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