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(mapping: inode->i_mapping, start: 0, end: -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);
35int filemap_invalidate_pages(struct address_space *mapping,
36 loff_t pos, loff_t end, bool nowait);
37
38int write_inode_now(struct inode *, int sync);
39int filemap_fdatawrite(struct address_space *);
40int filemap_flush(struct address_space *);
41int filemap_fdatawait_keep_errors(struct address_space *mapping);
42int filemap_fdatawait_range(struct address_space *, loff_t lstart, loff_t lend);
43int filemap_fdatawait_range_keep_errors(struct address_space *mapping,
44 loff_t start_byte, loff_t end_byte);
45int filemap_invalidate_inode(struct inode *inode, bool flush,
46 loff_t start, loff_t end);
47
48static inline int filemap_fdatawait(struct address_space *mapping)
49{
50 return filemap_fdatawait_range(mapping, lstart: 0, LLONG_MAX);
51}
52
53bool filemap_range_has_page(struct address_space *, loff_t lstart, loff_t lend);
54int filemap_write_and_wait_range(struct address_space *mapping,
55 loff_t lstart, loff_t lend);
56int __filemap_fdatawrite_range(struct address_space *mapping,
57 loff_t start, loff_t end, int sync_mode);
58int filemap_fdatawrite_range(struct address_space *mapping,
59 loff_t start, loff_t end);
60int filemap_check_errors(struct address_space *mapping);
61void __filemap_set_wb_err(struct address_space *mapping, int err);
62int filemap_fdatawrite_wbc(struct address_space *mapping,
63 struct writeback_control *wbc);
64int kiocb_write_and_wait(struct kiocb *iocb, size_t count);
65
66static inline int filemap_write_and_wait(struct address_space *mapping)
67{
68 return filemap_write_and_wait_range(mapping, lstart: 0, LLONG_MAX);
69}
70
71/**
72 * filemap_set_wb_err - set a writeback error on an address_space
73 * @mapping: mapping in which to set writeback error
74 * @err: error to be set in mapping
75 *
76 * When writeback fails in some way, we must record that error so that
77 * userspace can be informed when fsync and the like are called. We endeavor
78 * to report errors on any file that was open at the time of the error. Some
79 * internal callers also need to know when writeback errors have occurred.
80 *
81 * When a writeback error occurs, most filesystems will want to call
82 * filemap_set_wb_err to record the error in the mapping so that it will be
83 * automatically reported whenever fsync is called on the file.
84 */
85static inline void filemap_set_wb_err(struct address_space *mapping, int err)
86{
87 /* Fastpath for common case of no error */
88 if (unlikely(err))
89 __filemap_set_wb_err(mapping, err);
90}
91
92/**
93 * filemap_check_wb_err - has an error occurred since the mark was sampled?
94 * @mapping: mapping to check for writeback errors
95 * @since: previously-sampled errseq_t
96 *
97 * Grab the errseq_t value from the mapping, and see if it has changed "since"
98 * the given value was sampled.
99 *
100 * If it has then report the latest error set, otherwise return 0.
101 */
102static inline int filemap_check_wb_err(struct address_space *mapping,
103 errseq_t since)
104{
105 return errseq_check(eseq: &mapping->wb_err, since);
106}
107
108/**
109 * filemap_sample_wb_err - sample the current errseq_t to test for later errors
110 * @mapping: mapping to be sampled
111 *
112 * Writeback errors are always reported relative to a particular sample point
113 * in the past. This function provides those sample points.
114 */
115static inline errseq_t filemap_sample_wb_err(struct address_space *mapping)
116{
117 return errseq_sample(eseq: &mapping->wb_err);
118}
119
120/**
121 * file_sample_sb_err - sample the current errseq_t to test for later errors
122 * @file: file pointer to be sampled
123 *
124 * Grab the most current superblock-level errseq_t value for the given
125 * struct file.
126 */
127static inline errseq_t file_sample_sb_err(struct file *file)
128{
129 return errseq_sample(eseq: &file->f_path.dentry->d_sb->s_wb_err);
130}
131
132/*
133 * Flush file data before changing attributes. Caller must hold any locks
134 * required to prevent further writes to this file until we're done setting
135 * flags.
136 */
137static inline int inode_drain_writes(struct inode *inode)
138{
139 inode_dio_wait(inode);
140 return filemap_write_and_wait(mapping: inode->i_mapping);
141}
142
143static inline bool mapping_empty(const struct address_space *mapping)
144{
145 return xa_empty(xa: &mapping->i_pages);
146}
147
148/*
149 * mapping_shrinkable - test if page cache state allows inode reclaim
150 * @mapping: the page cache mapping
151 *
152 * This checks the mapping's cache state for the pupose of inode
153 * reclaim and LRU management.
154 *
155 * The caller is expected to hold the i_lock, but is not required to
156 * hold the i_pages lock, which usually protects cache state. That's
157 * because the i_lock and the list_lru lock that protect the inode and
158 * its LRU state don't nest inside the irq-safe i_pages lock.
159 *
160 * Cache deletions are performed under the i_lock, which ensures that
161 * when an inode goes empty, it will reliably get queued on the LRU.
162 *
163 * Cache additions do not acquire the i_lock and may race with this
164 * check, in which case we'll report the inode as shrinkable when it
165 * has cache pages. This is okay: the shrinker also checks the
166 * refcount and the referenced bit, which will be elevated or set in
167 * the process of adding new cache pages to an inode.
168 */
169static inline bool mapping_shrinkable(const struct address_space *mapping)
170{
171 void *head;
172
173 /*
174 * On highmem systems, there could be lowmem pressure from the
175 * inodes before there is highmem pressure from the page
176 * cache. Make inodes shrinkable regardless of cache state.
177 */
178 if (IS_ENABLED(CONFIG_HIGHMEM))
179 return true;
180
181 /* Cache completely empty? Shrink away. */
182 head = rcu_access_pointer(mapping->i_pages.xa_head);
183 if (!head)
184 return true;
185
186 /*
187 * The xarray stores single offset-0 entries directly in the
188 * head pointer, which allows non-resident page cache entries
189 * to escape the shadow shrinker's list of xarray nodes. The
190 * inode shrinker needs to pick them up under memory pressure.
191 */
192 if (!xa_is_node(entry: head) && xa_is_value(entry: head))
193 return true;
194
195 return false;
196}
197
198/*
199 * Bits in mapping->flags.
200 */
201enum mapping_flags {
202 AS_EIO = 0, /* IO error on async write */
203 AS_ENOSPC = 1, /* ENOSPC on async write */
204 AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */
205 AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */
206 AS_EXITING = 4, /* final truncate in progress */
207 /* writeback related tags are not used */
208 AS_NO_WRITEBACK_TAGS = 5,
209 AS_RELEASE_ALWAYS = 6, /* Call ->release_folio(), even if no private data */
210 AS_STABLE_WRITES = 7, /* must wait for writeback before modifying
211 folio contents */
212 AS_INACCESSIBLE = 8, /* Do not attempt direct R/W access to the mapping */
213 AS_WRITEBACK_MAY_DEADLOCK_ON_RECLAIM = 9,
214 AS_KERNEL_FILE = 10, /* mapping for a fake kernel file that shouldn't
215 account usage to user cgroups */
216 /* Bits 16-25 are used for FOLIO_ORDER */
217 AS_FOLIO_ORDER_BITS = 5,
218 AS_FOLIO_ORDER_MIN = 16,
219 AS_FOLIO_ORDER_MAX = AS_FOLIO_ORDER_MIN + AS_FOLIO_ORDER_BITS,
220};
221
222#define AS_FOLIO_ORDER_BITS_MASK ((1u << AS_FOLIO_ORDER_BITS) - 1)
223#define AS_FOLIO_ORDER_MIN_MASK (AS_FOLIO_ORDER_BITS_MASK << AS_FOLIO_ORDER_MIN)
224#define AS_FOLIO_ORDER_MAX_MASK (AS_FOLIO_ORDER_BITS_MASK << AS_FOLIO_ORDER_MAX)
225#define AS_FOLIO_ORDER_MASK (AS_FOLIO_ORDER_MIN_MASK | AS_FOLIO_ORDER_MAX_MASK)
226
227/**
228 * mapping_set_error - record a writeback error in the address_space
229 * @mapping: the mapping in which an error should be set
230 * @error: the error to set in the mapping
231 *
232 * When writeback fails in some way, we must record that error so that
233 * userspace can be informed when fsync and the like are called. We endeavor
234 * to report errors on any file that was open at the time of the error. Some
235 * internal callers also need to know when writeback errors have occurred.
236 *
237 * When a writeback error occurs, most filesystems will want to call
238 * mapping_set_error to record the error in the mapping so that it can be
239 * reported when the application calls fsync(2).
240 */
241static inline void mapping_set_error(struct address_space *mapping, int error)
242{
243 if (likely(!error))
244 return;
245
246 /* Record in wb_err for checkers using errseq_t based tracking */
247 __filemap_set_wb_err(mapping, err: error);
248
249 /* Record it in superblock */
250 if (mapping->host)
251 errseq_set(eseq: &mapping->host->i_sb->s_wb_err, err: error);
252
253 /* Record it in flags for now, for legacy callers */
254 if (error == -ENOSPC)
255 set_bit(nr: AS_ENOSPC, addr: &mapping->flags);
256 else
257 set_bit(nr: AS_EIO, addr: &mapping->flags);
258}
259
260static inline void mapping_set_unevictable(struct address_space *mapping)
261{
262 set_bit(nr: AS_UNEVICTABLE, addr: &mapping->flags);
263}
264
265static inline void mapping_clear_unevictable(struct address_space *mapping)
266{
267 clear_bit(nr: AS_UNEVICTABLE, addr: &mapping->flags);
268}
269
270static inline bool mapping_unevictable(const struct address_space *mapping)
271{
272 return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
273}
274
275static inline void mapping_set_exiting(struct address_space *mapping)
276{
277 set_bit(nr: AS_EXITING, addr: &mapping->flags);
278}
279
280static inline int mapping_exiting(const struct address_space *mapping)
281{
282 return test_bit(AS_EXITING, &mapping->flags);
283}
284
285static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
286{
287 set_bit(nr: AS_NO_WRITEBACK_TAGS, addr: &mapping->flags);
288}
289
290static inline int mapping_use_writeback_tags(const struct address_space *mapping)
291{
292 return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
293}
294
295static inline bool mapping_release_always(const struct address_space *mapping)
296{
297 return test_bit(AS_RELEASE_ALWAYS, &mapping->flags);
298}
299
300static inline void mapping_set_release_always(struct address_space *mapping)
301{
302 set_bit(nr: AS_RELEASE_ALWAYS, addr: &mapping->flags);
303}
304
305static inline void mapping_clear_release_always(struct address_space *mapping)
306{
307 clear_bit(nr: AS_RELEASE_ALWAYS, addr: &mapping->flags);
308}
309
310static inline bool mapping_stable_writes(const struct address_space *mapping)
311{
312 return test_bit(AS_STABLE_WRITES, &mapping->flags);
313}
314
315static inline void mapping_set_stable_writes(struct address_space *mapping)
316{
317 set_bit(nr: AS_STABLE_WRITES, addr: &mapping->flags);
318}
319
320static inline void mapping_clear_stable_writes(struct address_space *mapping)
321{
322 clear_bit(nr: AS_STABLE_WRITES, addr: &mapping->flags);
323}
324
325static inline void mapping_set_inaccessible(struct address_space *mapping)
326{
327 /*
328 * It's expected inaccessible mappings are also unevictable. Compaction
329 * migrate scanner (isolate_migratepages_block()) relies on this to
330 * reduce page locking.
331 */
332 set_bit(nr: AS_UNEVICTABLE, addr: &mapping->flags);
333 set_bit(nr: AS_INACCESSIBLE, addr: &mapping->flags);
334}
335
336static inline bool mapping_inaccessible(const struct address_space *mapping)
337{
338 return test_bit(AS_INACCESSIBLE, &mapping->flags);
339}
340
341static inline void mapping_set_writeback_may_deadlock_on_reclaim(struct address_space *mapping)
342{
343 set_bit(nr: AS_WRITEBACK_MAY_DEADLOCK_ON_RECLAIM, addr: &mapping->flags);
344}
345
346static inline bool mapping_writeback_may_deadlock_on_reclaim(const struct address_space *mapping)
347{
348 return test_bit(AS_WRITEBACK_MAY_DEADLOCK_ON_RECLAIM, &mapping->flags);
349}
350
351static inline gfp_t mapping_gfp_mask(const struct address_space *mapping)
352{
353 return mapping->gfp_mask;
354}
355
356/* Restricts the given gfp_mask to what the mapping allows. */
357static inline gfp_t mapping_gfp_constraint(const struct address_space *mapping,
358 gfp_t gfp_mask)
359{
360 return mapping_gfp_mask(mapping) & gfp_mask;
361}
362
363/*
364 * This is non-atomic. Only to be used before the mapping is activated.
365 * Probably needs a barrier...
366 */
367static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
368{
369 m->gfp_mask = mask;
370}
371
372/*
373 * There are some parts of the kernel which assume that PMD entries
374 * are exactly HPAGE_PMD_ORDER. Those should be fixed, but until then,
375 * limit the maximum allocation order to PMD size. I'm not aware of any
376 * assumptions about maximum order if THP are disabled, but 8 seems like
377 * a good order (that's 1MB if you're using 4kB pages)
378 */
379#ifdef CONFIG_TRANSPARENT_HUGEPAGE
380#define PREFERRED_MAX_PAGECACHE_ORDER HPAGE_PMD_ORDER
381#else
382#define PREFERRED_MAX_PAGECACHE_ORDER 8
383#endif
384
385/*
386 * xas_split_alloc() does not support arbitrary orders. This implies no
387 * 512MB THP on ARM64 with 64KB base page size.
388 */
389#define MAX_XAS_ORDER (XA_CHUNK_SHIFT * 2 - 1)
390#define MAX_PAGECACHE_ORDER min(MAX_XAS_ORDER, PREFERRED_MAX_PAGECACHE_ORDER)
391
392/*
393 * mapping_max_folio_size_supported() - Check the max folio size supported
394 *
395 * The filesystem should call this function at mount time if there is a
396 * requirement on the folio mapping size in the page cache.
397 */
398static inline size_t mapping_max_folio_size_supported(void)
399{
400 if (IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE))
401 return 1U << (PAGE_SHIFT + MAX_PAGECACHE_ORDER);
402 return PAGE_SIZE;
403}
404
405/*
406 * mapping_set_folio_order_range() - Set the orders supported by a file.
407 * @mapping: The address space of the file.
408 * @min: Minimum folio order (between 0-MAX_PAGECACHE_ORDER inclusive).
409 * @max: Maximum folio order (between @min-MAX_PAGECACHE_ORDER inclusive).
410 *
411 * The filesystem should call this function in its inode constructor to
412 * indicate which base size (min) and maximum size (max) of folio the VFS
413 * can use to cache the contents of the file. This should only be used
414 * if the filesystem needs special handling of folio sizes (ie there is
415 * something the core cannot know).
416 * Do not tune it based on, eg, i_size.
417 *
418 * Context: This should not be called while the inode is active as it
419 * is non-atomic.
420 */
421static inline void mapping_set_folio_order_range(struct address_space *mapping,
422 unsigned int min,
423 unsigned int max)
424{
425 if (!IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE))
426 return;
427
428 if (min > MAX_PAGECACHE_ORDER)
429 min = MAX_PAGECACHE_ORDER;
430
431 if (max > MAX_PAGECACHE_ORDER)
432 max = MAX_PAGECACHE_ORDER;
433
434 if (max < min)
435 max = min;
436
437 mapping->flags = (mapping->flags & ~AS_FOLIO_ORDER_MASK) |
438 (min << AS_FOLIO_ORDER_MIN) | (max << AS_FOLIO_ORDER_MAX);
439}
440
441static inline void mapping_set_folio_min_order(struct address_space *mapping,
442 unsigned int min)
443{
444 mapping_set_folio_order_range(mapping, min, MAX_PAGECACHE_ORDER);
445}
446
447/**
448 * mapping_set_large_folios() - Indicate the file supports large folios.
449 * @mapping: The address space of the file.
450 *
451 * The filesystem should call this function in its inode constructor to
452 * indicate that the VFS can use large folios to cache the contents of
453 * the file.
454 *
455 * Context: This should not be called while the inode is active as it
456 * is non-atomic.
457 */
458static inline void mapping_set_large_folios(struct address_space *mapping)
459{
460 mapping_set_folio_order_range(mapping, min: 0, MAX_PAGECACHE_ORDER);
461}
462
463static inline unsigned int
464mapping_max_folio_order(const struct address_space *mapping)
465{
466 if (!IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE))
467 return 0;
468 return (mapping->flags & AS_FOLIO_ORDER_MAX_MASK) >> AS_FOLIO_ORDER_MAX;
469}
470
471static inline unsigned int
472mapping_min_folio_order(const struct address_space *mapping)
473{
474 if (!IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE))
475 return 0;
476 return (mapping->flags & AS_FOLIO_ORDER_MIN_MASK) >> AS_FOLIO_ORDER_MIN;
477}
478
479static inline unsigned long
480mapping_min_folio_nrpages(const struct address_space *mapping)
481{
482 return 1UL << mapping_min_folio_order(mapping);
483}
484
485static inline unsigned long
486mapping_min_folio_nrbytes(const struct address_space *mapping)
487{
488 return mapping_min_folio_nrpages(mapping) << PAGE_SHIFT;
489}
490
491/**
492 * mapping_align_index() - Align index for this mapping.
493 * @mapping: The address_space.
494 * @index: The page index.
495 *
496 * The index of a folio must be naturally aligned. If you are adding a
497 * new folio to the page cache and need to know what index to give it,
498 * call this function.
499 */
500static inline pgoff_t mapping_align_index(const struct address_space *mapping,
501 pgoff_t index)
502{
503 return round_down(index, mapping_min_folio_nrpages(mapping));
504}
505
506/*
507 * Large folio support currently depends on THP. These dependencies are
508 * being worked on but are not yet fixed.
509 */
510static inline bool mapping_large_folio_support(const struct address_space *mapping)
511{
512 /* AS_FOLIO_ORDER is only reasonable for pagecache folios */
513 VM_WARN_ONCE((unsigned long)mapping & FOLIO_MAPPING_ANON,
514 "Anonymous mapping always supports large folio");
515
516 return mapping_max_folio_order(mapping) > 0;
517}
518
519/* Return the maximum folio size for this pagecache mapping, in bytes. */
520static inline size_t mapping_max_folio_size(const struct address_space *mapping)
521{
522 return PAGE_SIZE << mapping_max_folio_order(mapping);
523}
524
525static inline int filemap_nr_thps(const struct address_space *mapping)
526{
527#ifdef CONFIG_READ_ONLY_THP_FOR_FS
528 return atomic_read(&mapping->nr_thps);
529#else
530 return 0;
531#endif
532}
533
534static inline void filemap_nr_thps_inc(struct address_space *mapping)
535{
536#ifdef CONFIG_READ_ONLY_THP_FOR_FS
537 if (!mapping_large_folio_support(mapping))
538 atomic_inc(&mapping->nr_thps);
539#else
540 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
541#endif
542}
543
544static inline void filemap_nr_thps_dec(struct address_space *mapping)
545{
546#ifdef CONFIG_READ_ONLY_THP_FOR_FS
547 if (!mapping_large_folio_support(mapping))
548 atomic_dec(&mapping->nr_thps);
549#else
550 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
551#endif
552}
553
554struct address_space *folio_mapping(const struct folio *folio);
555
556/**
557 * folio_flush_mapping - Find the file mapping this folio belongs to.
558 * @folio: The folio.
559 *
560 * For folios which are in the page cache, return the mapping that this
561 * page belongs to. Anonymous folios return NULL, even if they're in
562 * the swap cache. Other kinds of folio also return NULL.
563 *
564 * This is ONLY used by architecture cache flushing code. If you aren't
565 * writing cache flushing code, you want either folio_mapping() or
566 * folio_file_mapping().
567 */
568static inline struct address_space *folio_flush_mapping(struct folio *folio)
569{
570 if (unlikely(folio_test_swapcache(folio)))
571 return NULL;
572
573 return folio_mapping(folio);
574}
575
576/**
577 * folio_inode - Get the host inode for this folio.
578 * @folio: The folio.
579 *
580 * For folios which are in the page cache, return the inode that this folio
581 * belongs to.
582 *
583 * Do not call this for folios which aren't in the page cache.
584 */
585static inline struct inode *folio_inode(struct folio *folio)
586{
587 return folio->mapping->host;
588}
589
590/**
591 * folio_attach_private - Attach private data to a folio.
592 * @folio: Folio to attach data to.
593 * @data: Data to attach to folio.
594 *
595 * Attaching private data to a folio increments the page's reference count.
596 * The data must be detached before the folio will be freed.
597 */
598static inline void folio_attach_private(struct folio *folio, void *data)
599{
600 folio_get(folio);
601 folio->private = data;
602 folio_set_private(folio);
603}
604
605/**
606 * folio_change_private - Change private data on a folio.
607 * @folio: Folio to change the data on.
608 * @data: Data to set on the folio.
609 *
610 * Change the private data attached to a folio and return the old
611 * data. The page must previously have had data attached and the data
612 * must be detached before the folio will be freed.
613 *
614 * Return: Data that was previously attached to the folio.
615 */
616static inline void *folio_change_private(struct folio *folio, void *data)
617{
618 void *old = folio_get_private(folio);
619
620 folio->private = data;
621 return old;
622}
623
624/**
625 * folio_detach_private - Detach private data from a folio.
626 * @folio: Folio to detach data from.
627 *
628 * Removes the data that was previously attached to the folio and decrements
629 * the refcount on the page.
630 *
631 * Return: Data that was attached to the folio.
632 */
633static inline void *folio_detach_private(struct folio *folio)
634{
635 void *data = folio_get_private(folio);
636
637 if (!folio_test_private(folio))
638 return NULL;
639 folio_clear_private(folio);
640 folio->private = NULL;
641 folio_put(folio);
642
643 return data;
644}
645
646static inline void attach_page_private(struct page *page, void *data)
647{
648 folio_attach_private(page_folio(page), data);
649}
650
651static inline void *detach_page_private(struct page *page)
652{
653 return folio_detach_private(page_folio(page));
654}
655
656#ifdef CONFIG_NUMA
657struct folio *filemap_alloc_folio_noprof(gfp_t gfp, unsigned int order);
658#else
659static inline struct folio *filemap_alloc_folio_noprof(gfp_t gfp, unsigned int order)
660{
661 return folio_alloc_noprof(gfp, order);
662}
663#endif
664
665#define filemap_alloc_folio(...) \
666 alloc_hooks(filemap_alloc_folio_noprof(__VA_ARGS__))
667
668static inline struct page *__page_cache_alloc(gfp_t gfp)
669{
670 return &filemap_alloc_folio(gfp, 0)->page;
671}
672
673static inline gfp_t readahead_gfp_mask(struct address_space *x)
674{
675 return mapping_gfp_mask(mapping: x) | __GFP_NORETRY | __GFP_NOWARN;
676}
677
678typedef int filler_t(struct file *, struct folio *);
679
680pgoff_t page_cache_next_miss(struct address_space *mapping,
681 pgoff_t index, unsigned long max_scan);
682pgoff_t page_cache_prev_miss(struct address_space *mapping,
683 pgoff_t index, unsigned long max_scan);
684
685/**
686 * typedef fgf_t - Flags for getting folios from the page cache.
687 *
688 * Most users of the page cache will not need to use these flags;
689 * there are convenience functions such as filemap_get_folio() and
690 * filemap_lock_folio(). For users which need more control over exactly
691 * what is done with the folios, these flags to __filemap_get_folio()
692 * are available.
693 *
694 * * %FGP_ACCESSED - The folio will be marked accessed.
695 * * %FGP_LOCK - The folio is returned locked.
696 * * %FGP_CREAT - If no folio is present then a new folio is allocated,
697 * added to the page cache and the VM's LRU list. The folio is
698 * returned locked.
699 * * %FGP_FOR_MMAP - The caller wants to do its own locking dance if the
700 * folio is already in cache. If the folio was allocated, unlock it
701 * before returning so the caller can do the same dance.
702 * * %FGP_WRITE - The folio will be written to by the caller.
703 * * %FGP_NOFS - __GFP_FS will get cleared in gfp.
704 * * %FGP_NOWAIT - Don't block on the folio lock.
705 * * %FGP_STABLE - Wait for the folio to be stable (finished writeback)
706 * * %FGP_DONTCACHE - Uncached buffered IO
707 * * %FGP_WRITEBEGIN - The flags to use in a filesystem write_begin()
708 * implementation.
709 */
710typedef unsigned int __bitwise fgf_t;
711
712#define FGP_ACCESSED ((__force fgf_t)0x00000001)
713#define FGP_LOCK ((__force fgf_t)0x00000002)
714#define FGP_CREAT ((__force fgf_t)0x00000004)
715#define FGP_WRITE ((__force fgf_t)0x00000008)
716#define FGP_NOFS ((__force fgf_t)0x00000010)
717#define FGP_NOWAIT ((__force fgf_t)0x00000020)
718#define FGP_FOR_MMAP ((__force fgf_t)0x00000040)
719#define FGP_STABLE ((__force fgf_t)0x00000080)
720#define FGP_DONTCACHE ((__force fgf_t)0x00000100)
721#define FGF_GET_ORDER(fgf) (((__force unsigned)fgf) >> 26) /* top 6 bits */
722
723#define FGP_WRITEBEGIN (FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE)
724
725static inline unsigned int filemap_get_order(size_t size)
726{
727 unsigned int shift = ilog2(size);
728
729 if (shift <= PAGE_SHIFT)
730 return 0;
731
732 return shift - PAGE_SHIFT;
733}
734
735/**
736 * fgf_set_order - Encode a length in the fgf_t flags.
737 * @size: The suggested size of the folio to create.
738 *
739 * The caller of __filemap_get_folio() can use this to suggest a preferred
740 * size for the folio that is created. If there is already a folio at
741 * the index, it will be returned, no matter what its size. If a folio
742 * is freshly created, it may be of a different size than requested
743 * due to alignment constraints, memory pressure, or the presence of
744 * other folios at nearby indices.
745 */
746static inline fgf_t fgf_set_order(size_t size)
747{
748 unsigned int order = filemap_get_order(size);
749
750 if (!order)
751 return 0;
752 return (__force fgf_t)(order << 26);
753}
754
755void *filemap_get_entry(struct address_space *mapping, pgoff_t index);
756struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
757 fgf_t fgp_flags, gfp_t gfp);
758struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
759 fgf_t fgp_flags, gfp_t gfp);
760
761/**
762 * write_begin_get_folio - Get folio for write_begin with flags.
763 * @iocb: The kiocb passed from write_begin (may be NULL).
764 * @mapping: The address space to search.
765 * @index: The page cache index.
766 * @len: Length of data being written.
767 *
768 * This is a helper for filesystem write_begin() implementations.
769 * It wraps __filemap_get_folio(), setting appropriate flags in
770 * the write begin context.
771 *
772 * Return: A folio or an ERR_PTR.
773 */
774static inline struct folio *write_begin_get_folio(const struct kiocb *iocb,
775 struct address_space *mapping, pgoff_t index, size_t len)
776{
777 fgf_t fgp_flags = FGP_WRITEBEGIN;
778
779 fgp_flags |= fgf_set_order(size: len);
780
781 if (iocb && iocb->ki_flags & IOCB_DONTCACHE)
782 fgp_flags |= FGP_DONTCACHE;
783
784 return __filemap_get_folio(mapping, index, fgp_flags,
785 gfp: mapping_gfp_mask(mapping));
786}
787
788/**
789 * filemap_get_folio - Find and get a folio.
790 * @mapping: The address_space to search.
791 * @index: The page index.
792 *
793 * Looks up the page cache entry at @mapping & @index. If a folio is
794 * present, it is returned with an increased refcount.
795 *
796 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
797 * this index. Will not return a shadow, swap or DAX entry.
798 */
799static inline struct folio *filemap_get_folio(struct address_space *mapping,
800 pgoff_t index)
801{
802 return __filemap_get_folio(mapping, index, fgp_flags: 0, gfp: 0);
803}
804
805/**
806 * filemap_lock_folio - Find and lock a folio.
807 * @mapping: The address_space to search.
808 * @index: The page index.
809 *
810 * Looks up the page cache entry at @mapping & @index. If a folio is
811 * present, it is returned locked with an increased refcount.
812 *
813 * Context: May sleep.
814 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
815 * this index. Will not return a shadow, swap or DAX entry.
816 */
817static inline struct folio *filemap_lock_folio(struct address_space *mapping,
818 pgoff_t index)
819{
820 return __filemap_get_folio(mapping, index, FGP_LOCK, gfp: 0);
821}
822
823/**
824 * filemap_grab_folio - grab a folio from the page cache
825 * @mapping: The address space to search
826 * @index: The page index
827 *
828 * Looks up the page cache entry at @mapping & @index. If no folio is found,
829 * a new folio is created. The folio is locked, marked as accessed, and
830 * returned.
831 *
832 * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found
833 * and failed to create a folio.
834 */
835static inline struct folio *filemap_grab_folio(struct address_space *mapping,
836 pgoff_t index)
837{
838 return __filemap_get_folio(mapping, index,
839 FGP_LOCK | FGP_ACCESSED | FGP_CREAT,
840 gfp: mapping_gfp_mask(mapping));
841}
842
843/**
844 * find_get_page - find and get a page reference
845 * @mapping: the address_space to search
846 * @offset: the page index
847 *
848 * Looks up the page cache slot at @mapping & @offset. If there is a
849 * page cache page, it is returned with an increased refcount.
850 *
851 * Otherwise, %NULL is returned.
852 */
853static inline struct page *find_get_page(struct address_space *mapping,
854 pgoff_t offset)
855{
856 return pagecache_get_page(mapping, index: offset, fgp_flags: 0, gfp: 0);
857}
858
859static inline struct page *find_get_page_flags(struct address_space *mapping,
860 pgoff_t offset, fgf_t fgp_flags)
861{
862 return pagecache_get_page(mapping, index: offset, fgp_flags, gfp: 0);
863}
864
865/**
866 * find_lock_page - locate, pin and lock a pagecache page
867 * @mapping: the address_space to search
868 * @index: the page index
869 *
870 * Looks up the page cache entry at @mapping & @index. If there is a
871 * page cache page, it is returned locked and with an increased
872 * refcount.
873 *
874 * Context: May sleep.
875 * Return: A struct page or %NULL if there is no page in the cache for this
876 * index.
877 */
878static inline struct page *find_lock_page(struct address_space *mapping,
879 pgoff_t index)
880{
881 return pagecache_get_page(mapping, index, FGP_LOCK, gfp: 0);
882}
883
884/**
885 * find_or_create_page - locate or add a pagecache page
886 * @mapping: the page's address_space
887 * @index: the page's index into the mapping
888 * @gfp_mask: page allocation mode
889 *
890 * Looks up the page cache slot at @mapping & @offset. If there is a
891 * page cache page, it is returned locked and with an increased
892 * refcount.
893 *
894 * If the page is not present, a new page is allocated using @gfp_mask
895 * and added to the page cache and the VM's LRU list. The page is
896 * returned locked and with an increased refcount.
897 *
898 * On memory exhaustion, %NULL is returned.
899 *
900 * find_or_create_page() may sleep, even if @gfp_flags specifies an
901 * atomic allocation!
902 */
903static inline struct page *find_or_create_page(struct address_space *mapping,
904 pgoff_t index, gfp_t gfp_mask)
905{
906 return pagecache_get_page(mapping, index,
907 FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
908 gfp: gfp_mask);
909}
910
911/**
912 * grab_cache_page_nowait - returns locked page at given index in given cache
913 * @mapping: target address_space
914 * @index: the page index
915 *
916 * Returns locked page at given index in given cache, creating it if
917 * needed, but do not wait if the page is locked or to reclaim memory.
918 * This is intended for speculative data generators, where the data can
919 * be regenerated if the page couldn't be grabbed. This routine should
920 * be safe to call while holding the lock for another page.
921 *
922 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
923 * and deadlock against the caller's locked page.
924 */
925static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
926 pgoff_t index)
927{
928 return pagecache_get_page(mapping, index,
929 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
930 gfp: mapping_gfp_mask(mapping));
931}
932
933/**
934 * folio_next_index - Get the index of the next folio.
935 * @folio: The current folio.
936 *
937 * Return: The index of the folio which follows this folio in the file.
938 */
939static inline pgoff_t folio_next_index(const struct folio *folio)
940{
941 return folio->index + folio_nr_pages(folio);
942}
943
944/**
945 * folio_file_page - The page for a particular index.
946 * @folio: The folio which contains this index.
947 * @index: The index we want to look up.
948 *
949 * Sometimes after looking up a folio in the page cache, we need to
950 * obtain the specific page for an index (eg a page fault).
951 *
952 * Return: The page containing the file data for this index.
953 */
954static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
955{
956 return folio_page(folio, index & (folio_nr_pages(folio) - 1));
957}
958
959/**
960 * folio_contains - Does this folio contain this index?
961 * @folio: The folio.
962 * @index: The page index within the file.
963 *
964 * Context: The caller should have the folio locked and ensure
965 * e.g., shmem did not move this folio to the swap cache.
966 * Return: true or false.
967 */
968static inline bool folio_contains(const struct folio *folio, pgoff_t index)
969{
970 VM_WARN_ON_ONCE_FOLIO(folio_test_swapcache(folio), folio);
971 return index - folio->index < folio_nr_pages(folio);
972}
973
974unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start,
975 pgoff_t end, struct folio_batch *fbatch);
976unsigned filemap_get_folios_contig(struct address_space *mapping,
977 pgoff_t *start, pgoff_t end, struct folio_batch *fbatch);
978unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start,
979 pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch);
980
981struct folio *read_cache_folio(struct address_space *, pgoff_t index,
982 filler_t *filler, struct file *file);
983struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index,
984 gfp_t flags);
985struct page *read_cache_page(struct address_space *, pgoff_t index,
986 filler_t *filler, struct file *file);
987extern struct page * read_cache_page_gfp(struct address_space *mapping,
988 pgoff_t index, gfp_t gfp_mask);
989
990static inline struct page *read_mapping_page(struct address_space *mapping,
991 pgoff_t index, struct file *file)
992{
993 return read_cache_page(mapping, index, NULL, file);
994}
995
996static inline struct folio *read_mapping_folio(struct address_space *mapping,
997 pgoff_t index, struct file *file)
998{
999 return read_cache_folio(mapping, index, NULL, file);
1000}
1001
1002/**
1003 * page_pgoff - Calculate the logical page offset of this page.
1004 * @folio: The folio containing this page.
1005 * @page: The page which we need the offset of.
1006 *
1007 * For file pages, this is the offset from the beginning of the file
1008 * in units of PAGE_SIZE. For anonymous pages, this is the offset from
1009 * the beginning of the anon_vma in units of PAGE_SIZE. This will
1010 * return nonsense for KSM pages.
1011 *
1012 * Context: Caller must have a reference on the folio or otherwise
1013 * prevent it from being split or freed.
1014 *
1015 * Return: The offset in units of PAGE_SIZE.
1016 */
1017static inline pgoff_t page_pgoff(const struct folio *folio,
1018 const struct page *page)
1019{
1020 return folio->index + folio_page_idx(folio, page);
1021}
1022
1023/**
1024 * folio_pos - Returns the byte position of this folio in its file.
1025 * @folio: The folio.
1026 */
1027static inline loff_t folio_pos(const struct folio *folio)
1028{
1029 return ((loff_t)folio->index) * PAGE_SIZE;
1030}
1031
1032/*
1033 * Return byte-offset into filesystem object for page.
1034 */
1035static inline loff_t page_offset(struct page *page)
1036{
1037 struct folio *folio = page_folio(page);
1038
1039 return folio_pos(folio) + folio_page_idx(folio, page) * PAGE_SIZE;
1040}
1041
1042/*
1043 * Get the offset in PAGE_SIZE (even for hugetlb folios).
1044 */
1045static inline pgoff_t folio_pgoff(const struct folio *folio)
1046{
1047 return folio->index;
1048}
1049
1050static inline pgoff_t linear_page_index(const struct vm_area_struct *vma,
1051 const unsigned long address)
1052{
1053 pgoff_t pgoff;
1054 pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
1055 pgoff += vma->vm_pgoff;
1056 return pgoff;
1057}
1058
1059struct wait_page_key {
1060 struct folio *folio;
1061 int bit_nr;
1062 int page_match;
1063};
1064
1065struct wait_page_queue {
1066 struct folio *folio;
1067 int bit_nr;
1068 wait_queue_entry_t wait;
1069};
1070
1071static inline bool wake_page_match(struct wait_page_queue *wait_page,
1072 struct wait_page_key *key)
1073{
1074 if (wait_page->folio != key->folio)
1075 return false;
1076 key->page_match = 1;
1077
1078 if (wait_page->bit_nr != key->bit_nr)
1079 return false;
1080
1081 return true;
1082}
1083
1084void __folio_lock(struct folio *folio);
1085int __folio_lock_killable(struct folio *folio);
1086vm_fault_t __folio_lock_or_retry(struct folio *folio, struct vm_fault *vmf);
1087void unlock_page(struct page *page);
1088void folio_unlock(struct folio *folio);
1089
1090/**
1091 * folio_trylock() - Attempt to lock a folio.
1092 * @folio: The folio to attempt to lock.
1093 *
1094 * Sometimes it is undesirable to wait for a folio to be unlocked (eg
1095 * when the locks are being taken in the wrong order, or if making
1096 * progress through a batch of folios is more important than processing
1097 * them in order). Usually folio_lock() is the correct function to call.
1098 *
1099 * Context: Any context.
1100 * Return: Whether the lock was successfully acquired.
1101 */
1102static inline bool folio_trylock(struct folio *folio)
1103{
1104 return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
1105}
1106
1107/*
1108 * Return true if the page was successfully locked
1109 */
1110static inline bool trylock_page(struct page *page)
1111{
1112 return folio_trylock(page_folio(page));
1113}
1114
1115/**
1116 * folio_lock() - Lock this folio.
1117 * @folio: The folio to lock.
1118 *
1119 * The folio lock protects against many things, probably more than it
1120 * should. It is primarily held while a folio is being brought uptodate,
1121 * either from its backing file or from swap. It is also held while a
1122 * folio is being truncated from its address_space, so holding the lock
1123 * is sufficient to keep folio->mapping stable.
1124 *
1125 * The folio lock is also held while write() is modifying the page to
1126 * provide POSIX atomicity guarantees (as long as the write does not
1127 * cross a page boundary). Other modifications to the data in the folio
1128 * do not hold the folio lock and can race with writes, eg DMA and stores
1129 * to mapped pages.
1130 *
1131 * Context: May sleep. If you need to acquire the locks of two or
1132 * more folios, they must be in order of ascending index, if they are
1133 * in the same address_space. If they are in different address_spaces,
1134 * acquire the lock of the folio which belongs to the address_space which
1135 * has the lowest address in memory first.
1136 */
1137static inline void folio_lock(struct folio *folio)
1138{
1139 might_sleep();
1140 if (!folio_trylock(folio))
1141 __folio_lock(folio);
1142}
1143
1144/**
1145 * lock_page() - Lock the folio containing this page.
1146 * @page: The page to lock.
1147 *
1148 * See folio_lock() for a description of what the lock protects.
1149 * This is a legacy function and new code should probably use folio_lock()
1150 * instead.
1151 *
1152 * Context: May sleep. Pages in the same folio share a lock, so do not
1153 * attempt to lock two pages which share a folio.
1154 */
1155static inline void lock_page(struct page *page)
1156{
1157 struct folio *folio;
1158 might_sleep();
1159
1160 folio = page_folio(page);
1161 if (!folio_trylock(folio))
1162 __folio_lock(folio);
1163}
1164
1165/**
1166 * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
1167 * @folio: The folio to lock.
1168 *
1169 * Attempts to lock the folio, like folio_lock(), except that the sleep
1170 * to acquire the lock is interruptible by a fatal signal.
1171 *
1172 * Context: May sleep; see folio_lock().
1173 * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
1174 */
1175static inline int folio_lock_killable(struct folio *folio)
1176{
1177 might_sleep();
1178 if (!folio_trylock(folio))
1179 return __folio_lock_killable(folio);
1180 return 0;
1181}
1182
1183/*
1184 * folio_lock_or_retry - Lock the folio, unless this would block and the
1185 * caller indicated that it can handle a retry.
1186 *
1187 * Return value and mmap_lock implications depend on flags; see
1188 * __folio_lock_or_retry().
1189 */
1190static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
1191 struct vm_fault *vmf)
1192{
1193 might_sleep();
1194 if (!folio_trylock(folio))
1195 return __folio_lock_or_retry(folio, vmf);
1196 return 0;
1197}
1198
1199/*
1200 * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
1201 * and should not be used directly.
1202 */
1203void folio_wait_bit(struct folio *folio, int bit_nr);
1204int folio_wait_bit_killable(struct folio *folio, int bit_nr);
1205
1206/*
1207 * Wait for a folio to be unlocked.
1208 *
1209 * This must be called with the caller "holding" the folio,
1210 * ie with increased folio reference count so that the folio won't
1211 * go away during the wait.
1212 */
1213static inline void folio_wait_locked(struct folio *folio)
1214{
1215 if (folio_test_locked(folio))
1216 folio_wait_bit(folio, bit_nr: PG_locked);
1217}
1218
1219static inline int folio_wait_locked_killable(struct folio *folio)
1220{
1221 if (!folio_test_locked(folio))
1222 return 0;
1223 return folio_wait_bit_killable(folio, bit_nr: PG_locked);
1224}
1225
1226void folio_end_read(struct folio *folio, bool success);
1227void wait_on_page_writeback(struct page *page);
1228void folio_wait_writeback(struct folio *folio);
1229int folio_wait_writeback_killable(struct folio *folio);
1230void end_page_writeback(struct page *page);
1231void folio_end_writeback(struct folio *folio);
1232void folio_end_writeback_no_dropbehind(struct folio *folio);
1233void folio_end_dropbehind(struct folio *folio);
1234void folio_wait_stable(struct folio *folio);
1235void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
1236void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
1237void __folio_cancel_dirty(struct folio *folio);
1238static inline void folio_cancel_dirty(struct folio *folio)
1239{
1240 /* Avoid atomic ops, locking, etc. when not actually needed. */
1241 if (folio_test_dirty(folio))
1242 __folio_cancel_dirty(folio);
1243}
1244bool folio_clear_dirty_for_io(struct folio *folio);
1245bool clear_page_dirty_for_io(struct page *page);
1246void folio_invalidate(struct folio *folio, size_t offset, size_t length);
1247bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
1248
1249#ifdef CONFIG_MIGRATION
1250int filemap_migrate_folio(struct address_space *mapping, struct folio *dst,
1251 struct folio *src, enum migrate_mode mode);
1252#else
1253#define filemap_migrate_folio NULL
1254#endif
1255void folio_end_private_2(struct folio *folio);
1256void folio_wait_private_2(struct folio *folio);
1257int folio_wait_private_2_killable(struct folio *folio);
1258
1259/*
1260 * Fault in userspace address range.
1261 */
1262size_t fault_in_writeable(char __user *uaddr, size_t size);
1263size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
1264size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
1265size_t fault_in_readable(const char __user *uaddr, size_t size);
1266
1267int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
1268 pgoff_t index, gfp_t gfp);
1269int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1270 pgoff_t index, gfp_t gfp);
1271void filemap_remove_folio(struct folio *folio);
1272void __filemap_remove_folio(struct folio *folio, void *shadow);
1273void replace_page_cache_folio(struct folio *old, struct folio *new);
1274void delete_from_page_cache_batch(struct address_space *mapping,
1275 struct folio_batch *fbatch);
1276bool filemap_release_folio(struct folio *folio, gfp_t gfp);
1277loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1278 int whence);
1279
1280/* Must be non-static for BPF error injection */
1281int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1282 pgoff_t index, gfp_t gfp, void **shadowp);
1283
1284bool filemap_range_has_writeback(struct address_space *mapping,
1285 loff_t start_byte, loff_t end_byte);
1286
1287/**
1288 * filemap_range_needs_writeback - check if range potentially needs writeback
1289 * @mapping: address space within which to check
1290 * @start_byte: offset in bytes where the range starts
1291 * @end_byte: offset in bytes where the range ends (inclusive)
1292 *
1293 * Find at least one page in the range supplied, usually used to check if
1294 * direct writing in this range will trigger a writeback. Used by O_DIRECT
1295 * read/write with IOCB_NOWAIT, to see if the caller needs to do
1296 * filemap_write_and_wait_range() before proceeding.
1297 *
1298 * Return: %true if the caller should do filemap_write_and_wait_range() before
1299 * doing O_DIRECT to a page in this range, %false otherwise.
1300 */
1301static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1302 loff_t start_byte,
1303 loff_t end_byte)
1304{
1305 if (!mapping->nrpages)
1306 return false;
1307 if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1308 !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1309 return false;
1310 return filemap_range_has_writeback(mapping, start_byte, end_byte);
1311}
1312
1313/**
1314 * struct readahead_control - Describes a readahead request.
1315 *
1316 * A readahead request is for consecutive pages. Filesystems which
1317 * implement the ->readahead method should call readahead_folio() or
1318 * __readahead_batch() in a loop and attempt to start reads into each
1319 * folio in the request.
1320 *
1321 * Most of the fields in this struct are private and should be accessed
1322 * by the functions below.
1323 *
1324 * @file: The file, used primarily by network filesystems for authentication.
1325 * May be NULL if invoked internally by the filesystem.
1326 * @mapping: Readahead this filesystem object.
1327 * @ra: File readahead state. May be NULL.
1328 */
1329struct readahead_control {
1330 struct file *file;
1331 struct address_space *mapping;
1332 struct file_ra_state *ra;
1333/* private: use the readahead_* accessors instead */
1334 pgoff_t _index;
1335 unsigned int _nr_pages;
1336 unsigned int _batch_count;
1337 bool dropbehind;
1338 bool _workingset;
1339 unsigned long _pflags;
1340};
1341
1342#define DEFINE_READAHEAD(ractl, f, r, m, i) \
1343 struct readahead_control ractl = { \
1344 .file = f, \
1345 .mapping = m, \
1346 .ra = r, \
1347 ._index = i, \
1348 }
1349
1350#define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
1351
1352void page_cache_ra_unbounded(struct readahead_control *,
1353 unsigned long nr_to_read, unsigned long lookahead_count);
1354void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
1355void page_cache_async_ra(struct readahead_control *, struct folio *,
1356 unsigned long req_count);
1357void readahead_expand(struct readahead_control *ractl,
1358 loff_t new_start, size_t new_len);
1359
1360/**
1361 * page_cache_sync_readahead - generic file readahead
1362 * @mapping: address_space which holds the pagecache and I/O vectors
1363 * @ra: file_ra_state which holds the readahead state
1364 * @file: Used by the filesystem for authentication.
1365 * @index: Index of first page to be read.
1366 * @req_count: Total number of pages being read by the caller.
1367 *
1368 * page_cache_sync_readahead() should be called when a cache miss happened:
1369 * it will submit the read. The readahead logic may decide to piggyback more
1370 * pages onto the read request if access patterns suggest it will improve
1371 * performance.
1372 */
1373static inline
1374void page_cache_sync_readahead(struct address_space *mapping,
1375 struct file_ra_state *ra, struct file *file, pgoff_t index,
1376 unsigned long req_count)
1377{
1378 DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1379 page_cache_sync_ra(&ractl, req_count);
1380}
1381
1382/**
1383 * page_cache_async_readahead - file readahead for marked pages
1384 * @mapping: address_space which holds the pagecache and I/O vectors
1385 * @ra: file_ra_state which holds the readahead state
1386 * @file: Used by the filesystem for authentication.
1387 * @folio: The folio which triggered the readahead call.
1388 * @req_count: Total number of pages being read by the caller.
1389 *
1390 * page_cache_async_readahead() should be called when a page is used which
1391 * is marked as PageReadahead; this is a marker to suggest that the application
1392 * has used up enough of the readahead window that we should start pulling in
1393 * more pages.
1394 */
1395static inline
1396void page_cache_async_readahead(struct address_space *mapping,
1397 struct file_ra_state *ra, struct file *file,
1398 struct folio *folio, unsigned long req_count)
1399{
1400 DEFINE_READAHEAD(ractl, file, ra, mapping, folio->index);
1401 page_cache_async_ra(&ractl, folio, req_count);
1402}
1403
1404static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1405{
1406 struct folio *folio;
1407
1408 BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1409 ractl->_nr_pages -= ractl->_batch_count;
1410 ractl->_index += ractl->_batch_count;
1411
1412 if (!ractl->_nr_pages) {
1413 ractl->_batch_count = 0;
1414 return NULL;
1415 }
1416
1417 folio = xa_load(&ractl->mapping->i_pages, index: ractl->_index);
1418 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1419 ractl->_batch_count = folio_nr_pages(folio);
1420
1421 return folio;
1422}
1423
1424/**
1425 * readahead_folio - Get the next folio to read.
1426 * @ractl: The current readahead request.
1427 *
1428 * Context: The folio is locked. The caller should unlock the folio once
1429 * all I/O to that folio has completed.
1430 * Return: A pointer to the next folio, or %NULL if we are done.
1431 */
1432static inline struct folio *readahead_folio(struct readahead_control *ractl)
1433{
1434 struct folio *folio = __readahead_folio(ractl);
1435
1436 if (folio)
1437 folio_put(folio);
1438 return folio;
1439}
1440
1441static inline unsigned int __readahead_batch(struct readahead_control *rac,
1442 struct page **array, unsigned int array_sz)
1443{
1444 unsigned int i = 0;
1445 XA_STATE(xas, &rac->mapping->i_pages, 0);
1446 struct folio *folio;
1447
1448 BUG_ON(rac->_batch_count > rac->_nr_pages);
1449 rac->_nr_pages -= rac->_batch_count;
1450 rac->_index += rac->_batch_count;
1451 rac->_batch_count = 0;
1452
1453 xas_set(xas: &xas, index: rac->_index);
1454 rcu_read_lock();
1455 xas_for_each(&xas, folio, rac->_index + rac->_nr_pages - 1) {
1456 if (xas_retry(xas: &xas, entry: folio))
1457 continue;
1458 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1459 array[i++] = folio_page(folio, 0);
1460 rac->_batch_count += folio_nr_pages(folio);
1461 if (i == array_sz)
1462 break;
1463 }
1464 rcu_read_unlock();
1465
1466 return i;
1467}
1468
1469/**
1470 * readahead_pos - The byte offset into the file of this readahead request.
1471 * @rac: The readahead request.
1472 */
1473static inline loff_t readahead_pos(const struct readahead_control *rac)
1474{
1475 return (loff_t)rac->_index * PAGE_SIZE;
1476}
1477
1478/**
1479 * readahead_length - The number of bytes in this readahead request.
1480 * @rac: The readahead request.
1481 */
1482static inline size_t readahead_length(const struct readahead_control *rac)
1483{
1484 return rac->_nr_pages * PAGE_SIZE;
1485}
1486
1487/**
1488 * readahead_index - The index of the first page in this readahead request.
1489 * @rac: The readahead request.
1490 */
1491static inline pgoff_t readahead_index(const struct readahead_control *rac)
1492{
1493 return rac->_index;
1494}
1495
1496/**
1497 * readahead_count - The number of pages in this readahead request.
1498 * @rac: The readahead request.
1499 */
1500static inline unsigned int readahead_count(const struct readahead_control *rac)
1501{
1502 return rac->_nr_pages;
1503}
1504
1505/**
1506 * readahead_batch_length - The number of bytes in the current batch.
1507 * @rac: The readahead request.
1508 */
1509static inline size_t readahead_batch_length(const struct readahead_control *rac)
1510{
1511 return rac->_batch_count * PAGE_SIZE;
1512}
1513
1514static inline unsigned long dir_pages(const struct inode *inode)
1515{
1516 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1517 PAGE_SHIFT;
1518}
1519
1520/**
1521 * folio_mkwrite_check_truncate - check if folio was truncated
1522 * @folio: the folio to check
1523 * @inode: the inode to check the folio against
1524 *
1525 * Return: the number of bytes in the folio up to EOF,
1526 * or -EFAULT if the folio was truncated.
1527 */
1528static inline ssize_t folio_mkwrite_check_truncate(const struct folio *folio,
1529 const struct inode *inode)
1530{
1531 loff_t size = i_size_read(inode);
1532 pgoff_t index = size >> PAGE_SHIFT;
1533 size_t offset = offset_in_folio(folio, size);
1534
1535 if (!folio->mapping)
1536 return -EFAULT;
1537
1538 /* folio is wholly inside EOF */
1539 if (folio_next_index(folio) - 1 < index)
1540 return folio_size(folio);
1541 /* folio is wholly past EOF */
1542 if (folio->index > index || !offset)
1543 return -EFAULT;
1544 /* folio is partially inside EOF */
1545 return offset;
1546}
1547
1548/**
1549 * i_blocks_per_folio - How many blocks fit in this folio.
1550 * @inode: The inode which contains the blocks.
1551 * @folio: The folio.
1552 *
1553 * If the block size is larger than the size of this folio, return zero.
1554 *
1555 * Context: The caller should hold a refcount on the folio to prevent it
1556 * from being split.
1557 * Return: The number of filesystem blocks covered by this folio.
1558 */
1559static inline
1560unsigned int i_blocks_per_folio(const struct inode *inode,
1561 const struct folio *folio)
1562{
1563 return folio_size(folio) >> inode->i_blkbits;
1564}
1565#endif /* _LINUX_PAGEMAP_H */
1566