, gfp_t gfp_mask);
596 extern int read_cache_pages(struct address_space *mapping,
597 struct list_head *pages, filler_t *filler, void *data);
599 static inline struct page *read_mapping_page(struct address_space *mapping,
600 pgoff_t index, void *data)
602 return read_cache_page(mapping, index, NULL, data);
606 * Get index of the page within radix-tree (but not for hugetlb pages).
607 * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
609 static inline pgoff_t page_to_index(struct page *page)
613 if (likely(!PageTransTail(page)))
616 head = compound_head(page);
618 * We don't initialize ->index for tail pages: calculate based on
621 return head->index + page - head;
624 extern pgoff_t hugetlb_basepage_index(struct page *page);
627 * Get the offset in PAGE_SIZE (even for hugetlb pages).
628 * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
630 static inline pgoff_t page_to_pgoff(struct page *page)
632 if (unlikely(PageHuge(page)))
633 return hugetlb_basepage_index(page);
634 return page_to_index(page);
638 * Return byte-offset into filesystem object for page.
640 static inline loff_t page_offset(struct page *page)
642 return ((loff_t)page->index) << PAGE_SHIFT;
645 static inline loff_t page_file_offset(struct page *page)
647 return ((loff_t)page_index(page)) << PAGE_SHIFT;
651 * folio_pos - Returns the byte position of this folio in its file.
654 static inline loff_t folio_pos(struct folio *folio)
656 return page_offset(&folio->page);
660 * folio_file_pos - Returns the byte position of this folio in its file.
663 * This differs from folio_pos() for folios which belong to a swap file.
664 * NFS is the only filesystem today which needs to use folio_file_pos().
666 static inline loff_t folio_file_pos(struct folio *folio)
668 return page_file_offset(&folio->page);
671 extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
672 unsigned long address);
674 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
675 unsigned long address)
678 if (unlikely(is_vm_hugetlb_page(vma)))
679 return linear_hugepage_index(vma, address);
680 pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
681 pgoff += vma->vm_pgoff;
685 struct wait_page_key {
691 struct wait_page_queue {
694 wait_queue_entry_t wait;
697 static inline bool wake_page_match(struct wait_page_queue *wait_page,
698 struct wait_page_key *key)
700 if (wait_page->folio != key->folio)
704 if (wait_page->bit_nr != key->bit_nr)
710 void __folio_lock(struct folio *folio);
711 int __folio_lock_killable(struct folio *folio);
712 bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
714 void unlock_page(struct page *page);
715 void folio_unlock(struct folio *folio);
717 static inline bool folio_trylock(struct folio *folio)
719 return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
723 * Return true if the page was successfully locked
725 static inline int trylock_page(struct page *page)
727 return folio_trylock(page_folio(page));
730 static inline void folio_lock(struct folio *folio)
733 if (!folio_trylock(folio))
738 * lock_page may only be called if we have the page's inode pinned.
740 static inline void lock_page(struct page *page)
745 folio = page_folio(page);
746 if (!folio_trylock(folio))
750 static inline int folio_lock_killable(struct folio *folio)
753 if (!folio_trylock(folio))
754 return __folio_lock_killable(folio);
759 * lock_page_killable is like lock_page but can be interrupted by fatal
760 * signals. It returns 0 if it locked the page and -EINTR if it was
761 * killed while waiting.
763 static inline int lock_page_killable(struct page *page)
765 return folio_lock_killable(page_folio(page));
769 * lock_page_or_retry - Lock the page, unless this would block and the
770 * caller indicated that it can handle a retry.
772 * Return value and mmap_lock implications depend on flags; see
773 * __folio_lock_or_retry().
775 static inline bool lock_page_or_retry(struct page *page, struct mm_struct *mm,
781 folio = page_folio(page);
782 return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
786 * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
787 * and should not be used directly.
789 void folio_wait_bit(struct folio *folio, int bit_nr);
790 int folio_wait_bit_killable(struct folio *folio, int bit_nr);
793 * Wait for a folio to be unlocked.
795 * This must be called with the caller "holding" the folio,
796 * ie with increased "page->count" so that the folio won't
797 * go away during the wait..
799 static inline void folio_wait_locked(struct folio *folio)
801 if (folio_test_locked(folio))
802 folio_wait_bit(folio, PG_locked);
805 static inline int folio_wait_locked_killable(struct folio *folio)
807 if (!folio_test_locked(folio))
809 return folio_wait_bit_killable(folio, PG_locked);
812 static inline void wait_on_page_locked(struct page *page)
814 folio_wait_locked(page_folio(page));
817 static inline int wait_on_page_locked_killable(struct page *page)
819 return folio_wait_locked_killable(page_folio(page));
822 int put_and_wait_on_page_locked(struct page *page, int state);
823 void wait_on_page_writeback(struct page *page);
824 void folio_wait_writeback(struct folio *folio);
825 int folio_wait_writeback_killable(struct folio *folio);
826 void end_page_writeback(struct page *page);
827 void folio_end_writeback(struct folio *folio);
828 void wait_for_stable_page(struct page *page);
829 void folio_wait_stable(struct folio *folio);
830 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
831 static inline void __set_page_dirty(struct page *page,
832 struct address_space *mapping, int warn)
834 __folio_mark_dirty(page_folio(page), mapping, warn);
836 void folio_account_cleaned(struct folio *folio, struct address_space *mapping,
837 struct bdi_writeback *wb);
838 static inline void account_page_cleaned(struct page *page,
839 struct address_space *mapping, struct bdi_writeback *wb)
841 return folio_account_cleaned(page_folio(page), mapping, wb);
843 void __folio_cancel_dirty(struct folio *folio);
844 static inline void folio_cancel_dirty(struct folio *folio)
846 /* Avoid atomic ops, locking, etc. when not actually needed. */
847 if (folio_test_dirty(folio))
848 __folio_cancel_dirty(folio);
850 static inline void cancel_dirty_page(struct page *page)
852 folio_cancel_dirty(page_folio(page));
854 bool folio_clear_dirty_for_io(struct folio *folio);
855 bool clear_page_dirty_for_io(struct page *page);
856 int __must_check folio_write_one(struct folio *folio);
857 static inline int __must_check write_one_page(struct page *page)
859 return folio_write_one(page_folio(page));
862 int __set_page_dirty_nobuffers(struct page *page);
863 int __set_page_dirty_no_writeback(struct page *page);
865 void page_endio(struct page *page, bool is_write, int err);
867 void folio_end_private_2(struct folio *folio);
868 void folio_wait_private_2(struct folio *folio);
869 int folio_wait_private_2_killable(struct folio *folio);
872 * Add an arbitrary waiter to a page's wait queue
874 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
877 * Fault in userspace address range.
879 size_t fault_in_writeable(char __user *uaddr, size_t size);
880 size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
881 size_t fault_in_readable(const char __user *uaddr, size_t size);
883 int add_to_page_cache_locked(struct page *page, struct address_space *mapping,
884 pgoff_t index, gfp_t gfp);
885 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
886 pgoff_t index, gfp_t gfp);
887 int filemap_add_folio(struct address_space *mapping, struct folio *folio,
888 pgoff_t index, gfp_t gfp);
889 extern void delete_from_page_cache(struct page *page);
890 extern void __delete_from_page_cache(struct page *page, void *shadow);
891 void replace_page_cache_page(struct page *old, struct page *new);
892 void delete_from_page_cache_batch(struct address_space *mapping,
893 struct pagevec *pvec);
894 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
898 * Like add_to_page_cache_locked, but used to add newly allocated pages:
899 * the page is new, so we can just run __SetPageLocked() against it.
901 static inline int add_to_page_cache(struct page *page,
902 struct address_space *mapping, pgoff_t offset, gfp_t gfp_mask)
906 __SetPageLocked(page);
907 error = add_to_page_cache_locked(page, mapping, offset, gfp_mask);
909 __ClearPageLocked(page);
913 /* Must be non-static for BPF error injection */
914 int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
915 pgoff_t index, gfp_t gfp, void **shadowp);
918 * struct readahead_control - Describes a readahead request.
920 * A readahead request is for consecutive pages. Filesystems which
921 * implement the ->readahead method should call readahead_page() or
922 * readahead_page_batch() in a loop and attempt to start I/O against
923 * each page in the request.
925 * Most of the fields in this struct are private and should be accessed
926 * by the functions below.
928 * @file: The file, used primarily by network filesystems for authentication.
929 * May be NULL if invoked internally by the filesystem.
930 * @mapping: Readahead this filesystem object.
931 * @ra: File readahead state. May be NULL.
933 struct readahead_control {
935 struct address_space *mapping;
936 struct file_ra_state *ra;
937 /* private: use the readahead_* accessors instead */
939 unsigned int _nr_pages;
940 unsigned int _batch_count;
943 #define DEFINE_READAHEAD(ractl, f, r, m, i) \
944 struct readahead_control ractl = { \
951 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
953 void page_cache_ra_unbounded(struct readahead_control *,
954 unsigned long nr_to_read, unsigned long lookahead_count);
955 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
956 void page_cache_async_ra(struct readahead_control *, struct page *,
957 unsigned long req_count);
958 void readahead_expand(struct readahead_control *ractl,
959 loff_t new_start, size_t new_len);
962 * page_cache_sync_readahead - generic file readahead
963 * @mapping: address_space which holds the pagecache and I/O vectors
964 * @ra: file_ra_state which holds the readahead state
965 * @file: Used by the filesystem for authentication.
966 * @index: Index of first page to be read.
967 * @req_count: Total number of pages being read by the caller.
969 * page_cache_sync_readahead() should be called when a cache miss happened:
970 * it will submit the read. The readahead logic may decide to piggyback more
971 * pages onto the read request if access patterns suggest it will improve
975 void page_cache_sync_readahead(struct address_space *mapping,
976 struct file_ra_state *ra, struct file *file, pgoff_t index,
977 unsigned long req_count)
979 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
980 page_cache_sync_ra(&ractl, req_count);
984 * page_cache_async_readahead - file readahead for marked pages
985 * @mapping: address_space which holds the pagecache and I/O vectors
986 * @ra: file_ra_state which holds the readahead state
987 * @file: Used by the filesystem for authentication.
988 * @page: The page at @index which triggered the readahead call.
989 * @index: Index of first page to be read.
990 * @req_count: Total number of pages being read by the caller.
992 * page_cache_async_readahead() should be called when a page is used which
993 * is marked as PageReadahead; this is a marker to suggest that the application
994 * has used up enough of the readahead window that we should start pulling in
998 void page_cache_async_readahead(struct address_space *mapping,
999 struct file_ra_state *ra, struct file *file,
1000 struct page *page, pgoff_t index, unsigned long req_count)
1002 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1003 page_cache_async_ra(&ractl, page, req_count);
1006 static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1008 struct folio *folio;
1010 BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1011 ractl->_nr_pages -= ractl->_batch_count;
1012 ractl->_index += ractl->_batch_count;
1014 if (!ractl->_nr_pages) {
1015 ractl->_batch_count = 0;
1019 folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1020 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1021 ractl->_batch_count = folio_nr_pages(folio);
1027 * readahead_page - Get the next page to read.
1028 * @ractl: The current readahead request.
1030 * Context: The page is locked and has an elevated refcount. The caller
1031 * should decreases the refcount once the page has been submitted for I/O
1032 * and unlock the page once all I/O to that page has completed.
1033 * Return: A pointer to the next page, or %NULL if we are done.
1035 static inline struct page *readahead_page(struct readahead_control *ractl)
1037 struct folio *folio = __readahead_folio(ractl);
1039 return &folio->page;
1043 * readahead_folio - Get the next folio to read.
1044 * @ractl: The current readahead request.
1046 * Context: The folio is locked. The caller should unlock the folio once
1047 * all I/O to that folio has completed.
1048 * Return: A pointer to the next folio, or %NULL if we are done.
1050 static inline struct folio *readahead_folio(struct readahead_control *ractl)
1052 struct folio *folio = __readahead_folio(ractl);
1059 static inline unsigned int __readahead_batch(struct readahead_control *rac,
1060 struct page **array, unsigned int array_sz)
1063 XA_STATE(xas, &rac->mapping->i_pages, 0);
1066 BUG_ON(rac->_batch_count > rac->_nr_pages);
1067 rac->_nr_pages -= rac->_batch_count;
1068 rac->_index += rac->_batch_count;
1069 rac->_batch_count = 0;
1071 xas_set(&xas, rac->_index);
1073 xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1074 if (xas_retry(&xas, page))
1076 VM_BUG_ON_PAGE(!PageLocked(page), page);
1077 VM_BUG_ON_PAGE(PageTail(page), page);
1079 rac->_batch_count += thp_nr_pages(page);
1082 * The page cache isn't using multi-index entries yet,
1083 * so the xas cursor needs to be manually moved to the
1084 * next index. This can be removed once the page cache
1088 xas_set(&xas, rac->_index + rac->_batch_count);
1099 * readahead_page_batch - Get a batch of pages to read.
1100 * @rac: The current readahead request.
1101 * @array: An array of pointers to struct page.
1103 * Context: The pages are locked and have an elevated refcount. The caller
1104 * should decreases the refcount once the page has been submitted for I/O
1105 * and unlock the page once all I/O to that page has completed.
1106 * Return: The number of pages placed in the array. 0 indicates the request
1109 #define readahead_page_batch(rac, array) \
1110 __readahead_batch(rac, array, ARRAY_SIZE(array))
1113 * readahead_pos - The byte offset into the file of this readahead request.
1114 * @rac: The readahead request.
1116 static inline loff_t readahead_pos(struct readahead_control *rac)
1118 return (loff_t)rac->_index * PAGE_SIZE;
1122 * readahead_length - The number of bytes in this readahead request.
1123 * @rac: The readahead request.
1125 static inline size_t readahead_length(struct readahead_control *rac)
1127 return rac->_nr_pages * PAGE_SIZE;
1131 * readahead_index - The index of the first page in this readahead request.
1132 * @rac: The readahead request.
1134 static inline pgoff_t readahead_index(struct readahead_control *rac)
1140 * readahead_count - The number of pages in this readahead request.
1141 * @rac: The readahead request.
1143 static inline unsigned int readahead_count(struct readahead_control *rac)
1145 return rac->_nr_pages;
1149 * readahead_batch_length - The number of bytes in the current batch.
1150 * @rac: The readahead request.
1152 static inline size_t readahead_batch_length(struct readahead_control *rac)
1154 return rac->_batch_count * PAGE_SIZE;
1157 static inline unsigned long dir_pages(struct inode *inode)
1159 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1164 * folio_mkwrite_check_truncate - check if folio was truncated
1165 * @folio: the folio to check
1166 * @inode: the inode to check the folio against
1168 * Return: the number of bytes in the folio up to EOF,
1169 * or -EFAULT if the folio was truncated.
1171 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1172 struct inode *inode)
1174 loff_t size = i_size_read(inode);
1175 pgoff_t index = size >> PAGE_SHIFT;
1176 size_t offset = offset_in_folio(folio, size);
1178 if (!folio->mapping)
1181 /* folio is wholly inside EOF */
1182 if (folio_next_index(folio) - 1 < index)
1183 return folio_size(folio);
1184 /* folio is wholly past EOF */
1185 if (folio->index > index || !offset)
1187 /* folio is partially inside EOF */
1192 * page_mkwrite_check_truncate - check if page was truncated
1193 * @page: the page to check
1194 * @inode: the inode to check the page against
1196 * Returns the number of bytes in the page up to EOF,
1197 * or -EFAULT if the page was truncated.
1199 static inline int page_mkwrite_check_truncate(struct page *page,
1200 struct inode *inode)
1202 loff_t size = i_size_read(inode);
1203 pgoff_t index = size >> PAGE_SHIFT;
1204 int offset = offset_in_page(size);
1206 if (page->mapping != inode->i_mapping)
1209 /* page is wholly inside EOF */
1210 if (page->index < index)
1212 /* page is wholly past EOF */
1213 if (page->index > index || !offset)
1215 /* page is partially inside EOF */
1220 * i_blocks_per_folio - How many blocks fit in this folio.
1221 * @inode: The inode which contains the blocks.
1222 * @folio: The folio.
1224 * If the block size is larger than the size of this folio, return zero.
1226 * Context: The caller should hold a refcount on the folio to prevent it
1228 * Return: The number of filesystem blocks covered by this folio.
1231 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1233 return folio_size(folio) >> inode->i_blkbits;
1237 unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1239 return i_blocks_per_folio(inode, page_folio(page));
1241 #endif /* _LINUX_PAGEMAP_H */