Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef _LINUX_PAGEMAP_H
3#define _LINUX_PAGEMAP_H
4
5/*
6 * Copyright 1995 Linus Torvalds
7 */
8#include <linux/mm.h>
9#include <linux/fs.h>
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>
18
19struct folio_batch;
20
21unsigned long invalidate_mapping_pages(struct address_space *mapping,
22 pgoff_t start, pgoff_t end);
23
24static inline void invalidate_remote_inode(struct inode *inode)
25{
26 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
27 S_ISLNK(inode->i_mode))
28 invalidate_mapping_pages(inode->i_mapping, 0, -1);
29}
30int invalidate_inode_pages2(struct address_space *mapping);
31int invalidate_inode_pages2_range(struct address_space *mapping,
32 pgoff_t start, pgoff_t end);
33int kiocb_invalidate_pages(struct kiocb *iocb, size_t count);
34void kiocb_invalidate_post_direct_write(struct kiocb *iocb, size_t count);
35
36int write_inode_now(struct inode *, int sync);
37int filemap_fdatawrite(struct address_space *);
38int filemap_flush(struct address_space *);
39int filemap_fdatawait_keep_errors(struct address_space *mapping);
40int filemap_fdatawait_range(struct address_space *, loff_t lstart, loff_t lend);
41int filemap_fdatawait_range_keep_errors(struct address_space *mapping,
42 loff_t start_byte, loff_t end_byte);
43int filemap_invalidate_inode(struct inode *inode, bool flush,
44 loff_t start, loff_t end);
45
46static inline int filemap_fdatawait(struct address_space *mapping)
47{
48 return filemap_fdatawait_range(mapping, 0, LLONG_MAX);
49}
50
51bool filemap_range_has_page(struct address_space *, loff_t lstart, loff_t lend);
52int filemap_write_and_wait_range(struct address_space *mapping,
53 loff_t lstart, loff_t lend);
54int __filemap_fdatawrite_range(struct address_space *mapping,
55 loff_t start, loff_t end, int sync_mode);
56int filemap_fdatawrite_range(struct address_space *mapping,
57 loff_t start, loff_t end);
58int filemap_check_errors(struct address_space *mapping);
59void __filemap_set_wb_err(struct address_space *mapping, int err);
60int filemap_fdatawrite_wbc(struct address_space *mapping,
61 struct writeback_control *wbc);
62int kiocb_write_and_wait(struct kiocb *iocb, size_t count);
63
64static inline int filemap_write_and_wait(struct address_space *mapping)
65{
66 return filemap_write_and_wait_range(mapping, 0, LLONG_MAX);
67}
68
69/**
70 * filemap_set_wb_err - set a writeback error on an address_space
71 * @mapping: mapping in which to set writeback error
72 * @err: error to be set in mapping
73 *
74 * When writeback fails in some way, we must record that error so that
75 * userspace can be informed when fsync and the like are called. We endeavor
76 * to report errors on any file that was open at the time of the error. Some
77 * internal callers also need to know when writeback errors have occurred.
78 *
79 * When a writeback error occurs, most filesystems will want to call
80 * filemap_set_wb_err to record the error in the mapping so that it will be
81 * automatically reported whenever fsync is called on the file.
82 */
83static inline void filemap_set_wb_err(struct address_space *mapping, int err)
84{
85 /* Fastpath for common case of no error */
86 if (unlikely(err))
87 __filemap_set_wb_err(mapping, err);
88}
89
90/**
91 * filemap_check_wb_err - has an error occurred since the mark was sampled?
92 * @mapping: mapping to check for writeback errors
93 * @since: previously-sampled errseq_t
94 *
95 * Grab the errseq_t value from the mapping, and see if it has changed "since"
96 * the given value was sampled.
97 *
98 * If it has then report the latest error set, otherwise return 0.
99 */
100static inline int filemap_check_wb_err(struct address_space *mapping,
101 errseq_t since)
102{
103 return errseq_check(&mapping->wb_err, since);
104}
105
106/**
107 * filemap_sample_wb_err - sample the current errseq_t to test for later errors
108 * @mapping: mapping to be sampled
109 *
110 * Writeback errors are always reported relative to a particular sample point
111 * in the past. This function provides those sample points.
112 */
113static inline errseq_t filemap_sample_wb_err(struct address_space *mapping)
114{
115 return errseq_sample(&mapping->wb_err);
116}
117
118/**
119 * file_sample_sb_err - sample the current errseq_t to test for later errors
120 * @file: file pointer to be sampled
121 *
122 * Grab the most current superblock-level errseq_t value for the given
123 * struct file.
124 */
125static inline errseq_t file_sample_sb_err(struct file *file)
126{
127 return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err);
128}
129
130/*
131 * Flush file data before changing attributes. Caller must hold any locks
132 * required to prevent further writes to this file until we're done setting
133 * flags.
134 */
135static inline int inode_drain_writes(struct inode *inode)
136{
137 inode_dio_wait(inode);
138 return filemap_write_and_wait(inode->i_mapping);
139}
140
141static inline bool mapping_empty(struct address_space *mapping)
142{
143 return xa_empty(&mapping->i_pages);
144}
145
146/*
147 * mapping_shrinkable - test if page cache state allows inode reclaim
148 * @mapping: the page cache mapping
149 *
150 * This checks the mapping's cache state for the pupose of inode
151 * reclaim and LRU management.
152 *
153 * The caller is expected to hold the i_lock, but is not required to
154 * hold the i_pages lock, which usually protects cache state. That's
155 * because the i_lock and the list_lru lock that protect the inode and
156 * its LRU state don't nest inside the irq-safe i_pages lock.
157 *
158 * Cache deletions are performed under the i_lock, which ensures that
159 * when an inode goes empty, it will reliably get queued on the LRU.
160 *
161 * Cache additions do not acquire the i_lock and may race with this
162 * check, in which case we'll report the inode as shrinkable when it
163 * has cache pages. This is okay: the shrinker also checks the
164 * refcount and the referenced bit, which will be elevated or set in
165 * the process of adding new cache pages to an inode.
166 */
167static inline bool mapping_shrinkable(struct address_space *mapping)
168{
169 void *head;
170
171 /*
172 * On highmem systems, there could be lowmem pressure from the
173 * inodes before there is highmem pressure from the page
174 * cache. Make inodes shrinkable regardless of cache state.
175 */
176 if (IS_ENABLED(CONFIG_HIGHMEM))
177 return true;
178
179 /* Cache completely empty? Shrink away. */
180 head = rcu_access_pointer(mapping->i_pages.xa_head);
181 if (!head)
182 return true;
183
184 /*
185 * The xarray stores single offset-0 entries directly in the
186 * head pointer, which allows non-resident page cache entries
187 * to escape the shadow shrinker's list of xarray nodes. The
188 * inode shrinker needs to pick them up under memory pressure.
189 */
190 if (!xa_is_node(head) && xa_is_value(head))
191 return true;
192
193 return false;
194}
195
196/*
197 * Bits in mapping->flags.
198 */
199enum mapping_flags {
200 AS_EIO = 0, /* IO error on async write */
201 AS_ENOSPC = 1, /* ENOSPC on async write */
202 AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */
203 AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */
204 AS_EXITING = 4, /* final truncate in progress */
205 /* writeback related tags are not used */
206 AS_NO_WRITEBACK_TAGS = 5,
207 AS_LARGE_FOLIO_SUPPORT = 6,
208 AS_RELEASE_ALWAYS, /* Call ->release_folio(), even if no private data */
209 AS_STABLE_WRITES, /* must wait for writeback before modifying
210 folio contents */
211 AS_UNMOVABLE, /* The mapping cannot be moved, ever */
212};
213
214/**
215 * mapping_set_error - record a writeback error in the address_space
216 * @mapping: the mapping in which an error should be set
217 * @error: the error to set in the mapping
218 *
219 * When writeback fails in some way, we must record that error so that
220 * userspace can be informed when fsync and the like are called. We endeavor
221 * to report errors on any file that was open at the time of the error. Some
222 * internal callers also need to know when writeback errors have occurred.
223 *
224 * When a writeback error occurs, most filesystems will want to call
225 * mapping_set_error to record the error in the mapping so that it can be
226 * reported when the application calls fsync(2).
227 */
228static inline void mapping_set_error(struct address_space *mapping, int error)
229{
230 if (likely(!error))
231 return;
232
233 /* Record in wb_err for checkers using errseq_t based tracking */
234 __filemap_set_wb_err(mapping, error);
235
236 /* Record it in superblock */
237 if (mapping->host)
238 errseq_set(&mapping->host->i_sb->s_wb_err, error);
239
240 /* Record it in flags for now, for legacy callers */
241 if (error == -ENOSPC)
242 set_bit(AS_ENOSPC, &mapping->flags);
243 else
244 set_bit(AS_EIO, &mapping->flags);
245}
246
247static inline void mapping_set_unevictable(struct address_space *mapping)
248{
249 set_bit(AS_UNEVICTABLE, &mapping->flags);
250}
251
252static inline void mapping_clear_unevictable(struct address_space *mapping)
253{
254 clear_bit(AS_UNEVICTABLE, &mapping->flags);
255}
256
257static inline bool mapping_unevictable(struct address_space *mapping)
258{
259 return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
260}
261
262static inline void mapping_set_exiting(struct address_space *mapping)
263{
264 set_bit(AS_EXITING, &mapping->flags);
265}
266
267static inline int mapping_exiting(struct address_space *mapping)
268{
269 return test_bit(AS_EXITING, &mapping->flags);
270}
271
272static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
273{
274 set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
275}
276
277static inline int mapping_use_writeback_tags(struct address_space *mapping)
278{
279 return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
280}
281
282static inline bool mapping_release_always(const struct address_space *mapping)
283{
284 return test_bit(AS_RELEASE_ALWAYS, &mapping->flags);
285}
286
287static inline void mapping_set_release_always(struct address_space *mapping)
288{
289 set_bit(AS_RELEASE_ALWAYS, &mapping->flags);
290}
291
292static inline void mapping_clear_release_always(struct address_space *mapping)
293{
294 clear_bit(AS_RELEASE_ALWAYS, &mapping->flags);
295}
296
297static inline bool mapping_stable_writes(const struct address_space *mapping)
298{
299 return test_bit(AS_STABLE_WRITES, &mapping->flags);
300}
301
302static inline void mapping_set_stable_writes(struct address_space *mapping)
303{
304 set_bit(AS_STABLE_WRITES, &mapping->flags);
305}
306
307static inline void mapping_clear_stable_writes(struct address_space *mapping)
308{
309 clear_bit(AS_STABLE_WRITES, &mapping->flags);
310}
311
312static inline void mapping_set_unmovable(struct address_space *mapping)
313{
314 /*
315 * It's expected unmovable mappings are also unevictable. Compaction
316 * migrate scanner (isolate_migratepages_block()) relies on this to
317 * reduce page locking.
318 */
319 set_bit(AS_UNEVICTABLE, &mapping->flags);
320 set_bit(AS_UNMOVABLE, &mapping->flags);
321}
322
323static inline bool mapping_unmovable(struct address_space *mapping)
324{
325 return test_bit(AS_UNMOVABLE, &mapping->flags);
326}
327
328static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
329{
330 return mapping->gfp_mask;
331}
332
333/* Restricts the given gfp_mask to what the mapping allows. */
334static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
335 gfp_t gfp_mask)
336{
337 return mapping_gfp_mask(mapping) & gfp_mask;
338}
339
340/*
341 * This is non-atomic. Only to be used before the mapping is activated.
342 * Probably needs a barrier...
343 */
344static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
345{
346 m->gfp_mask = mask;
347}
348
349/*
350 * There are some parts of the kernel which assume that PMD entries
351 * are exactly HPAGE_PMD_ORDER. Those should be fixed, but until then,
352 * limit the maximum allocation order to PMD size. I'm not aware of any
353 * assumptions about maximum order if THP are disabled, but 8 seems like
354 * a good order (that's 1MB if you're using 4kB pages)
355 */
356#ifdef CONFIG_TRANSPARENT_HUGEPAGE
357#define PREFERRED_MAX_PAGECACHE_ORDER HPAGE_PMD_ORDER
358#else
359#define PREFERRED_MAX_PAGECACHE_ORDER 8
360#endif
361
362/*
363 * xas_split_alloc() does not support arbitrary orders. This implies no
364 * 512MB THP on ARM64 with 64KB base page size.
365 */
366#define MAX_XAS_ORDER (XA_CHUNK_SHIFT * 2 - 1)
367#define MAX_PAGECACHE_ORDER min(MAX_XAS_ORDER, PREFERRED_MAX_PAGECACHE_ORDER)
368
369/**
370 * mapping_set_large_folios() - Indicate the file supports large folios.
371 * @mapping: The file.
372 *
373 * The filesystem should call this function in its inode constructor to
374 * indicate that the VFS can use large folios to cache the contents of
375 * the file.
376 *
377 * Context: This should not be called while the inode is active as it
378 * is non-atomic.
379 */
380static inline void mapping_set_large_folios(struct address_space *mapping)
381{
382 __set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
383}
384
385/*
386 * Large folio support currently depends on THP. These dependencies are
387 * being worked on but are not yet fixed.
388 */
389static inline bool mapping_large_folio_support(struct address_space *mapping)
390{
391 /* AS_LARGE_FOLIO_SUPPORT is only reasonable for pagecache folios */
392 VM_WARN_ONCE((unsigned long)mapping & PAGE_MAPPING_ANON,
393 "Anonymous mapping always supports large folio");
394
395 return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) &&
396 test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
397}
398
399/* Return the maximum folio size for this pagecache mapping, in bytes. */
400static inline size_t mapping_max_folio_size(struct address_space *mapping)
401{
402 if (mapping_large_folio_support(mapping))
403 return PAGE_SIZE << MAX_PAGECACHE_ORDER;
404 return PAGE_SIZE;
405}
406
407static inline int filemap_nr_thps(struct address_space *mapping)
408{
409#ifdef CONFIG_READ_ONLY_THP_FOR_FS
410 return atomic_read(&mapping->nr_thps);
411#else
412 return 0;
413#endif
414}
415
416static inline void filemap_nr_thps_inc(struct address_space *mapping)
417{
418#ifdef CONFIG_READ_ONLY_THP_FOR_FS
419 if (!mapping_large_folio_support(mapping))
420 atomic_inc(&mapping->nr_thps);
421#else
422 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
423#endif
424}
425
426static inline void filemap_nr_thps_dec(struct address_space *mapping)
427{
428#ifdef CONFIG_READ_ONLY_THP_FOR_FS
429 if (!mapping_large_folio_support(mapping))
430 atomic_dec(&mapping->nr_thps);
431#else
432 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
433#endif
434}
435
436struct address_space *page_mapping(struct page *);
437struct address_space *folio_mapping(struct folio *);
438struct address_space *swapcache_mapping(struct folio *);
439
440/**
441 * folio_file_mapping - Find the mapping this folio belongs to.
442 * @folio: The folio.
443 *
444 * For folios which are in the page cache, return the mapping that this
445 * page belongs to. Folios in the swap cache return the mapping of the
446 * swap file or swap device where the data is stored. This is different
447 * from the mapping returned by folio_mapping(). The only reason to
448 * use it is if, like NFS, you return 0 from ->activate_swapfile.
449 *
450 * Do not call this for folios which aren't in the page cache or swap cache.
451 */
452static inline struct address_space *folio_file_mapping(struct folio *folio)
453{
454 if (unlikely(folio_test_swapcache(folio)))
455 return swapcache_mapping(folio);
456
457 return folio->mapping;
458}
459
460/**
461 * folio_flush_mapping - Find the file mapping this folio belongs to.
462 * @folio: The folio.
463 *
464 * For folios which are in the page cache, return the mapping that this
465 * page belongs to. Anonymous folios return NULL, even if they're in
466 * the swap cache. Other kinds of folio also return NULL.
467 *
468 * This is ONLY used by architecture cache flushing code. If you aren't
469 * writing cache flushing code, you want either folio_mapping() or
470 * folio_file_mapping().
471 */
472static inline struct address_space *folio_flush_mapping(struct folio *folio)
473{
474 if (unlikely(folio_test_swapcache(folio)))
475 return NULL;
476
477 return folio_mapping(folio);
478}
479
480static inline struct address_space *page_file_mapping(struct page *page)
481{
482 return folio_file_mapping(page_folio(page));
483}
484
485/**
486 * folio_inode - Get the host inode for this folio.
487 * @folio: The folio.
488 *
489 * For folios which are in the page cache, return the inode that this folio
490 * belongs to.
491 *
492 * Do not call this for folios which aren't in the page cache.
493 */
494static inline struct inode *folio_inode(struct folio *folio)
495{
496 return folio->mapping->host;
497}
498
499/**
500 * folio_attach_private - Attach private data to a folio.
501 * @folio: Folio to attach data to.
502 * @data: Data to attach to folio.
503 *
504 * Attaching private data to a folio increments the page's reference count.
505 * The data must be detached before the folio will be freed.
506 */
507static inline void folio_attach_private(struct folio *folio, void *data)
508{
509 folio_get(folio);
510 folio->private = data;
511 folio_set_private(folio);
512}
513
514/**
515 * folio_change_private - Change private data on a folio.
516 * @folio: Folio to change the data on.
517 * @data: Data to set on the folio.
518 *
519 * Change the private data attached to a folio and return the old
520 * data. The page must previously have had data attached and the data
521 * must be detached before the folio will be freed.
522 *
523 * Return: Data that was previously attached to the folio.
524 */
525static inline void *folio_change_private(struct folio *folio, void *data)
526{
527 void *old = folio_get_private(folio);
528
529 folio->private = data;
530 return old;
531}
532
533/**
534 * folio_detach_private - Detach private data from a folio.
535 * @folio: Folio to detach data from.
536 *
537 * Removes the data that was previously attached to the folio and decrements
538 * the refcount on the page.
539 *
540 * Return: Data that was attached to the folio.
541 */
542static inline void *folio_detach_private(struct folio *folio)
543{
544 void *data = folio_get_private(folio);
545
546 if (!folio_test_private(folio))
547 return NULL;
548 folio_clear_private(folio);
549 folio->private = NULL;
550 folio_put(folio);
551
552 return data;
553}
554
555static inline void attach_page_private(struct page *page, void *data)
556{
557 folio_attach_private(page_folio(page), data);
558}
559
560static inline void *detach_page_private(struct page *page)
561{
562 return folio_detach_private(page_folio(page));
563}
564
565#ifdef CONFIG_NUMA
566struct folio *filemap_alloc_folio_noprof(gfp_t gfp, unsigned int order);
567#else
568static inline struct folio *filemap_alloc_folio_noprof(gfp_t gfp, unsigned int order)
569{
570 return folio_alloc_noprof(gfp, order);
571}
572#endif
573
574#define filemap_alloc_folio(...) \
575 alloc_hooks(filemap_alloc_folio_noprof(__VA_ARGS__))
576
577static inline struct page *__page_cache_alloc(gfp_t gfp)
578{
579 return &filemap_alloc_folio(gfp, 0)->page;
580}
581
582static inline gfp_t readahead_gfp_mask(struct address_space *x)
583{
584 return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
585}
586
587typedef int filler_t(struct file *, struct folio *);
588
589pgoff_t page_cache_next_miss(struct address_space *mapping,
590 pgoff_t index, unsigned long max_scan);
591pgoff_t page_cache_prev_miss(struct address_space *mapping,
592 pgoff_t index, unsigned long max_scan);
593
594/**
595 * typedef fgf_t - Flags for getting folios from the page cache.
596 *
597 * Most users of the page cache will not need to use these flags;
598 * there are convenience functions such as filemap_get_folio() and
599 * filemap_lock_folio(). For users which need more control over exactly
600 * what is done with the folios, these flags to __filemap_get_folio()
601 * are available.
602 *
603 * * %FGP_ACCESSED - The folio will be marked accessed.
604 * * %FGP_LOCK - The folio is returned locked.
605 * * %FGP_CREAT - If no folio is present then a new folio is allocated,
606 * added to the page cache and the VM's LRU list. The folio is
607 * returned locked.
608 * * %FGP_FOR_MMAP - The caller wants to do its own locking dance if the
609 * folio is already in cache. If the folio was allocated, unlock it
610 * before returning so the caller can do the same dance.
611 * * %FGP_WRITE - The folio will be written to by the caller.
612 * * %FGP_NOFS - __GFP_FS will get cleared in gfp.
613 * * %FGP_NOWAIT - Don't block on the folio lock.
614 * * %FGP_STABLE - Wait for the folio to be stable (finished writeback)
615 * * %FGP_WRITEBEGIN - The flags to use in a filesystem write_begin()
616 * implementation.
617 */
618typedef unsigned int __bitwise fgf_t;
619
620#define FGP_ACCESSED ((__force fgf_t)0x00000001)
621#define FGP_LOCK ((__force fgf_t)0x00000002)
622#define FGP_CREAT ((__force fgf_t)0x00000004)
623#define FGP_WRITE ((__force fgf_t)0x00000008)
624#define FGP_NOFS ((__force fgf_t)0x00000010)
625#define FGP_NOWAIT ((__force fgf_t)0x00000020)
626#define FGP_FOR_MMAP ((__force fgf_t)0x00000040)
627#define FGP_STABLE ((__force fgf_t)0x00000080)
628#define FGF_GET_ORDER(fgf) (((__force unsigned)fgf) >> 26) /* top 6 bits */
629
630#define FGP_WRITEBEGIN (FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE)
631
632/**
633 * fgf_set_order - Encode a length in the fgf_t flags.
634 * @size: The suggested size of the folio to create.
635 *
636 * The caller of __filemap_get_folio() can use this to suggest a preferred
637 * size for the folio that is created. If there is already a folio at
638 * the index, it will be returned, no matter what its size. If a folio
639 * is freshly created, it may be of a different size than requested
640 * due to alignment constraints, memory pressure, or the presence of
641 * other folios at nearby indices.
642 */
643static inline fgf_t fgf_set_order(size_t size)
644{
645 unsigned int shift = ilog2(size);
646
647 if (shift <= PAGE_SHIFT)
648 return 0;
649 return (__force fgf_t)((shift - PAGE_SHIFT) << 26);
650}
651
652void *filemap_get_entry(struct address_space *mapping, pgoff_t index);
653struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
654 fgf_t fgp_flags, gfp_t gfp);
655struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
656 fgf_t fgp_flags, gfp_t gfp);
657
658/**
659 * filemap_get_folio - Find and get a folio.
660 * @mapping: The address_space to search.
661 * @index: The page index.
662 *
663 * Looks up the page cache entry at @mapping & @index. If a folio is
664 * present, it is returned with an increased refcount.
665 *
666 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
667 * this index. Will not return a shadow, swap or DAX entry.
668 */
669static inline struct folio *filemap_get_folio(struct address_space *mapping,
670 pgoff_t index)
671{
672 return __filemap_get_folio(mapping, index, 0, 0);
673}
674
675/**
676 * filemap_lock_folio - Find and lock a folio.
677 * @mapping: The address_space to search.
678 * @index: The page index.
679 *
680 * Looks up the page cache entry at @mapping & @index. If a folio is
681 * present, it is returned locked with an increased refcount.
682 *
683 * Context: May sleep.
684 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
685 * this index. Will not return a shadow, swap or DAX entry.
686 */
687static inline struct folio *filemap_lock_folio(struct address_space *mapping,
688 pgoff_t index)
689{
690 return __filemap_get_folio(mapping, index, FGP_LOCK, 0);
691}
692
693/**
694 * filemap_grab_folio - grab a folio from the page cache
695 * @mapping: The address space to search
696 * @index: The page index
697 *
698 * Looks up the page cache entry at @mapping & @index. If no folio is found,
699 * a new folio is created. The folio is locked, marked as accessed, and
700 * returned.
701 *
702 * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found
703 * and failed to create a folio.
704 */
705static inline struct folio *filemap_grab_folio(struct address_space *mapping,
706 pgoff_t index)
707{
708 return __filemap_get_folio(mapping, index,
709 FGP_LOCK | FGP_ACCESSED | FGP_CREAT,
710 mapping_gfp_mask(mapping));
711}
712
713/**
714 * find_get_page - find and get a page reference
715 * @mapping: the address_space to search
716 * @offset: the page index
717 *
718 * Looks up the page cache slot at @mapping & @offset. If there is a
719 * page cache page, it is returned with an increased refcount.
720 *
721 * Otherwise, %NULL is returned.
722 */
723static inline struct page *find_get_page(struct address_space *mapping,
724 pgoff_t offset)
725{
726 return pagecache_get_page(mapping, offset, 0, 0);
727}
728
729static inline struct page *find_get_page_flags(struct address_space *mapping,
730 pgoff_t offset, fgf_t fgp_flags)
731{
732 return pagecache_get_page(mapping, offset, fgp_flags, 0);
733}
734
735/**
736 * find_lock_page - locate, pin and lock a pagecache page
737 * @mapping: the address_space to search
738 * @index: the page index
739 *
740 * Looks up the page cache entry at @mapping & @index. If there is a
741 * page cache page, it is returned locked and with an increased
742 * refcount.
743 *
744 * Context: May sleep.
745 * Return: A struct page or %NULL if there is no page in the cache for this
746 * index.
747 */
748static inline struct page *find_lock_page(struct address_space *mapping,
749 pgoff_t index)
750{
751 return pagecache_get_page(mapping, index, FGP_LOCK, 0);
752}
753
754/**
755 * find_or_create_page - locate or add a pagecache page
756 * @mapping: the page's address_space
757 * @index: the page's index into the mapping
758 * @gfp_mask: page allocation mode
759 *
760 * Looks up the page cache slot at @mapping & @offset. If there is a
761 * page cache page, it is returned locked and with an increased
762 * refcount.
763 *
764 * If the page is not present, a new page is allocated using @gfp_mask
765 * and added to the page cache and the VM's LRU list. The page is
766 * returned locked and with an increased refcount.
767 *
768 * On memory exhaustion, %NULL is returned.
769 *
770 * find_or_create_page() may sleep, even if @gfp_flags specifies an
771 * atomic allocation!
772 */
773static inline struct page *find_or_create_page(struct address_space *mapping,
774 pgoff_t index, gfp_t gfp_mask)
775{
776 return pagecache_get_page(mapping, index,
777 FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
778 gfp_mask);
779}
780
781/**
782 * grab_cache_page_nowait - returns locked page at given index in given cache
783 * @mapping: target address_space
784 * @index: the page index
785 *
786 * Same as grab_cache_page(), but do not wait if the page is unavailable.
787 * This is intended for speculative data generators, where the data can
788 * be regenerated if the page couldn't be grabbed. This routine should
789 * be safe to call while holding the lock for another page.
790 *
791 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
792 * and deadlock against the caller's locked page.
793 */
794static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
795 pgoff_t index)
796{
797 return pagecache_get_page(mapping, index,
798 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
799 mapping_gfp_mask(mapping));
800}
801
802#define swapcache_index(folio) __page_file_index(&(folio)->page)
803
804/**
805 * folio_index - File index of a folio.
806 * @folio: The folio.
807 *
808 * For a folio which is either in the page cache or the swap cache,
809 * return its index within the address_space it belongs to. If you know
810 * the page is definitely in the page cache, you can look at the folio's
811 * index directly.
812 *
813 * Return: The index (offset in units of pages) of a folio in its file.
814 */
815static inline pgoff_t folio_index(struct folio *folio)
816{
817 if (unlikely(folio_test_swapcache(folio)))
818 return swapcache_index(folio);
819 return folio->index;
820}
821
822/**
823 * folio_next_index - Get the index of the next folio.
824 * @folio: The current folio.
825 *
826 * Return: The index of the folio which follows this folio in the file.
827 */
828static inline pgoff_t folio_next_index(struct folio *folio)
829{
830 return folio->index + folio_nr_pages(folio);
831}
832
833/**
834 * folio_file_page - The page for a particular index.
835 * @folio: The folio which contains this index.
836 * @index: The index we want to look up.
837 *
838 * Sometimes after looking up a folio in the page cache, we need to
839 * obtain the specific page for an index (eg a page fault).
840 *
841 * Return: The page containing the file data for this index.
842 */
843static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
844{
845 return folio_page(folio, index & (folio_nr_pages(folio) - 1));
846}
847
848/**
849 * folio_contains - Does this folio contain this index?
850 * @folio: The folio.
851 * @index: The page index within the file.
852 *
853 * Context: The caller should have the page locked in order to prevent
854 * (eg) shmem from moving the page between the page cache and swap cache
855 * and changing its index in the middle of the operation.
856 * Return: true or false.
857 */
858static inline bool folio_contains(struct folio *folio, pgoff_t index)
859{
860 return index - folio_index(folio) < folio_nr_pages(folio);
861}
862
863/*
864 * Given the page we found in the page cache, return the page corresponding
865 * to this index in the file
866 */
867static inline struct page *find_subpage(struct page *head, pgoff_t index)
868{
869 /* HugeTLBfs wants the head page regardless */
870 if (PageHuge(head))
871 return head;
872
873 return head + (index & (thp_nr_pages(head) - 1));
874}
875
876unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start,
877 pgoff_t end, struct folio_batch *fbatch);
878unsigned filemap_get_folios_contig(struct address_space *mapping,
879 pgoff_t *start, pgoff_t end, struct folio_batch *fbatch);
880unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start,
881 pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch);
882
883struct page *grab_cache_page_write_begin(struct address_space *mapping,
884 pgoff_t index);
885
886/*
887 * Returns locked page at given index in given cache, creating it if needed.
888 */
889static inline struct page *grab_cache_page(struct address_space *mapping,
890 pgoff_t index)
891{
892 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
893}
894
895struct folio *read_cache_folio(struct address_space *, pgoff_t index,
896 filler_t *filler, struct file *file);
897struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index,
898 gfp_t flags);
899struct page *read_cache_page(struct address_space *, pgoff_t index,
900 filler_t *filler, struct file *file);
901extern struct page * read_cache_page_gfp(struct address_space *mapping,
902 pgoff_t index, gfp_t gfp_mask);
903
904static inline struct page *read_mapping_page(struct address_space *mapping,
905 pgoff_t index, struct file *file)
906{
907 return read_cache_page(mapping, index, NULL, file);
908}
909
910static inline struct folio *read_mapping_folio(struct address_space *mapping,
911 pgoff_t index, struct file *file)
912{
913 return read_cache_folio(mapping, index, NULL, file);
914}
915
916/*
917 * Get the offset in PAGE_SIZE (even for hugetlb pages).
918 */
919static inline pgoff_t page_to_pgoff(struct page *page)
920{
921 struct page *head;
922
923 if (likely(!PageTransTail(page)))
924 return page->index;
925
926 head = compound_head(page);
927 /*
928 * We don't initialize ->index for tail pages: calculate based on
929 * head page
930 */
931 return head->index + page - head;
932}
933
934/*
935 * Return byte-offset into filesystem object for page.
936 */
937static inline loff_t page_offset(struct page *page)
938{
939 return ((loff_t)page->index) << PAGE_SHIFT;
940}
941
942static inline loff_t page_file_offset(struct page *page)
943{
944 return ((loff_t)page_index(page)) << PAGE_SHIFT;
945}
946
947/**
948 * folio_pos - Returns the byte position of this folio in its file.
949 * @folio: The folio.
950 */
951static inline loff_t folio_pos(struct folio *folio)
952{
953 return page_offset(&folio->page);
954}
955
956/**
957 * folio_file_pos - Returns the byte position of this folio in its file.
958 * @folio: The folio.
959 *
960 * This differs from folio_pos() for folios which belong to a swap file.
961 * NFS is the only filesystem today which needs to use folio_file_pos().
962 */
963static inline loff_t folio_file_pos(struct folio *folio)
964{
965 return page_file_offset(&folio->page);
966}
967
968/*
969 * Get the offset in PAGE_SIZE (even for hugetlb folios).
970 */
971static inline pgoff_t folio_pgoff(struct folio *folio)
972{
973 return folio->index;
974}
975
976static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
977 unsigned long address)
978{
979 pgoff_t pgoff;
980 pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
981 pgoff += vma->vm_pgoff;
982 return pgoff;
983}
984
985struct wait_page_key {
986 struct folio *folio;
987 int bit_nr;
988 int page_match;
989};
990
991struct wait_page_queue {
992 struct folio *folio;
993 int bit_nr;
994 wait_queue_entry_t wait;
995};
996
997static inline bool wake_page_match(struct wait_page_queue *wait_page,
998 struct wait_page_key *key)
999{
1000 if (wait_page->folio != key->folio)
1001 return false;
1002 key->page_match = 1;
1003
1004 if (wait_page->bit_nr != key->bit_nr)
1005 return false;
1006
1007 return true;
1008}
1009
1010void __folio_lock(struct folio *folio);
1011int __folio_lock_killable(struct folio *folio);
1012vm_fault_t __folio_lock_or_retry(struct folio *folio, struct vm_fault *vmf);
1013void unlock_page(struct page *page);
1014void folio_unlock(struct folio *folio);
1015
1016/**
1017 * folio_trylock() - Attempt to lock a folio.
1018 * @folio: The folio to attempt to lock.
1019 *
1020 * Sometimes it is undesirable to wait for a folio to be unlocked (eg
1021 * when the locks are being taken in the wrong order, or if making
1022 * progress through a batch of folios is more important than processing
1023 * them in order). Usually folio_lock() is the correct function to call.
1024 *
1025 * Context: Any context.
1026 * Return: Whether the lock was successfully acquired.
1027 */
1028static inline bool folio_trylock(struct folio *folio)
1029{
1030 return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
1031}
1032
1033/*
1034 * Return true if the page was successfully locked
1035 */
1036static inline bool trylock_page(struct page *page)
1037{
1038 return folio_trylock(page_folio(page));
1039}
1040
1041/**
1042 * folio_lock() - Lock this folio.
1043 * @folio: The folio to lock.
1044 *
1045 * The folio lock protects against many things, probably more than it
1046 * should. It is primarily held while a folio is being brought uptodate,
1047 * either from its backing file or from swap. It is also held while a
1048 * folio is being truncated from its address_space, so holding the lock
1049 * is sufficient to keep folio->mapping stable.
1050 *
1051 * The folio lock is also held while write() is modifying the page to
1052 * provide POSIX atomicity guarantees (as long as the write does not
1053 * cross a page boundary). Other modifications to the data in the folio
1054 * do not hold the folio lock and can race with writes, eg DMA and stores
1055 * to mapped pages.
1056 *
1057 * Context: May sleep. If you need to acquire the locks of two or
1058 * more folios, they must be in order of ascending index, if they are
1059 * in the same address_space. If they are in different address_spaces,
1060 * acquire the lock of the folio which belongs to the address_space which
1061 * has the lowest address in memory first.
1062 */
1063static inline void folio_lock(struct folio *folio)
1064{
1065 might_sleep();
1066 if (!folio_trylock(folio))
1067 __folio_lock(folio);
1068}
1069
1070/**
1071 * lock_page() - Lock the folio containing this page.
1072 * @page: The page to lock.
1073 *
1074 * See folio_lock() for a description of what the lock protects.
1075 * This is a legacy function and new code should probably use folio_lock()
1076 * instead.
1077 *
1078 * Context: May sleep. Pages in the same folio share a lock, so do not
1079 * attempt to lock two pages which share a folio.
1080 */
1081static inline void lock_page(struct page *page)
1082{
1083 struct folio *folio;
1084 might_sleep();
1085
1086 folio = page_folio(page);
1087 if (!folio_trylock(folio))
1088 __folio_lock(folio);
1089}
1090
1091/**
1092 * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
1093 * @folio: The folio to lock.
1094 *
1095 * Attempts to lock the folio, like folio_lock(), except that the sleep
1096 * to acquire the lock is interruptible by a fatal signal.
1097 *
1098 * Context: May sleep; see folio_lock().
1099 * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
1100 */
1101static inline int folio_lock_killable(struct folio *folio)
1102{
1103 might_sleep();
1104 if (!folio_trylock(folio))
1105 return __folio_lock_killable(folio);
1106 return 0;
1107}
1108
1109/*
1110 * folio_lock_or_retry - Lock the folio, unless this would block and the
1111 * caller indicated that it can handle a retry.
1112 *
1113 * Return value and mmap_lock implications depend on flags; see
1114 * __folio_lock_or_retry().
1115 */
1116static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
1117 struct vm_fault *vmf)
1118{
1119 might_sleep();
1120 if (!folio_trylock(folio))
1121 return __folio_lock_or_retry(folio, vmf);
1122 return 0;
1123}
1124
1125/*
1126 * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
1127 * and should not be used directly.
1128 */
1129void folio_wait_bit(struct folio *folio, int bit_nr);
1130int folio_wait_bit_killable(struct folio *folio, int bit_nr);
1131
1132/*
1133 * Wait for a folio to be unlocked.
1134 *
1135 * This must be called with the caller "holding" the folio,
1136 * ie with increased folio reference count so that the folio won't
1137 * go away during the wait.
1138 */
1139static inline void folio_wait_locked(struct folio *folio)
1140{
1141 if (folio_test_locked(folio))
1142 folio_wait_bit(folio, PG_locked);
1143}
1144
1145static inline int folio_wait_locked_killable(struct folio *folio)
1146{
1147 if (!folio_test_locked(folio))
1148 return 0;
1149 return folio_wait_bit_killable(folio, PG_locked);
1150}
1151
1152static inline void wait_on_page_locked(struct page *page)
1153{
1154 folio_wait_locked(page_folio(page));
1155}
1156
1157void folio_end_read(struct folio *folio, bool success);
1158void wait_on_page_writeback(struct page *page);
1159void folio_wait_writeback(struct folio *folio);
1160int folio_wait_writeback_killable(struct folio *folio);
1161void end_page_writeback(struct page *page);
1162void folio_end_writeback(struct folio *folio);
1163void wait_for_stable_page(struct page *page);
1164void folio_wait_stable(struct folio *folio);
1165void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
1166void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
1167void __folio_cancel_dirty(struct folio *folio);
1168static inline void folio_cancel_dirty(struct folio *folio)
1169{
1170 /* Avoid atomic ops, locking, etc. when not actually needed. */
1171 if (folio_test_dirty(folio))
1172 __folio_cancel_dirty(folio);
1173}
1174bool folio_clear_dirty_for_io(struct folio *folio);
1175bool clear_page_dirty_for_io(struct page *page);
1176void folio_invalidate(struct folio *folio, size_t offset, size_t length);
1177bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
1178
1179#ifdef CONFIG_MIGRATION
1180int filemap_migrate_folio(struct address_space *mapping, struct folio *dst,
1181 struct folio *src, enum migrate_mode mode);
1182#else
1183#define filemap_migrate_folio NULL
1184#endif
1185void folio_end_private_2(struct folio *folio);
1186void folio_wait_private_2(struct folio *folio);
1187int folio_wait_private_2_killable(struct folio *folio);
1188
1189/*
1190 * Add an arbitrary waiter to a page's wait queue
1191 */
1192void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
1193
1194/*
1195 * Fault in userspace address range.
1196 */
1197size_t fault_in_writeable(char __user *uaddr, size_t size);
1198size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
1199size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
1200size_t fault_in_readable(const char __user *uaddr, size_t size);
1201
1202int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
1203 pgoff_t index, gfp_t gfp);
1204int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1205 pgoff_t index, gfp_t gfp);
1206void filemap_remove_folio(struct folio *folio);
1207void __filemap_remove_folio(struct folio *folio, void *shadow);
1208void replace_page_cache_folio(struct folio *old, struct folio *new);
1209void delete_from_page_cache_batch(struct address_space *mapping,
1210 struct folio_batch *fbatch);
1211bool filemap_release_folio(struct folio *folio, gfp_t gfp);
1212loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1213 int whence);
1214
1215/* Must be non-static for BPF error injection */
1216int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1217 pgoff_t index, gfp_t gfp, void **shadowp);
1218
1219bool filemap_range_has_writeback(struct address_space *mapping,
1220 loff_t start_byte, loff_t end_byte);
1221
1222/**
1223 * filemap_range_needs_writeback - check if range potentially needs writeback
1224 * @mapping: address space within which to check
1225 * @start_byte: offset in bytes where the range starts
1226 * @end_byte: offset in bytes where the range ends (inclusive)
1227 *
1228 * Find at least one page in the range supplied, usually used to check if
1229 * direct writing in this range will trigger a writeback. Used by O_DIRECT
1230 * read/write with IOCB_NOWAIT, to see if the caller needs to do
1231 * filemap_write_and_wait_range() before proceeding.
1232 *
1233 * Return: %true if the caller should do filemap_write_and_wait_range() before
1234 * doing O_DIRECT to a page in this range, %false otherwise.
1235 */
1236static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1237 loff_t start_byte,
1238 loff_t end_byte)
1239{
1240 if (!mapping->nrpages)
1241 return false;
1242 if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1243 !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1244 return false;
1245 return filemap_range_has_writeback(mapping, start_byte, end_byte);
1246}
1247
1248/**
1249 * struct readahead_control - Describes a readahead request.
1250 *
1251 * A readahead request is for consecutive pages. Filesystems which
1252 * implement the ->readahead method should call readahead_page() or
1253 * readahead_page_batch() in a loop and attempt to start I/O against
1254 * each page in the request.
1255 *
1256 * Most of the fields in this struct are private and should be accessed
1257 * by the functions below.
1258 *
1259 * @file: The file, used primarily by network filesystems for authentication.
1260 * May be NULL if invoked internally by the filesystem.
1261 * @mapping: Readahead this filesystem object.
1262 * @ra: File readahead state. May be NULL.
1263 */
1264struct readahead_control {
1265 struct file *file;
1266 struct address_space *mapping;
1267 struct file_ra_state *ra;
1268/* private: use the readahead_* accessors instead */
1269 pgoff_t _index;
1270 unsigned int _nr_pages;
1271 unsigned int _batch_count;
1272 bool _workingset;
1273 unsigned long _pflags;
1274};
1275
1276#define DEFINE_READAHEAD(ractl, f, r, m, i) \
1277 struct readahead_control ractl = { \
1278 .file = f, \
1279 .mapping = m, \
1280 .ra = r, \
1281 ._index = i, \
1282 }
1283
1284#define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
1285
1286void page_cache_ra_unbounded(struct readahead_control *,
1287 unsigned long nr_to_read, unsigned long lookahead_count);
1288void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
1289void page_cache_async_ra(struct readahead_control *, struct folio *,
1290 unsigned long req_count);
1291void readahead_expand(struct readahead_control *ractl,
1292 loff_t new_start, size_t new_len);
1293
1294/**
1295 * page_cache_sync_readahead - generic file readahead
1296 * @mapping: address_space which holds the pagecache and I/O vectors
1297 * @ra: file_ra_state which holds the readahead state
1298 * @file: Used by the filesystem for authentication.
1299 * @index: Index of first page to be read.
1300 * @req_count: Total number of pages being read by the caller.
1301 *
1302 * page_cache_sync_readahead() should be called when a cache miss happened:
1303 * it will submit the read. The readahead logic may decide to piggyback more
1304 * pages onto the read request if access patterns suggest it will improve
1305 * performance.
1306 */
1307static inline
1308void page_cache_sync_readahead(struct address_space *mapping,
1309 struct file_ra_state *ra, struct file *file, pgoff_t index,
1310 unsigned long req_count)
1311{
1312 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1313 page_cache_sync_ra(&ractl, req_count);
1314}
1315
1316/**
1317 * page_cache_async_readahead - file readahead for marked pages
1318 * @mapping: address_space which holds the pagecache and I/O vectors
1319 * @ra: file_ra_state which holds the readahead state
1320 * @file: Used by the filesystem for authentication.
1321 * @folio: The folio at @index which triggered the readahead call.
1322 * @index: Index of first page to be read.
1323 * @req_count: Total number of pages being read by the caller.
1324 *
1325 * page_cache_async_readahead() should be called when a page is used which
1326 * is marked as PageReadahead; this is a marker to suggest that the application
1327 * has used up enough of the readahead window that we should start pulling in
1328 * more pages.
1329 */
1330static inline
1331void page_cache_async_readahead(struct address_space *mapping,
1332 struct file_ra_state *ra, struct file *file,
1333 struct folio *folio, pgoff_t index, unsigned long req_count)
1334{
1335 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1336 page_cache_async_ra(&ractl, folio, req_count);
1337}
1338
1339static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1340{
1341 struct folio *folio;
1342
1343 BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1344 ractl->_nr_pages -= ractl->_batch_count;
1345 ractl->_index += ractl->_batch_count;
1346
1347 if (!ractl->_nr_pages) {
1348 ractl->_batch_count = 0;
1349 return NULL;
1350 }
1351
1352 folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1353 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1354 ractl->_batch_count = folio_nr_pages(folio);
1355
1356 return folio;
1357}
1358
1359/**
1360 * readahead_page - Get the next page to read.
1361 * @ractl: The current readahead request.
1362 *
1363 * Context: The page is locked and has an elevated refcount. The caller
1364 * should decreases the refcount once the page has been submitted for I/O
1365 * and unlock the page once all I/O to that page has completed.
1366 * Return: A pointer to the next page, or %NULL if we are done.
1367 */
1368static inline struct page *readahead_page(struct readahead_control *ractl)
1369{
1370 struct folio *folio = __readahead_folio(ractl);
1371
1372 return &folio->page;
1373}
1374
1375/**
1376 * readahead_folio - Get the next folio to read.
1377 * @ractl: The current readahead request.
1378 *
1379 * Context: The folio is locked. The caller should unlock the folio once
1380 * all I/O to that folio has completed.
1381 * Return: A pointer to the next folio, or %NULL if we are done.
1382 */
1383static inline struct folio *readahead_folio(struct readahead_control *ractl)
1384{
1385 struct folio *folio = __readahead_folio(ractl);
1386
1387 if (folio)
1388 folio_put(folio);
1389 return folio;
1390}
1391
1392static inline unsigned int __readahead_batch(struct readahead_control *rac,
1393 struct page **array, unsigned int array_sz)
1394{
1395 unsigned int i = 0;
1396 XA_STATE(xas, &rac->mapping->i_pages, 0);
1397 struct page *page;
1398
1399 BUG_ON(rac->_batch_count > rac->_nr_pages);
1400 rac->_nr_pages -= rac->_batch_count;
1401 rac->_index += rac->_batch_count;
1402 rac->_batch_count = 0;
1403
1404 xas_set(&xas, rac->_index);
1405 rcu_read_lock();
1406 xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1407 if (xas_retry(&xas, page))
1408 continue;
1409 VM_BUG_ON_PAGE(!PageLocked(page), page);
1410 VM_BUG_ON_PAGE(PageTail(page), page);
1411 array[i++] = page;
1412 rac->_batch_count += thp_nr_pages(page);
1413 if (i == array_sz)
1414 break;
1415 }
1416 rcu_read_unlock();
1417
1418 return i;
1419}
1420
1421/**
1422 * readahead_page_batch - Get a batch of pages to read.
1423 * @rac: The current readahead request.
1424 * @array: An array of pointers to struct page.
1425 *
1426 * Context: The pages are locked and have an elevated refcount. The caller
1427 * should decreases the refcount once the page has been submitted for I/O
1428 * and unlock the page once all I/O to that page has completed.
1429 * Return: The number of pages placed in the array. 0 indicates the request
1430 * is complete.
1431 */
1432#define readahead_page_batch(rac, array) \
1433 __readahead_batch(rac, array, ARRAY_SIZE(array))
1434
1435/**
1436 * readahead_pos - The byte offset into the file of this readahead request.
1437 * @rac: The readahead request.
1438 */
1439static inline loff_t readahead_pos(struct readahead_control *rac)
1440{
1441 return (loff_t)rac->_index * PAGE_SIZE;
1442}
1443
1444/**
1445 * readahead_length - The number of bytes in this readahead request.
1446 * @rac: The readahead request.
1447 */
1448static inline size_t readahead_length(struct readahead_control *rac)
1449{
1450 return rac->_nr_pages * PAGE_SIZE;
1451}
1452
1453/**
1454 * readahead_index - The index of the first page in this readahead request.
1455 * @rac: The readahead request.
1456 */
1457static inline pgoff_t readahead_index(struct readahead_control *rac)
1458{
1459 return rac->_index;
1460}
1461
1462/**
1463 * readahead_count - The number of pages in this readahead request.
1464 * @rac: The readahead request.
1465 */
1466static inline unsigned int readahead_count(struct readahead_control *rac)
1467{
1468 return rac->_nr_pages;
1469}
1470
1471/**
1472 * readahead_batch_length - The number of bytes in the current batch.
1473 * @rac: The readahead request.
1474 */
1475static inline size_t readahead_batch_length(struct readahead_control *rac)
1476{
1477 return rac->_batch_count * PAGE_SIZE;
1478}
1479
1480static inline unsigned long dir_pages(struct inode *inode)
1481{
1482 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1483 PAGE_SHIFT;
1484}
1485
1486/**
1487 * folio_mkwrite_check_truncate - check if folio was truncated
1488 * @folio: the folio to check
1489 * @inode: the inode to check the folio against
1490 *
1491 * Return: the number of bytes in the folio up to EOF,
1492 * or -EFAULT if the folio was truncated.
1493 */
1494static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1495 struct inode *inode)
1496{
1497 loff_t size = i_size_read(inode);
1498 pgoff_t index = size >> PAGE_SHIFT;
1499 size_t offset = offset_in_folio(folio, size);
1500
1501 if (!folio->mapping)
1502 return -EFAULT;
1503
1504 /* folio is wholly inside EOF */
1505 if (folio_next_index(folio) - 1 < index)
1506 return folio_size(folio);
1507 /* folio is wholly past EOF */
1508 if (folio->index > index || !offset)
1509 return -EFAULT;
1510 /* folio is partially inside EOF */
1511 return offset;
1512}
1513
1514/**
1515 * page_mkwrite_check_truncate - check if page was truncated
1516 * @page: the page to check
1517 * @inode: the inode to check the page against
1518 *
1519 * Returns the number of bytes in the page up to EOF,
1520 * or -EFAULT if the page was truncated.
1521 */
1522static inline int page_mkwrite_check_truncate(struct page *page,
1523 struct inode *inode)
1524{
1525 loff_t size = i_size_read(inode);
1526 pgoff_t index = size >> PAGE_SHIFT;
1527 int offset = offset_in_page(size);
1528
1529 if (page->mapping != inode->i_mapping)
1530 return -EFAULT;
1531
1532 /* page is wholly inside EOF */
1533 if (page->index < index)
1534 return PAGE_SIZE;
1535 /* page is wholly past EOF */
1536 if (page->index > index || !offset)
1537 return -EFAULT;
1538 /* page is partially inside EOF */
1539 return offset;
1540}
1541
1542/**
1543 * i_blocks_per_folio - How many blocks fit in this folio.
1544 * @inode: The inode which contains the blocks.
1545 * @folio: The folio.
1546 *
1547 * If the block size is larger than the size of this folio, return zero.
1548 *
1549 * Context: The caller should hold a refcount on the folio to prevent it
1550 * from being split.
1551 * Return: The number of filesystem blocks covered by this folio.
1552 */
1553static inline
1554unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1555{
1556 return folio_size(folio) >> inode->i_blkbits;
1557}
1558
1559static inline
1560unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1561{
1562 return i_blocks_per_folio(inode, page_folio(page));
1563}
1564#endif /* _LINUX_PAGEMAP_H */