]>
git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - include/linux/pagemap.h
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
);
166 * For file cache pages, return the address_space, otherwise return NULL
168 static inline struct address_space
*page_mapping_file(struct page
*page
)
170 if (unlikely(PageSwapCache(page
)))
172 return page_mapping(page
);
176 * speculatively take a reference to a page.
177 * If the page is free (_refcount == 0), then _refcount is untouched, and 0
178 * is returned. Otherwise, _refcount is incremented by 1 and 1 is returned.
180 * This function must be called inside the same rcu_read_lock() section as has
181 * been used to lookup the page in the pagecache radix-tree (or page table):
182 * this allows allocators to use a synchronize_rcu() to stabilize _refcount.
184 * Unless an RCU grace period has passed, the count of all pages coming out
185 * of the allocator must be considered unstable. page_count may return higher
186 * than expected, and put_page must be able to do the right thing when the
187 * page has been finished with, no matter what it is subsequently allocated
188 * for (because put_page is what is used here to drop an invalid speculative
191 * This is the interesting part of the lockless pagecache (and lockless
192 * get_user_pages) locking protocol, where the lookup-side (eg. find_get_page)
193 * has the following pattern:
194 * 1. find page in radix tree
195 * 2. conditionally increment refcount
196 * 3. check the page is still in pagecache (if no, goto 1)
198 * Remove-side that cares about stability of _refcount (eg. reclaim) has the
199 * following (with the i_pages lock held):
200 * A. atomically check refcount is correct and set it to 0 (atomic_cmpxchg)
201 * B. remove page from pagecache
204 * There are 2 critical interleavings that matter:
205 * - 2 runs before A: in this case, A sees elevated refcount and bails out
206 * - A runs before 2: in this case, 2 sees zero refcount and retries;
207 * subsequently, B will complete and 1 will find no page, causing the
208 * lookup to return NULL.
210 * It is possible that between 1 and 2, the page is removed then the exact same
211 * page is inserted into the same position in pagecache. That's OK: the
212 * old find_get_page using a lock could equally have run before or after
213 * such a re-insertion, depending on order that locks are granted.
215 * Lookups racing against pagecache insertion isn't a big problem: either 1
216 * will find the page or it will not. Likewise, the old find_get_page could run
217 * either before the insertion or afterwards, depending on timing.
219 static inline int __page_cache_add_speculative(struct page
*page
, int count
)
221 #ifdef CONFIG_TINY_RCU
222 # ifdef CONFIG_PREEMPT_COUNT
223 VM_BUG_ON(!in_atomic() && !irqs_disabled());
226 * Preempt must be disabled here - we rely on rcu_read_lock doing
229 * Pagecache won't be truncated from interrupt context, so if we have
230 * found a page in the radix tree here, we have pinned its refcount by
231 * disabling preempt, and hence no need for the "speculative get" that
234 VM_BUG_ON_PAGE(page_count(page
) == 0, page
);
235 page_ref_add(page
, count
);
238 if (unlikely(!page_ref_add_unless(page
, count
, 0))) {
240 * Either the page has been freed, or will be freed.
241 * In either case, retry here and the caller should
242 * do the right thing (see comments above).
247 VM_BUG_ON_PAGE(PageTail(page
), page
);
252 static inline int page_cache_get_speculative(struct page
*page
)
254 return __page_cache_add_speculative(page
, 1);
257 static inline int page_cache_add_speculative(struct page
*page
, int count
)
259 return __page_cache_add_speculative(page
, count
);
263 * attach_page_private - Attach private data to a page.
264 * @page: Page to attach data to.
265 * @data: Data to attach to page.
267 * Attaching private data to a page increments the page's reference count.
268 * The data must be detached before the page will be freed.
270 static inline void attach_page_private(struct page
*page
, void *data
)
273 set_page_private(page
, (unsigned long)data
);
274 SetPagePrivate(page
);
278 * detach_page_private - Detach private data from a page.
279 * @page: Page to detach data from.
281 * Removes the data that was previously attached to the page and decrements
282 * the refcount on the page.
284 * Return: Data that was attached to the page.
286 static inline void *detach_page_private(struct page
*page
)
288 void *data
= (void *)page_private(page
);
290 if (!PagePrivate(page
))
292 ClearPagePrivate(page
);
293 set_page_private(page
, 0);
300 extern struct page
*__page_cache_alloc(gfp_t gfp
);
302 static inline struct page
*__page_cache_alloc(gfp_t gfp
)
304 return alloc_pages(gfp
, 0);
308 static inline struct page
*page_cache_alloc(struct address_space
*x
)
310 return __page_cache_alloc(mapping_gfp_mask(x
));
313 static inline gfp_t
readahead_gfp_mask(struct address_space
*x
)
315 return mapping_gfp_mask(x
) | __GFP_NORETRY
| __GFP_NOWARN
;
318 typedef int filler_t(void *, struct page
*);
320 pgoff_t
page_cache_next_miss(struct address_space
*mapping
,
321 pgoff_t index
, unsigned long max_scan
);
322 pgoff_t
page_cache_prev_miss(struct address_space
*mapping
,
323 pgoff_t index
, unsigned long max_scan
);
325 #define FGP_ACCESSED 0x00000001
326 #define FGP_LOCK 0x00000002
327 #define FGP_CREAT 0x00000004
328 #define FGP_WRITE 0x00000008
329 #define FGP_NOFS 0x00000010
330 #define FGP_NOWAIT 0x00000020
331 #define FGP_FOR_MMAP 0x00000040
332 #define FGP_HEAD 0x00000080
333 #define FGP_ENTRY 0x00000100
335 struct page
*pagecache_get_page(struct address_space
*mapping
, pgoff_t offset
,
336 int fgp_flags
, gfp_t cache_gfp_mask
);
339 * find_get_page - find and get a page reference
340 * @mapping: the address_space to search
341 * @offset: the page index
343 * Looks up the page cache slot at @mapping & @offset. If there is a
344 * page cache page, it is returned with an increased refcount.
346 * Otherwise, %NULL is returned.
348 static inline struct page
*find_get_page(struct address_space
*mapping
,
351 return pagecache_get_page(mapping
, offset
, 0, 0);
354 static inline struct page
*find_get_page_flags(struct address_space
*mapping
,
355 pgoff_t offset
, int fgp_flags
)
357 return pagecache_get_page(mapping
, offset
, fgp_flags
, 0);
361 * find_lock_page - locate, pin and lock a pagecache page
362 * @mapping: the address_space to search
363 * @index: the page index
365 * Looks up the page cache entry at @mapping & @index. If there is a
366 * page cache page, it is returned locked and with an increased
369 * Context: May sleep.
370 * Return: A struct page or %NULL if there is no page in the cache for this
373 static inline struct page
*find_lock_page(struct address_space
*mapping
,
376 return pagecache_get_page(mapping
, index
, FGP_LOCK
, 0);
380 * find_lock_head - Locate, pin and lock a pagecache page.
381 * @mapping: The address_space to search.
382 * @index: The page index.
384 * Looks up the page cache entry at @mapping & @index. If there is a
385 * page cache page, its head page is returned locked and with an increased
388 * Context: May sleep.
389 * Return: A struct page which is !PageTail, or %NULL if there is no page
390 * in the cache for this index.
392 static inline struct page
*find_lock_head(struct address_space
*mapping
,
395 return pagecache_get_page(mapping
, index
, FGP_LOCK
| FGP_HEAD
, 0);
399 * find_or_create_page - locate or add a pagecache page
400 * @mapping: the page's address_space
401 * @index: the page's index into the mapping
402 * @gfp_mask: page allocation mode
404 * Looks up the page cache slot at @mapping & @offset. If there is a
405 * page cache page, it is returned locked and with an increased
408 * If the page is not present, a new page is allocated using @gfp_mask
409 * and added to the page cache and the VM's LRU list. The page is
410 * returned locked and with an increased refcount.
412 * On memory exhaustion, %NULL is returned.
414 * find_or_create_page() may sleep, even if @gfp_flags specifies an
417 static inline struct page
*find_or_create_page(struct address_space
*mapping
,
418 pgoff_t index
, gfp_t gfp_mask
)
420 return pagecache_get_page(mapping
, index
,
421 FGP_LOCK
|FGP_ACCESSED
|FGP_CREAT
,
426 * grab_cache_page_nowait - returns locked page at given index in given cache
427 * @mapping: target address_space
428 * @index: the page index
430 * Same as grab_cache_page(), but do not wait if the page is unavailable.
431 * This is intended for speculative data generators, where the data can
432 * be regenerated if the page couldn't be grabbed. This routine should
433 * be safe to call while holding the lock for another page.
435 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
436 * and deadlock against the caller's locked page.
438 static inline struct page
*grab_cache_page_nowait(struct address_space
*mapping
,
441 return pagecache_get_page(mapping
, index
,
442 FGP_LOCK
|FGP_CREAT
|FGP_NOFS
|FGP_NOWAIT
,
443 mapping_gfp_mask(mapping
));
446 /* Does this page contain this index? */
447 static inline bool thp_contains(struct page
*head
, pgoff_t index
)
449 /* HugeTLBfs indexes the page cache in units of hpage_size */
451 return head
->index
== index
;
452 return page_index(head
) == (index
& ~(thp_nr_pages(head
) - 1UL));
456 * Given the page we found in the page cache, return the page corresponding
457 * to this index in the file
459 static inline struct page
*find_subpage(struct page
*head
, pgoff_t index
)
461 /* HugeTLBfs wants the head page regardless */
465 return head
+ (index
& (thp_nr_pages(head
) - 1));
468 unsigned find_get_entries(struct address_space
*mapping
, pgoff_t start
,
469 pgoff_t end
, struct pagevec
*pvec
, pgoff_t
*indices
);
470 unsigned find_get_pages_range(struct address_space
*mapping
, pgoff_t
*start
,
471 pgoff_t end
, unsigned int nr_pages
,
472 struct page
**pages
);
473 static inline unsigned find_get_pages(struct address_space
*mapping
,
474 pgoff_t
*start
, unsigned int nr_pages
,
477 return find_get_pages_range(mapping
, start
, (pgoff_t
)-1, nr_pages
,
480 unsigned find_get_pages_contig(struct address_space
*mapping
, pgoff_t start
,
481 unsigned int nr_pages
, struct page
**pages
);
482 unsigned find_get_pages_range_tag(struct address_space
*mapping
, pgoff_t
*index
,
483 pgoff_t end
, xa_mark_t tag
, unsigned int nr_pages
,
484 struct page
**pages
);
485 static inline unsigned find_get_pages_tag(struct address_space
*mapping
,
486 pgoff_t
*index
, xa_mark_t tag
, unsigned int nr_pages
,
489 return find_get_pages_range_tag(mapping
, index
, (pgoff_t
)-1, tag
,
493 struct page
*grab_cache_page_write_begin(struct address_space
*mapping
,
494 pgoff_t index
, unsigned flags
);
497 * Returns locked page at given index in given cache, creating it if needed.
499 static inline struct page
*grab_cache_page(struct address_space
*mapping
,
502 return find_or_create_page(mapping
, index
, mapping_gfp_mask(mapping
));
505 extern struct page
* read_cache_page(struct address_space
*mapping
,
506 pgoff_t index
, filler_t
*filler
, void *data
);
507 extern struct page
* read_cache_page_gfp(struct address_space
*mapping
,
508 pgoff_t index
, gfp_t gfp_mask
);
509 extern int read_cache_pages(struct address_space
*mapping
,
510 struct list_head
*pages
, filler_t
*filler
, void *data
);
512 static inline struct page
*read_mapping_page(struct address_space
*mapping
,
513 pgoff_t index
, void *data
)
515 return read_cache_page(mapping
, index
, NULL
, data
);
519 * Get index of the page within radix-tree (but not for hugetlb pages).
520 * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
522 static inline pgoff_t
page_to_index(struct page
*page
)
526 if (likely(!PageTransTail(page
)))
530 * We don't initialize ->index for tail pages: calculate based on
533 pgoff
= compound_head(page
)->index
;
534 pgoff
+= page
- compound_head(page
);
538 extern pgoff_t
hugetlb_basepage_index(struct page
*page
);
541 * Get the offset in PAGE_SIZE (even for hugetlb pages).
542 * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
544 static inline pgoff_t
page_to_pgoff(struct page
*page
)
546 if (unlikely(PageHuge(page
)))
547 return hugetlb_basepage_index(page
);
548 return page_to_index(page
);
552 * Return byte-offset into filesystem object for page.
554 static inline loff_t
page_offset(struct page
*page
)
556 return ((loff_t
)page
->index
) << PAGE_SHIFT
;
559 static inline loff_t
page_file_offset(struct page
*page
)
561 return ((loff_t
)page_index(page
)) << PAGE_SHIFT
;
564 extern pgoff_t
linear_hugepage_index(struct vm_area_struct
*vma
,
565 unsigned long address
);
567 static inline pgoff_t
linear_page_index(struct vm_area_struct
*vma
,
568 unsigned long address
)
571 if (unlikely(is_vm_hugetlb_page(vma
)))
572 return linear_hugepage_index(vma
, address
);
573 pgoff
= (address
- vma
->vm_start
) >> PAGE_SHIFT
;
574 pgoff
+= vma
->vm_pgoff
;
578 struct wait_page_key
{
584 struct wait_page_queue
{
587 wait_queue_entry_t wait
;
590 static inline bool wake_page_match(struct wait_page_queue
*wait_page
,
591 struct wait_page_key
*key
)
593 if (wait_page
->page
!= key
->page
)
597 if (wait_page
->bit_nr
!= key
->bit_nr
)
603 extern void __lock_page(struct page
*page
);
604 extern int __lock_page_killable(struct page
*page
);
605 extern int __lock_page_async(struct page
*page
, struct wait_page_queue
*wait
);
606 extern int __lock_page_or_retry(struct page
*page
, struct mm_struct
*mm
,
608 extern void unlock_page(struct page
*page
);
611 * Return true if the page was successfully locked
613 static inline int trylock_page(struct page
*page
)
615 page
= compound_head(page
);
616 return (likely(!test_and_set_bit_lock(PG_locked
, &page
->flags
)));
620 * lock_page may only be called if we have the page's inode pinned.
622 static inline void lock_page(struct page
*page
)
625 if (!trylock_page(page
))
630 * lock_page_killable is like lock_page but can be interrupted by fatal
631 * signals. It returns 0 if it locked the page and -EINTR if it was
632 * killed while waiting.
634 static inline int lock_page_killable(struct page
*page
)
637 if (!trylock_page(page
))
638 return __lock_page_killable(page
);
643 * lock_page_async - Lock the page, unless this would block. If the page
644 * is already locked, then queue a callback when the page becomes unlocked.
645 * This callback can then retry the operation.
647 * Returns 0 if the page is locked successfully, or -EIOCBQUEUED if the page
648 * was already locked and the callback defined in 'wait' was queued.
650 static inline int lock_page_async(struct page
*page
,
651 struct wait_page_queue
*wait
)
653 if (!trylock_page(page
))
654 return __lock_page_async(page
, wait
);
659 * lock_page_or_retry - Lock the page, unless this would block and the
660 * caller indicated that it can handle a retry.
662 * Return value and mmap_lock implications depend on flags; see
663 * __lock_page_or_retry().
665 static inline int lock_page_or_retry(struct page
*page
, struct mm_struct
*mm
,
669 return trylock_page(page
) || __lock_page_or_retry(page
, mm
, flags
);
673 * This is exported only for wait_on_page_locked/wait_on_page_writeback, etc.,
674 * and should not be used directly.
676 extern void wait_on_page_bit(struct page
*page
, int bit_nr
);
677 extern int wait_on_page_bit_killable(struct page
*page
, int bit_nr
);
680 * Wait for a page to be unlocked.
682 * This must be called with the caller "holding" the page,
683 * ie with increased "page->count" so that the page won't
684 * go away during the wait..
686 static inline void wait_on_page_locked(struct page
*page
)
688 if (PageLocked(page
))
689 wait_on_page_bit(compound_head(page
), PG_locked
);
692 static inline int wait_on_page_locked_killable(struct page
*page
)
694 if (!PageLocked(page
))
696 return wait_on_page_bit_killable(compound_head(page
), PG_locked
);
699 int put_and_wait_on_page_locked(struct page
*page
, int state
);
700 void wait_on_page_writeback(struct page
*page
);
701 int wait_on_page_writeback_killable(struct page
*page
);
702 extern void end_page_writeback(struct page
*page
);
703 void wait_for_stable_page(struct page
*page
);
705 void __set_page_dirty(struct page
*, struct address_space
*, int warn
);
706 int __set_page_dirty_nobuffers(struct page
*page
);
707 int __set_page_dirty_no_writeback(struct page
*page
);
709 void page_endio(struct page
*page
, bool is_write
, int err
);
712 * set_page_private_2 - Set PG_private_2 on a page and take a ref
715 * Set the PG_private_2 flag on a page and take the reference needed for the VM
716 * to handle its lifetime correctly. This sets the flag and takes the
717 * reference unconditionally, so care must be taken not to set the flag again
718 * if it's already set.
720 static inline void set_page_private_2(struct page
*page
)
722 page
= compound_head(page
);
724 SetPagePrivate2(page
);
727 void end_page_private_2(struct page
*page
);
728 void wait_on_page_private_2(struct page
*page
);
729 int wait_on_page_private_2_killable(struct page
*page
);
732 * Add an arbitrary waiter to a page's wait queue
734 extern void add_page_wait_queue(struct page
*page
, wait_queue_entry_t
*waiter
);
737 * Fault everything in given userspace address range in.
739 static inline int fault_in_pages_writeable(char __user
*uaddr
, size_t size
)
741 char __user
*end
= uaddr
+ size
- 1;
743 if (unlikely(size
== 0))
746 if (unlikely(uaddr
> end
))
749 * Writing zeroes into userspace here is OK, because we know that if
750 * the zero gets there, we'll be overwriting it.
753 if (unlikely(__put_user(0, uaddr
) != 0))
756 } while (uaddr
<= end
);
758 /* Check whether the range spilled into the next page. */
759 if (((unsigned long)uaddr
& PAGE_MASK
) ==
760 ((unsigned long)end
& PAGE_MASK
))
761 return __put_user(0, end
);
766 static inline int fault_in_pages_readable(const char __user
*uaddr
, size_t size
)
769 const char __user
*end
= uaddr
+ size
- 1;
771 if (unlikely(size
== 0))
774 if (unlikely(uaddr
> end
))
778 if (unlikely(__get_user(c
, uaddr
) != 0))
781 } while (uaddr
<= end
);
783 /* Check whether the range spilled into the next page. */
784 if (((unsigned long)uaddr
& PAGE_MASK
) ==
785 ((unsigned long)end
& PAGE_MASK
)) {
786 return __get_user(c
, end
);
793 int add_to_page_cache_locked(struct page
*page
, struct address_space
*mapping
,
794 pgoff_t index
, gfp_t gfp_mask
);
795 int add_to_page_cache_lru(struct page
*page
, struct address_space
*mapping
,
796 pgoff_t index
, gfp_t gfp_mask
);
797 extern void delete_from_page_cache(struct page
*page
);
798 extern void __delete_from_page_cache(struct page
*page
, void *shadow
);
799 void replace_page_cache_page(struct page
*old
, struct page
*new);
800 void delete_from_page_cache_batch(struct address_space
*mapping
,
801 struct pagevec
*pvec
);
802 loff_t
mapping_seek_hole_data(struct address_space
*, loff_t start
, loff_t end
,
806 * Like add_to_page_cache_locked, but used to add newly allocated pages:
807 * the page is new, so we can just run __SetPageLocked() against it.
809 static inline int add_to_page_cache(struct page
*page
,
810 struct address_space
*mapping
, pgoff_t offset
, gfp_t gfp_mask
)
814 __SetPageLocked(page
);
815 error
= add_to_page_cache_locked(page
, mapping
, offset
, gfp_mask
);
817 __ClearPageLocked(page
);
822 * struct readahead_control - Describes a readahead request.
824 * A readahead request is for consecutive pages. Filesystems which
825 * implement the ->readahead method should call readahead_page() or
826 * readahead_page_batch() in a loop and attempt to start I/O against
827 * each page in the request.
829 * Most of the fields in this struct are private and should be accessed
830 * by the functions below.
832 * @file: The file, used primarily by network filesystems for authentication.
833 * May be NULL if invoked internally by the filesystem.
834 * @mapping: Readahead this filesystem object.
835 * @ra: File readahead state. May be NULL.
837 struct readahead_control
{
839 struct address_space
*mapping
;
840 struct file_ra_state
*ra
;
841 /* private: use the readahead_* accessors instead */
843 unsigned int _nr_pages
;
844 unsigned int _batch_count
;
847 #define DEFINE_READAHEAD(ractl, f, r, m, i) \
848 struct readahead_control ractl = { \
855 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
857 void page_cache_ra_unbounded(struct readahead_control
*,
858 unsigned long nr_to_read
, unsigned long lookahead_count
);
859 void page_cache_sync_ra(struct readahead_control
*, unsigned long req_count
);
860 void page_cache_async_ra(struct readahead_control
*, struct page
*,
861 unsigned long req_count
);
862 void readahead_expand(struct readahead_control
*ractl
,
863 loff_t new_start
, size_t new_len
);
866 * page_cache_sync_readahead - generic file readahead
867 * @mapping: address_space which holds the pagecache and I/O vectors
868 * @ra: file_ra_state which holds the readahead state
869 * @file: Used by the filesystem for authentication.
870 * @index: Index of first page to be read.
871 * @req_count: Total number of pages being read by the caller.
873 * page_cache_sync_readahead() should be called when a cache miss happened:
874 * it will submit the read. The readahead logic may decide to piggyback more
875 * pages onto the read request if access patterns suggest it will improve
879 void page_cache_sync_readahead(struct address_space
*mapping
,
880 struct file_ra_state
*ra
, struct file
*file
, pgoff_t index
,
881 unsigned long req_count
)
883 DEFINE_READAHEAD(ractl
, file
, ra
, mapping
, index
);
884 page_cache_sync_ra(&ractl
, req_count
);
888 * page_cache_async_readahead - file readahead for marked pages
889 * @mapping: address_space which holds the pagecache and I/O vectors
890 * @ra: file_ra_state which holds the readahead state
891 * @file: Used by the filesystem for authentication.
892 * @page: The page at @index which triggered the readahead call.
893 * @index: Index of first page to be read.
894 * @req_count: Total number of pages being read by the caller.
896 * page_cache_async_readahead() should be called when a page is used which
897 * is marked as PageReadahead; this is a marker to suggest that the application
898 * has used up enough of the readahead window that we should start pulling in
902 void page_cache_async_readahead(struct address_space
*mapping
,
903 struct file_ra_state
*ra
, struct file
*file
,
904 struct page
*page
, pgoff_t index
, unsigned long req_count
)
906 DEFINE_READAHEAD(ractl
, file
, ra
, mapping
, index
);
907 page_cache_async_ra(&ractl
, page
, req_count
);
911 * readahead_page - Get the next page to read.
912 * @rac: The current readahead request.
914 * Context: The page is locked and has an elevated refcount. The caller
915 * should decreases the refcount once the page has been submitted for I/O
916 * and unlock the page once all I/O to that page has completed.
917 * Return: A pointer to the next page, or %NULL if we are done.
919 static inline struct page
*readahead_page(struct readahead_control
*rac
)
923 BUG_ON(rac
->_batch_count
> rac
->_nr_pages
);
924 rac
->_nr_pages
-= rac
->_batch_count
;
925 rac
->_index
+= rac
->_batch_count
;
927 if (!rac
->_nr_pages
) {
928 rac
->_batch_count
= 0;
932 page
= xa_load(&rac
->mapping
->i_pages
, rac
->_index
);
933 VM_BUG_ON_PAGE(!PageLocked(page
), page
);
934 rac
->_batch_count
= thp_nr_pages(page
);
939 static inline unsigned int __readahead_batch(struct readahead_control
*rac
,
940 struct page
**array
, unsigned int array_sz
)
943 XA_STATE(xas
, &rac
->mapping
->i_pages
, 0);
946 BUG_ON(rac
->_batch_count
> rac
->_nr_pages
);
947 rac
->_nr_pages
-= rac
->_batch_count
;
948 rac
->_index
+= rac
->_batch_count
;
949 rac
->_batch_count
= 0;
951 xas_set(&xas
, rac
->_index
);
953 xas_for_each(&xas
, page
, rac
->_index
+ rac
->_nr_pages
- 1) {
954 if (xas_retry(&xas
, page
))
956 VM_BUG_ON_PAGE(!PageLocked(page
), page
);
957 VM_BUG_ON_PAGE(PageTail(page
), page
);
959 rac
->_batch_count
+= thp_nr_pages(page
);
962 * The page cache isn't using multi-index entries yet,
963 * so the xas cursor needs to be manually moved to the
964 * next index. This can be removed once the page cache
968 xas_set(&xas
, rac
->_index
+ rac
->_batch_count
);
979 * readahead_page_batch - Get a batch of pages to read.
980 * @rac: The current readahead request.
981 * @array: An array of pointers to struct page.
983 * Context: The pages are locked and have an elevated refcount. The caller
984 * should decreases the refcount once the page has been submitted for I/O
985 * and unlock the page once all I/O to that page has completed.
986 * Return: The number of pages placed in the array. 0 indicates the request
989 #define readahead_page_batch(rac, array) \
990 __readahead_batch(rac, array, ARRAY_SIZE(array))
993 * readahead_pos - The byte offset into the file of this readahead request.
994 * @rac: The readahead request.
996 static inline loff_t
readahead_pos(struct readahead_control
*rac
)
998 return (loff_t
)rac
->_index
* PAGE_SIZE
;
1002 * readahead_length - The number of bytes in this readahead request.
1003 * @rac: The readahead request.
1005 static inline size_t readahead_length(struct readahead_control
*rac
)
1007 return rac
->_nr_pages
* PAGE_SIZE
;
1011 * readahead_index - The index of the first page in this readahead request.
1012 * @rac: The readahead request.
1014 static inline pgoff_t
readahead_index(struct readahead_control
*rac
)
1020 * readahead_count - The number of pages in this readahead request.
1021 * @rac: The readahead request.
1023 static inline unsigned int readahead_count(struct readahead_control
*rac
)
1025 return rac
->_nr_pages
;
1029 * readahead_batch_length - The number of bytes in the current batch.
1030 * @rac: The readahead request.
1032 static inline size_t readahead_batch_length(struct readahead_control
*rac
)
1034 return rac
->_batch_count
* PAGE_SIZE
;
1037 static inline unsigned long dir_pages(struct inode
*inode
)
1039 return (unsigned long)(inode
->i_size
+ PAGE_SIZE
- 1) >>
1044 * page_mkwrite_check_truncate - check if page was truncated
1045 * @page: the page to check
1046 * @inode: the inode to check the page against
1048 * Returns the number of bytes in the page up to EOF,
1049 * or -EFAULT if the page was truncated.
1051 static inline int page_mkwrite_check_truncate(struct page
*page
,
1052 struct inode
*inode
)
1054 loff_t size
= i_size_read(inode
);
1055 pgoff_t index
= size
>> PAGE_SHIFT
;
1056 int offset
= offset_in_page(size
);
1058 if (page
->mapping
!= inode
->i_mapping
)
1061 /* page is wholly inside EOF */
1062 if (page
->index
< index
)
1064 /* page is wholly past EOF */
1065 if (page
->index
> index
|| !offset
)
1067 /* page is partially inside EOF */
1072 * i_blocks_per_page - How many blocks fit in this page.
1073 * @inode: The inode which contains the blocks.
1074 * @page: The page (head page if the page is a THP).
1076 * If the block size is larger than the size of this page, return zero.
1078 * Context: The caller should hold a refcount on the page to prevent it
1080 * Return: The number of filesystem blocks covered by this page.
1083 unsigned int i_blocks_per_page(struct inode
*inode
, struct page
*page
)
1085 return thp_size(page
) >> inode
->i_blkbits
;
1087 #endif /* _LINUX_PAGEMAP_H */