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