xref: /linux/include/linux/pagemap.h (revision e4c07ec89ef5299c7bebea6640ac82bc9f7e1c95)
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 int filemap_invalidate_inode(struct inode *inode, bool flush,
44 			     loff_t start, loff_t end);
45 
filemap_fdatawait(struct address_space * mapping)46 static inline int filemap_fdatawait(struct address_space *mapping)
47 {
48 	return filemap_fdatawait_range(mapping, 0, LLONG_MAX);
49 }
50 
51 bool filemap_range_has_page(struct address_space *, loff_t lstart, loff_t lend);
52 int filemap_write_and_wait_range(struct address_space *mapping,
53 		loff_t lstart, loff_t lend);
54 int __filemap_fdatawrite_range(struct address_space *mapping,
55 		loff_t start, loff_t end, int sync_mode);
56 int filemap_fdatawrite_range(struct address_space *mapping,
57 		loff_t start, loff_t end);
58 int filemap_check_errors(struct address_space *mapping);
59 void __filemap_set_wb_err(struct address_space *mapping, int err);
60 int filemap_fdatawrite_wbc(struct address_space *mapping,
61 			   struct writeback_control *wbc);
62 int kiocb_write_and_wait(struct kiocb *iocb, size_t count);
63 
filemap_write_and_wait(struct address_space * mapping)64 static inline int filemap_write_and_wait(struct address_space *mapping)
65 {
66 	return filemap_write_and_wait_range(mapping, 0, LLONG_MAX);
67 }
68 
69 /**
70  * filemap_set_wb_err - set a writeback error on an address_space
71  * @mapping: mapping in which to set writeback error
72  * @err: error to be set in mapping
73  *
74  * When writeback fails in some way, we must record that error so that
75  * userspace can be informed when fsync and the like are called.  We endeavor
76  * to report errors on any file that was open at the time of the error.  Some
77  * internal callers also need to know when writeback errors have occurred.
78  *
79  * When a writeback error occurs, most filesystems will want to call
80  * filemap_set_wb_err to record the error in the mapping so that it will be
81  * automatically reported whenever fsync is called on the file.
82  */
filemap_set_wb_err(struct address_space * mapping,int err)83 static inline void filemap_set_wb_err(struct address_space *mapping, int err)
84 {
85 	/* Fastpath for common case of no error */
86 	if (unlikely(err))
87 		__filemap_set_wb_err(mapping, err);
88 }
89 
90 /**
91  * filemap_check_wb_err - has an error occurred since the mark was sampled?
92  * @mapping: mapping to check for writeback errors
93  * @since: previously-sampled errseq_t
94  *
95  * Grab the errseq_t value from the mapping, and see if it has changed "since"
96  * the given value was sampled.
97  *
98  * If it has then report the latest error set, otherwise return 0.
99  */
filemap_check_wb_err(struct address_space * mapping,errseq_t since)100 static inline int filemap_check_wb_err(struct address_space *mapping,
101 					errseq_t since)
102 {
103 	return errseq_check(&mapping->wb_err, since);
104 }
105 
106 /**
107  * filemap_sample_wb_err - sample the current errseq_t to test for later errors
108  * @mapping: mapping to be sampled
109  *
110  * Writeback errors are always reported relative to a particular sample point
111  * in the past. This function provides those sample points.
112  */
filemap_sample_wb_err(struct address_space * mapping)113 static inline errseq_t filemap_sample_wb_err(struct address_space *mapping)
114 {
115 	return errseq_sample(&mapping->wb_err);
116 }
117 
118 /**
119  * file_sample_sb_err - sample the current errseq_t to test for later errors
120  * @file: file pointer to be sampled
121  *
122  * Grab the most current superblock-level errseq_t value for the given
123  * struct file.
124  */
file_sample_sb_err(struct file * file)125 static inline errseq_t file_sample_sb_err(struct file *file)
126 {
127 	return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err);
128 }
129 
130 /*
131  * Flush file data before changing attributes.  Caller must hold any locks
132  * required to prevent further writes to this file until we're done setting
133  * flags.
134  */
inode_drain_writes(struct inode * inode)135 static inline int inode_drain_writes(struct inode *inode)
136 {
137 	inode_dio_wait(inode);
138 	return filemap_write_and_wait(inode->i_mapping);
139 }
140 
mapping_empty(struct address_space * mapping)141 static inline bool mapping_empty(struct address_space *mapping)
142 {
143 	return xa_empty(&mapping->i_pages);
144 }
145 
146 /*
147  * mapping_shrinkable - test if page cache state allows inode reclaim
148  * @mapping: the page cache mapping
149  *
150  * This checks the mapping's cache state for the pupose of inode
151  * reclaim and LRU management.
152  *
153  * The caller is expected to hold the i_lock, but is not required to
154  * hold the i_pages lock, which usually protects cache state. That's
155  * because the i_lock and the list_lru lock that protect the inode and
156  * its LRU state don't nest inside the irq-safe i_pages lock.
157  *
158  * Cache deletions are performed under the i_lock, which ensures that
159  * when an inode goes empty, it will reliably get queued on the LRU.
160  *
161  * Cache additions do not acquire the i_lock and may race with this
162  * check, in which case we'll report the inode as shrinkable when it
163  * has cache pages. This is okay: the shrinker also checks the
164  * refcount and the referenced bit, which will be elevated or set in
165  * the process of adding new cache pages to an inode.
166  */
mapping_shrinkable(struct address_space * mapping)167 static inline bool mapping_shrinkable(struct address_space *mapping)
168 {
169 	void *head;
170 
171 	/*
172 	 * On highmem systems, there could be lowmem pressure from the
173 	 * inodes before there is highmem pressure from the page
174 	 * cache. Make inodes shrinkable regardless of cache state.
175 	 */
176 	if (IS_ENABLED(CONFIG_HIGHMEM))
177 		return true;
178 
179 	/* Cache completely empty? Shrink away. */
180 	head = rcu_access_pointer(mapping->i_pages.xa_head);
181 	if (!head)
182 		return true;
183 
184 	/*
185 	 * The xarray stores single offset-0 entries directly in the
186 	 * head pointer, which allows non-resident page cache entries
187 	 * to escape the shadow shrinker's list of xarray nodes. The
188 	 * inode shrinker needs to pick them up under memory pressure.
189 	 */
190 	if (!xa_is_node(head) && xa_is_value(head))
191 		return true;
192 
193 	return false;
194 }
195 
196 /*
197  * Bits in mapping->flags.
198  */
199 enum mapping_flags {
200 	AS_EIO		= 0,	/* IO error on async write */
201 	AS_ENOSPC	= 1,	/* ENOSPC on async write */
202 	AS_MM_ALL_LOCKS	= 2,	/* under mm_take_all_locks() */
203 	AS_UNEVICTABLE	= 3,	/* e.g., ramdisk, SHM_LOCK */
204 	AS_EXITING	= 4, 	/* final truncate in progress */
205 	/* writeback related tags are not used */
206 	AS_NO_WRITEBACK_TAGS = 5,
207 	AS_LARGE_FOLIO_SUPPORT = 6,
208 	AS_RELEASE_ALWAYS,	/* Call ->release_folio(), even if no private data */
209 	AS_STABLE_WRITES,	/* must wait for writeback before modifying
210 				   folio contents */
211 	AS_UNMOVABLE,		/* The mapping cannot be moved, ever */
212 };
213 
214 /**
215  * mapping_set_error - record a writeback error in the address_space
216  * @mapping: the mapping in which an error should be set
217  * @error: the error to set in the mapping
218  *
219  * When writeback fails in some way, we must record that error so that
220  * userspace can be informed when fsync and the like are called.  We endeavor
221  * to report errors on any file that was open at the time of the error.  Some
222  * internal callers also need to know when writeback errors have occurred.
223  *
224  * When a writeback error occurs, most filesystems will want to call
225  * mapping_set_error to record the error in the mapping so that it can be
226  * reported when the application calls fsync(2).
227  */
mapping_set_error(struct address_space * mapping,int error)228 static inline void mapping_set_error(struct address_space *mapping, int error)
229 {
230 	if (likely(!error))
231 		return;
232 
233 	/* Record in wb_err for checkers using errseq_t based tracking */
234 	__filemap_set_wb_err(mapping, error);
235 
236 	/* Record it in superblock */
237 	if (mapping->host)
238 		errseq_set(&mapping->host->i_sb->s_wb_err, error);
239 
240 	/* Record it in flags for now, for legacy callers */
241 	if (error == -ENOSPC)
242 		set_bit(AS_ENOSPC, &mapping->flags);
243 	else
244 		set_bit(AS_EIO, &mapping->flags);
245 }
246 
mapping_set_unevictable(struct address_space * mapping)247 static inline void mapping_set_unevictable(struct address_space *mapping)
248 {
249 	set_bit(AS_UNEVICTABLE, &mapping->flags);
250 }
251 
mapping_clear_unevictable(struct address_space * mapping)252 static inline void mapping_clear_unevictable(struct address_space *mapping)
253 {
254 	clear_bit(AS_UNEVICTABLE, &mapping->flags);
255 }
256 
mapping_unevictable(struct address_space * mapping)257 static inline bool mapping_unevictable(struct address_space *mapping)
258 {
259 	return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
260 }
261 
mapping_set_exiting(struct address_space * mapping)262 static inline void mapping_set_exiting(struct address_space *mapping)
263 {
264 	set_bit(AS_EXITING, &mapping->flags);
265 }
266 
mapping_exiting(struct address_space * mapping)267 static inline int mapping_exiting(struct address_space *mapping)
268 {
269 	return test_bit(AS_EXITING, &mapping->flags);
270 }
271 
mapping_set_no_writeback_tags(struct address_space * mapping)272 static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
273 {
274 	set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
275 }
276 
mapping_use_writeback_tags(struct address_space * mapping)277 static inline int mapping_use_writeback_tags(struct address_space *mapping)
278 {
279 	return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
280 }
281 
mapping_release_always(const struct address_space * mapping)282 static inline bool mapping_release_always(const struct address_space *mapping)
283 {
284 	return test_bit(AS_RELEASE_ALWAYS, &mapping->flags);
285 }
286 
mapping_set_release_always(struct address_space * mapping)287 static inline void mapping_set_release_always(struct address_space *mapping)
288 {
289 	set_bit(AS_RELEASE_ALWAYS, &mapping->flags);
290 }
291 
mapping_clear_release_always(struct address_space * mapping)292 static inline void mapping_clear_release_always(struct address_space *mapping)
293 {
294 	clear_bit(AS_RELEASE_ALWAYS, &mapping->flags);
295 }
296 
mapping_stable_writes(const struct address_space * mapping)297 static inline bool mapping_stable_writes(const struct address_space *mapping)
298 {
299 	return test_bit(AS_STABLE_WRITES, &mapping->flags);
300 }
301 
mapping_set_stable_writes(struct address_space * mapping)302 static inline void mapping_set_stable_writes(struct address_space *mapping)
303 {
304 	set_bit(AS_STABLE_WRITES, &mapping->flags);
305 }
306 
mapping_clear_stable_writes(struct address_space * mapping)307 static inline void mapping_clear_stable_writes(struct address_space *mapping)
308 {
309 	clear_bit(AS_STABLE_WRITES, &mapping->flags);
310 }
311 
mapping_set_unmovable(struct address_space * mapping)312 static inline void mapping_set_unmovable(struct address_space *mapping)
313 {
314 	/*
315 	 * It's expected unmovable mappings are also unevictable. Compaction
316 	 * migrate scanner (isolate_migratepages_block()) relies on this to
317 	 * reduce page locking.
318 	 */
319 	set_bit(AS_UNEVICTABLE, &mapping->flags);
320 	set_bit(AS_UNMOVABLE, &mapping->flags);
321 }
322 
mapping_unmovable(struct address_space * mapping)323 static inline bool mapping_unmovable(struct address_space *mapping)
324 {
325 	return test_bit(AS_UNMOVABLE, &mapping->flags);
326 }
327 
mapping_gfp_mask(struct address_space * mapping)328 static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
329 {
330 	return mapping->gfp_mask;
331 }
332 
333 /* Restricts the given gfp_mask to what the mapping allows. */
mapping_gfp_constraint(struct address_space * mapping,gfp_t gfp_mask)334 static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
335 		gfp_t gfp_mask)
336 {
337 	return mapping_gfp_mask(mapping) & gfp_mask;
338 }
339 
340 /*
341  * This is non-atomic.  Only to be used before the mapping is activated.
342  * Probably needs a barrier...
343  */
mapping_set_gfp_mask(struct address_space * m,gfp_t mask)344 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
345 {
346 	m->gfp_mask = mask;
347 }
348 
349 /*
350  * There are some parts of the kernel which assume that PMD entries
351  * are exactly HPAGE_PMD_ORDER.  Those should be fixed, but until then,
352  * limit the maximum allocation order to PMD size.  I'm not aware of any
353  * assumptions about maximum order if THP are disabled, but 8 seems like
354  * a good order (that's 1MB if you're using 4kB pages)
355  */
356 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
357 #define MAX_PAGECACHE_ORDER	HPAGE_PMD_ORDER
358 #else
359 #define MAX_PAGECACHE_ORDER	8
360 #endif
361 
362 /**
363  * mapping_set_large_folios() - Indicate the file supports large folios.
364  * @mapping: The file.
365  *
366  * The filesystem should call this function in its inode constructor to
367  * indicate that the VFS can use large folios to cache the contents of
368  * the file.
369  *
370  * Context: This should not be called while the inode is active as it
371  * is non-atomic.
372  */
mapping_set_large_folios(struct address_space * mapping)373 static inline void mapping_set_large_folios(struct address_space *mapping)
374 {
375 	__set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
376 }
377 
378 /*
379  * Large folio support currently depends on THP.  These dependencies are
380  * being worked on but are not yet fixed.
381  */
mapping_large_folio_support(struct address_space * mapping)382 static inline bool mapping_large_folio_support(struct address_space *mapping)
383 {
384 	return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) &&
385 		test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
386 }
387 
388 /* Return the maximum folio size for this pagecache mapping, in bytes. */
mapping_max_folio_size(struct address_space * mapping)389 static inline size_t mapping_max_folio_size(struct address_space *mapping)
390 {
391 	if (mapping_large_folio_support(mapping))
392 		return PAGE_SIZE << MAX_PAGECACHE_ORDER;
393 	return PAGE_SIZE;
394 }
395 
filemap_nr_thps(struct address_space * mapping)396 static inline int filemap_nr_thps(struct address_space *mapping)
397 {
398 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
399 	return atomic_read(&mapping->nr_thps);
400 #else
401 	return 0;
402 #endif
403 }
404 
filemap_nr_thps_inc(struct address_space * mapping)405 static inline void filemap_nr_thps_inc(struct address_space *mapping)
406 {
407 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
408 	if (!mapping_large_folio_support(mapping))
409 		atomic_inc(&mapping->nr_thps);
410 #else
411 	WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
412 #endif
413 }
414 
filemap_nr_thps_dec(struct address_space * mapping)415 static inline void filemap_nr_thps_dec(struct address_space *mapping)
416 {
417 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
418 	if (!mapping_large_folio_support(mapping))
419 		atomic_dec(&mapping->nr_thps);
420 #else
421 	WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
422 #endif
423 }
424 
425 struct address_space *page_mapping(struct page *);
426 struct address_space *folio_mapping(struct folio *);
427 struct address_space *swapcache_mapping(struct folio *);
428 
429 /**
430  * folio_file_mapping - Find the mapping this folio belongs to.
431  * @folio: The folio.
432  *
433  * For folios which are in the page cache, return the mapping that this
434  * page belongs to.  Folios in the swap cache return the mapping of the
435  * swap file or swap device where the data is stored.  This is different
436  * from the mapping returned by folio_mapping().  The only reason to
437  * use it is if, like NFS, you return 0 from ->activate_swapfile.
438  *
439  * Do not call this for folios which aren't in the page cache or swap cache.
440  */
folio_file_mapping(struct folio * folio)441 static inline struct address_space *folio_file_mapping(struct folio *folio)
442 {
443 	if (unlikely(folio_test_swapcache(folio)))
444 		return swapcache_mapping(folio);
445 
446 	return folio->mapping;
447 }
448 
449 /**
450  * folio_flush_mapping - Find the file mapping this folio belongs to.
451  * @folio: The folio.
452  *
453  * For folios which are in the page cache, return the mapping that this
454  * page belongs to.  Anonymous folios return NULL, even if they're in
455  * the swap cache.  Other kinds of folio also return NULL.
456  *
457  * This is ONLY used by architecture cache flushing code.  If you aren't
458  * writing cache flushing code, you want either folio_mapping() or
459  * folio_file_mapping().
460  */
folio_flush_mapping(struct folio * folio)461 static inline struct address_space *folio_flush_mapping(struct folio *folio)
462 {
463 	if (unlikely(folio_test_swapcache(folio)))
464 		return NULL;
465 
466 	return folio_mapping(folio);
467 }
468 
page_file_mapping(struct page * page)469 static inline struct address_space *page_file_mapping(struct page *page)
470 {
471 	return folio_file_mapping(page_folio(page));
472 }
473 
474 /**
475  * folio_inode - Get the host inode for this folio.
476  * @folio: The folio.
477  *
478  * For folios which are in the page cache, return the inode that this folio
479  * belongs to.
480  *
481  * Do not call this for folios which aren't in the page cache.
482  */
folio_inode(struct folio * folio)483 static inline struct inode *folio_inode(struct folio *folio)
484 {
485 	return folio->mapping->host;
486 }
487 
488 /**
489  * folio_attach_private - Attach private data to a folio.
490  * @folio: Folio to attach data to.
491  * @data: Data to attach to folio.
492  *
493  * Attaching private data to a folio increments the page's reference count.
494  * The data must be detached before the folio will be freed.
495  */
folio_attach_private(struct folio * folio,void * data)496 static inline void folio_attach_private(struct folio *folio, void *data)
497 {
498 	folio_get(folio);
499 	folio->private = data;
500 	folio_set_private(folio);
501 }
502 
503 /**
504  * folio_change_private - Change private data on a folio.
505  * @folio: Folio to change the data on.
506  * @data: Data to set on the folio.
507  *
508  * Change the private data attached to a folio and return the old
509  * data.  The page must previously have had data attached and the data
510  * must be detached before the folio will be freed.
511  *
512  * Return: Data that was previously attached to the folio.
513  */
folio_change_private(struct folio * folio,void * data)514 static inline void *folio_change_private(struct folio *folio, void *data)
515 {
516 	void *old = folio_get_private(folio);
517 
518 	folio->private = data;
519 	return old;
520 }
521 
522 /**
523  * folio_detach_private - Detach private data from a folio.
524  * @folio: Folio to detach data from.
525  *
526  * Removes the data that was previously attached to the folio and decrements
527  * the refcount on the page.
528  *
529  * Return: Data that was attached to the folio.
530  */
folio_detach_private(struct folio * folio)531 static inline void *folio_detach_private(struct folio *folio)
532 {
533 	void *data = folio_get_private(folio);
534 
535 	if (!folio_test_private(folio))
536 		return NULL;
537 	folio_clear_private(folio);
538 	folio->private = NULL;
539 	folio_put(folio);
540 
541 	return data;
542 }
543 
attach_page_private(struct page * page,void * data)544 static inline void attach_page_private(struct page *page, void *data)
545 {
546 	folio_attach_private(page_folio(page), data);
547 }
548 
detach_page_private(struct page * page)549 static inline void *detach_page_private(struct page *page)
550 {
551 	return folio_detach_private(page_folio(page));
552 }
553 
554 #ifdef CONFIG_NUMA
555 struct folio *filemap_alloc_folio_noprof(gfp_t gfp, unsigned int order);
556 #else
filemap_alloc_folio_noprof(gfp_t gfp,unsigned int order)557 static inline struct folio *filemap_alloc_folio_noprof(gfp_t gfp, unsigned int order)
558 {
559 	return folio_alloc_noprof(gfp, order);
560 }
561 #endif
562 
563 #define filemap_alloc_folio(...)				\
564 	alloc_hooks(filemap_alloc_folio_noprof(__VA_ARGS__))
565 
__page_cache_alloc(gfp_t gfp)566 static inline struct page *__page_cache_alloc(gfp_t gfp)
567 {
568 	return &filemap_alloc_folio(gfp, 0)->page;
569 }
570 
readahead_gfp_mask(struct address_space * x)571 static inline gfp_t readahead_gfp_mask(struct address_space *x)
572 {
573 	return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
574 }
575 
576 typedef int filler_t(struct file *, struct folio *);
577 
578 pgoff_t page_cache_next_miss(struct address_space *mapping,
579 			     pgoff_t index, unsigned long max_scan);
580 pgoff_t page_cache_prev_miss(struct address_space *mapping,
581 			     pgoff_t index, unsigned long max_scan);
582 
583 /**
584  * typedef fgf_t - Flags for getting folios from the page cache.
585  *
586  * Most users of the page cache will not need to use these flags;
587  * there are convenience functions such as filemap_get_folio() and
588  * filemap_lock_folio().  For users which need more control over exactly
589  * what is done with the folios, these flags to __filemap_get_folio()
590  * are available.
591  *
592  * * %FGP_ACCESSED - The folio will be marked accessed.
593  * * %FGP_LOCK - The folio is returned locked.
594  * * %FGP_CREAT - If no folio is present then a new folio is allocated,
595  *   added to the page cache and the VM's LRU list.  The folio is
596  *   returned locked.
597  * * %FGP_FOR_MMAP - The caller wants to do its own locking dance if the
598  *   folio is already in cache.  If the folio was allocated, unlock it
599  *   before returning so the caller can do the same dance.
600  * * %FGP_WRITE - The folio will be written to by the caller.
601  * * %FGP_NOFS - __GFP_FS will get cleared in gfp.
602  * * %FGP_NOWAIT - Don't block on the folio lock.
603  * * %FGP_STABLE - Wait for the folio to be stable (finished writeback)
604  * * %FGP_WRITEBEGIN - The flags to use in a filesystem write_begin()
605  *   implementation.
606  */
607 typedef unsigned int __bitwise fgf_t;
608 
609 #define FGP_ACCESSED		((__force fgf_t)0x00000001)
610 #define FGP_LOCK		((__force fgf_t)0x00000002)
611 #define FGP_CREAT		((__force fgf_t)0x00000004)
612 #define FGP_WRITE		((__force fgf_t)0x00000008)
613 #define FGP_NOFS		((__force fgf_t)0x00000010)
614 #define FGP_NOWAIT		((__force fgf_t)0x00000020)
615 #define FGP_FOR_MMAP		((__force fgf_t)0x00000040)
616 #define FGP_STABLE		((__force fgf_t)0x00000080)
617 #define FGF_GET_ORDER(fgf)	(((__force unsigned)fgf) >> 26)	/* top 6 bits */
618 
619 #define FGP_WRITEBEGIN		(FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE)
620 
621 /**
622  * fgf_set_order - Encode a length in the fgf_t flags.
623  * @size: The suggested size of the folio to create.
624  *
625  * The caller of __filemap_get_folio() can use this to suggest a preferred
626  * size for the folio that is created.  If there is already a folio at
627  * the index, it will be returned, no matter what its size.  If a folio
628  * is freshly created, it may be of a different size than requested
629  * due to alignment constraints, memory pressure, or the presence of
630  * other folios at nearby indices.
631  */
fgf_set_order(size_t size)632 static inline fgf_t fgf_set_order(size_t size)
633 {
634 	unsigned int shift = ilog2(size);
635 
636 	if (shift <= PAGE_SHIFT)
637 		return 0;
638 	return (__force fgf_t)((shift - PAGE_SHIFT) << 26);
639 }
640 
641 void *filemap_get_entry(struct address_space *mapping, pgoff_t index);
642 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
643 		fgf_t fgp_flags, gfp_t gfp);
644 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
645 		fgf_t fgp_flags, gfp_t gfp);
646 
647 /**
648  * filemap_get_folio - Find and get a folio.
649  * @mapping: The address_space to search.
650  * @index: The page index.
651  *
652  * Looks up the page cache entry at @mapping & @index.  If a folio is
653  * present, it is returned with an increased refcount.
654  *
655  * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
656  * this index.  Will not return a shadow, swap or DAX entry.
657  */
filemap_get_folio(struct address_space * mapping,pgoff_t index)658 static inline struct folio *filemap_get_folio(struct address_space *mapping,
659 					pgoff_t index)
660 {
661 	return __filemap_get_folio(mapping, index, 0, 0);
662 }
663 
664 /**
665  * filemap_lock_folio - Find and lock a folio.
666  * @mapping: The address_space to search.
667  * @index: The page index.
668  *
669  * Looks up the page cache entry at @mapping & @index.  If a folio is
670  * present, it is returned locked with an increased refcount.
671  *
672  * Context: May sleep.
673  * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
674  * this index.  Will not return a shadow, swap or DAX entry.
675  */
filemap_lock_folio(struct address_space * mapping,pgoff_t index)676 static inline struct folio *filemap_lock_folio(struct address_space *mapping,
677 					pgoff_t index)
678 {
679 	return __filemap_get_folio(mapping, index, FGP_LOCK, 0);
680 }
681 
682 /**
683  * filemap_grab_folio - grab a folio from the page cache
684  * @mapping: The address space to search
685  * @index: The page index
686  *
687  * Looks up the page cache entry at @mapping & @index. If no folio is found,
688  * a new folio is created. The folio is locked, marked as accessed, and
689  * returned.
690  *
691  * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found
692  * and failed to create a folio.
693  */
filemap_grab_folio(struct address_space * mapping,pgoff_t index)694 static inline struct folio *filemap_grab_folio(struct address_space *mapping,
695 					pgoff_t index)
696 {
697 	return __filemap_get_folio(mapping, index,
698 			FGP_LOCK | FGP_ACCESSED | FGP_CREAT,
699 			mapping_gfp_mask(mapping));
700 }
701 
702 /**
703  * find_get_page - find and get a page reference
704  * @mapping: the address_space to search
705  * @offset: the page index
706  *
707  * Looks up the page cache slot at @mapping & @offset.  If there is a
708  * page cache page, it is returned with an increased refcount.
709  *
710  * Otherwise, %NULL is returned.
711  */
find_get_page(struct address_space * mapping,pgoff_t offset)712 static inline struct page *find_get_page(struct address_space *mapping,
713 					pgoff_t offset)
714 {
715 	return pagecache_get_page(mapping, offset, 0, 0);
716 }
717 
find_get_page_flags(struct address_space * mapping,pgoff_t offset,fgf_t fgp_flags)718 static inline struct page *find_get_page_flags(struct address_space *mapping,
719 					pgoff_t offset, fgf_t fgp_flags)
720 {
721 	return pagecache_get_page(mapping, offset, fgp_flags, 0);
722 }
723 
724 /**
725  * find_lock_page - locate, pin and lock a pagecache page
726  * @mapping: the address_space to search
727  * @index: the page index
728  *
729  * Looks up the page cache entry at @mapping & @index.  If there is a
730  * page cache page, it is returned locked and with an increased
731  * refcount.
732  *
733  * Context: May sleep.
734  * Return: A struct page or %NULL if there is no page in the cache for this
735  * index.
736  */
find_lock_page(struct address_space * mapping,pgoff_t index)737 static inline struct page *find_lock_page(struct address_space *mapping,
738 					pgoff_t index)
739 {
740 	return pagecache_get_page(mapping, index, FGP_LOCK, 0);
741 }
742 
743 /**
744  * find_or_create_page - locate or add a pagecache page
745  * @mapping: the page's address_space
746  * @index: the page's index into the mapping
747  * @gfp_mask: page allocation mode
748  *
749  * Looks up the page cache slot at @mapping & @offset.  If there is a
750  * page cache page, it is returned locked and with an increased
751  * refcount.
752  *
753  * If the page is not present, a new page is allocated using @gfp_mask
754  * and added to the page cache and the VM's LRU list.  The page is
755  * returned locked and with an increased refcount.
756  *
757  * On memory exhaustion, %NULL is returned.
758  *
759  * find_or_create_page() may sleep, even if @gfp_flags specifies an
760  * atomic allocation!
761  */
find_or_create_page(struct address_space * mapping,pgoff_t index,gfp_t gfp_mask)762 static inline struct page *find_or_create_page(struct address_space *mapping,
763 					pgoff_t index, gfp_t gfp_mask)
764 {
765 	return pagecache_get_page(mapping, index,
766 					FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
767 					gfp_mask);
768 }
769 
770 /**
771  * grab_cache_page_nowait - returns locked page at given index in given cache
772  * @mapping: target address_space
773  * @index: the page index
774  *
775  * Same as grab_cache_page(), but do not wait if the page is unavailable.
776  * This is intended for speculative data generators, where the data can
777  * be regenerated if the page couldn't be grabbed.  This routine should
778  * be safe to call while holding the lock for another page.
779  *
780  * Clear __GFP_FS when allocating the page to avoid recursion into the fs
781  * and deadlock against the caller's locked page.
782  */
grab_cache_page_nowait(struct address_space * mapping,pgoff_t index)783 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
784 				pgoff_t index)
785 {
786 	return pagecache_get_page(mapping, index,
787 			FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
788 			mapping_gfp_mask(mapping));
789 }
790 
791 #define swapcache_index(folio)	__page_file_index(&(folio)->page)
792 
793 /**
794  * folio_index - File index of a folio.
795  * @folio: The folio.
796  *
797  * For a folio which is either in the page cache or the swap cache,
798  * return its index within the address_space it belongs to.  If you know
799  * the page is definitely in the page cache, you can look at the folio's
800  * index directly.
801  *
802  * Return: The index (offset in units of pages) of a folio in its file.
803  */
folio_index(struct folio * folio)804 static inline pgoff_t folio_index(struct folio *folio)
805 {
806         if (unlikely(folio_test_swapcache(folio)))
807                 return swapcache_index(folio);
808         return folio->index;
809 }
810 
811 /**
812  * folio_next_index - Get the index of the next folio.
813  * @folio: The current folio.
814  *
815  * Return: The index of the folio which follows this folio in the file.
816  */
folio_next_index(struct folio * folio)817 static inline pgoff_t folio_next_index(struct folio *folio)
818 {
819 	return folio->index + folio_nr_pages(folio);
820 }
821 
822 /**
823  * folio_file_page - The page for a particular index.
824  * @folio: The folio which contains this index.
825  * @index: The index we want to look up.
826  *
827  * Sometimes after looking up a folio in the page cache, we need to
828  * obtain the specific page for an index (eg a page fault).
829  *
830  * Return: The page containing the file data for this index.
831  */
folio_file_page(struct folio * folio,pgoff_t index)832 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
833 {
834 	return folio_page(folio, index & (folio_nr_pages(folio) - 1));
835 }
836 
837 /**
838  * folio_contains - Does this folio contain this index?
839  * @folio: The folio.
840  * @index: The page index within the file.
841  *
842  * Context: The caller should have the page locked in order to prevent
843  * (eg) shmem from moving the page between the page cache and swap cache
844  * and changing its index in the middle of the operation.
845  * Return: true or false.
846  */
folio_contains(struct folio * folio,pgoff_t index)847 static inline bool folio_contains(struct folio *folio, pgoff_t index)
848 {
849 	return index - folio_index(folio) < folio_nr_pages(folio);
850 }
851 
852 /*
853  * Given the page we found in the page cache, return the page corresponding
854  * to this index in the file
855  */
find_subpage(struct page * head,pgoff_t index)856 static inline struct page *find_subpage(struct page *head, pgoff_t index)
857 {
858 	/* HugeTLBfs wants the head page regardless */
859 	if (PageHuge(head))
860 		return head;
861 
862 	return head + (index & (thp_nr_pages(head) - 1));
863 }
864 
865 unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start,
866 		pgoff_t end, struct folio_batch *fbatch);
867 unsigned filemap_get_folios_contig(struct address_space *mapping,
868 		pgoff_t *start, pgoff_t end, struct folio_batch *fbatch);
869 unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start,
870 		pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch);
871 
872 struct page *grab_cache_page_write_begin(struct address_space *mapping,
873 			pgoff_t index);
874 
875 /*
876  * Returns locked page at given index in given cache, creating it if needed.
877  */
grab_cache_page(struct address_space * mapping,pgoff_t index)878 static inline struct page *grab_cache_page(struct address_space *mapping,
879 								pgoff_t index)
880 {
881 	return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
882 }
883 
884 struct folio *read_cache_folio(struct address_space *, pgoff_t index,
885 		filler_t *filler, struct file *file);
886 struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index,
887 		gfp_t flags);
888 struct page *read_cache_page(struct address_space *, pgoff_t index,
889 		filler_t *filler, struct file *file);
890 extern struct page * read_cache_page_gfp(struct address_space *mapping,
891 				pgoff_t index, gfp_t gfp_mask);
892 
read_mapping_page(struct address_space * mapping,pgoff_t index,struct file * file)893 static inline struct page *read_mapping_page(struct address_space *mapping,
894 				pgoff_t index, struct file *file)
895 {
896 	return read_cache_page(mapping, index, NULL, file);
897 }
898 
read_mapping_folio(struct address_space * mapping,pgoff_t index,struct file * file)899 static inline struct folio *read_mapping_folio(struct address_space *mapping,
900 				pgoff_t index, struct file *file)
901 {
902 	return read_cache_folio(mapping, index, NULL, file);
903 }
904 
905 /*
906  * Get the offset in PAGE_SIZE (even for hugetlb pages).
907  */
page_to_pgoff(struct page * page)908 static inline pgoff_t page_to_pgoff(struct page *page)
909 {
910 	struct page *head;
911 
912 	if (likely(!PageTransTail(page)))
913 		return page->index;
914 
915 	head = compound_head(page);
916 	/*
917 	 *  We don't initialize ->index for tail pages: calculate based on
918 	 *  head page
919 	 */
920 	return head->index + page - head;
921 }
922 
923 /*
924  * Return byte-offset into filesystem object for page.
925  */
page_offset(struct page * page)926 static inline loff_t page_offset(struct page *page)
927 {
928 	return ((loff_t)page->index) << PAGE_SHIFT;
929 }
930 
page_file_offset(struct page * page)931 static inline loff_t page_file_offset(struct page *page)
932 {
933 	return ((loff_t)page_index(page)) << PAGE_SHIFT;
934 }
935 
936 /**
937  * folio_pos - Returns the byte position of this folio in its file.
938  * @folio: The folio.
939  */
folio_pos(struct folio * folio)940 static inline loff_t folio_pos(struct folio *folio)
941 {
942 	return page_offset(&folio->page);
943 }
944 
945 /**
946  * folio_file_pos - Returns the byte position of this folio in its file.
947  * @folio: The folio.
948  *
949  * This differs from folio_pos() for folios which belong to a swap file.
950  * NFS is the only filesystem today which needs to use folio_file_pos().
951  */
folio_file_pos(struct folio * folio)952 static inline loff_t folio_file_pos(struct folio *folio)
953 {
954 	return page_file_offset(&folio->page);
955 }
956 
957 /*
958  * Get the offset in PAGE_SIZE (even for hugetlb folios).
959  */
folio_pgoff(struct folio * folio)960 static inline pgoff_t folio_pgoff(struct folio *folio)
961 {
962 	return folio->index;
963 }
964 
linear_page_index(struct vm_area_struct * vma,unsigned long address)965 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
966 					unsigned long address)
967 {
968 	pgoff_t pgoff;
969 	pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
970 	pgoff += vma->vm_pgoff;
971 	return pgoff;
972 }
973 
974 struct wait_page_key {
975 	struct folio *folio;
976 	int bit_nr;
977 	int page_match;
978 };
979 
980 struct wait_page_queue {
981 	struct folio *folio;
982 	int bit_nr;
983 	wait_queue_entry_t wait;
984 };
985 
wake_page_match(struct wait_page_queue * wait_page,struct wait_page_key * key)986 static inline bool wake_page_match(struct wait_page_queue *wait_page,
987 				  struct wait_page_key *key)
988 {
989 	if (wait_page->folio != key->folio)
990 	       return false;
991 	key->page_match = 1;
992 
993 	if (wait_page->bit_nr != key->bit_nr)
994 		return false;
995 
996 	return true;
997 }
998 
999 void __folio_lock(struct folio *folio);
1000 int __folio_lock_killable(struct folio *folio);
1001 vm_fault_t __folio_lock_or_retry(struct folio *folio, struct vm_fault *vmf);
1002 void unlock_page(struct page *page);
1003 void folio_unlock(struct folio *folio);
1004 
1005 /**
1006  * folio_trylock() - Attempt to lock a folio.
1007  * @folio: The folio to attempt to lock.
1008  *
1009  * Sometimes it is undesirable to wait for a folio to be unlocked (eg
1010  * when the locks are being taken in the wrong order, or if making
1011  * progress through a batch of folios is more important than processing
1012  * them in order).  Usually folio_lock() is the correct function to call.
1013  *
1014  * Context: Any context.
1015  * Return: Whether the lock was successfully acquired.
1016  */
folio_trylock(struct folio * folio)1017 static inline bool folio_trylock(struct folio *folio)
1018 {
1019 	return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
1020 }
1021 
1022 /*
1023  * Return true if the page was successfully locked
1024  */
trylock_page(struct page * page)1025 static inline bool trylock_page(struct page *page)
1026 {
1027 	return folio_trylock(page_folio(page));
1028 }
1029 
1030 /**
1031  * folio_lock() - Lock this folio.
1032  * @folio: The folio to lock.
1033  *
1034  * The folio lock protects against many things, probably more than it
1035  * should.  It is primarily held while a folio is being brought uptodate,
1036  * either from its backing file or from swap.  It is also held while a
1037  * folio is being truncated from its address_space, so holding the lock
1038  * is sufficient to keep folio->mapping stable.
1039  *
1040  * The folio lock is also held while write() is modifying the page to
1041  * provide POSIX atomicity guarantees (as long as the write does not
1042  * cross a page boundary).  Other modifications to the data in the folio
1043  * do not hold the folio lock and can race with writes, eg DMA and stores
1044  * to mapped pages.
1045  *
1046  * Context: May sleep.  If you need to acquire the locks of two or
1047  * more folios, they must be in order of ascending index, if they are
1048  * in the same address_space.  If they are in different address_spaces,
1049  * acquire the lock of the folio which belongs to the address_space which
1050  * has the lowest address in memory first.
1051  */
folio_lock(struct folio * folio)1052 static inline void folio_lock(struct folio *folio)
1053 {
1054 	might_sleep();
1055 	if (!folio_trylock(folio))
1056 		__folio_lock(folio);
1057 }
1058 
1059 /**
1060  * lock_page() - Lock the folio containing this page.
1061  * @page: The page to lock.
1062  *
1063  * See folio_lock() for a description of what the lock protects.
1064  * This is a legacy function and new code should probably use folio_lock()
1065  * instead.
1066  *
1067  * Context: May sleep.  Pages in the same folio share a lock, so do not
1068  * attempt to lock two pages which share a folio.
1069  */
lock_page(struct page * page)1070 static inline void lock_page(struct page *page)
1071 {
1072 	struct folio *folio;
1073 	might_sleep();
1074 
1075 	folio = page_folio(page);
1076 	if (!folio_trylock(folio))
1077 		__folio_lock(folio);
1078 }
1079 
1080 /**
1081  * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
1082  * @folio: The folio to lock.
1083  *
1084  * Attempts to lock the folio, like folio_lock(), except that the sleep
1085  * to acquire the lock is interruptible by a fatal signal.
1086  *
1087  * Context: May sleep; see folio_lock().
1088  * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
1089  */
folio_lock_killable(struct folio * folio)1090 static inline int folio_lock_killable(struct folio *folio)
1091 {
1092 	might_sleep();
1093 	if (!folio_trylock(folio))
1094 		return __folio_lock_killable(folio);
1095 	return 0;
1096 }
1097 
1098 /*
1099  * folio_lock_or_retry - Lock the folio, unless this would block and the
1100  * caller indicated that it can handle a retry.
1101  *
1102  * Return value and mmap_lock implications depend on flags; see
1103  * __folio_lock_or_retry().
1104  */
folio_lock_or_retry(struct folio * folio,struct vm_fault * vmf)1105 static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
1106 					     struct vm_fault *vmf)
1107 {
1108 	might_sleep();
1109 	if (!folio_trylock(folio))
1110 		return __folio_lock_or_retry(folio, vmf);
1111 	return 0;
1112 }
1113 
1114 /*
1115  * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
1116  * and should not be used directly.
1117  */
1118 void folio_wait_bit(struct folio *folio, int bit_nr);
1119 int folio_wait_bit_killable(struct folio *folio, int bit_nr);
1120 
1121 /*
1122  * Wait for a folio to be unlocked.
1123  *
1124  * This must be called with the caller "holding" the folio,
1125  * ie with increased folio reference count so that the folio won't
1126  * go away during the wait.
1127  */
folio_wait_locked(struct folio * folio)1128 static inline void folio_wait_locked(struct folio *folio)
1129 {
1130 	if (folio_test_locked(folio))
1131 		folio_wait_bit(folio, PG_locked);
1132 }
1133 
folio_wait_locked_killable(struct folio * folio)1134 static inline int folio_wait_locked_killable(struct folio *folio)
1135 {
1136 	if (!folio_test_locked(folio))
1137 		return 0;
1138 	return folio_wait_bit_killable(folio, PG_locked);
1139 }
1140 
wait_on_page_locked(struct page * page)1141 static inline void wait_on_page_locked(struct page *page)
1142 {
1143 	folio_wait_locked(page_folio(page));
1144 }
1145 
1146 void folio_end_read(struct folio *folio, bool success);
1147 void wait_on_page_writeback(struct page *page);
1148 void folio_wait_writeback(struct folio *folio);
1149 int folio_wait_writeback_killable(struct folio *folio);
1150 void end_page_writeback(struct page *page);
1151 void folio_end_writeback(struct folio *folio);
1152 void wait_for_stable_page(struct page *page);
1153 void folio_wait_stable(struct folio *folio);
1154 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
1155 void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
1156 void __folio_cancel_dirty(struct folio *folio);
folio_cancel_dirty(struct folio * folio)1157 static inline void folio_cancel_dirty(struct folio *folio)
1158 {
1159 	/* Avoid atomic ops, locking, etc. when not actually needed. */
1160 	if (folio_test_dirty(folio))
1161 		__folio_cancel_dirty(folio);
1162 }
1163 bool folio_clear_dirty_for_io(struct folio *folio);
1164 bool clear_page_dirty_for_io(struct page *page);
1165 void folio_invalidate(struct folio *folio, size_t offset, size_t length);
1166 bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
1167 
1168 #ifdef CONFIG_MIGRATION
1169 int filemap_migrate_folio(struct address_space *mapping, struct folio *dst,
1170 		struct folio *src, enum migrate_mode mode);
1171 #else
1172 #define filemap_migrate_folio NULL
1173 #endif
1174 void folio_end_private_2(struct folio *folio);
1175 void folio_wait_private_2(struct folio *folio);
1176 int folio_wait_private_2_killable(struct folio *folio);
1177 
1178 /*
1179  * Add an arbitrary waiter to a page's wait queue
1180  */
1181 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
1182 
1183 /*
1184  * Fault in userspace address range.
1185  */
1186 size_t fault_in_writeable(char __user *uaddr, size_t size);
1187 size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
1188 size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
1189 size_t fault_in_readable(const char __user *uaddr, size_t size);
1190 
1191 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
1192 		pgoff_t index, gfp_t gfp);
1193 int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1194 		pgoff_t index, gfp_t gfp);
1195 void filemap_remove_folio(struct folio *folio);
1196 void __filemap_remove_folio(struct folio *folio, void *shadow);
1197 void replace_page_cache_folio(struct folio *old, struct folio *new);
1198 void delete_from_page_cache_batch(struct address_space *mapping,
1199 				  struct folio_batch *fbatch);
1200 bool filemap_release_folio(struct folio *folio, gfp_t gfp);
1201 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1202 		int whence);
1203 
1204 /* Must be non-static for BPF error injection */
1205 int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1206 		pgoff_t index, gfp_t gfp, void **shadowp);
1207 
1208 bool filemap_range_has_writeback(struct address_space *mapping,
1209 				 loff_t start_byte, loff_t end_byte);
1210 
1211 /**
1212  * filemap_range_needs_writeback - check if range potentially needs writeback
1213  * @mapping:           address space within which to check
1214  * @start_byte:        offset in bytes where the range starts
1215  * @end_byte:          offset in bytes where the range ends (inclusive)
1216  *
1217  * Find at least one page in the range supplied, usually used to check if
1218  * direct writing in this range will trigger a writeback. Used by O_DIRECT
1219  * read/write with IOCB_NOWAIT, to see if the caller needs to do
1220  * filemap_write_and_wait_range() before proceeding.
1221  *
1222  * Return: %true if the caller should do filemap_write_and_wait_range() before
1223  * doing O_DIRECT to a page in this range, %false otherwise.
1224  */
filemap_range_needs_writeback(struct address_space * mapping,loff_t start_byte,loff_t end_byte)1225 static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1226 						 loff_t start_byte,
1227 						 loff_t end_byte)
1228 {
1229 	if (!mapping->nrpages)
1230 		return false;
1231 	if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1232 	    !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1233 		return false;
1234 	return filemap_range_has_writeback(mapping, start_byte, end_byte);
1235 }
1236 
1237 /**
1238  * struct readahead_control - Describes a readahead request.
1239  *
1240  * A readahead request is for consecutive pages.  Filesystems which
1241  * implement the ->readahead method should call readahead_page() or
1242  * readahead_page_batch() in a loop and attempt to start I/O against
1243  * each page in the request.
1244  *
1245  * Most of the fields in this struct are private and should be accessed
1246  * by the functions below.
1247  *
1248  * @file: The file, used primarily by network filesystems for authentication.
1249  *	  May be NULL if invoked internally by the filesystem.
1250  * @mapping: Readahead this filesystem object.
1251  * @ra: File readahead state.  May be NULL.
1252  */
1253 struct readahead_control {
1254 	struct file *file;
1255 	struct address_space *mapping;
1256 	struct file_ra_state *ra;
1257 /* private: use the readahead_* accessors instead */
1258 	pgoff_t _index;
1259 	unsigned int _nr_pages;
1260 	unsigned int _batch_count;
1261 	bool _workingset;
1262 	unsigned long _pflags;
1263 };
1264 
1265 #define DEFINE_READAHEAD(ractl, f, r, m, i)				\
1266 	struct readahead_control ractl = {				\
1267 		.file = f,						\
1268 		.mapping = m,						\
1269 		.ra = r,						\
1270 		._index = i,						\
1271 	}
1272 
1273 #define VM_READAHEAD_PAGES	(SZ_128K / PAGE_SIZE)
1274 
1275 void page_cache_ra_unbounded(struct readahead_control *,
1276 		unsigned long nr_to_read, unsigned long lookahead_count);
1277 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
1278 void page_cache_async_ra(struct readahead_control *, struct folio *,
1279 		unsigned long req_count);
1280 void readahead_expand(struct readahead_control *ractl,
1281 		      loff_t new_start, size_t new_len);
1282 
1283 /**
1284  * page_cache_sync_readahead - generic file readahead
1285  * @mapping: address_space which holds the pagecache and I/O vectors
1286  * @ra: file_ra_state which holds the readahead state
1287  * @file: Used by the filesystem for authentication.
1288  * @index: Index of first page to be read.
1289  * @req_count: Total number of pages being read by the caller.
1290  *
1291  * page_cache_sync_readahead() should be called when a cache miss happened:
1292  * it will submit the read.  The readahead logic may decide to piggyback more
1293  * pages onto the read request if access patterns suggest it will improve
1294  * performance.
1295  */
1296 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)1297 void page_cache_sync_readahead(struct address_space *mapping,
1298 		struct file_ra_state *ra, struct file *file, pgoff_t index,
1299 		unsigned long req_count)
1300 {
1301 	DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1302 	page_cache_sync_ra(&ractl, req_count);
1303 }
1304 
1305 /**
1306  * page_cache_async_readahead - file readahead for marked pages
1307  * @mapping: address_space which holds the pagecache and I/O vectors
1308  * @ra: file_ra_state which holds the readahead state
1309  * @file: Used by the filesystem for authentication.
1310  * @folio: The folio at @index which triggered the readahead call.
1311  * @index: Index of first page to be read.
1312  * @req_count: Total number of pages being read by the caller.
1313  *
1314  * page_cache_async_readahead() should be called when a page is used which
1315  * is marked as PageReadahead; this is a marker to suggest that the application
1316  * has used up enough of the readahead window that we should start pulling in
1317  * more pages.
1318  */
1319 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)1320 void page_cache_async_readahead(struct address_space *mapping,
1321 		struct file_ra_state *ra, struct file *file,
1322 		struct folio *folio, pgoff_t index, unsigned long req_count)
1323 {
1324 	DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1325 	page_cache_async_ra(&ractl, folio, req_count);
1326 }
1327 
__readahead_folio(struct readahead_control * ractl)1328 static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1329 {
1330 	struct folio *folio;
1331 
1332 	BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1333 	ractl->_nr_pages -= ractl->_batch_count;
1334 	ractl->_index += ractl->_batch_count;
1335 
1336 	if (!ractl->_nr_pages) {
1337 		ractl->_batch_count = 0;
1338 		return NULL;
1339 	}
1340 
1341 	folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1342 	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1343 	ractl->_batch_count = folio_nr_pages(folio);
1344 
1345 	return folio;
1346 }
1347 
1348 /**
1349  * readahead_page - Get the next page to read.
1350  * @ractl: The current readahead request.
1351  *
1352  * Context: The page is locked and has an elevated refcount.  The caller
1353  * should decreases the refcount once the page has been submitted for I/O
1354  * and unlock the page once all I/O to that page has completed.
1355  * Return: A pointer to the next page, or %NULL if we are done.
1356  */
readahead_page(struct readahead_control * ractl)1357 static inline struct page *readahead_page(struct readahead_control *ractl)
1358 {
1359 	struct folio *folio = __readahead_folio(ractl);
1360 
1361 	return &folio->page;
1362 }
1363 
1364 /**
1365  * readahead_folio - Get the next folio to read.
1366  * @ractl: The current readahead request.
1367  *
1368  * Context: The folio is locked.  The caller should unlock the folio once
1369  * all I/O to that folio has completed.
1370  * Return: A pointer to the next folio, or %NULL if we are done.
1371  */
readahead_folio(struct readahead_control * ractl)1372 static inline struct folio *readahead_folio(struct readahead_control *ractl)
1373 {
1374 	struct folio *folio = __readahead_folio(ractl);
1375 
1376 	if (folio)
1377 		folio_put(folio);
1378 	return folio;
1379 }
1380 
__readahead_batch(struct readahead_control * rac,struct page ** array,unsigned int array_sz)1381 static inline unsigned int __readahead_batch(struct readahead_control *rac,
1382 		struct page **array, unsigned int array_sz)
1383 {
1384 	unsigned int i = 0;
1385 	XA_STATE(xas, &rac->mapping->i_pages, 0);
1386 	struct page *page;
1387 
1388 	BUG_ON(rac->_batch_count > rac->_nr_pages);
1389 	rac->_nr_pages -= rac->_batch_count;
1390 	rac->_index += rac->_batch_count;
1391 	rac->_batch_count = 0;
1392 
1393 	xas_set(&xas, rac->_index);
1394 	rcu_read_lock();
1395 	xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1396 		if (xas_retry(&xas, page))
1397 			continue;
1398 		VM_BUG_ON_PAGE(!PageLocked(page), page);
1399 		VM_BUG_ON_PAGE(PageTail(page), page);
1400 		array[i++] = page;
1401 		rac->_batch_count += thp_nr_pages(page);
1402 		if (i == array_sz)
1403 			break;
1404 	}
1405 	rcu_read_unlock();
1406 
1407 	return i;
1408 }
1409 
1410 /**
1411  * readahead_page_batch - Get a batch of pages to read.
1412  * @rac: The current readahead request.
1413  * @array: An array of pointers to struct page.
1414  *
1415  * Context: The pages are locked and have an elevated refcount.  The caller
1416  * should decreases the refcount once the page has been submitted for I/O
1417  * and unlock the page once all I/O to that page has completed.
1418  * Return: The number of pages placed in the array.  0 indicates the request
1419  * is complete.
1420  */
1421 #define readahead_page_batch(rac, array)				\
1422 	__readahead_batch(rac, array, ARRAY_SIZE(array))
1423 
1424 /**
1425  * readahead_pos - The byte offset into the file of this readahead request.
1426  * @rac: The readahead request.
1427  */
readahead_pos(struct readahead_control * rac)1428 static inline loff_t readahead_pos(struct readahead_control *rac)
1429 {
1430 	return (loff_t)rac->_index * PAGE_SIZE;
1431 }
1432 
1433 /**
1434  * readahead_length - The number of bytes in this readahead request.
1435  * @rac: The readahead request.
1436  */
readahead_length(struct readahead_control * rac)1437 static inline size_t readahead_length(struct readahead_control *rac)
1438 {
1439 	return rac->_nr_pages * PAGE_SIZE;
1440 }
1441 
1442 /**
1443  * readahead_index - The index of the first page in this readahead request.
1444  * @rac: The readahead request.
1445  */
readahead_index(struct readahead_control * rac)1446 static inline pgoff_t readahead_index(struct readahead_control *rac)
1447 {
1448 	return rac->_index;
1449 }
1450 
1451 /**
1452  * readahead_count - The number of pages in this readahead request.
1453  * @rac: The readahead request.
1454  */
readahead_count(struct readahead_control * rac)1455 static inline unsigned int readahead_count(struct readahead_control *rac)
1456 {
1457 	return rac->_nr_pages;
1458 }
1459 
1460 /**
1461  * readahead_batch_length - The number of bytes in the current batch.
1462  * @rac: The readahead request.
1463  */
readahead_batch_length(struct readahead_control * rac)1464 static inline size_t readahead_batch_length(struct readahead_control *rac)
1465 {
1466 	return rac->_batch_count * PAGE_SIZE;
1467 }
1468 
dir_pages(struct inode * inode)1469 static inline unsigned long dir_pages(struct inode *inode)
1470 {
1471 	return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1472 			       PAGE_SHIFT;
1473 }
1474 
1475 /**
1476  * folio_mkwrite_check_truncate - check if folio was truncated
1477  * @folio: the folio to check
1478  * @inode: the inode to check the folio against
1479  *
1480  * Return: the number of bytes in the folio up to EOF,
1481  * or -EFAULT if the folio was truncated.
1482  */
folio_mkwrite_check_truncate(struct folio * folio,struct inode * inode)1483 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1484 					      struct inode *inode)
1485 {
1486 	loff_t size = i_size_read(inode);
1487 	pgoff_t index = size >> PAGE_SHIFT;
1488 	size_t offset = offset_in_folio(folio, size);
1489 
1490 	if (!folio->mapping)
1491 		return -EFAULT;
1492 
1493 	/* folio is wholly inside EOF */
1494 	if (folio_next_index(folio) - 1 < index)
1495 		return folio_size(folio);
1496 	/* folio is wholly past EOF */
1497 	if (folio->index > index || !offset)
1498 		return -EFAULT;
1499 	/* folio is partially inside EOF */
1500 	return offset;
1501 }
1502 
1503 /**
1504  * page_mkwrite_check_truncate - check if page was truncated
1505  * @page: the page to check
1506  * @inode: the inode to check the page against
1507  *
1508  * Returns the number of bytes in the page up to EOF,
1509  * or -EFAULT if the page was truncated.
1510  */
page_mkwrite_check_truncate(struct page * page,struct inode * inode)1511 static inline int page_mkwrite_check_truncate(struct page *page,
1512 					      struct inode *inode)
1513 {
1514 	loff_t size = i_size_read(inode);
1515 	pgoff_t index = size >> PAGE_SHIFT;
1516 	int offset = offset_in_page(size);
1517 
1518 	if (page->mapping != inode->i_mapping)
1519 		return -EFAULT;
1520 
1521 	/* page is wholly inside EOF */
1522 	if (page->index < index)
1523 		return PAGE_SIZE;
1524 	/* page is wholly past EOF */
1525 	if (page->index > index || !offset)
1526 		return -EFAULT;
1527 	/* page is partially inside EOF */
1528 	return offset;
1529 }
1530 
1531 /**
1532  * i_blocks_per_folio - How many blocks fit in this folio.
1533  * @inode: The inode which contains the blocks.
1534  * @folio: The folio.
1535  *
1536  * If the block size is larger than the size of this folio, return zero.
1537  *
1538  * Context: The caller should hold a refcount on the folio to prevent it
1539  * from being split.
1540  * Return: The number of filesystem blocks covered by this folio.
1541  */
1542 static inline
i_blocks_per_folio(struct inode * inode,struct folio * folio)1543 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1544 {
1545 	return folio_size(folio) >> inode->i_blkbits;
1546 }
1547 
1548 static inline
i_blocks_per_page(struct inode * inode,struct page * page)1549 unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1550 {
1551 	return i_blocks_per_folio(inode, page_folio(page));
1552 }
1553 #endif /* _LINUX_PAGEMAP_H */
1554