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>
22 * Bits in mapping->flags.
25 AS_EIO = 0, /* IO error on async write */
26 AS_ENOSPC = 1, /* ENOSPC on async write */
27 AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */
28 AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */
29 AS_EXITING = 4, /* final truncate in progress */
30 /* writeback related tags are not used */
31 AS_NO_WRITEBACK_TAGS = 5,
32 AS_THP_SUPPORT = 6, /* THPs supported */
36 * mapping_set_error - record a writeback error in the address_space
37 * @mapping: the mapping in which an error should be set
38 * @error: the error to set in the mapping
40 * When writeback fails in some way, we must record that error so that
41 * userspace can be informed when fsync and the like are called. We endeavor
42 * to report errors on any file that was open at the time of the error. Some
43 * internal callers also need to know when writeback errors have occurred.
45 * When a writeback error occurs, most filesystems will want to call
46 * mapping_set_error to record the error in the mapping so that it can be
47 * reported when the application calls fsync(2).
49 static inline void mapping_set_error(struct address_space *mapping, int error)
54 /* Record in wb_err for checkers using errseq_t based tracking */
55 __filemap_set_wb_err(mapping, error);
57 /* Record it in superblock */
59 errseq_set(&mapping->host->i_sb->s_wb_err, error);
61 /* Record it in flags for now, for legacy callers */
63 set_bit(AS_ENOSPC, &mapping->flags);
65 set_bit(AS_EIO, &mapping->flags);
68 static inline void mapping_set_unevictable(struct address_space *mapping)
70 set_bit(AS_UNEVICTABLE, &mapping->flags);
73 static inline void mapping_clear_unevictable(struct address_space *mapping)
75 clear_bit(AS_UNEVICTABLE, &mapping->flags);
78 static inline bool mapping_unevictable(struct address_space *mapping)
80 return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
83 static inline void mapping_set_exiting(struct address_space *mapping)
85 set_bit(AS_EXITING, &mapping->flags);
88 static inline int mapping_exiting(struct address_space *mapping)
90 return test_bit(AS_EXITING, &mapping->flags);
93 static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
95 set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
98 static inline int mapping_use_writeback_tags(struct address_space *mapping)
100 return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
103 static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
105 return mapping->gfp_mask;
108 /* Restricts the given gfp_mask to what the mapping allows. */
109 static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
112 return mapping_gfp_mask(mapping) & gfp_mask;
116 * This is non-atomic. Only to be used before the mapping is activated.
117 * Probably needs a barrier...
119 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
124 static inline bool mapping_thp_support(struct address_space *mapping)
126 return test_bit(AS_THP_SUPPORT, &mapping->flags);
129 static inline int filemap_nr_thps(struct address_space *mapping)
131 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
132 return atomic_read(&mapping->nr_thps);
138 static inline void filemap_nr_thps_inc(struct address_space *mapping)
140 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
141 if (!mapping_thp_support(mapping))
142 atomic_inc(&mapping->nr_thps);
148 static inline void filemap_nr_thps_dec(struct address_space *mapping)
150 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
151 if (!mapping_thp_support(mapping))
152 atomic_dec(&mapping->nr_thps);
158 void release_pages(struct page **pages, int nr);
161 * speculatively take a reference to a page.
162 * If the page is free (_refcount == 0), then _refcount is untouched, and 0
163 * is returned. Otherwise, _refcount is incremented by 1 and 1 is returned.
165 * This function must be called inside the same rcu_read_lock() section as has
166 * been used to lookup the page in the pagecache radix-tree (or page table):
167 * this allows allocators to use a synchronize_rcu() to stabilize _refcount.
169 * Unless an RCU grace period has passed, the count of all pages coming out
170 * of the allocator must be considered unstable. page_count may return higher
171 * than expected, and put_page must be able to do the right thing when the
172 * page has been finished with, no matter what it is subsequently allocated
173 * for (because put_page is what is used here to drop an invalid speculative
176 * This is the interesting part of the lockless pagecache (and lockless
177 * get_user_pages) locking protocol, where the lookup-side (eg. find_get_page)
178 * has the following pattern:
179 * 1. find page in radix tree
180 * 2. conditionally increment refcount
181 * 3. check the page is still in pagecache (if no, goto 1)
183 * Remove-side that cares about stability of _refcount (eg. reclaim) has the
184 * following (with the i_pages lock held):
185 * A. atomically check refcount is correct and set it to 0 (atomic_cmpxchg)
186 * B. remove page from pagecache
189 * There are 2 critical interleavings that matter:
190 * - 2 runs before A: in this case, A sees elevated refcount and bails out
191 * - A runs before 2: in this case, 2 sees zero refcount and retries;
192 * subsequently, B will complete and 1 will find no page, causing the
193 * lookup to return NULL.
195 * It is possible that between 1 and 2, the page is removed then the exact same
196 * page is inserted into the same position in pagecache. That's OK: the
197 * old find_get_page using a lock could equally have run before or after
198 * such a re-insertion, depending on order that locks are granted.
200 * Lookups racing against pagecache insertion isn't a big problem: either 1
201 * will find the page or it will not. Likewise, the old find_get_page could run
202 * either before the insertion or afterwards, depending on timing.
204 static inline int __page_cache_add_speculative(struct page *page, int count)
206 #ifdef CONFIG_TINY_RCU
207 # ifdef CONFIG_PREEMPT_COUNT
208 VM_BUG_ON(!in_atomic() && !irqs_disabled());
211 * Preempt must be disabled here - we rely on rcu_read_lock doing
214 * Pagecache won't be truncated from interrupt context, so if we have
215 * found a page in the radix tree here, we have pinned its refcount by
216 * disabling preempt, and hence no need for the "speculative get" that
219 VM_BUG_ON_PAGE(page_count(page) == 0, page);
220 page_ref_add(page, count);
223 if (unlikely(!page_ref_add_unless(page, count, 0))) {
225 * Either the page has been freed, or will be freed.
226 * In either case, retry here and the caller should
227 * do the right thing (see comments above).
232 VM_BUG_ON_PAGE(PageTail(page), page);
237 static inline int page_cache_get_speculative(struct page *page)
239 return __page_cache_add_speculative(page, 1);
242 static inline int page_cache_add_speculative(struct page *page, int count)
244 return __page_cache_add_speculative(page, count);
248 * attach_page_private - Attach private data to a page.
249 * @page: Page to attach data to.
250 * @data: Data to attach to page.
252 * Attaching private data to a page increments the page's reference count.
253 * The data must be detached before the page will be freed.
255 static inline void attach_page_private(struct page *page, void *data)
258 set_page_private(page, (unsigned long)data);
259 SetPagePrivate(page);
263 * detach_page_private - Detach private data from a page.
264 * @page: Page to detach data from.
266 * Removes the data that was previously attached to the page and decrements
267 * the refcount on the page.
269 * Return: Data that was attached to the page.
271 static inline void *detach_page_private(struct page *page)
273 void *data = (void *)page_private(page);
275 if (!PagePrivate(page))
277 ClearPagePrivate(page);
278 set_page_private(page, 0);
285 extern struct page *__page_cache_alloc(gfp_t gfp);
287 static inline struct page *__page_cache_alloc(gfp_t gfp)
289 return alloc_pages(gfp, 0);
293 static inline struct page *page_cache_alloc(struct address_space *x)
295 return __page_cache_alloc(mapping_gfp_mask(x));
298 static inline gfp_t readahead_gfp_mask(struct address_space *x)
300 return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
303 typedef int filler_t(void *, struct page *);
305 pgoff_t page_cache_next_miss(struct address_space *mapping,
306 pgoff_t index, unsigned long max_scan);
307 pgoff_t page_cache_prev_miss(struct address_space *mapping,
308 pgoff_t index, unsigned long max_scan);
310 #define FGP_ACCESSED 0x00000001
311 #define FGP_LOCK 0x00000002
312 #define FGP_CREAT 0x00000004
313 #define FGP_WRITE 0x00000008
314 #define FGP_NOFS 0x00000010
315 #define FGP_NOWAIT 0x00000020
316 #define FGP_FOR_MMAP 0x00000040
317 #define FGP_HEAD 0x00000080
318 #define FGP_ENTRY 0x00000100
320 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t offset,
321 int fgp_flags, gfp_t cache_gfp_mask);
324 * find_get_page - find and get a page reference
325 * @mapping: the address_space to search
326 * @offset: the page index
328 * Looks up the page cache slot at @mapping & @offset. If there is a
329 * page cache page, it is returned with an increased refcount.
331 * Otherwise, %NULL is returned.
333 static inline struct page *find_get_page(struct address_space *mapping,
336 return pagecache_get_page(mapping, offset, 0, 0);
339 static inline struct page *find_get_page_flags(struct address_space *mapping,
340 pgoff_t offset, int fgp_flags)
342 return pagecache_get_page(mapping, offset, fgp_flags, 0);
346 * find_lock_page - locate, pin and lock a pagecache page
347 * @mapping: the address_space to search
348 * @index: the page index
350 * Looks up the page cache entry at @mapping & @index. If there is a
351 * page cache page, it is returned locked and with an increased
354 * Context: May sleep.
355 * Return: A struct page or %NULL if there is no page in the cache for this
358 static inline struct page *find_lock_page(struct address_space *mapping,
361 return pagecache_get_page(mapping, index, FGP_LOCK, 0);
365 * find_lock_head - Locate, pin and lock a pagecache page.
366 * @mapping: The address_space to search.
367 * @index: The page index.
369 * Looks up the page cache entry at @mapping & @index. If there is a
370 * page cache page, its head page is returned locked and with an increased
373 * Context: May sleep.
374 * Return: A struct page which is !PageTail, or %NULL if there is no page
375 * in the cache for this index.
377 static inline struct page *find_lock_head(struct address_space *mapping,
380 return pagecache_get_page(mapping, index, FGP_LOCK | FGP_HEAD, 0);
384 * find_or_create_page - locate or add a pagecache page
385 * @mapping: the page's address_space
386 * @index: the page's index into the mapping
387 * @gfp_mask: page allocation mode
389 * Looks up the page cache slot at @mapping & @offset. If there is a
390 * page cache page, it is returned locked and with an increased
393 * If the page is not present, a new page is allocated using @gfp_mask
394 * and added to the page cache and the VM's LRU list. The page is
395 * returned locked and with an increased refcount.
397 * On memory exhaustion, %NULL is returned.
399 * find_or_create_page() may sleep, even if @gfp_flags specifies an
402 static inline struct page *find_or_create_page(struct address_space *mapping,
403 pgoff_t index, gfp_t gfp_mask)
405 return pagecache_get_page(mapping, index,
406 FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
411 * grab_cache_page_nowait - returns locked page at given index in given cache
412 * @mapping: target address_space
413 * @index: the page index
415 * Same as grab_cache_page(), but do not wait if the page is unavailable.
416 * This is intended for speculative data generators, where the data can
417 * be regenerated if the page couldn't be grabbed. This routine should
418 * be safe to call while holding the lock for another page.
420 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
421 * and deadlock against the caller's locked page.
423 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
426 return pagecache_get_page(mapping, index,
427 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
428 mapping_gfp_mask(mapping));
431 /* Does this page contain this index? */
432 static inline bool thp_contains(struct page *head, pgoff_t index)
434 /* HugeTLBfs indexes the page cache in units of hpage_size */
436 return head->index == index;
437 return page_index(head) == (index & ~(thp_nr_pages(head) - 1UL));
441 * Given the page we found in the page cache, return the page corresponding
442 * to this index in the file
444 static inline struct page *find_subpage(struct page *head, pgoff_t index)
446 /* HugeTLBfs wants the head page regardless */
450 return head + (index & (thp_nr_pages(head) - 1));
453 unsigned find_get_entries(struct address_space *mapping, pgoff_t start,
454 pgoff_t end, struct pagevec *pvec, pgoff_t *indices);
455 unsigned find_get_pages_range(struct address_space *mapping, pgoff_t *start,
456 pgoff_t end, unsigned int nr_pages,
457 struct page **pages);
458 static inline unsigned find_get_pages(struct address_space *mapping,
459 pgoff_t *start, unsigned int nr_pages,
462 return find_get_pages_range(mapping, start, (pgoff_t)-1, nr_pages,
465 unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t start,
466 unsigned int nr_pages, struct page **pages);
467 unsigned find_get_pages_range_tag(struct address_space *mapping, pgoff_t *index,
468 pgoff_t end, xa_mark_t tag, unsigned int nr_pages,
469 struct page **pages);
470 static inline unsigned find_get_pages_tag(struct address_space *mapping,
471 pgoff_t *index, xa_mark_t tag, unsigned int nr_pages,
474 return find_get_pages_range_tag(mapping, index, (pgoff_t)-1, tag,
478 struct page *grab_cache_page_write_begin(struct address_space *mapping,
479 pgoff_t index, unsigned flags);
482 * Returns locked page at given index in given cache, creating it if needed.
484 static inline struct page *grab_cache_page(struct address_space *mapping,
487 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
490 extern struct page * read_cache_page(struct address_space *mapping,
491 pgoff_t index, filler_t *filler, void *data);
492 extern struct page * read_cache_page_gfp(struct address_space *mapping,
493 pgoff_t index, gfp_t gfp_mask);
494 extern int read_cache_pages(struct address_space *mapping,
495 struct list_head *pages, filler_t *filler, void *data);
497 static inline struct page *read_mapping_page(struct address_space *mapping,
498 pgoff_t index, void *data)
500 return read_cache_page(mapping, index, NULL, data);
504 * Get index of the page with in radix-tree
505 * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
507 static inline pgoff_t page_to_index(struct page *page)
511 if (likely(!PageTransTail(page)))
515 * We don't initialize ->index for tail pages: calculate based on
518 pgoff = compound_head(page)->index;
519 pgoff += page - compound_head(page);
524 * Get the offset in PAGE_SIZE.
525 * (TODO: hugepage should have ->index in PAGE_SIZE)
527 static inline pgoff_t page_to_pgoff(struct page *page)
529 if (unlikely(PageHeadHuge(page)))
530 return page->index << compound_order(page);
532 return page_to_index(page);
536 * Return byte-offset into filesystem object for page.
538 static inline loff_t page_offset(struct page *page)
540 return ((loff_t)page->index) << PAGE_SHIFT;
543 static inline loff_t page_file_offset(struct page *page)
545 return ((loff_t)page_index(page)) << PAGE_SHIFT;
548 extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
549 unsigned long address);
551 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
552 unsigned long address)
555 if (unlikely(is_vm_hugetlb_page(vma)))
556 return linear_hugepage_index(vma, address);
557 pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
558 pgoff += vma->vm_pgoff;
562 struct wait_page_key {
568 struct wait_page_queue {
571 wait_queue_entry_t wait;
574 static inline bool wake_page_match(struct wait_page_queue *wait_page,
575 struct wait_page_key *key)
577 if (wait_page->page != key->page)
581 if (wait_page->bit_nr != key->bit_nr)
587 extern void __lock_page(struct page *page);
588 extern int __lock_page_killable(struct page *page);
589 extern int __lock_page_async(struct page *page, struct wait_page_queue *wait);
590 extern int __lock_page_or_retry(struct page *page, struct mm_struct *mm,
592 extern void unlock_page(struct page *page);
595 * Return true if the page was successfully locked
597 static inline int trylock_page(struct page *page)
599 page = compound_head(page);
600 return (likely(!test_and_set_bit_lock(PG_locked, &page->flags)));
604 * lock_page may only be called if we have the page's inode pinned.
606 static inline void lock_page(struct page *page)
609 if (!trylock_page(page))
614 * lock_page_killable is like lock_page but can be interrupted by fatal
615 * signals. It returns 0 if it locked the page and -EINTR if it was
616 * killed while waiting.
618 static inline int lock_page_killable(struct page *page)
621 if (!trylock_page(page))
622 return __lock_page_killable(page);
627 * lock_page_async - Lock the page, unless this would block. If the page
628 * is already locked, then queue a callback when the page becomes unlocked.
629 * This callback can then retry the operation.
631 * Returns 0 if the page is locked successfully, or -EIOCBQUEUED if the page
632 * was already locked and the callback defined in 'wait' was queued.
634 static inline int lock_page_async(struct page *page,
635 struct wait_page_queue *wait)
637 if (!trylock_page(page))
638 return __lock_page_async(page, wait);
643 * lock_page_or_retry - Lock the page, unless this would block and the
644 * caller indicated that it can handle a retry.
646 * Return value and mmap_lock implications depend on flags; see
647 * __lock_page_or_retry().
649 static inline int lock_page_or_retry(struct page *page, struct mm_struct *mm,
653 return trylock_page(page) || __lock_page_or_retry(page, mm, flags);
657 * This is exported only for wait_on_page_locked/wait_on_page_writeback, etc.,
658 * and should not be used directly.
660 extern void wait_on_page_bit(struct page *page, int bit_nr);
661 extern int wait_on_page_bit_killable(struct page *page, int bit_nr);
664 * Wait for a page to be unlocked.
666 * This must be called with the caller "holding" the page,
667 * ie with increased "page->count" so that the page won't
668 * go away during the wait..
670 static inline void wait_on_page_locked(struct page *page)
672 if (PageLocked(page))
673 wait_on_page_bit(compound_head(page), PG_locked);
676 static inline int wait_on_page_locked_killable(struct page *page)
678 if (!PageLocked(page))
680 return wait_on_page_bit_killable(compound_head(page), PG_locked);
683 int put_and_wait_on_page_locked(struct page *page, int state);
684 void wait_on_page_writeback(struct page *page);
685 int wait_on_page_writeback_killable(struct page *page);
686 extern void end_page_writeback(struct page *page);
687 void wait_for_stable_page(struct page *page);
689 void page_endio(struct page *page, bool is_write, int err);
692 * set_page_private_2 - Set PG_private_2 on a page and take a ref
695 * Set the PG_private_2 flag on a page and take the reference needed for the VM
696 * to handle its lifetime correctly. This sets the flag and takes the
697 * reference unconditionally, so care must be taken not to set the flag again
698 * if it's already set.
700 static inline void set_page_private_2(struct page *page)
702 page = compound_head(page);
704 SetPagePrivate2(page);
707 void end_page_private_2(struct page *page);
708 void wait_on_page_private_2(struct page *page);
709 int wait_on_page_private_2_killable(struct page *page);
712 * Add an arbitrary waiter to a page's wait queue
714 extern void add_page_wait_queue(struct page *page, wait_queue_entry_t *waiter);
717 * Fault everything in given userspace address range in.
719 static inline int fault_in_pages_writeable(char __user *uaddr, int size)
721 char __user *end = uaddr + size - 1;
723 if (unlikely(size == 0))
726 if (unlikely(uaddr > end))
729 * Writing zeroes into userspace here is OK, because we know that if
730 * the zero gets there, we'll be overwriting it.
733 if (unlikely(__put_user(0, uaddr) != 0))
736 } while (uaddr <= end);
738 /* Check whether the range spilled into the next page. */
739 if (((unsigned long)uaddr & PAGE_MASK) ==
740 ((unsigned long)end & PAGE_MASK))
741 return __put_user(0, end);
746 static inline int fault_in_pages_readable(const char __user *uaddr, int size)
749 const char __user *end = uaddr + size - 1;
751 if (unlikely(size == 0))
754 if (unlikely(uaddr > end))
758 if (unlikely(__get_user(c, uaddr) != 0))
761 } while (uaddr <= end);
763 /* Check whether the range spilled into the next page. */
764 if (((unsigned long)uaddr & PAGE_MASK) ==
765 ((unsigned long)end & PAGE_MASK)) {
766 return __get_user(c, end);
773 int add_to_page_cache_locked(struct page *page, struct address_space *mapping,
774 pgoff_t index, gfp_t gfp_mask);
775 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
776 pgoff_t index, gfp_t gfp_mask);
777 extern void delete_from_page_cache(struct page *page);
778 extern void __delete_from_page_cache(struct page *page, void *shadow);
779 void replace_page_cache_page(struct page *old, struct page *new);
780 void delete_from_page_cache_batch(struct address_space *mapping,
781 struct pagevec *pvec);
782 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
786 * Like add_to_page_cache_locked, but used to add newly allocated pages:
787 * the page is new, so we can just run __SetPageLocked() against it.
789 static inline int add_to_page_cache(struct page *page,
790 struct address_space *mapping, pgoff_t offset, gfp_t gfp_mask)
794 __SetPageLocked(page);
795 error = add_to_page_cache_locked(page, mapping, offset, gfp_mask);
797 __ClearPageLocked(page);
802 * struct readahead_control - Describes a readahead request.
804 * A readahead request is for consecutive pages. Filesystems which
805 * implement the ->readahead method should call readahead_page() or
806 * readahead_page_batch() in a loop and attempt to start I/O against
807 * each page in the request.
809 * Most of the fields in this struct are private and should be accessed
810 * by the functions below.
812 * @file: The file, used primarily by network filesystems for authentication.
813 * May be NULL if invoked internally by the filesystem.
814 * @mapping: Readahead this filesystem object.
815 * @ra: File readahead state. May be NULL.
817 struct readahead_control {
819 struct address_space *mapping;
820 struct file_ra_state *ra;
821 /* private: use the readahead_* accessors instead */
823 unsigned int _nr_pages;
824 unsigned int _batch_count;
827 #define DEFINE_READAHEAD(ractl, f, r, m, i) \
828 struct readahead_control ractl = { \
835 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
837 void page_cache_ra_unbounded(struct readahead_control *,
838 unsigned long nr_to_read, unsigned long lookahead_count);
839 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
840 void page_cache_async_ra(struct readahead_control *, struct page *,
841 unsigned long req_count);
842 void readahead_expand(struct readahead_control *ractl,
843 loff_t new_start, size_t new_len);
846 * page_cache_sync_readahead - generic file readahead
847 * @mapping: address_space which holds the pagecache and I/O vectors
848 * @ra: file_ra_state which holds the readahead state
849 * @file: Used by the filesystem for authentication.
850 * @index: Index of first page to be read.
851 * @req_count: Total number of pages being read by the caller.
853 * page_cache_sync_readahead() should be called when a cache miss happened:
854 * it will submit the read. The readahead logic may decide to piggyback more
855 * pages onto the read request if access patterns suggest it will improve
859 void page_cache_sync_readahead(struct address_space *mapping,
860 struct file_ra_state *ra, struct file *file, pgoff_t index,
861 unsigned long req_count)
863 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
864 page_cache_sync_ra(&ractl, req_count);
868 * page_cache_async_readahead - file readahead for marked pages
869 * @mapping: address_space which holds the pagecache and I/O vectors
870 * @ra: file_ra_state which holds the readahead state
871 * @file: Used by the filesystem for authentication.
872 * @page: The page at @index which triggered the readahead call.
873 * @index: Index of first page to be read.
874 * @req_count: Total number of pages being read by the caller.
876 * page_cache_async_readahead() should be called when a page is used which
877 * is marked as PageReadahead; this is a marker to suggest that the application
878 * has used up enough of the readahead window that we should start pulling in
882 void page_cache_async_readahead(struct address_space *mapping,
883 struct file_ra_state *ra, struct file *file,
884 struct page *page, pgoff_t index, unsigned long req_count)
886 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
887 page_cache_async_ra(&ractl, page, req_count);
891 * readahead_page - Get the next page to read.
892 * @rac: The current readahead request.
894 * Context: The page is locked and has an elevated refcount. The caller
895 * should decreases the refcount once the page has been submitted for I/O
896 * and unlock the page once all I/O to that page has completed.
897 * Return: A pointer to the next page, or %NULL if we are done.
899 static inline struct page *readahead_page(struct readahead_control *rac)
903 BUG_ON(rac->_batch_count > rac->_nr_pages);
904 rac->_nr_pages -= rac->_batch_count;
905 rac->_index += rac->_batch_count;
907 if (!rac->_nr_pages) {
908 rac->_batch_count = 0;
912 page = xa_load(&rac->mapping->i_pages, rac->_index);
913 VM_BUG_ON_PAGE(!PageLocked(page), page);
914 rac->_batch_count = thp_nr_pages(page);
919 static inline unsigned int __readahead_batch(struct readahead_control *rac,
920 struct page **array, unsigned int array_sz)
923 XA_STATE(xas, &rac->mapping->i_pages, 0);
926 BUG_ON(rac->_batch_count > rac->_nr_pages);
927 rac->_nr_pages -= rac->_batch_count;
928 rac->_index += rac->_batch_count;
929 rac->_batch_count = 0;
931 xas_set(&xas, rac->_index);
933 xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
934 if (xas_retry(&xas, page))
936 VM_BUG_ON_PAGE(!PageLocked(page), page);
937 VM_BUG_ON_PAGE(PageTail(page), page);
939 rac->_batch_count += thp_nr_pages(page);
942 * The page cache isn't using multi-index entries yet,
943 * so the xas cursor needs to be manually moved to the
944 * next index. This can be removed once the page cache
948 xas_set(&xas, rac->_index + rac->_batch_count);
959 * readahead_page_batch - Get a batch of pages to read.
960 * @rac: The current readahead request.
961 * @array: An array of pointers to struct page.
963 * Context: The pages are locked and have an elevated refcount. The caller
964 * should decreases the refcount once the page has been submitted for I/O
965 * and unlock the page once all I/O to that page has completed.
966 * Return: The number of pages placed in the array. 0 indicates the request
969 #define readahead_page_batch(rac, array) \
970 __readahead_batch(rac, array, ARRAY_SIZE(array))
973 * readahead_pos - The byte offset into the file of this readahead request.
974 * @rac: The readahead request.
976 static inline loff_t readahead_pos(struct readahead_control *rac)
978 return (loff_t)rac->_index * PAGE_SIZE;
982 * readahead_length - The number of bytes in this readahead request.
983 * @rac: The readahead request.
985 static inline loff_t readahead_length(struct readahead_control *rac)
987 return (loff_t)rac->_nr_pages * PAGE_SIZE;
991 * readahead_index - The index of the first page in this readahead request.
992 * @rac: The readahead request.
994 static inline pgoff_t readahead_index(struct readahead_control *rac)
1000 * readahead_count - The number of pages in this readahead request.
1001 * @rac: The readahead request.
1003 static inline unsigned int readahead_count(struct readahead_control *rac)
1005 return rac->_nr_pages;
1009 * readahead_batch_length - The number of bytes in the current batch.
1010 * @rac: The readahead request.
1012 static inline loff_t readahead_batch_length(struct readahead_control *rac)
1014 return rac->_batch_count * PAGE_SIZE;
1017 static inline unsigned long dir_pages(struct inode *inode)
1019 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1024 * page_mkwrite_check_truncate - check if page was truncated
1025 * @page: the page to check
1026 * @inode: the inode to check the page against
1028 * Returns the number of bytes in the page up to EOF,
1029 * or -EFAULT if the page was truncated.
1031 static inline int page_mkwrite_check_truncate(struct page *page,
1032 struct inode *inode)
1034 loff_t size = i_size_read(inode);
1035 pgoff_t index = size >> PAGE_SHIFT;
1036 int offset = offset_in_page(size);
1038 if (page->mapping != inode->i_mapping)
1041 /* page is wholly inside EOF */
1042 if (page->index < index)
1044 /* page is wholly past EOF */
1045 if (page->index > index || !offset)
1047 /* page is partially inside EOF */
1052 * i_blocks_per_page - How many blocks fit in this page.
1053 * @inode: The inode which contains the blocks.
1054 * @page: The page (head page if the page is a THP).
1056 * If the block size is larger than the size of this page, return zero.
1058 * Context: The caller should hold a refcount on the page to prevent it
1060 * Return: The number of filesystem blocks covered by this page.
1063 unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1065 return thp_size(page) >> inode->i_blkbits;
1067 #endif /* _LINUX_PAGEMAP_H */