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