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 kiocb_invalidate_pages(struct kiocb *iocb, size_t count);
34 : void kiocb_invalidate_post_direct_write(struct kiocb *iocb, size_t count);
35 :
36 : int write_inode_now(struct inode *, int sync);
37 : int filemap_fdatawrite(struct address_space *);
38 : int filemap_flush(struct address_space *);
39 : int filemap_fdatawait_keep_errors(struct address_space *mapping);
40 : int filemap_fdatawait_range(struct address_space *, loff_t lstart, loff_t lend);
41 : int filemap_fdatawait_range_keep_errors(struct address_space *mapping,
42 : loff_t start_byte, loff_t end_byte);
43 :
44 : static inline int filemap_fdatawait(struct address_space *mapping)
45 : {
46 100081 : return filemap_fdatawait_range(mapping, 0, LLONG_MAX);
47 : }
48 :
49 : bool filemap_range_has_page(struct address_space *, loff_t lstart, loff_t lend);
50 : int filemap_write_and_wait_range(struct address_space *mapping,
51 : loff_t lstart, loff_t lend);
52 : int __filemap_fdatawrite_range(struct address_space *mapping,
53 : loff_t start, loff_t end, int sync_mode);
54 : int filemap_fdatawrite_range(struct address_space *mapping,
55 : loff_t start, loff_t end);
56 : int filemap_check_errors(struct address_space *mapping);
57 : void __filemap_set_wb_err(struct address_space *mapping, int err);
58 : int filemap_fdatawrite_wbc(struct address_space *mapping,
59 : struct writeback_control *wbc);
60 : int kiocb_write_and_wait(struct kiocb *iocb, size_t count);
61 :
62 : static inline int filemap_write_and_wait(struct address_space *mapping)
63 : {
64 607433 : return filemap_write_and_wait_range(mapping, 0, LLONG_MAX);
65 : }
66 :
67 : /**
68 : * filemap_set_wb_err - set a writeback error on an address_space
69 : * @mapping: mapping in which to set writeback error
70 : * @err: error to be set in mapping
71 : *
72 : * When writeback fails in some way, we must record that error so that
73 : * userspace can be informed when fsync and the like are called. We endeavor
74 : * to report errors on any file that was open at the time of the error. Some
75 : * internal callers also need to know when writeback errors have occurred.
76 : *
77 : * When a writeback error occurs, most filesystems will want to call
78 : * filemap_set_wb_err to record the error in the mapping so that it will be
79 : * automatically reported whenever fsync is called on the file.
80 : */
81 : static inline void filemap_set_wb_err(struct address_space *mapping, int err)
82 : {
83 : /* Fastpath for common case of no error */
84 : if (unlikely(err))
85 : __filemap_set_wb_err(mapping, err);
86 : }
87 :
88 : /**
89 : * filemap_check_wb_err - has an error occurred since the mark was sampled?
90 : * @mapping: mapping to check for writeback errors
91 : * @since: previously-sampled errseq_t
92 : *
93 : * Grab the errseq_t value from the mapping, and see if it has changed "since"
94 : * the given value was sampled.
95 : *
96 : * If it has then report the latest error set, otherwise return 0.
97 : */
98 : static inline int filemap_check_wb_err(struct address_space *mapping,
99 : errseq_t since)
100 : {
101 19828 : return errseq_check(&mapping->wb_err, since);
102 : }
103 :
104 : /**
105 : * filemap_sample_wb_err - sample the current errseq_t to test for later errors
106 : * @mapping: mapping to be sampled
107 : *
108 : * Writeback errors are always reported relative to a particular sample point
109 : * in the past. This function provides those sample points.
110 : */
111 : static inline errseq_t filemap_sample_wb_err(struct address_space *mapping)
112 : {
113 1597526589 : return errseq_sample(&mapping->wb_err);
114 : }
115 :
116 : /**
117 : * file_sample_sb_err - sample the current errseq_t to test for later errors
118 : * @file: file pointer to be sampled
119 : *
120 : * Grab the most current superblock-level errseq_t value for the given
121 : * struct file.
122 : */
123 : static inline errseq_t file_sample_sb_err(struct file *file)
124 : {
125 1597965497 : return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err);
126 : }
127 :
128 : /*
129 : * Flush file data before changing attributes. Caller must hold any locks
130 : * required to prevent further writes to this file until we're done setting
131 : * flags.
132 : */
133 : static inline int inode_drain_writes(struct inode *inode)
134 : {
135 : inode_dio_wait(inode);
136 : return filemap_write_and_wait(inode->i_mapping);
137 : }
138 :
139 : static inline bool mapping_empty(struct address_space *mapping)
140 : {
141 2649950158 : return xa_empty(&mapping->i_pages);
142 : }
143 :
144 : /*
145 : * mapping_shrinkable - test if page cache state allows inode reclaim
146 : * @mapping: the page cache mapping
147 : *
148 : * This checks the mapping's cache state for the pupose of inode
149 : * reclaim and LRU management.
150 : *
151 : * The caller is expected to hold the i_lock, but is not required to
152 : * hold the i_pages lock, which usually protects cache state. That's
153 : * because the i_lock and the list_lru lock that protect the inode and
154 : * its LRU state don't nest inside the irq-safe i_pages lock.
155 : *
156 : * Cache deletions are performed under the i_lock, which ensures that
157 : * when an inode goes empty, it will reliably get queued on the LRU.
158 : *
159 : * Cache additions do not acquire the i_lock and may race with this
160 : * check, in which case we'll report the inode as shrinkable when it
161 : * has cache pages. This is okay: the shrinker also checks the
162 : * refcount and the referenced bit, which will be elevated or set in
163 : * the process of adding new cache pages to an inode.
164 : */
165 67035416109 : static inline bool mapping_shrinkable(struct address_space *mapping)
166 : {
167 67035416109 : void *head;
168 :
169 : /*
170 : * On highmem systems, there could be lowmem pressure from the
171 : * inodes before there is highmem pressure from the page
172 : * cache. Make inodes shrinkable regardless of cache state.
173 : */
174 67035416109 : if (IS_ENABLED(CONFIG_HIGHMEM))
175 : return true;
176 :
177 : /* Cache completely empty? Shrink away. */
178 67035416109 : head = rcu_access_pointer(mapping->i_pages.xa_head);
179 67035416109 : if (!head)
180 : return true;
181 :
182 : /*
183 : * The xarray stores single offset-0 entries directly in the
184 : * head pointer, which allows non-resident page cache entries
185 : * to escape the shadow shrinker's list of xarray nodes. The
186 : * inode shrinker needs to pick them up under memory pressure.
187 : */
188 8741163788 : if (!xa_is_node(head) && xa_is_value(head))
189 566541 : return true;
190 :
191 : return false;
192 : }
193 :
194 : /*
195 : * Bits in mapping->flags.
196 : */
197 : enum mapping_flags {
198 : AS_EIO = 0, /* IO error on async write */
199 : AS_ENOSPC = 1, /* ENOSPC on async write */
200 : AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */
201 : AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */
202 : AS_EXITING = 4, /* final truncate in progress */
203 : /* writeback related tags are not used */
204 : AS_NO_WRITEBACK_TAGS = 5,
205 : AS_LARGE_FOLIO_SUPPORT = 6,
206 : };
207 :
208 : /**
209 : * mapping_set_error - record a writeback error in the address_space
210 : * @mapping: the mapping in which an error should be set
211 : * @error: the error to set in the mapping
212 : *
213 : * When writeback fails in some way, we must record that error so that
214 : * userspace can be informed when fsync and the like are called. We endeavor
215 : * to report errors on any file that was open at the time of the error. Some
216 : * internal callers also need to know when writeback errors have occurred.
217 : *
218 : * When a writeback error occurs, most filesystems will want to call
219 : * mapping_set_error to record the error in the mapping so that it can be
220 : * reported when the application calls fsync(2).
221 : */
222 695946514 : static inline void mapping_set_error(struct address_space *mapping, int error)
223 : {
224 695946514 : if (likely(!error))
225 : return;
226 :
227 : /* Record in wb_err for checkers using errseq_t based tracking */
228 5378562 : __filemap_set_wb_err(mapping, error);
229 :
230 : /* Record it in superblock */
231 5378565 : if (mapping->host)
232 5378565 : errseq_set(&mapping->host->i_sb->s_wb_err, error);
233 :
234 : /* Record it in flags for now, for legacy callers */
235 5378559 : if (error == -ENOSPC)
236 2773181 : set_bit(AS_ENOSPC, &mapping->flags);
237 : else
238 2605378 : set_bit(AS_EIO, &mapping->flags);
239 : }
240 :
241 : static inline void mapping_set_unevictable(struct address_space *mapping)
242 : {
243 : set_bit(AS_UNEVICTABLE, &mapping->flags);
244 : }
245 :
246 : static inline void mapping_clear_unevictable(struct address_space *mapping)
247 : {
248 : clear_bit(AS_UNEVICTABLE, &mapping->flags);
249 : }
250 :
251 : static inline bool mapping_unevictable(struct address_space *mapping)
252 : {
253 : return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
254 : }
255 :
256 : static inline void mapping_set_exiting(struct address_space *mapping)
257 : {
258 1184360273 : set_bit(AS_EXITING, &mapping->flags);
259 : }
260 :
261 : static inline int mapping_exiting(struct address_space *mapping)
262 : {
263 0 : return test_bit(AS_EXITING, &mapping->flags);
264 : }
265 :
266 : static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
267 : {
268 : set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
269 : }
270 :
271 1612736642 : static inline int mapping_use_writeback_tags(struct address_space *mapping)
272 : {
273 1612736642 : return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
274 : }
275 :
276 : static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
277 : {
278 7272565428 : return mapping->gfp_mask;
279 : }
280 :
281 : /* Restricts the given gfp_mask to what the mapping allows. */
282 : static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
283 : gfp_t gfp_mask)
284 : {
285 246126109 : return mapping_gfp_mask(mapping) & gfp_mask;
286 : }
287 :
288 : /*
289 : * This is non-atomic. Only to be used before the mapping is activated.
290 : * Probably needs a barrier...
291 : */
292 : static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
293 : {
294 2740534069 : m->gfp_mask = mask;
295 : }
296 :
297 : /**
298 : * mapping_set_large_folios() - Indicate the file supports large folios.
299 : * @mapping: The file.
300 : *
301 : * The filesystem should call this function in its inode constructor to
302 : * indicate that the VFS can use large folios to cache the contents of
303 : * the file.
304 : *
305 : * Context: This should not be called while the inode is active as it
306 : * is non-atomic.
307 : */
308 : static inline void mapping_set_large_folios(struct address_space *mapping)
309 : {
310 1116029474 : __set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
311 : }
312 :
313 : /*
314 : * Large folio support currently depends on THP. These dependencies are
315 : * being worked on but are not yet fixed.
316 : */
317 : static inline bool mapping_large_folio_support(struct address_space *mapping)
318 : {
319 1928068559 : return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) &&
320 1927879978 : test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
321 : }
322 :
323 : static inline int filemap_nr_thps(struct address_space *mapping)
324 : {
325 : #ifdef CONFIG_READ_ONLY_THP_FOR_FS
326 : return atomic_read(&mapping->nr_thps);
327 : #else
328 : return 0;
329 : #endif
330 : }
331 :
332 : static inline void filemap_nr_thps_inc(struct address_space *mapping)
333 : {
334 : #ifdef CONFIG_READ_ONLY_THP_FOR_FS
335 : if (!mapping_large_folio_support(mapping))
336 : atomic_inc(&mapping->nr_thps);
337 : #else
338 : WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
339 : #endif
340 : }
341 :
342 188581 : static inline void filemap_nr_thps_dec(struct address_space *mapping)
343 : {
344 : #ifdef CONFIG_READ_ONLY_THP_FOR_FS
345 : if (!mapping_large_folio_support(mapping))
346 : atomic_dec(&mapping->nr_thps);
347 : #else
348 377162 : WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
349 : #endif
350 188581 : }
351 :
352 : struct address_space *page_mapping(struct page *);
353 : struct address_space *folio_mapping(struct folio *);
354 : struct address_space *swapcache_mapping(struct folio *);
355 :
356 : /**
357 : * folio_file_mapping - Find the mapping this folio belongs to.
358 : * @folio: The folio.
359 : *
360 : * For folios which are in the page cache, return the mapping that this
361 : * page belongs to. Folios in the swap cache return the mapping of the
362 : * swap file or swap device where the data is stored. This is different
363 : * from the mapping returned by folio_mapping(). The only reason to
364 : * use it is if, like NFS, you return 0 from ->activate_swapfile.
365 : *
366 : * Do not call this for folios which aren't in the page cache or swap cache.
367 : */
368 : static inline struct address_space *folio_file_mapping(struct folio *folio)
369 : {
370 : if (unlikely(folio_test_swapcache(folio)))
371 : return swapcache_mapping(folio);
372 :
373 : return folio->mapping;
374 : }
375 :
376 : static inline struct address_space *page_file_mapping(struct page *page)
377 : {
378 : return folio_file_mapping(page_folio(page));
379 : }
380 :
381 : /*
382 : * For file cache pages, return the address_space, otherwise return NULL
383 : */
384 : static inline struct address_space *page_mapping_file(struct page *page)
385 : {
386 : struct folio *folio = page_folio(page);
387 :
388 : if (unlikely(folio_test_swapcache(folio)))
389 : return NULL;
390 : return folio_mapping(folio);
391 : }
392 :
393 : /**
394 : * folio_inode - Get the host inode for this folio.
395 : * @folio: The folio.
396 : *
397 : * For folios which are in the page cache, return the inode that this folio
398 : * belongs to.
399 : *
400 : * Do not call this for folios which aren't in the page cache.
401 : */
402 : static inline struct inode *folio_inode(struct folio *folio)
403 : {
404 1505548350 : return folio->mapping->host;
405 : }
406 :
407 : /**
408 : * folio_attach_private - Attach private data to a folio.
409 : * @folio: Folio to attach data to.
410 : * @data: Data to attach to folio.
411 : *
412 : * Attaching private data to a folio increments the page's reference count.
413 : * The data must be detached before the folio will be freed.
414 : */
415 402192038 : static inline void folio_attach_private(struct folio *folio, void *data)
416 : {
417 402192038 : folio_get(folio);
418 402247352 : folio->private = data;
419 402247352 : folio_set_private(folio);
420 402237942 : }
421 :
422 : /**
423 : * folio_change_private - Change private data on a folio.
424 : * @folio: Folio to change the data on.
425 : * @data: Data to set on the folio.
426 : *
427 : * Change the private data attached to a folio and return the old
428 : * data. The page must previously have had data attached and the data
429 : * must be detached before the folio will be freed.
430 : *
431 : * Return: Data that was previously attached to the folio.
432 : */
433 : static inline void *folio_change_private(struct folio *folio, void *data)
434 : {
435 : void *old = folio_get_private(folio);
436 :
437 : folio->private = data;
438 : return old;
439 : }
440 :
441 : /**
442 : * folio_detach_private - Detach private data from a folio.
443 : * @folio: Folio to detach data from.
444 : *
445 : * Removes the data that was previously attached to the folio and decrements
446 : * the refcount on the page.
447 : *
448 : * Return: Data that was attached to the folio.
449 : */
450 402287914 : static inline void *folio_detach_private(struct folio *folio)
451 : {
452 402287914 : void *data = folio_get_private(folio);
453 :
454 402287914 : if (!folio_test_private(folio))
455 : return NULL;
456 402287914 : folio_clear_private(folio);
457 402309633 : folio->private = NULL;
458 402309633 : folio_put(folio);
459 :
460 402309633 : return data;
461 : }
462 :
463 220496438 : static inline void attach_page_private(struct page *page, void *data)
464 : {
465 220496438 : folio_attach_private(page_folio(page), data);
466 220524355 : }
467 :
468 220548777 : static inline void *detach_page_private(struct page *page)
469 : {
470 220548777 : return folio_detach_private(page_folio(page));
471 : }
472 :
473 : #ifdef CONFIG_NUMA
474 : struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order);
475 : #else
476 : static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order)
477 : {
478 : return folio_alloc(gfp, order);
479 : }
480 : #endif
481 :
482 : static inline struct page *__page_cache_alloc(gfp_t gfp)
483 : {
484 538 : return &filemap_alloc_folio(gfp, 0)->page;
485 : }
486 :
487 : static inline struct page *page_cache_alloc(struct address_space *x)
488 : {
489 : return __page_cache_alloc(mapping_gfp_mask(x));
490 : }
491 :
492 : static inline gfp_t readahead_gfp_mask(struct address_space *x)
493 : {
494 45671774 : return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
495 : }
496 :
497 : typedef int filler_t(struct file *, struct folio *);
498 :
499 : pgoff_t page_cache_next_miss(struct address_space *mapping,
500 : pgoff_t index, unsigned long max_scan);
501 : pgoff_t page_cache_prev_miss(struct address_space *mapping,
502 : pgoff_t index, unsigned long max_scan);
503 :
504 : #define FGP_ACCESSED 0x00000001
505 : #define FGP_LOCK 0x00000002
506 : #define FGP_CREAT 0x00000004
507 : #define FGP_WRITE 0x00000008
508 : #define FGP_NOFS 0x00000010
509 : #define FGP_NOWAIT 0x00000020
510 : #define FGP_FOR_MMAP 0x00000040
511 : #define FGP_STABLE 0x00000080
512 :
513 : #define FGP_WRITEBEGIN (FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE)
514 :
515 : void *filemap_get_entry(struct address_space *mapping, pgoff_t index);
516 : struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
517 : int fgp_flags, gfp_t gfp);
518 : struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
519 : int fgp_flags, gfp_t gfp);
520 :
521 : /**
522 : * filemap_get_folio - Find and get a folio.
523 : * @mapping: The address_space to search.
524 : * @index: The page index.
525 : *
526 : * Looks up the page cache entry at @mapping & @index. If a folio is
527 : * present, it is returned with an increased refcount.
528 : *
529 : * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
530 : * this index. Will not return a shadow, swap or DAX entry.
531 : */
532 : static inline struct folio *filemap_get_folio(struct address_space *mapping,
533 : pgoff_t index)
534 : {
535 4348673965 : return __filemap_get_folio(mapping, index, 0, 0);
536 : }
537 :
538 : /**
539 : * filemap_lock_folio - Find and lock a folio.
540 : * @mapping: The address_space to search.
541 : * @index: The page index.
542 : *
543 : * Looks up the page cache entry at @mapping & @index. If a folio is
544 : * present, it is returned locked with an increased refcount.
545 : *
546 : * Context: May sleep.
547 : * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
548 : * this index. Will not return a shadow, swap or DAX entry.
549 : */
550 : static inline struct folio *filemap_lock_folio(struct address_space *mapping,
551 : pgoff_t index)
552 : {
553 9594 : return __filemap_get_folio(mapping, index, FGP_LOCK, 0);
554 : }
555 :
556 : /**
557 : * filemap_grab_folio - grab a folio from the page cache
558 : * @mapping: The address space to search
559 : * @index: The page index
560 : *
561 : * Looks up the page cache entry at @mapping & @index. If no folio is found,
562 : * a new folio is created. The folio is locked, marked as accessed, and
563 : * returned.
564 : *
565 : * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found
566 : * and failed to create a folio.
567 : */
568 : static inline struct folio *filemap_grab_folio(struct address_space *mapping,
569 : pgoff_t index)
570 : {
571 116 : return __filemap_get_folio(mapping, index,
572 : FGP_LOCK | FGP_ACCESSED | FGP_CREAT,
573 : mapping_gfp_mask(mapping));
574 : }
575 :
576 : /**
577 : * find_get_page - find and get a page reference
578 : * @mapping: the address_space to search
579 : * @offset: the page index
580 : *
581 : * Looks up the page cache slot at @mapping & @offset. If there is a
582 : * page cache page, it is returned with an increased refcount.
583 : *
584 : * Otherwise, %NULL is returned.
585 : */
586 : static inline struct page *find_get_page(struct address_space *mapping,
587 : pgoff_t offset)
588 : {
589 67375878 : return pagecache_get_page(mapping, offset, 0, 0);
590 : }
591 :
592 : static inline struct page *find_get_page_flags(struct address_space *mapping,
593 : pgoff_t offset, int fgp_flags)
594 : {
595 43117578 : return pagecache_get_page(mapping, offset, fgp_flags, 0);
596 : }
597 :
598 : /**
599 : * find_lock_page - locate, pin and lock a pagecache page
600 : * @mapping: the address_space to search
601 : * @index: the page index
602 : *
603 : * Looks up the page cache entry at @mapping & @index. If there is a
604 : * page cache page, it is returned locked and with an increased
605 : * refcount.
606 : *
607 : * Context: May sleep.
608 : * Return: A struct page or %NULL if there is no page in the cache for this
609 : * index.
610 : */
611 : static inline struct page *find_lock_page(struct address_space *mapping,
612 : pgoff_t index)
613 : {
614 10913332 : return pagecache_get_page(mapping, index, FGP_LOCK, 0);
615 : }
616 :
617 : /**
618 : * find_or_create_page - locate or add a pagecache page
619 : * @mapping: the page's address_space
620 : * @index: the page's index into the mapping
621 : * @gfp_mask: page allocation mode
622 : *
623 : * Looks up the page cache slot at @mapping & @offset. If there is a
624 : * page cache page, it is returned locked and with an increased
625 : * refcount.
626 : *
627 : * If the page is not present, a new page is allocated using @gfp_mask
628 : * and added to the page cache and the VM's LRU list. The page is
629 : * returned locked and with an increased refcount.
630 : *
631 : * On memory exhaustion, %NULL is returned.
632 : *
633 : * find_or_create_page() may sleep, even if @gfp_flags specifies an
634 : * atomic allocation!
635 : */
636 : static inline struct page *find_or_create_page(struct address_space *mapping,
637 : pgoff_t index, gfp_t gfp_mask)
638 : {
639 44646936 : return pagecache_get_page(mapping, index,
640 : FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
641 : gfp_mask);
642 : }
643 :
644 : /**
645 : * grab_cache_page_nowait - returns locked page at given index in given cache
646 : * @mapping: target address_space
647 : * @index: the page index
648 : *
649 : * Same as grab_cache_page(), but do not wait if the page is unavailable.
650 : * This is intended for speculative data generators, where the data can
651 : * be regenerated if the page couldn't be grabbed. This routine should
652 : * be safe to call while holding the lock for another page.
653 : *
654 : * Clear __GFP_FS when allocating the page to avoid recursion into the fs
655 : * and deadlock against the caller's locked page.
656 : */
657 : static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
658 : pgoff_t index)
659 : {
660 : return pagecache_get_page(mapping, index,
661 : FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
662 : mapping_gfp_mask(mapping));
663 : }
664 :
665 : #define swapcache_index(folio) __page_file_index(&(folio)->page)
666 :
667 : /**
668 : * folio_index - File index of a folio.
669 : * @folio: The folio.
670 : *
671 : * For a folio which is either in the page cache or the swap cache,
672 : * return its index within the address_space it belongs to. If you know
673 : * the page is definitely in the page cache, you can look at the folio's
674 : * index directly.
675 : *
676 : * Return: The index (offset in units of pages) of a folio in its file.
677 : */
678 2581717919 : static inline pgoff_t folio_index(struct folio *folio)
679 : {
680 2581717919 : if (unlikely(folio_test_swapcache(folio)))
681 0 : return swapcache_index(folio);
682 2581717919 : return folio->index;
683 : }
684 :
685 : /**
686 : * folio_next_index - Get the index of the next folio.
687 : * @folio: The current folio.
688 : *
689 : * Return: The index of the folio which follows this folio in the file.
690 : */
691 : static inline pgoff_t folio_next_index(struct folio *folio)
692 : {
693 10643 : return folio->index + folio_nr_pages(folio);
694 : }
695 :
696 : /**
697 : * folio_file_page - The page for a particular index.
698 : * @folio: The folio which contains this index.
699 : * @index: The index we want to look up.
700 : *
701 : * Sometimes after looking up a folio in the page cache, we need to
702 : * obtain the specific page for an index (eg a page fault).
703 : *
704 : * Return: The page containing the file data for this index.
705 : */
706 36809532779 : static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
707 : {
708 : /* HugeTLBfs indexes the page cache in units of hpage_size */
709 36809532779 : if (folio_test_hugetlb(folio))
710 0 : return &folio->page;
711 36897352655 : return folio_page(folio, index & (folio_nr_pages(folio) - 1));
712 : }
713 :
714 : /**
715 : * folio_contains - Does this folio contain this index?
716 : * @folio: The folio.
717 : * @index: The page index within the file.
718 : *
719 : * Context: The caller should have the page locked in order to prevent
720 : * (eg) shmem from moving the page between the page cache and swap cache
721 : * and changing its index in the middle of the operation.
722 : * Return: true or false.
723 : */
724 : static inline bool folio_contains(struct folio *folio, pgoff_t index)
725 : {
726 : /* HugeTLBfs indexes the page cache in units of hpage_size */
727 : if (folio_test_hugetlb(folio))
728 : return folio->index == index;
729 : return index - folio_index(folio) < folio_nr_pages(folio);
730 : }
731 :
732 : /*
733 : * Given the page we found in the page cache, return the page corresponding
734 : * to this index in the file
735 : */
736 : static inline struct page *find_subpage(struct page *head, pgoff_t index)
737 : {
738 : /* HugeTLBfs wants the head page regardless */
739 : if (PageHuge(head))
740 : return head;
741 :
742 : return head + (index & (thp_nr_pages(head) - 1));
743 : }
744 :
745 : unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start,
746 : pgoff_t end, struct folio_batch *fbatch);
747 : unsigned filemap_get_folios_contig(struct address_space *mapping,
748 : pgoff_t *start, pgoff_t end, struct folio_batch *fbatch);
749 : unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start,
750 : pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch);
751 :
752 : struct page *grab_cache_page_write_begin(struct address_space *mapping,
753 : pgoff_t index);
754 :
755 : /*
756 : * Returns locked page at given index in given cache, creating it if needed.
757 : */
758 : static inline struct page *grab_cache_page(struct address_space *mapping,
759 : pgoff_t index)
760 : {
761 : return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
762 : }
763 :
764 : struct folio *read_cache_folio(struct address_space *, pgoff_t index,
765 : filler_t *filler, struct file *file);
766 : struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index,
767 : gfp_t flags);
768 : struct page *read_cache_page(struct address_space *, pgoff_t index,
769 : filler_t *filler, struct file *file);
770 : extern struct page * read_cache_page_gfp(struct address_space *mapping,
771 : pgoff_t index, gfp_t gfp_mask);
772 :
773 : static inline struct page *read_mapping_page(struct address_space *mapping,
774 : pgoff_t index, struct file *file)
775 : {
776 30862 : return read_cache_page(mapping, index, NULL, file);
777 : }
778 :
779 : static inline struct folio *read_mapping_folio(struct address_space *mapping,
780 : pgoff_t index, struct file *file)
781 : {
782 3859806596 : return read_cache_folio(mapping, index, NULL, file);
783 : }
784 :
785 : /*
786 : * Get index of the page within radix-tree (but not for hugetlb pages).
787 : * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
788 : */
789 : static inline pgoff_t page_to_index(struct page *page)
790 : {
791 : struct page *head;
792 :
793 : if (likely(!PageTransTail(page)))
794 : return page->index;
795 :
796 : head = compound_head(page);
797 : /*
798 : * We don't initialize ->index for tail pages: calculate based on
799 : * head page
800 : */
801 : return head->index + page - head;
802 : }
803 :
804 : extern pgoff_t hugetlb_basepage_index(struct page *page);
805 :
806 : /*
807 : * Get the offset in PAGE_SIZE (even for hugetlb pages).
808 : * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
809 : */
810 : static inline pgoff_t page_to_pgoff(struct page *page)
811 : {
812 : if (unlikely(PageHuge(page)))
813 : return hugetlb_basepage_index(page);
814 : return page_to_index(page);
815 : }
816 :
817 : /*
818 : * Return byte-offset into filesystem object for page.
819 : */
820 : static inline loff_t page_offset(struct page *page)
821 : {
822 8237155592 : return ((loff_t)page->index) << PAGE_SHIFT;
823 : }
824 :
825 : static inline loff_t page_file_offset(struct page *page)
826 : {
827 : return ((loff_t)page_index(page)) << PAGE_SHIFT;
828 : }
829 :
830 : /**
831 : * folio_pos - Returns the byte position of this folio in its file.
832 : * @folio: The folio.
833 : */
834 : static inline loff_t folio_pos(struct folio *folio)
835 : {
836 7828195820 : return page_offset(&folio->page);
837 : }
838 :
839 : /**
840 : * folio_file_pos - Returns the byte position of this folio in its file.
841 : * @folio: The folio.
842 : *
843 : * This differs from folio_pos() for folios which belong to a swap file.
844 : * NFS is the only filesystem today which needs to use folio_file_pos().
845 : */
846 : static inline loff_t folio_file_pos(struct folio *folio)
847 : {
848 : return page_file_offset(&folio->page);
849 : }
850 :
851 : /*
852 : * Get the offset in PAGE_SIZE (even for hugetlb folios).
853 : * (TODO: hugetlb folios should have ->index in PAGE_SIZE)
854 : */
855 0 : static inline pgoff_t folio_pgoff(struct folio *folio)
856 : {
857 0 : if (unlikely(folio_test_hugetlb(folio)))
858 0 : return hugetlb_basepage_index(&folio->page);
859 0 : return folio->index;
860 : }
861 :
862 : extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
863 : unsigned long address);
864 :
865 0 : static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
866 : unsigned long address)
867 : {
868 0 : pgoff_t pgoff;
869 0 : if (unlikely(is_vm_hugetlb_page(vma)))
870 0 : return linear_hugepage_index(vma, address);
871 0 : pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
872 0 : pgoff += vma->vm_pgoff;
873 0 : return pgoff;
874 : }
875 :
876 : struct wait_page_key {
877 : struct folio *folio;
878 : int bit_nr;
879 : int page_match;
880 : };
881 :
882 : struct wait_page_queue {
883 : struct folio *folio;
884 : int bit_nr;
885 : wait_queue_entry_t wait;
886 : };
887 :
888 : static inline bool wake_page_match(struct wait_page_queue *wait_page,
889 : struct wait_page_key *key)
890 : {
891 70212890 : if (wait_page->folio != key->folio)
892 : return false;
893 70050346 : key->page_match = 1;
894 :
895 70050346 : if (wait_page->bit_nr != key->bit_nr)
896 : return false;
897 :
898 : return true;
899 : }
900 :
901 : void __folio_lock(struct folio *folio);
902 : int __folio_lock_killable(struct folio *folio);
903 : bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
904 : unsigned int flags);
905 : void unlock_page(struct page *page);
906 : void folio_unlock(struct folio *folio);
907 :
908 : /**
909 : * folio_trylock() - Attempt to lock a folio.
910 : * @folio: The folio to attempt to lock.
911 : *
912 : * Sometimes it is undesirable to wait for a folio to be unlocked (eg
913 : * when the locks are being taken in the wrong order, or if making
914 : * progress through a batch of folios is more important than processing
915 : * them in order). Usually folio_lock() is the correct function to call.
916 : *
917 : * Context: Any context.
918 : * Return: Whether the lock was successfully acquired.
919 : */
920 44976085212 : static inline bool folio_trylock(struct folio *folio)
921 : {
922 44976085212 : return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
923 : }
924 :
925 : /*
926 : * Return true if the page was successfully locked
927 : */
928 : static inline int trylock_page(struct page *page)
929 : {
930 : return folio_trylock(page_folio(page));
931 : }
932 :
933 : /**
934 : * folio_lock() - Lock this folio.
935 : * @folio: The folio to lock.
936 : *
937 : * The folio lock protects against many things, probably more than it
938 : * should. It is primarily held while a folio is being brought uptodate,
939 : * either from its backing file or from swap. It is also held while a
940 : * folio is being truncated from its address_space, so holding the lock
941 : * is sufficient to keep folio->mapping stable.
942 : *
943 : * The folio lock is also held while write() is modifying the page to
944 : * provide POSIX atomicity guarantees (as long as the write does not
945 : * cross a page boundary). Other modifications to the data in the folio
946 : * do not hold the folio lock and can race with writes, eg DMA and stores
947 : * to mapped pages.
948 : *
949 : * Context: May sleep. If you need to acquire the locks of two or
950 : * more folios, they must be in order of ascending index, if they are
951 : * in the same address_space. If they are in different address_spaces,
952 : * acquire the lock of the folio which belongs to the address_space which
953 : * has the lowest address in memory first.
954 : */
955 6159801598 : static inline void folio_lock(struct folio *folio)
956 : {
957 6159801598 : might_sleep();
958 6159905377 : if (!folio_trylock(folio))
959 2691535 : __folio_lock(folio);
960 6160885373 : }
961 :
962 : /**
963 : * lock_page() - Lock the folio containing this page.
964 : * @page: The page to lock.
965 : *
966 : * See folio_lock() for a description of what the lock protects.
967 : * This is a legacy function and new code should probably use folio_lock()
968 : * instead.
969 : *
970 : * Context: May sleep. Pages in the same folio share a lock, so do not
971 : * attempt to lock two pages which share a folio.
972 : */
973 295139127 : static inline void lock_page(struct page *page)
974 : {
975 295139127 : struct folio *folio;
976 295139127 : might_sleep();
977 :
978 295127091 : folio = page_folio(page);
979 295081768 : if (!folio_trylock(folio))
980 169580 : __folio_lock(folio);
981 295195495 : }
982 :
983 : /**
984 : * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
985 : * @folio: The folio to lock.
986 : *
987 : * Attempts to lock the folio, like folio_lock(), except that the sleep
988 : * to acquire the lock is interruptible by a fatal signal.
989 : *
990 : * Context: May sleep; see folio_lock().
991 : * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
992 : */
993 : static inline int folio_lock_killable(struct folio *folio)
994 : {
995 : might_sleep();
996 : if (!folio_trylock(folio))
997 : return __folio_lock_killable(folio);
998 : return 0;
999 : }
1000 :
1001 : /*
1002 : * folio_lock_or_retry - Lock the folio, unless this would block and the
1003 : * caller indicated that it can handle a retry.
1004 : *
1005 : * Return value and mmap_lock implications depend on flags; see
1006 : * __folio_lock_or_retry().
1007 : */
1008 : static inline bool folio_lock_or_retry(struct folio *folio,
1009 : struct mm_struct *mm, unsigned int flags)
1010 : {
1011 : might_sleep();
1012 : return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
1013 : }
1014 :
1015 : /*
1016 : * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
1017 : * and should not be used directly.
1018 : */
1019 : void folio_wait_bit(struct folio *folio, int bit_nr);
1020 : int folio_wait_bit_killable(struct folio *folio, int bit_nr);
1021 :
1022 : /*
1023 : * Wait for a folio to be unlocked.
1024 : *
1025 : * This must be called with the caller "holding" the folio,
1026 : * ie with increased folio reference count so that the folio won't
1027 : * go away during the wait.
1028 : */
1029 650048 : static inline void folio_wait_locked(struct folio *folio)
1030 : {
1031 650048 : if (folio_test_locked(folio))
1032 527287 : folio_wait_bit(folio, PG_locked);
1033 650048 : }
1034 :
1035 1660427757 : static inline int folio_wait_locked_killable(struct folio *folio)
1036 : {
1037 1660427757 : if (!folio_test_locked(folio))
1038 : return 0;
1039 18768307 : return folio_wait_bit_killable(folio, PG_locked);
1040 : }
1041 :
1042 650048 : static inline void wait_on_page_locked(struct page *page)
1043 : {
1044 650048 : folio_wait_locked(page_folio(page));
1045 650048 : }
1046 :
1047 : static inline int wait_on_page_locked_killable(struct page *page)
1048 : {
1049 : return folio_wait_locked_killable(page_folio(page));
1050 : }
1051 :
1052 : void wait_on_page_writeback(struct page *page);
1053 : void folio_wait_writeback(struct folio *folio);
1054 : int folio_wait_writeback_killable(struct folio *folio);
1055 : void end_page_writeback(struct page *page);
1056 : void folio_end_writeback(struct folio *folio);
1057 : void wait_for_stable_page(struct page *page);
1058 : void folio_wait_stable(struct folio *folio);
1059 : void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
1060 : static inline void __set_page_dirty(struct page *page,
1061 : struct address_space *mapping, int warn)
1062 : {
1063 : __folio_mark_dirty(page_folio(page), mapping, warn);
1064 : }
1065 : void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
1066 : void __folio_cancel_dirty(struct folio *folio);
1067 3551195043 : static inline void folio_cancel_dirty(struct folio *folio)
1068 : {
1069 : /* Avoid atomic ops, locking, etc. when not actually needed. */
1070 3551195043 : if (folio_test_dirty(folio))
1071 321820284 : __folio_cancel_dirty(folio);
1072 3551491340 : }
1073 : bool folio_clear_dirty_for_io(struct folio *folio);
1074 : bool clear_page_dirty_for_io(struct page *page);
1075 : void folio_invalidate(struct folio *folio, size_t offset, size_t length);
1076 : int __set_page_dirty_nobuffers(struct page *page);
1077 : bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
1078 :
1079 : #ifdef CONFIG_MIGRATION
1080 : int filemap_migrate_folio(struct address_space *mapping, struct folio *dst,
1081 : struct folio *src, enum migrate_mode mode);
1082 : #else
1083 : #define filemap_migrate_folio NULL
1084 : #endif
1085 : void folio_end_private_2(struct folio *folio);
1086 : void folio_wait_private_2(struct folio *folio);
1087 : int folio_wait_private_2_killable(struct folio *folio);
1088 :
1089 : /*
1090 : * Add an arbitrary waiter to a page's wait queue
1091 : */
1092 : void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
1093 :
1094 : /*
1095 : * Fault in userspace address range.
1096 : */
1097 : size_t fault_in_writeable(char __user *uaddr, size_t size);
1098 : size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
1099 : size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
1100 : size_t fault_in_readable(const char __user *uaddr, size_t size);
1101 :
1102 : int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
1103 : pgoff_t index, gfp_t gfp);
1104 : int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1105 : pgoff_t index, gfp_t gfp);
1106 : void filemap_remove_folio(struct folio *folio);
1107 : void __filemap_remove_folio(struct folio *folio, void *shadow);
1108 : void replace_page_cache_folio(struct folio *old, struct folio *new);
1109 : void delete_from_page_cache_batch(struct address_space *mapping,
1110 : struct folio_batch *fbatch);
1111 : bool filemap_release_folio(struct folio *folio, gfp_t gfp);
1112 : loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1113 : int whence);
1114 :
1115 : /* Must be non-static for BPF error injection */
1116 : int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1117 : pgoff_t index, gfp_t gfp, void **shadowp);
1118 :
1119 : bool filemap_range_has_writeback(struct address_space *mapping,
1120 : loff_t start_byte, loff_t end_byte);
1121 :
1122 : /**
1123 : * filemap_range_needs_writeback - check if range potentially needs writeback
1124 : * @mapping: address space within which to check
1125 : * @start_byte: offset in bytes where the range starts
1126 : * @end_byte: offset in bytes where the range ends (inclusive)
1127 : *
1128 : * Find at least one page in the range supplied, usually used to check if
1129 : * direct writing in this range will trigger a writeback. Used by O_DIRECT
1130 : * read/write with IOCB_NOWAIT, to see if the caller needs to do
1131 : * filemap_write_and_wait_range() before proceeding.
1132 : *
1133 : * Return: %true if the caller should do filemap_write_and_wait_range() before
1134 : * doing O_DIRECT to a page in this range, %false otherwise.
1135 : */
1136 0 : static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1137 : loff_t start_byte,
1138 : loff_t end_byte)
1139 : {
1140 0 : if (!mapping->nrpages)
1141 : return false;
1142 0 : if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1143 : !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1144 : return false;
1145 0 : return filemap_range_has_writeback(mapping, start_byte, end_byte);
1146 : }
1147 :
1148 : /**
1149 : * struct readahead_control - Describes a readahead request.
1150 : *
1151 : * A readahead request is for consecutive pages. Filesystems which
1152 : * implement the ->readahead method should call readahead_page() or
1153 : * readahead_page_batch() in a loop and attempt to start I/O against
1154 : * each page in the request.
1155 : *
1156 : * Most of the fields in this struct are private and should be accessed
1157 : * by the functions below.
1158 : *
1159 : * @file: The file, used primarily by network filesystems for authentication.
1160 : * May be NULL if invoked internally by the filesystem.
1161 : * @mapping: Readahead this filesystem object.
1162 : * @ra: File readahead state. May be NULL.
1163 : */
1164 : struct readahead_control {
1165 : struct file *file;
1166 : struct address_space *mapping;
1167 : struct file_ra_state *ra;
1168 : /* private: use the readahead_* accessors instead */
1169 : pgoff_t _index;
1170 : unsigned int _nr_pages;
1171 : unsigned int _batch_count;
1172 : bool _workingset;
1173 : unsigned long _pflags;
1174 : };
1175 :
1176 : #define DEFINE_READAHEAD(ractl, f, r, m, i) \
1177 : struct readahead_control ractl = { \
1178 : .file = f, \
1179 : .mapping = m, \
1180 : .ra = r, \
1181 : ._index = i, \
1182 : }
1183 :
1184 : #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
1185 :
1186 : void page_cache_ra_unbounded(struct readahead_control *,
1187 : unsigned long nr_to_read, unsigned long lookahead_count);
1188 : void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
1189 : void page_cache_async_ra(struct readahead_control *, struct folio *,
1190 : unsigned long req_count);
1191 : void readahead_expand(struct readahead_control *ractl,
1192 : loff_t new_start, size_t new_len);
1193 :
1194 : /**
1195 : * page_cache_sync_readahead - generic file readahead
1196 : * @mapping: address_space which holds the pagecache and I/O vectors
1197 : * @ra: file_ra_state which holds the readahead state
1198 : * @file: Used by the filesystem for authentication.
1199 : * @index: Index of first page to be read.
1200 : * @req_count: Total number of pages being read by the caller.
1201 : *
1202 : * page_cache_sync_readahead() should be called when a cache miss happened:
1203 : * it will submit the read. The readahead logic may decide to piggyback more
1204 : * pages onto the read request if access patterns suggest it will improve
1205 : * performance.
1206 : */
1207 : static inline
1208 30048139 : void page_cache_sync_readahead(struct address_space *mapping,
1209 : struct file_ra_state *ra, struct file *file, pgoff_t index,
1210 : unsigned long req_count)
1211 : {
1212 30048139 : DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1213 30048139 : page_cache_sync_ra(&ractl, req_count);
1214 30052033 : }
1215 :
1216 : /**
1217 : * page_cache_async_readahead - file readahead for marked pages
1218 : * @mapping: address_space which holds the pagecache and I/O vectors
1219 : * @ra: file_ra_state which holds the readahead state
1220 : * @file: Used by the filesystem for authentication.
1221 : * @folio: The folio at @index which triggered the readahead call.
1222 : * @index: Index of first page to be read.
1223 : * @req_count: Total number of pages being read by the caller.
1224 : *
1225 : * page_cache_async_readahead() should be called when a page is used which
1226 : * is marked as PageReadahead; this is a marker to suggest that the application
1227 : * has used up enough of the readahead window that we should start pulling in
1228 : * more pages.
1229 : */
1230 : static inline
1231 1089 : void page_cache_async_readahead(struct address_space *mapping,
1232 : struct file_ra_state *ra, struct file *file,
1233 : struct folio *folio, pgoff_t index, unsigned long req_count)
1234 : {
1235 1089 : DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1236 1089 : page_cache_async_ra(&ractl, folio, req_count);
1237 1089 : }
1238 :
1239 1192812767 : static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1240 : {
1241 1192812767 : struct folio *folio;
1242 :
1243 1192812767 : BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1244 1192812767 : ractl->_nr_pages -= ractl->_batch_count;
1245 1192812767 : ractl->_index += ractl->_batch_count;
1246 :
1247 1192812767 : if (!ractl->_nr_pages) {
1248 61223515 : ractl->_batch_count = 0;
1249 61223515 : return NULL;
1250 : }
1251 :
1252 1131589252 : folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1253 1131606826 : VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1254 1131606826 : ractl->_batch_count = folio_nr_pages(folio);
1255 :
1256 1131606826 : return folio;
1257 : }
1258 :
1259 : /**
1260 : * readahead_page - Get the next page to read.
1261 : * @ractl: The current readahead request.
1262 : *
1263 : * Context: The page is locked and has an elevated refcount. The caller
1264 : * should decreases the refcount once the page has been submitted for I/O
1265 : * and unlock the page once all I/O to that page has completed.
1266 : * Return: A pointer to the next page, or %NULL if we are done.
1267 : */
1268 : static inline struct page *readahead_page(struct readahead_control *ractl)
1269 : {
1270 : struct folio *folio = __readahead_folio(ractl);
1271 :
1272 : return &folio->page;
1273 : }
1274 :
1275 : /**
1276 : * readahead_folio - Get the next folio to read.
1277 : * @ractl: The current readahead request.
1278 : *
1279 : * Context: The folio is locked. The caller should unlock the folio once
1280 : * all I/O to that folio has completed.
1281 : * Return: A pointer to the next folio, or %NULL if we are done.
1282 : */
1283 1192814429 : static inline struct folio *readahead_folio(struct readahead_control *ractl)
1284 : {
1285 1192814429 : struct folio *folio = __readahead_folio(ractl);
1286 :
1287 1192810713 : if (folio)
1288 1131587828 : folio_put(folio);
1289 1192845661 : return folio;
1290 : }
1291 :
1292 4606197 : static inline unsigned int __readahead_batch(struct readahead_control *rac,
1293 : struct page **array, unsigned int array_sz)
1294 : {
1295 4606197 : unsigned int i = 0;
1296 4606197 : XA_STATE(xas, &rac->mapping->i_pages, 0);
1297 4606197 : struct page *page;
1298 :
1299 4606197 : BUG_ON(rac->_batch_count > rac->_nr_pages);
1300 4606197 : rac->_nr_pages -= rac->_batch_count;
1301 4606197 : rac->_index += rac->_batch_count;
1302 4606197 : rac->_batch_count = 0;
1303 :
1304 4606197 : xas_set(&xas, rac->_index);
1305 4606197 : rcu_read_lock();
1306 46286890 : xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1307 44160760 : if (xas_retry(&xas, page))
1308 0 : continue;
1309 44160760 : VM_BUG_ON_PAGE(!PageLocked(page), page);
1310 44160760 : VM_BUG_ON_PAGE(PageTail(page), page);
1311 44160760 : array[i++] = page;
1312 44160760 : rac->_batch_count += thp_nr_pages(page);
1313 44160760 : if (i == array_sz)
1314 : break;
1315 : }
1316 4606164 : rcu_read_unlock();
1317 :
1318 4606155 : return i;
1319 : }
1320 :
1321 : /**
1322 : * readahead_page_batch - Get a batch of pages to read.
1323 : * @rac: The current readahead request.
1324 : * @array: An array of pointers to struct page.
1325 : *
1326 : * Context: The pages are locked and have an elevated refcount. The caller
1327 : * should decreases the refcount once the page has been submitted for I/O
1328 : * and unlock the page once all I/O to that page has completed.
1329 : * Return: The number of pages placed in the array. 0 indicates the request
1330 : * is complete.
1331 : */
1332 : #define readahead_page_batch(rac, array) \
1333 : __readahead_batch(rac, array, ARRAY_SIZE(array))
1334 :
1335 : /**
1336 : * readahead_pos - The byte offset into the file of this readahead request.
1337 : * @rac: The readahead request.
1338 : */
1339 : static inline loff_t readahead_pos(struct readahead_control *rac)
1340 : {
1341 41992399 : return (loff_t)rac->_index * PAGE_SIZE;
1342 : }
1343 :
1344 : /**
1345 : * readahead_length - The number of bytes in this readahead request.
1346 : * @rac: The readahead request.
1347 : */
1348 : static inline size_t readahead_length(struct readahead_control *rac)
1349 : {
1350 38755150 : return rac->_nr_pages * PAGE_SIZE;
1351 : }
1352 :
1353 : /**
1354 : * readahead_index - The index of the first page in this readahead request.
1355 : * @rac: The readahead request.
1356 : */
1357 : static inline pgoff_t readahead_index(struct readahead_control *rac)
1358 : {
1359 109488162 : return rac->_index;
1360 : }
1361 :
1362 : /**
1363 : * readahead_count - The number of pages in this readahead request.
1364 : * @rac: The readahead request.
1365 : */
1366 : static inline unsigned int readahead_count(struct readahead_control *rac)
1367 : {
1368 661246929 : return rac->_nr_pages;
1369 : }
1370 :
1371 : /**
1372 : * readahead_batch_length - The number of bytes in the current batch.
1373 : * @rac: The readahead request.
1374 : */
1375 : static inline size_t readahead_batch_length(struct readahead_control *rac)
1376 : {
1377 3237249 : return rac->_batch_count * PAGE_SIZE;
1378 : }
1379 :
1380 : static inline unsigned long dir_pages(struct inode *inode)
1381 : {
1382 : return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1383 : PAGE_SHIFT;
1384 : }
1385 :
1386 : /**
1387 : * folio_mkwrite_check_truncate - check if folio was truncated
1388 : * @folio: the folio to check
1389 : * @inode: the inode to check the folio against
1390 : *
1391 : * Return: the number of bytes in the folio up to EOF,
1392 : * or -EFAULT if the folio was truncated.
1393 : */
1394 83718606 : static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1395 : struct inode *inode)
1396 : {
1397 83718606 : loff_t size = i_size_read(inode);
1398 83718606 : pgoff_t index = size >> PAGE_SHIFT;
1399 83718606 : size_t offset = offset_in_folio(folio, size);
1400 :
1401 83740614 : if (!folio->mapping)
1402 : return -EFAULT;
1403 :
1404 : /* folio is wholly inside EOF */
1405 122509804 : if (folio_next_index(folio) - 1 < index)
1406 80765804 : return folio_size(folio);
1407 : /* folio is wholly past EOF */
1408 2960965 : if (folio->index > index || !offset)
1409 : return -EFAULT;
1410 : /* folio is partially inside EOF */
1411 2960965 : return offset;
1412 : }
1413 :
1414 : /**
1415 : * page_mkwrite_check_truncate - check if page was truncated
1416 : * @page: the page to check
1417 : * @inode: the inode to check the page against
1418 : *
1419 : * Returns the number of bytes in the page up to EOF,
1420 : * or -EFAULT if the page was truncated.
1421 : */
1422 : static inline int page_mkwrite_check_truncate(struct page *page,
1423 : struct inode *inode)
1424 : {
1425 : loff_t size = i_size_read(inode);
1426 : pgoff_t index = size >> PAGE_SHIFT;
1427 : int offset = offset_in_page(size);
1428 :
1429 : if (page->mapping != inode->i_mapping)
1430 : return -EFAULT;
1431 :
1432 : /* page is wholly inside EOF */
1433 : if (page->index < index)
1434 : return PAGE_SIZE;
1435 : /* page is wholly past EOF */
1436 : if (page->index > index || !offset)
1437 : return -EFAULT;
1438 : /* page is partially inside EOF */
1439 : return offset;
1440 : }
1441 :
1442 : /**
1443 : * i_blocks_per_folio - How many blocks fit in this folio.
1444 : * @inode: The inode which contains the blocks.
1445 : * @folio: The folio.
1446 : *
1447 : * If the block size is larger than the size of this folio, return zero.
1448 : *
1449 : * Context: The caller should hold a refcount on the folio to prevent it
1450 : * from being split.
1451 : * Return: The number of filesystem blocks covered by this folio.
1452 : */
1453 : static inline
1454 7598479156 : unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1455 : {
1456 7598479156 : return folio_size(folio) >> inode->i_blkbits;
1457 : }
1458 :
1459 : static inline
1460 : unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1461 : {
1462 : return i_blocks_per_folio(inode, page_folio(page));
1463 : }
1464 : #endif /* _LINUX_PAGEMAP_H */
|
