Merge tag 'kbuild-fixes-v6.10' of git://git.kernel.org/pub/scm/linux/kernel/git/masah...
[linux-block.git] / include / linux / pagemap.h
CommitLineData
b2441318 1/* SPDX-License-Identifier: GPL-2.0 */
1da177e4
LT
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>
7c0f6ba6 13#include <linux/uaccess.h>
1da177e4 14#include <linux/gfp.h>
3e9f45bd 15#include <linux/bitops.h>
e286781d 16#include <linux/hardirq.h> /* for in_interrupt() */
8edf344c 17#include <linux/hugetlb_inline.h>
1da177e4 18
51dcbdac 19struct folio_batch;
aa65c29c 20
cbcc268b
MWO
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);
e003f74a 33int kiocb_invalidate_pages(struct kiocb *iocb, size_t count);
c402a9a9 34void kiocb_invalidate_post_direct_write(struct kiocb *iocb, size_t count);
3c435a0f 35
cbcc268b
MWO
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);
74e797d7
DH
43int filemap_invalidate_inode(struct inode *inode, bool flush,
44 loff_t start, loff_t end);
cbcc268b
MWO
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);
3c435a0f 62int kiocb_write_and_wait(struct kiocb *iocb, size_t count);
cbcc268b
MWO
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
7716506a
MWO
141static inline bool mapping_empty(struct address_space *mapping)
142{
143 return xa_empty(&mapping->i_pages);
144}
145
51b8c1fe
JW
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
1da177e4 196/*
9c5d760b 197 * Bits in mapping->flags.
1da177e4 198 */
9a896c9a 199enum mapping_flags {
9c5d760b
MH
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 */
371a096e 205 /* writeback related tags are not used */
9c5d760b 206 AS_NO_WRITEBACK_TAGS = 5,
ed2145c4 207 AS_LARGE_FOLIO_SUPPORT = 6,
b4fa966f 208 AS_RELEASE_ALWAYS, /* Call ->release_folio(), even if no private data */
762321da
CH
209 AS_STABLE_WRITES, /* must wait for writeback before modifying
210 folio contents */
13629252 211 AS_UNMOVABLE, /* The mapping cannot be moved, ever */
9a896c9a 212};
1da177e4 213
8ed1e46a
JL
214/**
215 * mapping_set_error - record a writeback error in the address_space
767e5ee5
MWO
216 * @mapping: the mapping in which an error should be set
217 * @error: the error to set in the mapping
8ed1e46a
JL
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 */
3e9f45bd
GC
228static inline void mapping_set_error(struct address_space *mapping, int error)
229{
8ed1e46a
JL
230 if (likely(!error))
231 return;
232
233 /* Record in wb_err for checkers using errseq_t based tracking */
735e4ae5
JL
234 __filemap_set_wb_err(mapping, error);
235
236 /* Record it in superblock */
8b7b2eb1
MK
237 if (mapping->host)
238 errseq_set(&mapping->host->i_sb->s_wb_err, error);
8ed1e46a
JL
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);
3e9f45bd
GC
245}
246
ba9ddf49
LS
247static inline void mapping_set_unevictable(struct address_space *mapping)
248{
249 set_bit(AS_UNEVICTABLE, &mapping->flags);
250}
251
89e004ea
LS
252static inline void mapping_clear_unevictable(struct address_space *mapping)
253{
254 clear_bit(AS_UNEVICTABLE, &mapping->flags);
255}
256
1eb6234e 257static inline bool mapping_unevictable(struct address_space *mapping)
ba9ddf49 258{
1eb6234e 259 return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
ba9ddf49 260}
ba9ddf49 261
91b0abe3
JW
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
371a096e
HY
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
b4fa966f
DH
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
762321da
CH
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
0003e2a4
SC
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
dd0fc66f 328static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
1da177e4 329{
9c5d760b 330 return mapping->gfp_mask;
1da177e4
LT
331}
332
c62d2555
MH
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
1da177e4
LT
340/*
341 * This is non-atomic. Only to be used before the mapping is activated.
342 * Probably needs a barrier...
343 */
260b2367 344static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
1da177e4 345{
9c5d760b 346 m->gfp_mask = mask;
1da177e4
LT
347}
348
79c13745
XY
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 MAX_PAGECACHE_ORDER HPAGE_PMD_ORDER
358#else
359#define MAX_PAGECACHE_ORDER 8
360#endif
361
ff36da69
MWO
362/**
363 * mapping_set_large_folios() - Indicate the file supports large folios.
364 * @mapping: The file.
365 *
366 * The filesystem should call this function in its inode constructor to
367 * indicate that the VFS can use large folios to cache the contents of
368 * the file.
369 *
370 * Context: This should not be called while the inode is active as it
371 * is non-atomic.
372 */
373static inline void mapping_set_large_folios(struct address_space *mapping)
374{
ed2145c4 375 __set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
ff36da69
MWO
376}
377
421f1ab4
MWO
378/*
379 * Large folio support currently depends on THP. These dependencies are
380 * being worked on but are not yet fixed.
381 */
ed2145c4 382static inline bool mapping_large_folio_support(struct address_space *mapping)
01c70267 383{
421f1ab4
MWO
384 return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) &&
385 test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
01c70267
MWO
386}
387
79c13745
XY
388/* Return the maximum folio size for this pagecache mapping, in bytes. */
389static inline size_t mapping_max_folio_size(struct address_space *mapping)
390{
391 if (mapping_large_folio_support(mapping))
392 return PAGE_SIZE << MAX_PAGECACHE_ORDER;
393 return PAGE_SIZE;
394}
395
6f4d2f97
MWO
396static inline int filemap_nr_thps(struct address_space *mapping)
397{
398#ifdef CONFIG_READ_ONLY_THP_FOR_FS
399 return atomic_read(&mapping->nr_thps);
400#else
401 return 0;
402#endif
403}
404
405static inline void filemap_nr_thps_inc(struct address_space *mapping)
406{
407#ifdef CONFIG_READ_ONLY_THP_FOR_FS
ed2145c4 408 if (!mapping_large_folio_support(mapping))
6f4d2f97
MWO
409 atomic_inc(&mapping->nr_thps);
410#else
06d44142 411 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
6f4d2f97
MWO
412#endif
413}
414
415static inline void filemap_nr_thps_dec(struct address_space *mapping)
416{
417#ifdef CONFIG_READ_ONLY_THP_FOR_FS
ed2145c4 418 if (!mapping_large_folio_support(mapping))
6f4d2f97
MWO
419 atomic_dec(&mapping->nr_thps);
420#else
06d44142 421 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
6f4d2f97
MWO
422#endif
423}
424
2f52578f
MWO
425struct address_space *page_mapping(struct page *);
426struct address_space *folio_mapping(struct folio *);
427struct address_space *swapcache_mapping(struct folio *);
428
429/**
430 * folio_file_mapping - Find the mapping this folio belongs to.
431 * @folio: The folio.
432 *
433 * For folios which are in the page cache, return the mapping that this
434 * page belongs to. Folios in the swap cache return the mapping of the
435 * swap file or swap device where the data is stored. This is different
436 * from the mapping returned by folio_mapping(). The only reason to
437 * use it is if, like NFS, you return 0 from ->activate_swapfile.
438 *
439 * Do not call this for folios which aren't in the page cache or swap cache.
440 */
441static inline struct address_space *folio_file_mapping(struct folio *folio)
442{
443 if (unlikely(folio_test_swapcache(folio)))
444 return swapcache_mapping(folio);
445
446 return folio->mapping;
447}
448
bc60abfb
MWO
449/**
450 * folio_flush_mapping - Find the file mapping this folio belongs to.
451 * @folio: The folio.
452 *
453 * For folios which are in the page cache, return the mapping that this
454 * page belongs to. Anonymous folios return NULL, even if they're in
455 * the swap cache. Other kinds of folio also return NULL.
456 *
457 * This is ONLY used by architecture cache flushing code. If you aren't
458 * writing cache flushing code, you want either folio_mapping() or
459 * folio_file_mapping().
842ca547 460 */
bc60abfb 461static inline struct address_space *folio_flush_mapping(struct folio *folio)
842ca547 462{
2f52578f 463 if (unlikely(folio_test_swapcache(folio)))
842ca547 464 return NULL;
bc60abfb 465
2f52578f 466 return folio_mapping(folio);
842ca547
MWO
467}
468
2f52578f
MWO
469static inline struct address_space *page_file_mapping(struct page *page)
470{
471 return folio_file_mapping(page_folio(page));
842ca547
MWO
472}
473
452c472e
DH
474/**
475 * folio_inode - Get the host inode for this folio.
476 * @folio: The folio.
477 *
478 * For folios which are in the page cache, return the inode that this folio
479 * belongs to.
480 *
481 * Do not call this for folios which aren't in the page cache.
482 */
483static inline struct inode *folio_inode(struct folio *folio)
484{
485 return folio->mapping->host;
486}
487
b03143ac 488/**
85d0a2ed
MWO
489 * folio_attach_private - Attach private data to a folio.
490 * @folio: Folio to attach data to.
491 * @data: Data to attach to folio.
b03143ac 492 *
85d0a2ed
MWO
493 * Attaching private data to a folio increments the page's reference count.
494 * The data must be detached before the folio will be freed.
b03143ac 495 */
85d0a2ed 496static inline void folio_attach_private(struct folio *folio, void *data)
b03143ac 497{
85d0a2ed
MWO
498 folio_get(folio);
499 folio->private = data;
500 folio_set_private(folio);
b03143ac
GJ
501}
502
a19672f6
DH
503/**
504 * folio_change_private - Change private data on a folio.
505 * @folio: Folio to change the data on.
506 * @data: Data to set on the folio.
507 *
508 * Change the private data attached to a folio and return the old
509 * data. The page must previously have had data attached and the data
510 * must be detached before the folio will be freed.
511 *
512 * Return: Data that was previously attached to the folio.
513 */
514static inline void *folio_change_private(struct folio *folio, void *data)
515{
516 void *old = folio_get_private(folio);
517
518 folio->private = data;
519 return old;
520}
521
b03143ac 522/**
85d0a2ed
MWO
523 * folio_detach_private - Detach private data from a folio.
524 * @folio: Folio to detach data from.
b03143ac 525 *
85d0a2ed 526 * Removes the data that was previously attached to the folio and decrements
b03143ac
GJ
527 * the refcount on the page.
528 *
85d0a2ed 529 * Return: Data that was attached to the folio.
b03143ac 530 */
85d0a2ed 531static inline void *folio_detach_private(struct folio *folio)
b03143ac 532{
85d0a2ed 533 void *data = folio_get_private(folio);
b03143ac 534
85d0a2ed 535 if (!folio_test_private(folio))
b03143ac 536 return NULL;
85d0a2ed
MWO
537 folio_clear_private(folio);
538 folio->private = NULL;
539 folio_put(folio);
b03143ac
GJ
540
541 return data;
542}
543
85d0a2ed
MWO
544static inline void attach_page_private(struct page *page, void *data)
545{
546 folio_attach_private(page_folio(page), data);
547}
548
549static inline void *detach_page_private(struct page *page)
550{
551 return folio_detach_private(page_folio(page));
552}
553
44110fe3 554#ifdef CONFIG_NUMA
b951aaff 555struct folio *filemap_alloc_folio_noprof(gfp_t gfp, unsigned int order);
44110fe3 556#else
b951aaff 557static inline struct folio *filemap_alloc_folio_noprof(gfp_t gfp, unsigned int order)
2ae88149 558{
b951aaff 559 return folio_alloc_noprof(gfp, order);
2ae88149
NP
560}
561#endif
562
b951aaff
SB
563#define filemap_alloc_folio(...) \
564 alloc_hooks(filemap_alloc_folio_noprof(__VA_ARGS__))
565
bb3c579e
MWO
566static inline struct page *__page_cache_alloc(gfp_t gfp)
567{
568 return &filemap_alloc_folio(gfp, 0)->page;
569}
570
8a5c743e 571static inline gfp_t readahead_gfp_mask(struct address_space *x)
7b1de586 572{
453f85d4 573 return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
7b1de586
WF
574}
575
e9b5b23e 576typedef int filler_t(struct file *, struct folio *);
1da177e4 577
0d3f9296 578pgoff_t page_cache_next_miss(struct address_space *mapping,
e7b563bb 579 pgoff_t index, unsigned long max_scan);
0d3f9296 580pgoff_t page_cache_prev_miss(struct address_space *mapping,
e7b563bb
JW
581 pgoff_t index, unsigned long max_scan);
582
ffc143db
MWO
583/**
584 * typedef fgf_t - Flags for getting folios from the page cache.
585 *
586 * Most users of the page cache will not need to use these flags;
587 * there are convenience functions such as filemap_get_folio() and
588 * filemap_lock_folio(). For users which need more control over exactly
589 * what is done with the folios, these flags to __filemap_get_folio()
590 * are available.
591 *
592 * * %FGP_ACCESSED - The folio will be marked accessed.
593 * * %FGP_LOCK - The folio is returned locked.
594 * * %FGP_CREAT - If no folio is present then a new folio is allocated,
595 * added to the page cache and the VM's LRU list. The folio is
596 * returned locked.
597 * * %FGP_FOR_MMAP - The caller wants to do its own locking dance if the
598 * folio is already in cache. If the folio was allocated, unlock it
599 * before returning so the caller can do the same dance.
600 * * %FGP_WRITE - The folio will be written to by the caller.
601 * * %FGP_NOFS - __GFP_FS will get cleared in gfp.
602 * * %FGP_NOWAIT - Don't block on the folio lock.
603 * * %FGP_STABLE - Wait for the folio to be stable (finished writeback)
604 * * %FGP_WRITEBEGIN - The flags to use in a filesystem write_begin()
605 * implementation.
606 */
607typedef unsigned int __bitwise fgf_t;
608
609#define FGP_ACCESSED ((__force fgf_t)0x00000001)
610#define FGP_LOCK ((__force fgf_t)0x00000002)
611#define FGP_CREAT ((__force fgf_t)0x00000004)
612#define FGP_WRITE ((__force fgf_t)0x00000008)
613#define FGP_NOFS ((__force fgf_t)0x00000010)
614#define FGP_NOWAIT ((__force fgf_t)0x00000020)
615#define FGP_FOR_MMAP ((__force fgf_t)0x00000040)
616#define FGP_STABLE ((__force fgf_t)0x00000080)
4f661701 617#define FGF_GET_ORDER(fgf) (((__force unsigned)fgf) >> 26) /* top 6 bits */
2457aec6 618
e999a5c5
MW
619#define FGP_WRITEBEGIN (FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE)
620
4f661701
MWO
621/**
622 * fgf_set_order - Encode a length in the fgf_t flags.
623 * @size: The suggested size of the folio to create.
624 *
625 * The caller of __filemap_get_folio() can use this to suggest a preferred
626 * size for the folio that is created. If there is already a folio at
627 * the index, it will be returned, no matter what its size. If a folio
628 * is freshly created, it may be of a different size than requested
629 * due to alignment constraints, memory pressure, or the presence of
630 * other folios at nearby indices.
631 */
632static inline fgf_t fgf_set_order(size_t size)
633{
634 unsigned int shift = ilog2(size);
635
636 if (shift <= PAGE_SHIFT)
637 return 0;
638 return (__force fgf_t)((shift - PAGE_SHIFT) << 26);
639}
640
263e721e 641void *filemap_get_entry(struct address_space *mapping, pgoff_t index);
3f0c6a07 642struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
ffc143db 643 fgf_t fgp_flags, gfp_t gfp);
3f0c6a07 644struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
ffc143db 645 fgf_t fgp_flags, gfp_t gfp);
3f0c6a07
MWO
646
647/**
648 * filemap_get_folio - Find and get a folio.
649 * @mapping: The address_space to search.
650 * @index: The page index.
651 *
652 * Looks up the page cache entry at @mapping & @index. If a folio is
653 * present, it is returned with an increased refcount.
654 *
66dabbb6
CH
655 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
656 * this index. Will not return a shadow, swap or DAX entry.
3f0c6a07
MWO
657 */
658static inline struct folio *filemap_get_folio(struct address_space *mapping,
659 pgoff_t index)
660{
661 return __filemap_get_folio(mapping, index, 0, 0);
662}
2457aec6 663
ccd16945
MWO
664/**
665 * filemap_lock_folio - Find and lock a folio.
666 * @mapping: The address_space to search.
667 * @index: The page index.
668 *
669 * Looks up the page cache entry at @mapping & @index. If a folio is
670 * present, it is returned locked with an increased refcount.
671 *
672 * Context: May sleep.
66dabbb6
CH
673 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
674 * this index. Will not return a shadow, swap or DAX entry.
ccd16945
MWO
675 */
676static inline struct folio *filemap_lock_folio(struct address_space *mapping,
677 pgoff_t index)
678{
679 return __filemap_get_folio(mapping, index, FGP_LOCK, 0);
680}
681
ee7a5906
VMO
682/**
683 * filemap_grab_folio - grab a folio from the page cache
684 * @mapping: The address space to search
685 * @index: The page index
686 *
687 * Looks up the page cache entry at @mapping & @index. If no folio is found,
688 * a new folio is created. The folio is locked, marked as accessed, and
689 * returned.
690 *
66dabbb6
CH
691 * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found
692 * and failed to create a folio.
ee7a5906
VMO
693 */
694static inline struct folio *filemap_grab_folio(struct address_space *mapping,
695 pgoff_t index)
696{
697 return __filemap_get_folio(mapping, index,
698 FGP_LOCK | FGP_ACCESSED | FGP_CREAT,
699 mapping_gfp_mask(mapping));
700}
701
2457aec6
MG
702/**
703 * find_get_page - find and get a page reference
704 * @mapping: the address_space to search
705 * @offset: the page index
706 *
707 * Looks up the page cache slot at @mapping & @offset. If there is a
708 * page cache page, it is returned with an increased refcount.
709 *
710 * Otherwise, %NULL is returned.
711 */
712static inline struct page *find_get_page(struct address_space *mapping,
713 pgoff_t offset)
714{
45f87de5 715 return pagecache_get_page(mapping, offset, 0, 0);
2457aec6
MG
716}
717
718static inline struct page *find_get_page_flags(struct address_space *mapping,
ffc143db 719 pgoff_t offset, fgf_t fgp_flags)
2457aec6 720{
45f87de5 721 return pagecache_get_page(mapping, offset, fgp_flags, 0);
2457aec6
MG
722}
723
724/**
725 * find_lock_page - locate, pin and lock a pagecache page
2457aec6 726 * @mapping: the address_space to search
89b42235 727 * @index: the page index
2457aec6 728 *
89b42235 729 * Looks up the page cache entry at @mapping & @index. If there is a
2457aec6
MG
730 * page cache page, it is returned locked and with an increased
731 * refcount.
732 *
a8cf7f27
MWO
733 * Context: May sleep.
734 * Return: A struct page or %NULL if there is no page in the cache for this
735 * index.
2457aec6
MG
736 */
737static inline struct page *find_lock_page(struct address_space *mapping,
a8cf7f27
MWO
738 pgoff_t index)
739{
740 return pagecache_get_page(mapping, index, FGP_LOCK, 0);
741}
742
2457aec6
MG
743/**
744 * find_or_create_page - locate or add a pagecache page
745 * @mapping: the page's address_space
746 * @index: the page's index into the mapping
747 * @gfp_mask: page allocation mode
748 *
749 * Looks up the page cache slot at @mapping & @offset. If there is a
750 * page cache page, it is returned locked and with an increased
751 * refcount.
752 *
753 * If the page is not present, a new page is allocated using @gfp_mask
754 * and added to the page cache and the VM's LRU list. The page is
755 * returned locked and with an increased refcount.
756 *
757 * On memory exhaustion, %NULL is returned.
758 *
759 * find_or_create_page() may sleep, even if @gfp_flags specifies an
760 * atomic allocation!
761 */
762static inline struct page *find_or_create_page(struct address_space *mapping,
767e5ee5 763 pgoff_t index, gfp_t gfp_mask)
2457aec6 764{
767e5ee5 765 return pagecache_get_page(mapping, index,
2457aec6 766 FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
45f87de5 767 gfp_mask);
2457aec6
MG
768}
769
770/**
771 * grab_cache_page_nowait - returns locked page at given index in given cache
772 * @mapping: target address_space
773 * @index: the page index
774 *
775 * Same as grab_cache_page(), but do not wait if the page is unavailable.
776 * This is intended for speculative data generators, where the data can
777 * be regenerated if the page couldn't be grabbed. This routine should
778 * be safe to call while holding the lock for another page.
779 *
780 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
781 * and deadlock against the caller's locked page.
782 */
783static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
784 pgoff_t index)
785{
786 return pagecache_get_page(mapping, index,
787 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
45f87de5 788 mapping_gfp_mask(mapping));
2457aec6
MG
789}
790
9257e156
MWO
791#define swapcache_index(folio) __page_file_index(&(folio)->page)
792
793/**
794 * folio_index - File index of a folio.
795 * @folio: The folio.
796 *
797 * For a folio which is either in the page cache or the swap cache,
798 * return its index within the address_space it belongs to. If you know
799 * the page is definitely in the page cache, you can look at the folio's
800 * index directly.
801 *
802 * Return: The index (offset in units of pages) of a folio in its file.
803 */
804static inline pgoff_t folio_index(struct folio *folio)
805{
806 if (unlikely(folio_test_swapcache(folio)))
807 return swapcache_index(folio);
808 return folio->index;
809}
810
f94b18f6
MWO
811/**
812 * folio_next_index - Get the index of the next folio.
813 * @folio: The current folio.
814 *
815 * Return: The index of the folio which follows this folio in the file.
816 */
817static inline pgoff_t folio_next_index(struct folio *folio)
818{
819 return folio->index + folio_nr_pages(folio);
820}
821
9257e156
MWO
822/**
823 * folio_file_page - The page for a particular index.
824 * @folio: The folio which contains this index.
825 * @index: The index we want to look up.
826 *
827 * Sometimes after looking up a folio in the page cache, we need to
828 * obtain the specific page for an index (eg a page fault).
829 *
830 * Return: The page containing the file data for this index.
831 */
832static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
833{
9257e156
MWO
834 return folio_page(folio, index & (folio_nr_pages(folio) - 1));
835}
836
837/**
838 * folio_contains - Does this folio contain this index?
839 * @folio: The folio.
840 * @index: The page index within the file.
841 *
842 * Context: The caller should have the page locked in order to prevent
843 * (eg) shmem from moving the page between the page cache and swap cache
844 * and changing its index in the middle of the operation.
845 * Return: true or false.
846 */
847static inline bool folio_contains(struct folio *folio, pgoff_t index)
848{
9257e156
MWO
849 return index - folio_index(folio) < folio_nr_pages(folio);
850}
851
ec848215
MWO
852/*
853 * Given the page we found in the page cache, return the page corresponding
854 * to this index in the file
855 */
856static inline struct page *find_subpage(struct page *head, pgoff_t index)
4101196b 857{
ec848215
MWO
858 /* HugeTLBfs wants the head page regardless */
859 if (PageHuge(head))
860 return head;
4101196b 861
6c357848 862 return head + (index & (thp_nr_pages(head) - 1));
4101196b
MWO
863}
864
be0ced5e
MWO
865unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start,
866 pgoff_t end, struct folio_batch *fbatch);
35b47146
VMO
867unsigned filemap_get_folios_contig(struct address_space *mapping,
868 pgoff_t *start, pgoff_t end, struct folio_batch *fbatch);
247f9e1f
VMO
869unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start,
870 pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch);
1da177e4 871
54566b2c 872struct page *grab_cache_page_write_begin(struct address_space *mapping,
b7446e7c 873 pgoff_t index);
afddba49 874
1da177e4
LT
875/*
876 * Returns locked page at given index in given cache, creating it if needed.
877 */
57f6b96c
FW
878static inline struct page *grab_cache_page(struct address_space *mapping,
879 pgoff_t index)
1da177e4
LT
880{
881 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
882}
883
539a3322 884struct folio *read_cache_folio(struct address_space *, pgoff_t index,
e9b5b23e 885 filler_t *filler, struct file *file);
3e629597
MWO
886struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index,
887 gfp_t flags);
539a3322 888struct page *read_cache_page(struct address_space *, pgoff_t index,
e9b5b23e 889 filler_t *filler, struct file *file);
0531b2aa
LT
890extern struct page * read_cache_page_gfp(struct address_space *mapping,
891 pgoff_t index, gfp_t gfp_mask);
1da177e4 892
090d2b18 893static inline struct page *read_mapping_page(struct address_space *mapping,
c14c6843 894 pgoff_t index, struct file *file)
090d2b18 895{
c14c6843 896 return read_cache_page(mapping, index, NULL, file);
090d2b18
PE
897}
898
539a3322 899static inline struct folio *read_mapping_folio(struct address_space *mapping,
c14c6843 900 pgoff_t index, struct file *file)
539a3322 901{
c14c6843 902 return read_cache_folio(mapping, index, NULL, file);
539a3322
MWO
903}
904
a0f7a756 905/*
a08c7193 906 * Get the offset in PAGE_SIZE (even for hugetlb pages).
a0f7a756 907 */
a08c7193 908static inline pgoff_t page_to_pgoff(struct page *page)
a0f7a756 909{
fe3df441 910 struct page *head;
e9b61f19 911
e9b61f19 912 if (likely(!PageTransTail(page)))
09cbfeaf 913 return page->index;
e9b61f19 914
fe3df441 915 head = compound_head(page);
e9b61f19
KS
916 /*
917 * We don't initialize ->index for tail pages: calculate based on
918 * head page
919 */
fe3df441 920 return head->index + page - head;
a0f7a756
NH
921}
922
1da177e4
LT
923/*
924 * Return byte-offset into filesystem object for page.
925 */
926static inline loff_t page_offset(struct page *page)
927{
09cbfeaf 928 return ((loff_t)page->index) << PAGE_SHIFT;
1da177e4
LT
929}
930
f981c595
MG
931static inline loff_t page_file_offset(struct page *page)
932{
8cd79788 933 return ((loff_t)page_index(page)) << PAGE_SHIFT;
f981c595
MG
934}
935
352b47a6
MWO
936/**
937 * folio_pos - Returns the byte position of this folio in its file.
938 * @folio: The folio.
939 */
940static inline loff_t folio_pos(struct folio *folio)
941{
942 return page_offset(&folio->page);
943}
944
945/**
946 * folio_file_pos - Returns the byte position of this folio in its file.
947 * @folio: The folio.
948 *
949 * This differs from folio_pos() for folios which belong to a swap file.
950 * NFS is the only filesystem today which needs to use folio_file_pos().
951 */
952static inline loff_t folio_file_pos(struct folio *folio)
953{
954 return page_file_offset(&folio->page);
955}
956
f087b903
MWO
957/*
958 * Get the offset in PAGE_SIZE (even for hugetlb folios).
f087b903
MWO
959 */
960static inline pgoff_t folio_pgoff(struct folio *folio)
961{
f087b903
MWO
962 return folio->index;
963}
964
1da177e4
LT
965static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
966 unsigned long address)
967{
0fe6e20b 968 pgoff_t pgoff;
0fe6e20b 969 pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
1da177e4 970 pgoff += vma->vm_pgoff;
09cbfeaf 971 return pgoff;
1da177e4
LT
972}
973
c7510ab2 974struct wait_page_key {
df4d4f12 975 struct folio *folio;
c7510ab2
JA
976 int bit_nr;
977 int page_match;
978};
979
980struct wait_page_queue {
df4d4f12 981 struct folio *folio;
c7510ab2
JA
982 int bit_nr;
983 wait_queue_entry_t wait;
984};
985
cdc8fcb4 986static inline bool wake_page_match(struct wait_page_queue *wait_page,
c7510ab2
JA
987 struct wait_page_key *key)
988{
df4d4f12 989 if (wait_page->folio != key->folio)
cdc8fcb4 990 return false;
c7510ab2
JA
991 key->page_match = 1;
992
993 if (wait_page->bit_nr != key->bit_nr)
cdc8fcb4 994 return false;
d1932dc3 995
cdc8fcb4 996 return true;
d1932dc3
JA
997}
998
7c23c782 999void __folio_lock(struct folio *folio);
af7f29d9 1000int __folio_lock_killable(struct folio *folio);
fdc724d6 1001vm_fault_t __folio_lock_or_retry(struct folio *folio, struct vm_fault *vmf);
4e136428
MWO
1002void unlock_page(struct page *page);
1003void folio_unlock(struct folio *folio);
1da177e4 1004
cd125eea
MWO
1005/**
1006 * folio_trylock() - Attempt to lock a folio.
1007 * @folio: The folio to attempt to lock.
1008 *
1009 * Sometimes it is undesirable to wait for a folio to be unlocked (eg
1010 * when the locks are being taken in the wrong order, or if making
1011 * progress through a batch of folios is more important than processing
1012 * them in order). Usually folio_lock() is the correct function to call.
1013 *
1014 * Context: Any context.
1015 * Return: Whether the lock was successfully acquired.
1016 */
7c23c782
MWO
1017static inline bool folio_trylock(struct folio *folio)
1018{
1019 return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
1020}
1021
f4458845
AM
1022/*
1023 * Return true if the page was successfully locked
1024 */
62087556 1025static inline bool trylock_page(struct page *page)
529ae9aa 1026{
7c23c782
MWO
1027 return folio_trylock(page_folio(page));
1028}
1029
cd125eea
MWO
1030/**
1031 * folio_lock() - Lock this folio.
1032 * @folio: The folio to lock.
1033 *
1034 * The folio lock protects against many things, probably more than it
1035 * should. It is primarily held while a folio is being brought uptodate,
1036 * either from its backing file or from swap. It is also held while a
1037 * folio is being truncated from its address_space, so holding the lock
1038 * is sufficient to keep folio->mapping stable.
1039 *
1040 * The folio lock is also held while write() is modifying the page to
1041 * provide POSIX atomicity guarantees (as long as the write does not
1042 * cross a page boundary). Other modifications to the data in the folio
1043 * do not hold the folio lock and can race with writes, eg DMA and stores
1044 * to mapped pages.
1045 *
1046 * Context: May sleep. If you need to acquire the locks of two or
1047 * more folios, they must be in order of ascending index, if they are
1048 * in the same address_space. If they are in different address_spaces,
1049 * acquire the lock of the folio which belongs to the address_space which
1050 * has the lowest address in memory first.
1051 */
7c23c782
MWO
1052static inline void folio_lock(struct folio *folio)
1053{
1054 might_sleep();
1055 if (!folio_trylock(folio))
1056 __folio_lock(folio);
529ae9aa
NP
1057}
1058
cd125eea
MWO
1059/**
1060 * lock_page() - Lock the folio containing this page.
1061 * @page: The page to lock.
1062 *
1063 * See folio_lock() for a description of what the lock protects.
1064 * This is a legacy function and new code should probably use folio_lock()
1065 * instead.
1066 *
1067 * Context: May sleep. Pages in the same folio share a lock, so do not
1068 * attempt to lock two pages which share a folio.
db37648c 1069 */
1da177e4
LT
1070static inline void lock_page(struct page *page)
1071{
7c23c782 1072 struct folio *folio;
1da177e4 1073 might_sleep();
7c23c782
MWO
1074
1075 folio = page_folio(page);
1076 if (!folio_trylock(folio))
1077 __folio_lock(folio);
1da177e4 1078}
db37648c 1079
cd125eea
MWO
1080/**
1081 * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
1082 * @folio: The folio to lock.
1083 *
1084 * Attempts to lock the folio, like folio_lock(), except that the sleep
1085 * to acquire the lock is interruptible by a fatal signal.
1086 *
1087 * Context: May sleep; see folio_lock().
1088 * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
1089 */
af7f29d9
MWO
1090static inline int folio_lock_killable(struct folio *folio)
1091{
1092 might_sleep();
1093 if (!folio_trylock(folio))
1094 return __folio_lock_killable(folio);
1095 return 0;
1096}
1097
d065bd81 1098/*
19672a9e 1099 * folio_lock_or_retry - Lock the folio, unless this would block and the
d065bd81 1100 * caller indicated that it can handle a retry.
9a95f3cf 1101 *
c1e8d7c6 1102 * Return value and mmap_lock implications depend on flags; see
9138e47e 1103 * __folio_lock_or_retry().
d065bd81 1104 */
fdc724d6
SB
1105static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
1106 struct vm_fault *vmf)
d065bd81
ML
1107{
1108 might_sleep();
fdc724d6
SB
1109 if (!folio_trylock(folio))
1110 return __folio_lock_or_retry(folio, vmf);
1111 return 0;
d065bd81
ML
1112}
1113
1da177e4 1114/*
101c0bf6 1115 * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
74d81bfa 1116 * and should not be used directly.
1da177e4 1117 */
101c0bf6
MWO
1118void folio_wait_bit(struct folio *folio, int bit_nr);
1119int folio_wait_bit_killable(struct folio *folio, int bit_nr);
a4796e37 1120
1da177e4 1121/*
6baa8d60 1122 * Wait for a folio to be unlocked.
1da177e4 1123 *
6baa8d60 1124 * This must be called with the caller "holding" the folio,
cd125eea
MWO
1125 * ie with increased folio reference count so that the folio won't
1126 * go away during the wait.
1da177e4 1127 */
6baa8d60
MWO
1128static inline void folio_wait_locked(struct folio *folio)
1129{
1130 if (folio_test_locked(folio))
101c0bf6 1131 folio_wait_bit(folio, PG_locked);
6baa8d60
MWO
1132}
1133
1134static inline int folio_wait_locked_killable(struct folio *folio)
1135{
1136 if (!folio_test_locked(folio))
1137 return 0;
101c0bf6 1138 return folio_wait_bit_killable(folio, PG_locked);
6baa8d60
MWO
1139}
1140
1da177e4
LT
1141static inline void wait_on_page_locked(struct page *page)
1142{
6baa8d60 1143 folio_wait_locked(page_folio(page));
1da177e4
LT
1144}
1145
0b237047 1146void folio_end_read(struct folio *folio, bool success);
19343b5b 1147void wait_on_page_writeback(struct page *page);
490e016f
MWO
1148void folio_wait_writeback(struct folio *folio);
1149int folio_wait_writeback_killable(struct folio *folio);
4268b480
MWO
1150void end_page_writeback(struct page *page);
1151void folio_end_writeback(struct folio *folio);
1d1d1a76 1152void wait_for_stable_page(struct page *page);
a49d0c50 1153void folio_wait_stable(struct folio *folio);
203a3151 1154void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
566d3362 1155void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
fdaf532a
MWO
1156void __folio_cancel_dirty(struct folio *folio);
1157static inline void folio_cancel_dirty(struct folio *folio)
1158{
1159 /* Avoid atomic ops, locking, etc. when not actually needed. */
1160 if (folio_test_dirty(folio))
1161 __folio_cancel_dirty(folio);
1162}
9350f20a
MWO
1163bool folio_clear_dirty_for_io(struct folio *folio);
1164bool clear_page_dirty_for_io(struct page *page);
5ad6b2bd 1165void folio_invalidate(struct folio *folio, size_t offset, size_t length);
46de8b97 1166bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
3a6b2162 1167
2ec810d5
MWO
1168#ifdef CONFIG_MIGRATION
1169int filemap_migrate_folio(struct address_space *mapping, struct folio *dst,
1170 struct folio *src, enum migrate_mode mode);
1171#else
1172#define filemap_migrate_folio NULL
1173#endif
b47393f8
MWO
1174void folio_end_private_2(struct folio *folio);
1175void folio_wait_private_2(struct folio *folio);
1176int folio_wait_private_2_killable(struct folio *folio);
73e10ded 1177
385e1ca5
DH
1178/*
1179 * Add an arbitrary waiter to a page's wait queue
1180 */
df4d4f12 1181void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
385e1ca5 1182
1da177e4 1183/*
bb523b40 1184 * Fault in userspace address range.
1da177e4 1185 */
bb523b40 1186size_t fault_in_writeable(char __user *uaddr, size_t size);
da32b581 1187size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
cdd591fc 1188size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
bb523b40 1189size_t fault_in_readable(const char __user *uaddr, size_t size);
f56f821f 1190
529ae9aa 1191int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
9dd3d069
MWO
1192 pgoff_t index, gfp_t gfp);
1193int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1194 pgoff_t index, gfp_t gfp);
452e9e69 1195void filemap_remove_folio(struct folio *folio);
452e9e69 1196void __filemap_remove_folio(struct folio *folio, void *shadow);
3720dd6d 1197void replace_page_cache_folio(struct folio *old, struct folio *new);
aa65c29c 1198void delete_from_page_cache_batch(struct address_space *mapping,
51dcbdac 1199 struct folio_batch *fbatch);
82c50f8b 1200bool filemap_release_folio(struct folio *folio, gfp_t gfp);
41139aa4
MWO
1201loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1202 int whence);
529ae9aa 1203
9dd3d069
MWO
1204/* Must be non-static for BPF error injection */
1205int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1206 pgoff_t index, gfp_t gfp, void **shadowp);
1207
4bdcd1dd
JA
1208bool filemap_range_has_writeback(struct address_space *mapping,
1209 loff_t start_byte, loff_t end_byte);
1210
1211/**
1212 * filemap_range_needs_writeback - check if range potentially needs writeback
1213 * @mapping: address space within which to check
1214 * @start_byte: offset in bytes where the range starts
1215 * @end_byte: offset in bytes where the range ends (inclusive)
1216 *
1217 * Find at least one page in the range supplied, usually used to check if
1218 * direct writing in this range will trigger a writeback. Used by O_DIRECT
1219 * read/write with IOCB_NOWAIT, to see if the caller needs to do
1220 * filemap_write_and_wait_range() before proceeding.
1221 *
1222 * Return: %true if the caller should do filemap_write_and_wait_range() before
1223 * doing O_DIRECT to a page in this range, %false otherwise.
1224 */
1225static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1226 loff_t start_byte,
1227 loff_t end_byte)
1228{
1229 if (!mapping->nrpages)
1230 return false;
1231 if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1232 !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1233 return false;
1234 return filemap_range_has_writeback(mapping, start_byte, end_byte);
1235}
1236
042124cc
MWO
1237/**
1238 * struct readahead_control - Describes a readahead request.
1239 *
1240 * A readahead request is for consecutive pages. Filesystems which
1241 * implement the ->readahead method should call readahead_page() or
1242 * readahead_page_batch() in a loop and attempt to start I/O against
1243 * each page in the request.
1244 *
1245 * Most of the fields in this struct are private and should be accessed
1246 * by the functions below.
1247 *
1248 * @file: The file, used primarily by network filesystems for authentication.
1249 * May be NULL if invoked internally by the filesystem.
1250 * @mapping: Readahead this filesystem object.
fcd9ae4f 1251 * @ra: File readahead state. May be NULL.
042124cc
MWO
1252 */
1253struct readahead_control {
1254 struct file *file;
1255 struct address_space *mapping;
fcd9ae4f 1256 struct file_ra_state *ra;
042124cc
MWO
1257/* private: use the readahead_* accessors instead */
1258 pgoff_t _index;
1259 unsigned int _nr_pages;
1260 unsigned int _batch_count;
17604240
CH
1261 bool _workingset;
1262 unsigned long _pflags;
042124cc
MWO
1263};
1264
fcd9ae4f
MWO
1265#define DEFINE_READAHEAD(ractl, f, r, m, i) \
1266 struct readahead_control ractl = { \
1aa83cfa
MWO
1267 .file = f, \
1268 .mapping = m, \
fcd9ae4f 1269 .ra = r, \
1aa83cfa
MWO
1270 ._index = i, \
1271 }
1272
fefa7c47
MWO
1273#define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
1274
1275void page_cache_ra_unbounded(struct readahead_control *,
1276 unsigned long nr_to_read, unsigned long lookahead_count);
fcd9ae4f 1277void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
7836d999 1278void page_cache_async_ra(struct readahead_control *, struct folio *,
fefa7c47 1279 unsigned long req_count);
3ca23644
DH
1280void readahead_expand(struct readahead_control *ractl,
1281 loff_t new_start, size_t new_len);
fefa7c47
MWO
1282
1283/**
1284 * page_cache_sync_readahead - generic file readahead
1285 * @mapping: address_space which holds the pagecache and I/O vectors
1286 * @ra: file_ra_state which holds the readahead state
1287 * @file: Used by the filesystem for authentication.
1288 * @index: Index of first page to be read.
1289 * @req_count: Total number of pages being read by the caller.
1290 *
1291 * page_cache_sync_readahead() should be called when a cache miss happened:
1292 * it will submit the read. The readahead logic may decide to piggyback more
1293 * pages onto the read request if access patterns suggest it will improve
1294 * performance.
1295 */
1296static inline
1297void page_cache_sync_readahead(struct address_space *mapping,
1298 struct file_ra_state *ra, struct file *file, pgoff_t index,
1299 unsigned long req_count)
1300{
fcd9ae4f
MWO
1301 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1302 page_cache_sync_ra(&ractl, req_count);
fefa7c47
MWO
1303}
1304
1305/**
1306 * page_cache_async_readahead - file readahead for marked pages
1307 * @mapping: address_space which holds the pagecache and I/O vectors
1308 * @ra: file_ra_state which holds the readahead state
1309 * @file: Used by the filesystem for authentication.
2ebdd1df 1310 * @folio: The folio at @index which triggered the readahead call.
fefa7c47
MWO
1311 * @index: Index of first page to be read.
1312 * @req_count: Total number of pages being read by the caller.
1313 *
1314 * page_cache_async_readahead() should be called when a page is used which
1315 * is marked as PageReadahead; this is a marker to suggest that the application
1316 * has used up enough of the readahead window that we should start pulling in
1317 * more pages.
1318 */
1319static inline
1320void page_cache_async_readahead(struct address_space *mapping,
1321 struct file_ra_state *ra, struct file *file,
2ebdd1df 1322 struct folio *folio, pgoff_t index, unsigned long req_count)
fefa7c47 1323{
fcd9ae4f 1324 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
2ebdd1df 1325 page_cache_async_ra(&ractl, folio, req_count);
fefa7c47
MWO
1326}
1327
9bf70167
MWO
1328static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1329{
1330 struct folio *folio;
1331
1332 BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1333 ractl->_nr_pages -= ractl->_batch_count;
1334 ractl->_index += ractl->_batch_count;
1335
1336 if (!ractl->_nr_pages) {
1337 ractl->_batch_count = 0;
1338 return NULL;
1339 }
1340
1341 folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1342 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1343 ractl->_batch_count = folio_nr_pages(folio);
1344
1345 return folio;
1346}
1347
042124cc
MWO
1348/**
1349 * readahead_page - Get the next page to read.
9bf70167 1350 * @ractl: The current readahead request.
042124cc
MWO
1351 *
1352 * Context: The page is locked and has an elevated refcount. The caller
1353 * should decreases the refcount once the page has been submitted for I/O
1354 * and unlock the page once all I/O to that page has completed.
1355 * Return: A pointer to the next page, or %NULL if we are done.
1356 */
9bf70167 1357static inline struct page *readahead_page(struct readahead_control *ractl)
042124cc 1358{
9bf70167 1359 struct folio *folio = __readahead_folio(ractl);
042124cc 1360
9bf70167
MWO
1361 return &folio->page;
1362}
042124cc 1363
9bf70167
MWO
1364/**
1365 * readahead_folio - Get the next folio to read.
1366 * @ractl: The current readahead request.
1367 *
1368 * Context: The folio is locked. The caller should unlock the folio once
1369 * all I/O to that folio has completed.
1370 * Return: A pointer to the next folio, or %NULL if we are done.
1371 */
1372static inline struct folio *readahead_folio(struct readahead_control *ractl)
1373{
1374 struct folio *folio = __readahead_folio(ractl);
042124cc 1375
9bf70167
MWO
1376 if (folio)
1377 folio_put(folio);
1378 return folio;
042124cc
MWO
1379}
1380
1381static inline unsigned int __readahead_batch(struct readahead_control *rac,
1382 struct page **array, unsigned int array_sz)
1383{
1384 unsigned int i = 0;
1385 XA_STATE(xas, &rac->mapping->i_pages, 0);
1386 struct page *page;
1387
1388 BUG_ON(rac->_batch_count > rac->_nr_pages);
1389 rac->_nr_pages -= rac->_batch_count;
1390 rac->_index += rac->_batch_count;
1391 rac->_batch_count = 0;
1392
1393 xas_set(&xas, rac->_index);
1394 rcu_read_lock();
1395 xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
4349a83a
MWO
1396 if (xas_retry(&xas, page))
1397 continue;
042124cc
MWO
1398 VM_BUG_ON_PAGE(!PageLocked(page), page);
1399 VM_BUG_ON_PAGE(PageTail(page), page);
1400 array[i++] = page;
6c357848 1401 rac->_batch_count += thp_nr_pages(page);
042124cc
MWO
1402 if (i == array_sz)
1403 break;
1404 }
1405 rcu_read_unlock();
1406
1407 return i;
1408}
1409
1410/**
1411 * readahead_page_batch - Get a batch of pages to read.
1412 * @rac: The current readahead request.
1413 * @array: An array of pointers to struct page.
1414 *
1415 * Context: The pages are locked and have an elevated refcount. The caller
1416 * should decreases the refcount once the page has been submitted for I/O
1417 * and unlock the page once all I/O to that page has completed.
1418 * Return: The number of pages placed in the array. 0 indicates the request
1419 * is complete.
1420 */
1421#define readahead_page_batch(rac, array) \
1422 __readahead_batch(rac, array, ARRAY_SIZE(array))
1423
1424/**
1425 * readahead_pos - The byte offset into the file of this readahead request.
1426 * @rac: The readahead request.
1427 */
1428static inline loff_t readahead_pos(struct readahead_control *rac)
1429{
1430 return (loff_t)rac->_index * PAGE_SIZE;
1431}
1432
1433/**
1434 * readahead_length - The number of bytes in this readahead request.
1435 * @rac: The readahead request.
1436 */
076171a6 1437static inline size_t readahead_length(struct readahead_control *rac)
042124cc 1438{
076171a6 1439 return rac->_nr_pages * PAGE_SIZE;
042124cc
MWO
1440}
1441
1442/**
1443 * readahead_index - The index of the first page in this readahead request.
1444 * @rac: The readahead request.
1445 */
1446static inline pgoff_t readahead_index(struct readahead_control *rac)
1447{
1448 return rac->_index;
1449}
1450
1451/**
1452 * readahead_count - The number of pages in this readahead request.
1453 * @rac: The readahead request.
1454 */
1455static inline unsigned int readahead_count(struct readahead_control *rac)
1456{
1457 return rac->_nr_pages;
1458}
1459
32c0a6bc
MWO
1460/**
1461 * readahead_batch_length - The number of bytes in the current batch.
1462 * @rac: The readahead request.
1463 */
076171a6 1464static inline size_t readahead_batch_length(struct readahead_control *rac)
32c0a6bc
MWO
1465{
1466 return rac->_batch_count * PAGE_SIZE;
1467}
1468
b57c2cb9
FF
1469static inline unsigned long dir_pages(struct inode *inode)
1470{
09cbfeaf
KS
1471 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1472 PAGE_SHIFT;
b57c2cb9
FF
1473}
1474
f705bf84
MWO
1475/**
1476 * folio_mkwrite_check_truncate - check if folio was truncated
1477 * @folio: the folio to check
1478 * @inode: the inode to check the folio against
1479 *
1480 * Return: the number of bytes in the folio up to EOF,
1481 * or -EFAULT if the folio was truncated.
1482 */
1483static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1484 struct inode *inode)
1485{
1486 loff_t size = i_size_read(inode);
1487 pgoff_t index = size >> PAGE_SHIFT;
1488 size_t offset = offset_in_folio(folio, size);
1489
1490 if (!folio->mapping)
1491 return -EFAULT;
1492
1493 /* folio is wholly inside EOF */
1494 if (folio_next_index(folio) - 1 < index)
1495 return folio_size(folio);
1496 /* folio is wholly past EOF */
1497 if (folio->index > index || !offset)
1498 return -EFAULT;
1499 /* folio is partially inside EOF */
1500 return offset;
1501}
1502
243145bc
AG
1503/**
1504 * page_mkwrite_check_truncate - check if page was truncated
1505 * @page: the page to check
1506 * @inode: the inode to check the page against
1507 *
1508 * Returns the number of bytes in the page up to EOF,
1509 * or -EFAULT if the page was truncated.
1510 */
1511static inline int page_mkwrite_check_truncate(struct page *page,
1512 struct inode *inode)
1513{
1514 loff_t size = i_size_read(inode);
1515 pgoff_t index = size >> PAGE_SHIFT;
1516 int offset = offset_in_page(size);
1517
1518 if (page->mapping != inode->i_mapping)
1519 return -EFAULT;
1520
1521 /* page is wholly inside EOF */
1522 if (page->index < index)
1523 return PAGE_SIZE;
1524 /* page is wholly past EOF */
1525 if (page->index > index || !offset)
1526 return -EFAULT;
1527 /* page is partially inside EOF */
1528 return offset;
1529}
1530
24addd84 1531/**
9eb7c76d 1532 * i_blocks_per_folio - How many blocks fit in this folio.
24addd84 1533 * @inode: The inode which contains the blocks.
9eb7c76d 1534 * @folio: The folio.
24addd84 1535 *
9eb7c76d 1536 * If the block size is larger than the size of this folio, return zero.
24addd84 1537 *
9eb7c76d 1538 * Context: The caller should hold a refcount on the folio to prevent it
24addd84 1539 * from being split.
9eb7c76d 1540 * Return: The number of filesystem blocks covered by this folio.
24addd84 1541 */
9eb7c76d
MWO
1542static inline
1543unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1544{
1545 return folio_size(folio) >> inode->i_blkbits;
1546}
1547
24addd84
MWO
1548static inline
1549unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1550{
9eb7c76d 1551 return i_blocks_per_folio(inode, page_folio(page));
24addd84 1552}
1da177e4 1553#endif /* _LINUX_PAGEMAP_H */