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 103740 : 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 587737 : 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 19944 : 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 1095221809 : 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 1095454734 : 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 2535332230 : 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 27775680165 : static inline bool mapping_shrinkable(struct address_space *mapping)
166 : {
167 27775680165 : 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 27775680165 : if (IS_ENABLED(CONFIG_HIGHMEM))
175 : return true;
176 :
177 : /* Cache completely empty? Shrink away. */
178 27775680165 : head = rcu_access_pointer(mapping->i_pages.xa_head);
179 27775680165 : 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 5721260768 : if (!xa_is_node(head) && xa_is_value(head))
189 355634 : 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 275895232 : static inline void mapping_set_error(struct address_space *mapping, int error)
223 : {
224 275895232 : if (likely(!error))
225 : return;
226 :
227 : /* Record in wb_err for checkers using errseq_t based tracking */
228 2398955 : __filemap_set_wb_err(mapping, error);
229 :
230 : /* Record it in superblock */
231 2398954 : if (mapping->host)
232 2398954 : errseq_set(&mapping->host->i_sb->s_wb_err, error);
233 :
234 : /* Record it in flags for now, for legacy callers */
235 2398954 : if (error == -ENOSPC)
236 1039599 : set_bit(AS_ENOSPC, &mapping->flags);
237 : else
238 1359355 : 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 1148517008 : set_bit(AS_EXITING, &mapping->flags);
259 : }
260 :
261 : static inline int mapping_exiting(struct address_space *mapping)
262 : {
263 : 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 778976878 : static inline int mapping_use_writeback_tags(struct address_space *mapping)
272 : {
273 778976878 : 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 4651126387 : 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 175385705 : 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 2237529304 : 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 1087426422 : __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 768986614 : return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) &&
320 768747291 : 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 239323 : 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 478646 : WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
349 : #endif
350 239323 : }
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 470847708 : 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 414429220 : static inline void folio_attach_private(struct folio *folio, void *data)
416 : {
417 414429220 : folio_get(folio);
418 414491649 : folio->private = data;
419 414491649 : folio_set_private(folio);
420 414544753 : }
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 414562232 : static inline void *folio_detach_private(struct folio *folio)
451 : {
452 414562232 : void *data = folio_get_private(folio);
453 :
454 414562232 : if (!folio_test_private(folio))
455 : return NULL;
456 414557606 : folio_clear_private(folio);
457 414587706 : folio->private = NULL;
458 414587706 : folio_put(folio);
459 :
460 414587706 : return data;
461 : }
462 :
463 224443245 : static inline void attach_page_private(struct page *page, void *data)
464 : {
465 224443245 : folio_attach_private(page_folio(page), data);
466 224497755 : }
467 :
468 224510081 : static inline void *detach_page_private(struct page *page)
469 : {
470 224510081 : return folio_detach_private(page_folio(page));
471 : }
472 :
473 : /*
474 : * There are some parts of the kernel which assume that PMD entries
475 : * are exactly HPAGE_PMD_ORDER. Those should be fixed, but until then,
476 : * limit the maximum allocation order to PMD size. I'm not aware of any
477 : * assumptions about maximum order if THP are disabled, but 8 seems like
478 : * a good order (that's 1MB if you're using 4kB pages)
479 : */
480 : #ifdef CONFIG_TRANSPARENT_HUGEPAGE
481 : #define MAX_PAGECACHE_ORDER HPAGE_PMD_ORDER
482 : #else
483 : #define MAX_PAGECACHE_ORDER 8
484 : #endif
485 :
486 : #ifdef CONFIG_NUMA
487 : struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order);
488 : #else
489 : static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order)
490 : {
491 : return folio_alloc(gfp, order);
492 : }
493 : #endif
494 :
495 : static inline struct page *__page_cache_alloc(gfp_t gfp)
496 : {
497 538 : return &filemap_alloc_folio(gfp, 0)->page;
498 : }
499 :
500 : static inline struct page *page_cache_alloc(struct address_space *x)
501 : {
502 : return __page_cache_alloc(mapping_gfp_mask(x));
503 : }
504 :
505 : static inline gfp_t readahead_gfp_mask(struct address_space *x)
506 : {
507 41095905 : return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
508 : }
509 :
510 : typedef int filler_t(struct file *, struct folio *);
511 :
512 : pgoff_t page_cache_next_miss(struct address_space *mapping,
513 : pgoff_t index, unsigned long max_scan);
514 : pgoff_t page_cache_prev_miss(struct address_space *mapping,
515 : pgoff_t index, unsigned long max_scan);
516 :
517 : /**
518 : * typedef fgf_t - Flags for getting folios from the page cache.
519 : *
520 : * Most users of the page cache will not need to use these flags;
521 : * there are convenience functions such as filemap_get_folio() and
522 : * filemap_lock_folio(). For users which need more control over exactly
523 : * what is done with the folios, these flags to __filemap_get_folio()
524 : * are available.
525 : *
526 : * * %FGP_ACCESSED - The folio will be marked accessed.
527 : * * %FGP_LOCK - The folio is returned locked.
528 : * * %FGP_CREAT - If no folio is present then a new folio is allocated,
529 : * added to the page cache and the VM's LRU list. The folio is
530 : * returned locked.
531 : * * %FGP_FOR_MMAP - The caller wants to do its own locking dance if the
532 : * folio is already in cache. If the folio was allocated, unlock it
533 : * before returning so the caller can do the same dance.
534 : * * %FGP_WRITE - The folio will be written to by the caller.
535 : * * %FGP_NOFS - __GFP_FS will get cleared in gfp.
536 : * * %FGP_NOWAIT - Don't block on the folio lock.
537 : * * %FGP_STABLE - Wait for the folio to be stable (finished writeback)
538 : * * %FGP_WRITEBEGIN - The flags to use in a filesystem write_begin()
539 : * implementation.
540 : */
541 : typedef unsigned int __bitwise fgf_t;
542 :
543 : #define FGP_ACCESSED ((__force fgf_t)0x00000001)
544 : #define FGP_LOCK ((__force fgf_t)0x00000002)
545 : #define FGP_CREAT ((__force fgf_t)0x00000004)
546 : #define FGP_WRITE ((__force fgf_t)0x00000008)
547 : #define FGP_NOFS ((__force fgf_t)0x00000010)
548 : #define FGP_NOWAIT ((__force fgf_t)0x00000020)
549 : #define FGP_FOR_MMAP ((__force fgf_t)0x00000040)
550 : #define FGP_STABLE ((__force fgf_t)0x00000080)
551 : #define FGF_GET_ORDER(fgf) (((__force unsigned)fgf) >> 26) /* top 6 bits */
552 :
553 : #define FGP_WRITEBEGIN (FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE)
554 :
555 : /**
556 : * fgf_set_order - Encode a length in the fgf_t flags.
557 : * @size: The suggested size of the folio to create.
558 : *
559 : * The caller of __filemap_get_folio() can use this to suggest a preferred
560 : * size for the folio that is created. If there is already a folio at
561 : * the index, it will be returned, no matter what its size. If a folio
562 : * is freshly created, it may be of a different size than requested
563 : * due to alignment constraints, memory pressure, or the presence of
564 : * other folios at nearby indices.
565 : */
566 426960478 : static inline fgf_t fgf_set_order(size_t size)
567 : {
568 426960478 : unsigned int shift = ilog2(size);
569 :
570 426960478 : if (shift <= PAGE_SHIFT)
571 : return 0;
572 134770356 : return (__force fgf_t)((shift - PAGE_SHIFT) << 26);
573 : }
574 :
575 : void *filemap_get_entry(struct address_space *mapping, pgoff_t index);
576 : struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
577 : fgf_t fgp_flags, gfp_t gfp);
578 : struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
579 : fgf_t fgp_flags, gfp_t gfp);
580 :
581 : /**
582 : * filemap_get_folio - Find and get a folio.
583 : * @mapping: The address_space to search.
584 : * @index: The page index.
585 : *
586 : * Looks up the page cache entry at @mapping & @index. If a folio is
587 : * present, it is returned with an increased refcount.
588 : *
589 : * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
590 : * this index. Will not return a shadow, swap or DAX entry.
591 : */
592 : static inline struct folio *filemap_get_folio(struct address_space *mapping,
593 : pgoff_t index)
594 : {
595 3275798211 : return __filemap_get_folio(mapping, index, 0, 0);
596 : }
597 :
598 : /**
599 : * filemap_lock_folio - Find and lock a folio.
600 : * @mapping: The address_space to search.
601 : * @index: The page index.
602 : *
603 : * Looks up the page cache entry at @mapping & @index. If a folio is
604 : * present, it is returned locked with an increased refcount.
605 : *
606 : * Context: May sleep.
607 : * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
608 : * this index. Will not return a shadow, swap or DAX entry.
609 : */
610 : static inline struct folio *filemap_lock_folio(struct address_space *mapping,
611 : pgoff_t index)
612 : {
613 27618 : return __filemap_get_folio(mapping, index, FGP_LOCK, 0);
614 : }
615 :
616 : /**
617 : * filemap_grab_folio - grab a folio from the page cache
618 : * @mapping: The address space to search
619 : * @index: The page index
620 : *
621 : * Looks up the page cache entry at @mapping & @index. If no folio is found,
622 : * a new folio is created. The folio is locked, marked as accessed, and
623 : * returned.
624 : *
625 : * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found
626 : * and failed to create a folio.
627 : */
628 : static inline struct folio *filemap_grab_folio(struct address_space *mapping,
629 : pgoff_t index)
630 : {
631 117 : return __filemap_get_folio(mapping, index,
632 : FGP_LOCK | FGP_ACCESSED | FGP_CREAT,
633 : mapping_gfp_mask(mapping));
634 : }
635 :
636 : /**
637 : * find_get_page - find and get a page reference
638 : * @mapping: the address_space to search
639 : * @offset: the page index
640 : *
641 : * Looks up the page cache slot at @mapping & @offset. If there is a
642 : * page cache page, it is returned with an increased refcount.
643 : *
644 : * Otherwise, %NULL is returned.
645 : */
646 : static inline struct page *find_get_page(struct address_space *mapping,
647 : pgoff_t offset)
648 : {
649 67095544 : return pagecache_get_page(mapping, offset, 0, 0);
650 : }
651 :
652 : static inline struct page *find_get_page_flags(struct address_space *mapping,
653 : pgoff_t offset, fgf_t fgp_flags)
654 : {
655 41139433 : return pagecache_get_page(mapping, offset, fgp_flags, 0);
656 : }
657 :
658 : /**
659 : * find_lock_page - locate, pin and lock a pagecache page
660 : * @mapping: the address_space to search
661 : * @index: the page index
662 : *
663 : * Looks up the page cache entry at @mapping & @index. If there is a
664 : * page cache page, it is returned locked and with an increased
665 : * refcount.
666 : *
667 : * Context: May sleep.
668 : * Return: A struct page or %NULL if there is no page in the cache for this
669 : * index.
670 : */
671 : static inline struct page *find_lock_page(struct address_space *mapping,
672 : pgoff_t index)
673 : {
674 10992804 : return pagecache_get_page(mapping, index, FGP_LOCK, 0);
675 : }
676 :
677 : /**
678 : * find_or_create_page - locate or add a pagecache page
679 : * @mapping: the page's address_space
680 : * @index: the page's index into the mapping
681 : * @gfp_mask: page allocation mode
682 : *
683 : * Looks up the page cache slot at @mapping & @offset. If there is a
684 : * page cache page, it is returned locked and with an increased
685 : * refcount.
686 : *
687 : * If the page is not present, a new page is allocated using @gfp_mask
688 : * and added to the page cache and the VM's LRU list. The page is
689 : * returned locked and with an increased refcount.
690 : *
691 : * On memory exhaustion, %NULL is returned.
692 : *
693 : * find_or_create_page() may sleep, even if @gfp_flags specifies an
694 : * atomic allocation!
695 : */
696 : static inline struct page *find_or_create_page(struct address_space *mapping,
697 : pgoff_t index, gfp_t gfp_mask)
698 : {
699 48019138 : return pagecache_get_page(mapping, index,
700 : FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
701 : gfp_mask);
702 : }
703 :
704 : /**
705 : * grab_cache_page_nowait - returns locked page at given index in given cache
706 : * @mapping: target address_space
707 : * @index: the page index
708 : *
709 : * Same as grab_cache_page(), but do not wait if the page is unavailable.
710 : * This is intended for speculative data generators, where the data can
711 : * be regenerated if the page couldn't be grabbed. This routine should
712 : * be safe to call while holding the lock for another page.
713 : *
714 : * Clear __GFP_FS when allocating the page to avoid recursion into the fs
715 : * and deadlock against the caller's locked page.
716 : */
717 : static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
718 : pgoff_t index)
719 : {
720 : return pagecache_get_page(mapping, index,
721 : FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
722 : mapping_gfp_mask(mapping));
723 : }
724 :
725 : #define swapcache_index(folio) __page_file_index(&(folio)->page)
726 :
727 : /**
728 : * folio_index - File index of a folio.
729 : * @folio: The folio.
730 : *
731 : * For a folio which is either in the page cache or the swap cache,
732 : * return its index within the address_space it belongs to. If you know
733 : * the page is definitely in the page cache, you can look at the folio's
734 : * index directly.
735 : *
736 : * Return: The index (offset in units of pages) of a folio in its file.
737 : */
738 4492114963 : static inline pgoff_t folio_index(struct folio *folio)
739 : {
740 8984207269 : if (unlikely(folio_test_swapcache(folio)))
741 0 : return swapcache_index(folio);
742 4492092306 : return folio->index;
743 : }
744 :
745 : /**
746 : * folio_next_index - Get the index of the next folio.
747 : * @folio: The current folio.
748 : *
749 : * Return: The index of the folio which follows this folio in the file.
750 : */
751 : static inline pgoff_t folio_next_index(struct folio *folio)
752 : {
753 27457 : return folio->index + folio_nr_pages(folio);
754 : }
755 :
756 : /**
757 : * folio_file_page - The page for a particular index.
758 : * @folio: The folio which contains this index.
759 : * @index: The index we want to look up.
760 : *
761 : * Sometimes after looking up a folio in the page cache, we need to
762 : * obtain the specific page for an index (eg a page fault).
763 : *
764 : * Return: The page containing the file data for this index.
765 : */
766 31157091385 : static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
767 : {
768 : /* HugeTLBfs indexes the page cache in units of hpage_size */
769 31157091385 : if (folio_test_hugetlb(folio))
770 0 : return &folio->page;
771 31159167850 : return folio_page(folio, index & (folio_nr_pages(folio) - 1));
772 : }
773 :
774 : /**
775 : * folio_contains - Does this folio contain this index?
776 : * @folio: The folio.
777 : * @index: The page index within the file.
778 : *
779 : * Context: The caller should have the page locked in order to prevent
780 : * (eg) shmem from moving the page between the page cache and swap cache
781 : * and changing its index in the middle of the operation.
782 : * Return: true or false.
783 : */
784 3252124416 : static inline bool folio_contains(struct folio *folio, pgoff_t index)
785 : {
786 : /* HugeTLBfs indexes the page cache in units of hpage_size */
787 3252124416 : if (folio_test_hugetlb(folio))
788 0 : return folio->index == index;
789 3252147368 : return index - folio_index(folio) < folio_nr_pages(folio);
790 : }
791 :
792 : /*
793 : * Given the page we found in the page cache, return the page corresponding
794 : * to this index in the file
795 : */
796 : static inline struct page *find_subpage(struct page *head, pgoff_t index)
797 : {
798 : /* HugeTLBfs wants the head page regardless */
799 : if (PageHuge(head))
800 : return head;
801 :
802 : return head + (index & (thp_nr_pages(head) - 1));
803 : }
804 :
805 : unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start,
806 : pgoff_t end, struct folio_batch *fbatch);
807 : unsigned filemap_get_folios_contig(struct address_space *mapping,
808 : pgoff_t *start, pgoff_t end, struct folio_batch *fbatch);
809 : unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start,
810 : pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch);
811 :
812 : struct page *grab_cache_page_write_begin(struct address_space *mapping,
813 : pgoff_t index);
814 :
815 : /*
816 : * Returns locked page at given index in given cache, creating it if needed.
817 : */
818 : static inline struct page *grab_cache_page(struct address_space *mapping,
819 : pgoff_t index)
820 : {
821 : return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
822 : }
823 :
824 : struct folio *read_cache_folio(struct address_space *, pgoff_t index,
825 : filler_t *filler, struct file *file);
826 : struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index,
827 : gfp_t flags);
828 : struct page *read_cache_page(struct address_space *, pgoff_t index,
829 : filler_t *filler, struct file *file);
830 : extern struct page * read_cache_page_gfp(struct address_space *mapping,
831 : pgoff_t index, gfp_t gfp_mask);
832 :
833 : static inline struct page *read_mapping_page(struct address_space *mapping,
834 : pgoff_t index, struct file *file)
835 : {
836 35317 : return read_cache_page(mapping, index, NULL, file);
837 : }
838 :
839 : static inline struct folio *read_mapping_folio(struct address_space *mapping,
840 : pgoff_t index, struct file *file)
841 : {
842 2834272318 : return read_cache_folio(mapping, index, NULL, file);
843 : }
844 :
845 : /*
846 : * Get index of the page within radix-tree (but not for hugetlb pages).
847 : * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
848 : */
849 : static inline pgoff_t page_to_index(struct page *page)
850 : {
851 : struct page *head;
852 :
853 : if (likely(!PageTransTail(page)))
854 : return page->index;
855 :
856 : head = compound_head(page);
857 : /*
858 : * We don't initialize ->index for tail pages: calculate based on
859 : * head page
860 : */
861 : return head->index + page - head;
862 : }
863 :
864 : extern pgoff_t hugetlb_basepage_index(struct page *page);
865 :
866 : /*
867 : * Get the offset in PAGE_SIZE (even for hugetlb pages).
868 : * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
869 : */
870 : static inline pgoff_t page_to_pgoff(struct page *page)
871 : {
872 : if (unlikely(PageHuge(page)))
873 : return hugetlb_basepage_index(page);
874 : return page_to_index(page);
875 : }
876 :
877 : /*
878 : * Return byte-offset into filesystem object for page.
879 : */
880 : static inline loff_t page_offset(struct page *page)
881 : {
882 5334514792 : return ((loff_t)page->index) << PAGE_SHIFT;
883 : }
884 :
885 : static inline loff_t page_file_offset(struct page *page)
886 : {
887 : return ((loff_t)page_index(page)) << PAGE_SHIFT;
888 : }
889 :
890 : /**
891 : * folio_pos - Returns the byte position of this folio in its file.
892 : * @folio: The folio.
893 : */
894 : static inline loff_t folio_pos(struct folio *folio)
895 : {
896 4931364978 : return page_offset(&folio->page);
897 : }
898 :
899 : /**
900 : * folio_file_pos - Returns the byte position of this folio in its file.
901 : * @folio: The folio.
902 : *
903 : * This differs from folio_pos() for folios which belong to a swap file.
904 : * NFS is the only filesystem today which needs to use folio_file_pos().
905 : */
906 : static inline loff_t folio_file_pos(struct folio *folio)
907 : {
908 : return page_file_offset(&folio->page);
909 : }
910 :
911 : /*
912 : * Get the offset in PAGE_SIZE (even for hugetlb folios).
913 : * (TODO: hugetlb folios should have ->index in PAGE_SIZE)
914 : */
915 0 : static inline pgoff_t folio_pgoff(struct folio *folio)
916 : {
917 0 : if (unlikely(folio_test_hugetlb(folio)))
918 0 : return hugetlb_basepage_index(&folio->page);
919 0 : return folio->index;
920 : }
921 :
922 : extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
923 : unsigned long address);
924 :
925 0 : static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
926 : unsigned long address)
927 : {
928 0 : pgoff_t pgoff;
929 0 : if (unlikely(is_vm_hugetlb_page(vma)))
930 0 : return linear_hugepage_index(vma, address);
931 0 : pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
932 0 : pgoff += vma->vm_pgoff;
933 0 : return pgoff;
934 : }
935 :
936 : struct wait_page_key {
937 : struct folio *folio;
938 : int bit_nr;
939 : int page_match;
940 : };
941 :
942 : struct wait_page_queue {
943 : struct folio *folio;
944 : int bit_nr;
945 : wait_queue_entry_t wait;
946 : };
947 :
948 : static inline bool wake_page_match(struct wait_page_queue *wait_page,
949 : struct wait_page_key *key)
950 : {
951 57864700 : if (wait_page->folio != key->folio)
952 : return false;
953 57749302 : key->page_match = 1;
954 :
955 57749302 : if (wait_page->bit_nr != key->bit_nr)
956 : return false;
957 :
958 : return true;
959 : }
960 :
961 : void __folio_lock(struct folio *folio);
962 : int __folio_lock_killable(struct folio *folio);
963 : bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
964 : unsigned int flags);
965 : void unlock_page(struct page *page);
966 : void folio_unlock(struct folio *folio);
967 :
968 : /**
969 : * folio_trylock() - Attempt to lock a folio.
970 : * @folio: The folio to attempt to lock.
971 : *
972 : * Sometimes it is undesirable to wait for a folio to be unlocked (eg
973 : * when the locks are being taken in the wrong order, or if making
974 : * progress through a batch of folios is more important than processing
975 : * them in order). Usually folio_lock() is the correct function to call.
976 : *
977 : * Context: Any context.
978 : * Return: Whether the lock was successfully acquired.
979 : */
980 37638435487 : static inline bool folio_trylock(struct folio *folio)
981 : {
982 37638435487 : return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
983 : }
984 :
985 : /*
986 : * Return true if the page was successfully locked
987 : */
988 : static inline int trylock_page(struct page *page)
989 : {
990 : return folio_trylock(page_folio(page));
991 : }
992 :
993 : /**
994 : * folio_lock() - Lock this folio.
995 : * @folio: The folio to lock.
996 : *
997 : * The folio lock protects against many things, probably more than it
998 : * should. It is primarily held while a folio is being brought uptodate,
999 : * either from its backing file or from swap. It is also held while a
1000 : * folio is being truncated from its address_space, so holding the lock
1001 : * is sufficient to keep folio->mapping stable.
1002 : *
1003 : * The folio lock is also held while write() is modifying the page to
1004 : * provide POSIX atomicity guarantees (as long as the write does not
1005 : * cross a page boundary). Other modifications to the data in the folio
1006 : * do not hold the folio lock and can race with writes, eg DMA and stores
1007 : * to mapped pages.
1008 : *
1009 : * Context: May sleep. If you need to acquire the locks of two or
1010 : * more folios, they must be in order of ascending index, if they are
1011 : * in the same address_space. If they are in different address_spaces,
1012 : * acquire the lock of the folio which belongs to the address_space which
1013 : * has the lowest address in memory first.
1014 : */
1015 3648596327 : static inline void folio_lock(struct folio *folio)
1016 : {
1017 3648596327 : might_sleep();
1018 3648564505 : if (!folio_trylock(folio))
1019 3844924 : __folio_lock(folio);
1020 3649906557 : }
1021 :
1022 : /**
1023 : * lock_page() - Lock the folio containing this page.
1024 : * @page: The page to lock.
1025 : *
1026 : * See folio_lock() for a description of what the lock protects.
1027 : * This is a legacy function and new code should probably use folio_lock()
1028 : * instead.
1029 : *
1030 : * Context: May sleep. Pages in the same folio share a lock, so do not
1031 : * attempt to lock two pages which share a folio.
1032 : */
1033 348975589 : static inline void lock_page(struct page *page)
1034 : {
1035 348975589 : struct folio *folio;
1036 348975589 : might_sleep();
1037 :
1038 348950525 : folio = page_folio(page);
1039 348948269 : if (!folio_trylock(folio))
1040 161144 : __folio_lock(folio);
1041 349012312 : }
1042 :
1043 : /**
1044 : * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
1045 : * @folio: The folio to lock.
1046 : *
1047 : * Attempts to lock the folio, like folio_lock(), except that the sleep
1048 : * to acquire the lock is interruptible by a fatal signal.
1049 : *
1050 : * Context: May sleep; see folio_lock().
1051 : * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
1052 : */
1053 : static inline int folio_lock_killable(struct folio *folio)
1054 : {
1055 : might_sleep();
1056 : if (!folio_trylock(folio))
1057 : return __folio_lock_killable(folio);
1058 : return 0;
1059 : }
1060 :
1061 : /*
1062 : * folio_lock_or_retry - Lock the folio, unless this would block and the
1063 : * caller indicated that it can handle a retry.
1064 : *
1065 : * Return value and mmap_lock implications depend on flags; see
1066 : * __folio_lock_or_retry().
1067 : */
1068 : static inline bool folio_lock_or_retry(struct folio *folio,
1069 : struct mm_struct *mm, unsigned int flags)
1070 : {
1071 : might_sleep();
1072 : return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
1073 : }
1074 :
1075 : /*
1076 : * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
1077 : * and should not be used directly.
1078 : */
1079 : void folio_wait_bit(struct folio *folio, int bit_nr);
1080 : int folio_wait_bit_killable(struct folio *folio, int bit_nr);
1081 :
1082 : /*
1083 : * Wait for a folio to be unlocked.
1084 : *
1085 : * This must be called with the caller "holding" the folio,
1086 : * ie with increased folio reference count so that the folio won't
1087 : * go away during the wait.
1088 : */
1089 645607 : static inline void folio_wait_locked(struct folio *folio)
1090 : {
1091 645607 : if (folio_test_locked(folio))
1092 526802 : folio_wait_bit(folio, PG_locked);
1093 645607 : }
1094 :
1095 1212812191 : static inline int folio_wait_locked_killable(struct folio *folio)
1096 : {
1097 1212812191 : if (!folio_test_locked(folio))
1098 : return 0;
1099 13234442 : return folio_wait_bit_killable(folio, PG_locked);
1100 : }
1101 :
1102 645607 : static inline void wait_on_page_locked(struct page *page)
1103 : {
1104 645607 : folio_wait_locked(page_folio(page));
1105 645607 : }
1106 :
1107 : static inline int wait_on_page_locked_killable(struct page *page)
1108 : {
1109 : return folio_wait_locked_killable(page_folio(page));
1110 : }
1111 :
1112 : void wait_on_page_writeback(struct page *page);
1113 : void folio_wait_writeback(struct folio *folio);
1114 : int folio_wait_writeback_killable(struct folio *folio);
1115 : void end_page_writeback(struct page *page);
1116 : void folio_end_writeback(struct folio *folio);
1117 : void wait_for_stable_page(struct page *page);
1118 : void folio_wait_stable(struct folio *folio);
1119 : void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
1120 : static inline void __set_page_dirty(struct page *page,
1121 : struct address_space *mapping, int warn)
1122 : {
1123 : __folio_mark_dirty(page_folio(page), mapping, warn);
1124 : }
1125 : void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
1126 : void __folio_cancel_dirty(struct folio *folio);
1127 2505641222 : static inline void folio_cancel_dirty(struct folio *folio)
1128 : {
1129 : /* Avoid atomic ops, locking, etc. when not actually needed. */
1130 2505641222 : if (folio_test_dirty(folio))
1131 79641364 : __folio_cancel_dirty(folio);
1132 2505649826 : }
1133 : bool folio_clear_dirty_for_io(struct folio *folio);
1134 : bool clear_page_dirty_for_io(struct page *page);
1135 : void folio_invalidate(struct folio *folio, size_t offset, size_t length);
1136 : int __set_page_dirty_nobuffers(struct page *page);
1137 : bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
1138 :
1139 : #ifdef CONFIG_MIGRATION
1140 : int filemap_migrate_folio(struct address_space *mapping, struct folio *dst,
1141 : struct folio *src, enum migrate_mode mode);
1142 : #else
1143 : #define filemap_migrate_folio NULL
1144 : #endif
1145 : void folio_end_private_2(struct folio *folio);
1146 : void folio_wait_private_2(struct folio *folio);
1147 : int folio_wait_private_2_killable(struct folio *folio);
1148 :
1149 : /*
1150 : * Add an arbitrary waiter to a page's wait queue
1151 : */
1152 : void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
1153 :
1154 : /*
1155 : * Fault in userspace address range.
1156 : */
1157 : size_t fault_in_writeable(char __user *uaddr, size_t size);
1158 : size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
1159 : size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
1160 : size_t fault_in_readable(const char __user *uaddr, size_t size);
1161 :
1162 : int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
1163 : pgoff_t index, gfp_t gfp);
1164 : int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1165 : pgoff_t index, gfp_t gfp);
1166 : void filemap_remove_folio(struct folio *folio);
1167 : void __filemap_remove_folio(struct folio *folio, void *shadow);
1168 : void replace_page_cache_folio(struct folio *old, struct folio *new);
1169 : void delete_from_page_cache_batch(struct address_space *mapping,
1170 : struct folio_batch *fbatch);
1171 : bool filemap_release_folio(struct folio *folio, gfp_t gfp);
1172 : loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1173 : int whence);
1174 :
1175 : /* Must be non-static for BPF error injection */
1176 : int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1177 : pgoff_t index, gfp_t gfp, void **shadowp);
1178 :
1179 : bool filemap_range_has_writeback(struct address_space *mapping,
1180 : loff_t start_byte, loff_t end_byte);
1181 :
1182 : /**
1183 : * filemap_range_needs_writeback - check if range potentially needs writeback
1184 : * @mapping: address space within which to check
1185 : * @start_byte: offset in bytes where the range starts
1186 : * @end_byte: offset in bytes where the range ends (inclusive)
1187 : *
1188 : * Find at least one page in the range supplied, usually used to check if
1189 : * direct writing in this range will trigger a writeback. Used by O_DIRECT
1190 : * read/write with IOCB_NOWAIT, to see if the caller needs to do
1191 : * filemap_write_and_wait_range() before proceeding.
1192 : *
1193 : * Return: %true if the caller should do filemap_write_and_wait_range() before
1194 : * doing O_DIRECT to a page in this range, %false otherwise.
1195 : */
1196 0 : static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1197 : loff_t start_byte,
1198 : loff_t end_byte)
1199 : {
1200 0 : if (!mapping->nrpages)
1201 : return false;
1202 0 : if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1203 : !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1204 : return false;
1205 0 : return filemap_range_has_writeback(mapping, start_byte, end_byte);
1206 : }
1207 :
1208 : /**
1209 : * struct readahead_control - Describes a readahead request.
1210 : *
1211 : * A readahead request is for consecutive pages. Filesystems which
1212 : * implement the ->readahead method should call readahead_page() or
1213 : * readahead_page_batch() in a loop and attempt to start I/O against
1214 : * each page in the request.
1215 : *
1216 : * Most of the fields in this struct are private and should be accessed
1217 : * by the functions below.
1218 : *
1219 : * @file: The file, used primarily by network filesystems for authentication.
1220 : * May be NULL if invoked internally by the filesystem.
1221 : * @mapping: Readahead this filesystem object.
1222 : * @ra: File readahead state. May be NULL.
1223 : */
1224 : struct readahead_control {
1225 : struct file *file;
1226 : struct address_space *mapping;
1227 : struct file_ra_state *ra;
1228 : /* private: use the readahead_* accessors instead */
1229 : pgoff_t _index;
1230 : unsigned int _nr_pages;
1231 : unsigned int _batch_count;
1232 : bool _workingset;
1233 : unsigned long _pflags;
1234 : };
1235 :
1236 : #define DEFINE_READAHEAD(ractl, f, r, m, i) \
1237 : struct readahead_control ractl = { \
1238 : .file = f, \
1239 : .mapping = m, \
1240 : .ra = r, \
1241 : ._index = i, \
1242 : }
1243 :
1244 : #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
1245 :
1246 : void page_cache_ra_unbounded(struct readahead_control *,
1247 : unsigned long nr_to_read, unsigned long lookahead_count);
1248 : void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
1249 : void page_cache_async_ra(struct readahead_control *, struct folio *,
1250 : unsigned long req_count);
1251 : void readahead_expand(struct readahead_control *ractl,
1252 : loff_t new_start, size_t new_len);
1253 :
1254 : /**
1255 : * page_cache_sync_readahead - generic file readahead
1256 : * @mapping: address_space which holds the pagecache and I/O vectors
1257 : * @ra: file_ra_state which holds the readahead state
1258 : * @file: Used by the filesystem for authentication.
1259 : * @index: Index of first page to be read.
1260 : * @req_count: Total number of pages being read by the caller.
1261 : *
1262 : * page_cache_sync_readahead() should be called when a cache miss happened:
1263 : * it will submit the read. The readahead logic may decide to piggyback more
1264 : * pages onto the read request if access patterns suggest it will improve
1265 : * performance.
1266 : */
1267 : static inline
1268 27416755 : void page_cache_sync_readahead(struct address_space *mapping,
1269 : struct file_ra_state *ra, struct file *file, pgoff_t index,
1270 : unsigned long req_count)
1271 : {
1272 27416755 : DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1273 27416755 : page_cache_sync_ra(&ractl, req_count);
1274 27421372 : }
1275 :
1276 : /**
1277 : * page_cache_async_readahead - file readahead for marked pages
1278 : * @mapping: address_space which holds the pagecache and I/O vectors
1279 : * @ra: file_ra_state which holds the readahead state
1280 : * @file: Used by the filesystem for authentication.
1281 : * @folio: The folio at @index which triggered the readahead call.
1282 : * @index: Index of first page to be read.
1283 : * @req_count: Total number of pages being read by the caller.
1284 : *
1285 : * page_cache_async_readahead() should be called when a page is used which
1286 : * is marked as PageReadahead; this is a marker to suggest that the application
1287 : * has used up enough of the readahead window that we should start pulling in
1288 : * more pages.
1289 : */
1290 : static inline
1291 1354 : void page_cache_async_readahead(struct address_space *mapping,
1292 : struct file_ra_state *ra, struct file *file,
1293 : struct folio *folio, pgoff_t index, unsigned long req_count)
1294 : {
1295 1354 : DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1296 1354 : page_cache_async_ra(&ractl, folio, req_count);
1297 1354 : }
1298 :
1299 969592124 : static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1300 : {
1301 969592124 : struct folio *folio;
1302 :
1303 969592124 : BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1304 969592124 : ractl->_nr_pages -= ractl->_batch_count;
1305 969592124 : ractl->_index += ractl->_batch_count;
1306 :
1307 969592124 : if (!ractl->_nr_pages) {
1308 54509158 : ractl->_batch_count = 0;
1309 54509158 : return NULL;
1310 : }
1311 :
1312 915082966 : folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1313 915108357 : VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1314 915058540 : ractl->_batch_count = folio_nr_pages(folio);
1315 :
1316 915023976 : return folio;
1317 : }
1318 :
1319 : /**
1320 : * readahead_page - Get the next page to read.
1321 : * @ractl: The current readahead request.
1322 : *
1323 : * Context: The page is locked and has an elevated refcount. The caller
1324 : * should decreases the refcount once the page has been submitted for I/O
1325 : * and unlock the page once all I/O to that page has completed.
1326 : * Return: A pointer to the next page, or %NULL if we are done.
1327 : */
1328 : static inline struct page *readahead_page(struct readahead_control *ractl)
1329 : {
1330 : struct folio *folio = __readahead_folio(ractl);
1331 :
1332 : return &folio->page;
1333 : }
1334 :
1335 : /**
1336 : * readahead_folio - Get the next folio to read.
1337 : * @ractl: The current readahead request.
1338 : *
1339 : * Context: The folio is locked. The caller should unlock the folio once
1340 : * all I/O to that folio has completed.
1341 : * Return: A pointer to the next folio, or %NULL if we are done.
1342 : */
1343 969603384 : static inline struct folio *readahead_folio(struct readahead_control *ractl)
1344 : {
1345 969603384 : struct folio *folio = __readahead_folio(ractl);
1346 :
1347 969535171 : if (folio)
1348 915026504 : folio_put(folio);
1349 969594538 : return folio;
1350 : }
1351 :
1352 4661877 : static inline unsigned int __readahead_batch(struct readahead_control *rac,
1353 : struct page **array, unsigned int array_sz)
1354 : {
1355 4661877 : unsigned int i = 0;
1356 4661877 : XA_STATE(xas, &rac->mapping->i_pages, 0);
1357 4661877 : struct page *page;
1358 :
1359 4661877 : BUG_ON(rac->_batch_count > rac->_nr_pages);
1360 4661877 : rac->_nr_pages -= rac->_batch_count;
1361 4661877 : rac->_index += rac->_batch_count;
1362 4661877 : rac->_batch_count = 0;
1363 :
1364 4661877 : xas_set(&xas, rac->_index);
1365 4661877 : rcu_read_lock();
1366 46955432 : xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1367 44810359 : if (xas_retry(&xas, page))
1368 0 : continue;
1369 89620354 : VM_BUG_ON_PAGE(!PageLocked(page), page);
1370 89620101 : VM_BUG_ON_PAGE(PageTail(page), page);
1371 44809896 : array[i++] = page;
1372 44809896 : rac->_batch_count += thp_nr_pages(page);
1373 44809674 : if (i == array_sz)
1374 : break;
1375 : }
1376 4661843 : rcu_read_unlock();
1377 :
1378 4661826 : return i;
1379 : }
1380 :
1381 : /**
1382 : * readahead_page_batch - Get a batch of pages to read.
1383 : * @rac: The current readahead request.
1384 : * @array: An array of pointers to struct page.
1385 : *
1386 : * Context: The pages are locked and have an elevated refcount. The caller
1387 : * should decreases the refcount once the page has been submitted for I/O
1388 : * and unlock the page once all I/O to that page has completed.
1389 : * Return: The number of pages placed in the array. 0 indicates the request
1390 : * is complete.
1391 : */
1392 : #define readahead_page_batch(rac, array) \
1393 : __readahead_batch(rac, array, ARRAY_SIZE(array))
1394 :
1395 : /**
1396 : * readahead_pos - The byte offset into the file of this readahead request.
1397 : * @rac: The readahead request.
1398 : */
1399 : static inline loff_t readahead_pos(struct readahead_control *rac)
1400 : {
1401 35279506 : return (loff_t)rac->_index * PAGE_SIZE;
1402 : }
1403 :
1404 : /**
1405 : * readahead_length - The number of bytes in this readahead request.
1406 : * @rac: The readahead request.
1407 : */
1408 : static inline size_t readahead_length(struct readahead_control *rac)
1409 : {
1410 31999691 : return rac->_nr_pages * PAGE_SIZE;
1411 : }
1412 :
1413 : /**
1414 : * readahead_index - The index of the first page in this readahead request.
1415 : * @rac: The readahead request.
1416 : */
1417 : static inline pgoff_t readahead_index(struct readahead_control *rac)
1418 : {
1419 99502368 : return rac->_index;
1420 : }
1421 :
1422 : /**
1423 : * readahead_count - The number of pages in this readahead request.
1424 : * @rac: The readahead request.
1425 : */
1426 : static inline unsigned int readahead_count(struct readahead_control *rac)
1427 : {
1428 532388079 : return rac->_nr_pages;
1429 : }
1430 :
1431 : /**
1432 : * readahead_batch_length - The number of bytes in the current batch.
1433 : * @rac: The readahead request.
1434 : */
1435 : static inline size_t readahead_batch_length(struct readahead_control *rac)
1436 : {
1437 3279815 : return rac->_batch_count * PAGE_SIZE;
1438 : }
1439 :
1440 : static inline unsigned long dir_pages(struct inode *inode)
1441 : {
1442 : return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1443 : PAGE_SHIFT;
1444 : }
1445 :
1446 : /**
1447 : * folio_mkwrite_check_truncate - check if folio was truncated
1448 : * @folio: the folio to check
1449 : * @inode: the inode to check the folio against
1450 : *
1451 : * Return: the number of bytes in the folio up to EOF,
1452 : * or -EFAULT if the folio was truncated.
1453 : */
1454 78279443 : static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1455 : struct inode *inode)
1456 : {
1457 78279443 : loff_t size = i_size_read(inode);
1458 78279443 : pgoff_t index = size >> PAGE_SHIFT;
1459 78279443 : size_t offset = offset_in_folio(folio, size);
1460 :
1461 78334149 : if (!folio->mapping)
1462 : return -EFAULT;
1463 :
1464 : /* folio is wholly inside EOF */
1465 78320858 : if (folio_next_index(folio) - 1 < index)
1466 75392583 : return folio_size(folio);
1467 : /* folio is wholly past EOF */
1468 2950597 : if (folio->index > index || !offset)
1469 : return -EFAULT;
1470 : /* folio is partially inside EOF */
1471 2950597 : return offset;
1472 : }
1473 :
1474 : /**
1475 : * page_mkwrite_check_truncate - check if page was truncated
1476 : * @page: the page to check
1477 : * @inode: the inode to check the page against
1478 : *
1479 : * Returns the number of bytes in the page up to EOF,
1480 : * or -EFAULT if the page was truncated.
1481 : */
1482 : static inline int page_mkwrite_check_truncate(struct page *page,
1483 : struct inode *inode)
1484 : {
1485 : loff_t size = i_size_read(inode);
1486 : pgoff_t index = size >> PAGE_SHIFT;
1487 : int offset = offset_in_page(size);
1488 :
1489 : if (page->mapping != inode->i_mapping)
1490 : return -EFAULT;
1491 :
1492 : /* page is wholly inside EOF */
1493 : if (page->index < index)
1494 : return PAGE_SIZE;
1495 : /* page is wholly past EOF */
1496 : if (page->index > index || !offset)
1497 : return -EFAULT;
1498 : /* page is partially inside EOF */
1499 : return offset;
1500 : }
1501 :
1502 : /**
1503 : * i_blocks_per_folio - How many blocks fit in this folio.
1504 : * @inode: The inode which contains the blocks.
1505 : * @folio: The folio.
1506 : *
1507 : * If the block size is larger than the size of this folio, return zero.
1508 : *
1509 : * Context: The caller should hold a refcount on the folio to prevent it
1510 : * from being split.
1511 : * Return: The number of filesystem blocks covered by this folio.
1512 : */
1513 : static inline
1514 4379605538 : unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1515 : {
1516 4379605538 : return folio_size(folio) >> inode->i_blkbits;
1517 : }
1518 :
1519 : static inline
1520 : unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1521 : {
1522 : return i_blocks_per_folio(inode, page_folio(page));
1523 : }
1524 : #endif /* _LINUX_PAGEMAP_H */
|
