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 write_inode_now(struct inode *, int sync);
34 int filemap_fdatawrite(struct address_space *);
35 int filemap_flush(struct address_space *);
36 int filemap_fdatawait_keep_errors(struct address_space *mapping);
37 int filemap_fdatawait_range(struct address_space *, loff_t lstart, loff_t lend);
38 int filemap_fdatawait_range_keep_errors(struct address_space *mapping,
39 		loff_t start_byte, loff_t end_byte);
40 
filemap_fdatawait(struct address_space * mapping)41 static inline int filemap_fdatawait(struct address_space *mapping)
42 {
43 	return filemap_fdatawait_range(mapping, 0, LLONG_MAX);
44 }
45 
46 bool filemap_range_has_page(struct address_space *, loff_t lstart, loff_t lend);
47 int filemap_write_and_wait_range(struct address_space *mapping,
48 		loff_t lstart, loff_t lend);
49 int __filemap_fdatawrite_range(struct address_space *mapping,
50 		loff_t start, loff_t end, int sync_mode);
51 int filemap_fdatawrite_range(struct address_space *mapping,
52 		loff_t start, loff_t end);
53 int filemap_check_errors(struct address_space *mapping);
54 void __filemap_set_wb_err(struct address_space *mapping, int err);
55 int filemap_fdatawrite_wbc(struct address_space *mapping,
56 			   struct writeback_control *wbc);
57 
filemap_write_and_wait(struct address_space * mapping)58 static inline int filemap_write_and_wait(struct address_space *mapping)
59 {
60 	return filemap_write_and_wait_range(mapping, 0, LLONG_MAX);
61 }
62 
63 /**
64  * filemap_set_wb_err - set a writeback error on an address_space
65  * @mapping: mapping in which to set writeback error
66  * @err: error to be set in mapping
67  *
68  * When writeback fails in some way, we must record that error so that
69  * userspace can be informed when fsync and the like are called.  We endeavor
70  * to report errors on any file that was open at the time of the error.  Some
71  * internal callers also need to know when writeback errors have occurred.
72  *
73  * When a writeback error occurs, most filesystems will want to call
74  * filemap_set_wb_err to record the error in the mapping so that it will be
75  * automatically reported whenever fsync is called on the file.
76  */
filemap_set_wb_err(struct address_space * mapping,int err)77 static inline void filemap_set_wb_err(struct address_space *mapping, int err)
78 {
79 	/* Fastpath for common case of no error */
80 	if (unlikely(err))
81 		__filemap_set_wb_err(mapping, err);
82 }
83 
84 /**
85  * filemap_check_wb_err - has an error occurred since the mark was sampled?
86  * @mapping: mapping to check for writeback errors
87  * @since: previously-sampled errseq_t
88  *
89  * Grab the errseq_t value from the mapping, and see if it has changed "since"
90  * the given value was sampled.
91  *
92  * If it has then report the latest error set, otherwise return 0.
93  */
filemap_check_wb_err(struct address_space * mapping,errseq_t since)94 static inline int filemap_check_wb_err(struct address_space *mapping,
95 					errseq_t since)
96 {
97 	return errseq_check(&mapping->wb_err, since);
98 }
99 
100 /**
101  * filemap_sample_wb_err - sample the current errseq_t to test for later errors
102  * @mapping: mapping to be sampled
103  *
104  * Writeback errors are always reported relative to a particular sample point
105  * in the past. This function provides those sample points.
106  */
filemap_sample_wb_err(struct address_space * mapping)107 static inline errseq_t filemap_sample_wb_err(struct address_space *mapping)
108 {
109 	return errseq_sample(&mapping->wb_err);
110 }
111 
112 /**
113  * file_sample_sb_err - sample the current errseq_t to test for later errors
114  * @file: file pointer to be sampled
115  *
116  * Grab the most current superblock-level errseq_t value for the given
117  * struct file.
118  */
file_sample_sb_err(struct file * file)119 static inline errseq_t file_sample_sb_err(struct file *file)
120 {
121 	return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err);
122 }
123 
124 /*
125  * Flush file data before changing attributes.  Caller must hold any locks
126  * required to prevent further writes to this file until we're done setting
127  * flags.
128  */
inode_drain_writes(struct inode * inode)129 static inline int inode_drain_writes(struct inode *inode)
130 {
131 	inode_dio_wait(inode);
132 	return filemap_write_and_wait(inode->i_mapping);
133 }
134 
mapping_empty(struct address_space * mapping)135 static inline bool mapping_empty(struct address_space *mapping)
136 {
137 	return xa_empty(&mapping->i_pages);
138 }
139 
140 /*
141  * mapping_shrinkable - test if page cache state allows inode reclaim
142  * @mapping: the page cache mapping
143  *
144  * This checks the mapping's cache state for the pupose of inode
145  * reclaim and LRU management.
146  *
147  * The caller is expected to hold the i_lock, but is not required to
148  * hold the i_pages lock, which usually protects cache state. That's
149  * because the i_lock and the list_lru lock that protect the inode and
150  * its LRU state don't nest inside the irq-safe i_pages lock.
151  *
152  * Cache deletions are performed under the i_lock, which ensures that
153  * when an inode goes empty, it will reliably get queued on the LRU.
154  *
155  * Cache additions do not acquire the i_lock and may race with this
156  * check, in which case we'll report the inode as shrinkable when it
157  * has cache pages. This is okay: the shrinker also checks the
158  * refcount and the referenced bit, which will be elevated or set in
159  * the process of adding new cache pages to an inode.
160  */
mapping_shrinkable(struct address_space * mapping)161 static inline bool mapping_shrinkable(struct address_space *mapping)
162 {
163 	void *head;
164 
165 	/*
166 	 * On highmem systems, there could be lowmem pressure from the
167 	 * inodes before there is highmem pressure from the page
168 	 * cache. Make inodes shrinkable regardless of cache state.
169 	 */
170 	if (IS_ENABLED(CONFIG_HIGHMEM))
171 		return true;
172 
173 	/* Cache completely empty? Shrink away. */
174 	head = rcu_access_pointer(mapping->i_pages.xa_head);
175 	if (!head)
176 		return true;
177 
178 	/*
179 	 * The xarray stores single offset-0 entries directly in the
180 	 * head pointer, which allows non-resident page cache entries
181 	 * to escape the shadow shrinker's list of xarray nodes. The
182 	 * inode shrinker needs to pick them up under memory pressure.
183 	 */
184 	if (!xa_is_node(head) && xa_is_value(head))
185 		return true;
186 
187 	return false;
188 }
189 
190 /*
191  * Bits in mapping->flags.
192  */
193 enum mapping_flags {
194 	AS_EIO		= 0,	/* IO error on async write */
195 	AS_ENOSPC	= 1,	/* ENOSPC on async write */
196 	AS_MM_ALL_LOCKS	= 2,	/* under mm_take_all_locks() */
197 	AS_UNEVICTABLE	= 3,	/* e.g., ramdisk, SHM_LOCK */
198 	AS_EXITING	= 4, 	/* final truncate in progress */
199 	/* writeback related tags are not used */
200 	AS_NO_WRITEBACK_TAGS = 5,
201 	AS_LARGE_FOLIO_SUPPORT = 6,
202 };
203 
204 /**
205  * mapping_set_error - record a writeback error in the address_space
206  * @mapping: the mapping in which an error should be set
207  * @error: the error to set in the mapping
208  *
209  * When writeback fails in some way, we must record that error so that
210  * userspace can be informed when fsync and the like are called.  We endeavor
211  * to report errors on any file that was open at the time of the error.  Some
212  * internal callers also need to know when writeback errors have occurred.
213  *
214  * When a writeback error occurs, most filesystems will want to call
215  * mapping_set_error to record the error in the mapping so that it can be
216  * reported when the application calls fsync(2).
217  */
mapping_set_error(struct address_space * mapping,int error)218 static inline void mapping_set_error(struct address_space *mapping, int error)
219 {
220 	if (likely(!error))
221 		return;
222 
223 	/* Record in wb_err for checkers using errseq_t based tracking */
224 	__filemap_set_wb_err(mapping, error);
225 
226 	/* Record it in superblock */
227 	if (mapping->host)
228 		errseq_set(&mapping->host->i_sb->s_wb_err, error);
229 
230 	/* Record it in flags for now, for legacy callers */
231 	if (error == -ENOSPC)
232 		set_bit(AS_ENOSPC, &mapping->flags);
233 	else
234 		set_bit(AS_EIO, &mapping->flags);
235 }
236 
mapping_set_unevictable(struct address_space * mapping)237 static inline void mapping_set_unevictable(struct address_space *mapping)
238 {
239 	set_bit(AS_UNEVICTABLE, &mapping->flags);
240 }
241 
mapping_clear_unevictable(struct address_space * mapping)242 static inline void mapping_clear_unevictable(struct address_space *mapping)
243 {
244 	clear_bit(AS_UNEVICTABLE, &mapping->flags);
245 }
246 
mapping_unevictable(struct address_space * mapping)247 static inline bool mapping_unevictable(struct address_space *mapping)
248 {
249 	return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
250 }
251 
mapping_set_exiting(struct address_space * mapping)252 static inline void mapping_set_exiting(struct address_space *mapping)
253 {
254 	set_bit(AS_EXITING, &mapping->flags);
255 }
256 
mapping_exiting(struct address_space * mapping)257 static inline int mapping_exiting(struct address_space *mapping)
258 {
259 	return test_bit(AS_EXITING, &mapping->flags);
260 }
261 
mapping_set_no_writeback_tags(struct address_space * mapping)262 static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
263 {
264 	set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
265 }
266 
mapping_use_writeback_tags(struct address_space * mapping)267 static inline int mapping_use_writeback_tags(struct address_space *mapping)
268 {
269 	return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
270 }
271 
mapping_gfp_mask(struct address_space * mapping)272 static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
273 {
274 	return mapping->gfp_mask;
275 }
276 
277 /* Restricts the given gfp_mask to what the mapping allows. */
mapping_gfp_constraint(struct address_space * mapping,gfp_t gfp_mask)278 static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
279 		gfp_t gfp_mask)
280 {
281 	return mapping_gfp_mask(mapping) & gfp_mask;
282 }
283 
284 /*
285  * This is non-atomic.  Only to be used before the mapping is activated.
286  * Probably needs a barrier...
287  */
mapping_set_gfp_mask(struct address_space * m,gfp_t mask)288 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
289 {
290 	m->gfp_mask = mask;
291 }
292 
293 /**
294  * mapping_set_large_folios() - Indicate the file supports large folios.
295  * @mapping: The file.
296  *
297  * The filesystem should call this function in its inode constructor to
298  * indicate that the VFS can use large folios to cache the contents of
299  * the file.
300  *
301  * Context: This should not be called while the inode is active as it
302  * is non-atomic.
303  */
mapping_set_large_folios(struct address_space * mapping)304 static inline void mapping_set_large_folios(struct address_space *mapping)
305 {
306 	__set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
307 }
308 
309 /*
310  * Large folio support currently depends on THP.  These dependencies are
311  * being worked on but are not yet fixed.
312  */
mapping_large_folio_support(struct address_space * mapping)313 static inline bool mapping_large_folio_support(struct address_space *mapping)
314 {
315 	return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) &&
316 		test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
317 }
318 
filemap_nr_thps(struct address_space * mapping)319 static inline int filemap_nr_thps(struct address_space *mapping)
320 {
321 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
322 	return atomic_read(&mapping->nr_thps);
323 #else
324 	return 0;
325 #endif
326 }
327 
filemap_nr_thps_inc(struct address_space * mapping)328 static inline void filemap_nr_thps_inc(struct address_space *mapping)
329 {
330 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
331 	if (!mapping_large_folio_support(mapping))
332 		atomic_inc(&mapping->nr_thps);
333 #else
334 	WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
335 #endif
336 }
337 
filemap_nr_thps_dec(struct address_space * mapping)338 static inline void filemap_nr_thps_dec(struct address_space *mapping)
339 {
340 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
341 	if (!mapping_large_folio_support(mapping))
342 		atomic_dec(&mapping->nr_thps);
343 #else
344 	WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
345 #endif
346 }
347 
348 void release_pages(struct page **pages, int nr);
349 
350 struct address_space *page_mapping(struct page *);
351 struct address_space *folio_mapping(struct folio *);
352 struct address_space *swapcache_mapping(struct folio *);
353 
354 /**
355  * folio_file_mapping - Find the mapping this folio belongs to.
356  * @folio: The folio.
357  *
358  * For folios which are in the page cache, return the mapping that this
359  * page belongs to.  Folios in the swap cache return the mapping of the
360  * swap file or swap device where the data is stored.  This is different
361  * from the mapping returned by folio_mapping().  The only reason to
362  * use it is if, like NFS, you return 0 from ->activate_swapfile.
363  *
364  * Do not call this for folios which aren't in the page cache or swap cache.
365  */
folio_file_mapping(struct folio * folio)366 static inline struct address_space *folio_file_mapping(struct folio *folio)
367 {
368 	if (unlikely(folio_test_swapcache(folio)))
369 		return swapcache_mapping(folio);
370 
371 	return folio->mapping;
372 }
373 
page_file_mapping(struct page * page)374 static inline struct address_space *page_file_mapping(struct page *page)
375 {
376 	return folio_file_mapping(page_folio(page));
377 }
378 
379 /*
380  * For file cache pages, return the address_space, otherwise return NULL
381  */
page_mapping_file(struct page * page)382 static inline struct address_space *page_mapping_file(struct page *page)
383 {
384 	struct folio *folio = page_folio(page);
385 
386 	if (unlikely(folio_test_swapcache(folio)))
387 		return NULL;
388 	return folio_mapping(folio);
389 }
390 
391 /**
392  * folio_inode - Get the host inode for this folio.
393  * @folio: The folio.
394  *
395  * For folios which are in the page cache, return the inode that this folio
396  * belongs to.
397  *
398  * Do not call this for folios which aren't in the page cache.
399  */
folio_inode(struct folio * folio)400 static inline struct inode *folio_inode(struct folio *folio)
401 {
402 	return folio->mapping->host;
403 }
404 
405 /**
406  * folio_attach_private - Attach private data to a folio.
407  * @folio: Folio to attach data to.
408  * @data: Data to attach to folio.
409  *
410  * Attaching private data to a folio increments the page's reference count.
411  * The data must be detached before the folio will be freed.
412  */
folio_attach_private(struct folio * folio,void * data)413 static inline void folio_attach_private(struct folio *folio, void *data)
414 {
415 	folio_get(folio);
416 	folio->private = data;
417 	folio_set_private(folio);
418 }
419 
420 /**
421  * folio_change_private - Change private data on a folio.
422  * @folio: Folio to change the data on.
423  * @data: Data to set on the folio.
424  *
425  * Change the private data attached to a folio and return the old
426  * data.  The page must previously have had data attached and the data
427  * must be detached before the folio will be freed.
428  *
429  * Return: Data that was previously attached to the folio.
430  */
folio_change_private(struct folio * folio,void * data)431 static inline void *folio_change_private(struct folio *folio, void *data)
432 {
433 	void *old = folio_get_private(folio);
434 
435 	folio->private = data;
436 	return old;
437 }
438 
439 /**
440  * folio_detach_private - Detach private data from a folio.
441  * @folio: Folio to detach data from.
442  *
443  * Removes the data that was previously attached to the folio and decrements
444  * the refcount on the page.
445  *
446  * Return: Data that was attached to the folio.
447  */
folio_detach_private(struct folio * folio)448 static inline void *folio_detach_private(struct folio *folio)
449 {
450 	void *data = folio_get_private(folio);
451 
452 	if (!folio_test_private(folio))
453 		return NULL;
454 	folio_clear_private(folio);
455 	folio->private = NULL;
456 	folio_put(folio);
457 
458 	return data;
459 }
460 
attach_page_private(struct page * page,void * data)461 static inline void attach_page_private(struct page *page, void *data)
462 {
463 	folio_attach_private(page_folio(page), data);
464 }
465 
detach_page_private(struct page * page)466 static inline void *detach_page_private(struct page *page)
467 {
468 	return folio_detach_private(page_folio(page));
469 }
470 
471 #ifdef CONFIG_NUMA
472 struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order);
473 #else
filemap_alloc_folio(gfp_t gfp,unsigned int order)474 static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order)
475 {
476 	return folio_alloc(gfp, order);
477 }
478 #endif
479 
__page_cache_alloc(gfp_t gfp)480 static inline struct page *__page_cache_alloc(gfp_t gfp)
481 {
482 	return &filemap_alloc_folio(gfp, 0)->page;
483 }
484 
page_cache_alloc(struct address_space * x)485 static inline struct page *page_cache_alloc(struct address_space *x)
486 {
487 	return __page_cache_alloc(mapping_gfp_mask(x));
488 }
489 
readahead_gfp_mask(struct address_space * x)490 static inline gfp_t readahead_gfp_mask(struct address_space *x)
491 {
492 	return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
493 }
494 
495 typedef int filler_t(struct file *, struct folio *);
496 
497 pgoff_t page_cache_next_miss(struct address_space *mapping,
498 			     pgoff_t index, unsigned long max_scan);
499 pgoff_t page_cache_prev_miss(struct address_space *mapping,
500 			     pgoff_t index, unsigned long max_scan);
501 
502 #define FGP_ACCESSED		0x00000001
503 #define FGP_LOCK		0x00000002
504 #define FGP_CREAT		0x00000004
505 #define FGP_WRITE		0x00000008
506 #define FGP_NOFS		0x00000010
507 #define FGP_NOWAIT		0x00000020
508 #define FGP_FOR_MMAP		0x00000040
509 #define FGP_HEAD		0x00000080
510 #define FGP_ENTRY		0x00000100
511 #define FGP_STABLE		0x00000200
512 
513 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
514 		int fgp_flags, gfp_t gfp);
515 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
516 		int fgp_flags, gfp_t gfp);
517 
518 /**
519  * filemap_get_folio - Find and get a folio.
520  * @mapping: The address_space to search.
521  * @index: The page index.
522  *
523  * Looks up the page cache entry at @mapping & @index.  If a folio is
524  * present, it is returned with an increased refcount.
525  *
526  * Otherwise, %NULL is returned.
527  */
filemap_get_folio(struct address_space * mapping,pgoff_t index)528 static inline struct folio *filemap_get_folio(struct address_space *mapping,
529 					pgoff_t index)
530 {
531 	return __filemap_get_folio(mapping, index, 0, 0);
532 }
533 
534 /**
535  * filemap_lock_folio - Find and lock a folio.
536  * @mapping: The address_space to search.
537  * @index: The page index.
538  *
539  * Looks up the page cache entry at @mapping & @index.  If a folio is
540  * present, it is returned locked with an increased refcount.
541  *
542  * Context: May sleep.
543  * Return: A folio or %NULL if there is no folio in the cache for this
544  * index.  Will not return a shadow, swap or DAX entry.
545  */
filemap_lock_folio(struct address_space * mapping,pgoff_t index)546 static inline struct folio *filemap_lock_folio(struct address_space *mapping,
547 					pgoff_t index)
548 {
549 	return __filemap_get_folio(mapping, index, FGP_LOCK, 0);
550 }
551 
552 /**
553  * find_get_page - find and get a page reference
554  * @mapping: the address_space to search
555  * @offset: the page index
556  *
557  * Looks up the page cache slot at @mapping & @offset.  If there is a
558  * page cache page, it is returned with an increased refcount.
559  *
560  * Otherwise, %NULL is returned.
561  */
find_get_page(struct address_space * mapping,pgoff_t offset)562 static inline struct page *find_get_page(struct address_space *mapping,
563 					pgoff_t offset)
564 {
565 	return pagecache_get_page(mapping, offset, 0, 0);
566 }
567 
find_get_page_flags(struct address_space * mapping,pgoff_t offset,int fgp_flags)568 static inline struct page *find_get_page_flags(struct address_space *mapping,
569 					pgoff_t offset, int fgp_flags)
570 {
571 	return pagecache_get_page(mapping, offset, fgp_flags, 0);
572 }
573 
574 /**
575  * find_lock_page - locate, pin and lock a pagecache page
576  * @mapping: the address_space to search
577  * @index: the page index
578  *
579  * Looks up the page cache entry at @mapping & @index.  If there is a
580  * page cache page, it is returned locked and with an increased
581  * refcount.
582  *
583  * Context: May sleep.
584  * Return: A struct page or %NULL if there is no page in the cache for this
585  * index.
586  */
find_lock_page(struct address_space * mapping,pgoff_t index)587 static inline struct page *find_lock_page(struct address_space *mapping,
588 					pgoff_t index)
589 {
590 	return pagecache_get_page(mapping, index, FGP_LOCK, 0);
591 }
592 
593 /**
594  * find_or_create_page - locate or add a pagecache page
595  * @mapping: the page's address_space
596  * @index: the page's index into the mapping
597  * @gfp_mask: page allocation mode
598  *
599  * Looks up the page cache slot at @mapping & @offset.  If there is a
600  * page cache page, it is returned locked and with an increased
601  * refcount.
602  *
603  * If the page is not present, a new page is allocated using @gfp_mask
604  * and added to the page cache and the VM's LRU list.  The page is
605  * returned locked and with an increased refcount.
606  *
607  * On memory exhaustion, %NULL is returned.
608  *
609  * find_or_create_page() may sleep, even if @gfp_flags specifies an
610  * atomic allocation!
611  */
find_or_create_page(struct address_space * mapping,pgoff_t index,gfp_t gfp_mask)612 static inline struct page *find_or_create_page(struct address_space *mapping,
613 					pgoff_t index, gfp_t gfp_mask)
614 {
615 	return pagecache_get_page(mapping, index,
616 					FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
617 					gfp_mask);
618 }
619 
620 /**
621  * grab_cache_page_nowait - returns locked page at given index in given cache
622  * @mapping: target address_space
623  * @index: the page index
624  *
625  * Same as grab_cache_page(), but do not wait if the page is unavailable.
626  * This is intended for speculative data generators, where the data can
627  * be regenerated if the page couldn't be grabbed.  This routine should
628  * be safe to call while holding the lock for another page.
629  *
630  * Clear __GFP_FS when allocating the page to avoid recursion into the fs
631  * and deadlock against the caller's locked page.
632  */
grab_cache_page_nowait(struct address_space * mapping,pgoff_t index)633 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
634 				pgoff_t index)
635 {
636 	return pagecache_get_page(mapping, index,
637 			FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
638 			mapping_gfp_mask(mapping));
639 }
640 
641 #define swapcache_index(folio)	__page_file_index(&(folio)->page)
642 
643 /**
644  * folio_index - File index of a folio.
645  * @folio: The folio.
646  *
647  * For a folio which is either in the page cache or the swap cache,
648  * return its index within the address_space it belongs to.  If you know
649  * the page is definitely in the page cache, you can look at the folio's
650  * index directly.
651  *
652  * Return: The index (offset in units of pages) of a folio in its file.
653  */
folio_index(struct folio * folio)654 static inline pgoff_t folio_index(struct folio *folio)
655 {
656         if (unlikely(folio_test_swapcache(folio)))
657                 return swapcache_index(folio);
658         return folio->index;
659 }
660 
661 /**
662  * folio_next_index - Get the index of the next folio.
663  * @folio: The current folio.
664  *
665  * Return: The index of the folio which follows this folio in the file.
666  */
folio_next_index(struct folio * folio)667 static inline pgoff_t folio_next_index(struct folio *folio)
668 {
669 	return folio->index + folio_nr_pages(folio);
670 }
671 
672 /**
673  * folio_file_page - The page for a particular index.
674  * @folio: The folio which contains this index.
675  * @index: The index we want to look up.
676  *
677  * Sometimes after looking up a folio in the page cache, we need to
678  * obtain the specific page for an index (eg a page fault).
679  *
680  * Return: The page containing the file data for this index.
681  */
folio_file_page(struct folio * folio,pgoff_t index)682 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
683 {
684 	/* HugeTLBfs indexes the page cache in units of hpage_size */
685 	if (folio_test_hugetlb(folio))
686 		return &folio->page;
687 	return folio_page(folio, index & (folio_nr_pages(folio) - 1));
688 }
689 
690 /**
691  * folio_contains - Does this folio contain this index?
692  * @folio: The folio.
693  * @index: The page index within the file.
694  *
695  * Context: The caller should have the page locked in order to prevent
696  * (eg) shmem from moving the page between the page cache and swap cache
697  * and changing its index in the middle of the operation.
698  * Return: true or false.
699  */
folio_contains(struct folio * folio,pgoff_t index)700 static inline bool folio_contains(struct folio *folio, pgoff_t index)
701 {
702 	/* HugeTLBfs indexes the page cache in units of hpage_size */
703 	if (folio_test_hugetlb(folio))
704 		return folio->index == index;
705 	return index - folio_index(folio) < folio_nr_pages(folio);
706 }
707 
708 /*
709  * Given the page we found in the page cache, return the page corresponding
710  * to this index in the file
711  */
find_subpage(struct page * head,pgoff_t index)712 static inline struct page *find_subpage(struct page *head, pgoff_t index)
713 {
714 	/* HugeTLBfs wants the head page regardless */
715 	if (PageHuge(head))
716 		return head;
717 
718 	return head + (index & (thp_nr_pages(head) - 1));
719 }
720 
721 unsigned find_get_pages_range(struct address_space *mapping, pgoff_t *start,
722 			pgoff_t end, unsigned int nr_pages,
723 			struct page **pages);
724 unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t start,
725 			       unsigned int nr_pages, struct page **pages);
726 unsigned find_get_pages_range_tag(struct address_space *mapping, pgoff_t *index,
727 			pgoff_t end, xa_mark_t tag, unsigned int nr_pages,
728 			struct page **pages);
find_get_pages_tag(struct address_space * mapping,pgoff_t * index,xa_mark_t tag,unsigned int nr_pages,struct page ** pages)729 static inline unsigned find_get_pages_tag(struct address_space *mapping,
730 			pgoff_t *index, xa_mark_t tag, unsigned int nr_pages,
731 			struct page **pages)
732 {
733 	return find_get_pages_range_tag(mapping, index, (pgoff_t)-1, tag,
734 					nr_pages, pages);
735 }
736 
737 struct page *grab_cache_page_write_begin(struct address_space *mapping,
738 			pgoff_t index);
739 
740 /*
741  * Returns locked page at given index in given cache, creating it if needed.
742  */
grab_cache_page(struct address_space * mapping,pgoff_t index)743 static inline struct page *grab_cache_page(struct address_space *mapping,
744 								pgoff_t index)
745 {
746 	return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
747 }
748 
749 struct folio *read_cache_folio(struct address_space *, pgoff_t index,
750 		filler_t *filler, struct file *file);
751 struct page *read_cache_page(struct address_space *, pgoff_t index,
752 		filler_t *filler, struct file *file);
753 extern struct page * read_cache_page_gfp(struct address_space *mapping,
754 				pgoff_t index, gfp_t gfp_mask);
755 
read_mapping_page(struct address_space * mapping,pgoff_t index,struct file * file)756 static inline struct page *read_mapping_page(struct address_space *mapping,
757 				pgoff_t index, struct file *file)
758 {
759 	return read_cache_page(mapping, index, NULL, file);
760 }
761 
read_mapping_folio(struct address_space * mapping,pgoff_t index,struct file * file)762 static inline struct folio *read_mapping_folio(struct address_space *mapping,
763 				pgoff_t index, struct file *file)
764 {
765 	return read_cache_folio(mapping, index, NULL, file);
766 }
767 
768 /*
769  * Get index of the page within radix-tree (but not for hugetlb pages).
770  * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
771  */
page_to_index(struct page * page)772 static inline pgoff_t page_to_index(struct page *page)
773 {
774 	struct page *head;
775 
776 	if (likely(!PageTransTail(page)))
777 		return page->index;
778 
779 	head = compound_head(page);
780 	/*
781 	 *  We don't initialize ->index for tail pages: calculate based on
782 	 *  head page
783 	 */
784 	return head->index + page - head;
785 }
786 
787 extern pgoff_t hugetlb_basepage_index(struct page *page);
788 
789 /*
790  * Get the offset in PAGE_SIZE (even for hugetlb pages).
791  * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
792  */
page_to_pgoff(struct page * page)793 static inline pgoff_t page_to_pgoff(struct page *page)
794 {
795 	if (unlikely(PageHuge(page)))
796 		return hugetlb_basepage_index(page);
797 	return page_to_index(page);
798 }
799 
800 /*
801  * Return byte-offset into filesystem object for page.
802  */
page_offset(struct page * page)803 static inline loff_t page_offset(struct page *page)
804 {
805 	return ((loff_t)page->index) << PAGE_SHIFT;
806 }
807 
page_file_offset(struct page * page)808 static inline loff_t page_file_offset(struct page *page)
809 {
810 	return ((loff_t)page_index(page)) << PAGE_SHIFT;
811 }
812 
813 /**
814  * folio_pos - Returns the byte position of this folio in its file.
815  * @folio: The folio.
816  */
folio_pos(struct folio * folio)817 static inline loff_t folio_pos(struct folio *folio)
818 {
819 	return page_offset(&folio->page);
820 }
821 
822 /**
823  * folio_file_pos - Returns the byte position of this folio in its file.
824  * @folio: The folio.
825  *
826  * This differs from folio_pos() for folios which belong to a swap file.
827  * NFS is the only filesystem today which needs to use folio_file_pos().
828  */
folio_file_pos(struct folio * folio)829 static inline loff_t folio_file_pos(struct folio *folio)
830 {
831 	return page_file_offset(&folio->page);
832 }
833 
834 /*
835  * Get the offset in PAGE_SIZE (even for hugetlb folios).
836  * (TODO: hugetlb folios should have ->index in PAGE_SIZE)
837  */
folio_pgoff(struct folio * folio)838 static inline pgoff_t folio_pgoff(struct folio *folio)
839 {
840 	if (unlikely(folio_test_hugetlb(folio)))
841 		return hugetlb_basepage_index(&folio->page);
842 	return folio->index;
843 }
844 
845 extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
846 				     unsigned long address);
847 
linear_page_index(struct vm_area_struct * vma,unsigned long address)848 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
849 					unsigned long address)
850 {
851 	pgoff_t pgoff;
852 	if (unlikely(is_vm_hugetlb_page(vma)))
853 		return linear_hugepage_index(vma, address);
854 	pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
855 	pgoff += vma->vm_pgoff;
856 	return pgoff;
857 }
858 
859 struct wait_page_key {
860 	struct folio *folio;
861 	int bit_nr;
862 	int page_match;
863 };
864 
865 struct wait_page_queue {
866 	struct folio *folio;
867 	int bit_nr;
868 	wait_queue_entry_t wait;
869 };
870 
wake_page_match(struct wait_page_queue * wait_page,struct wait_page_key * key)871 static inline bool wake_page_match(struct wait_page_queue *wait_page,
872 				  struct wait_page_key *key)
873 {
874 	if (wait_page->folio != key->folio)
875 	       return false;
876 	key->page_match = 1;
877 
878 	if (wait_page->bit_nr != key->bit_nr)
879 		return false;
880 
881 	return true;
882 }
883 
884 void __folio_lock(struct folio *folio);
885 int __folio_lock_killable(struct folio *folio);
886 bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
887 				unsigned int flags);
888 void unlock_page(struct page *page);
889 void folio_unlock(struct folio *folio);
890 
891 /**
892  * folio_trylock() - Attempt to lock a folio.
893  * @folio: The folio to attempt to lock.
894  *
895  * Sometimes it is undesirable to wait for a folio to be unlocked (eg
896  * when the locks are being taken in the wrong order, or if making
897  * progress through a batch of folios is more important than processing
898  * them in order).  Usually folio_lock() is the correct function to call.
899  *
900  * Context: Any context.
901  * Return: Whether the lock was successfully acquired.
902  */
folio_trylock(struct folio * folio)903 static inline bool folio_trylock(struct folio *folio)
904 {
905 	return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
906 }
907 
908 /*
909  * Return true if the page was successfully locked
910  */
trylock_page(struct page * page)911 static inline int trylock_page(struct page *page)
912 {
913 	return folio_trylock(page_folio(page));
914 }
915 
916 /**
917  * folio_lock() - Lock this folio.
918  * @folio: The folio to lock.
919  *
920  * The folio lock protects against many things, probably more than it
921  * should.  It is primarily held while a folio is being brought uptodate,
922  * either from its backing file or from swap.  It is also held while a
923  * folio is being truncated from its address_space, so holding the lock
924  * is sufficient to keep folio->mapping stable.
925  *
926  * The folio lock is also held while write() is modifying the page to
927  * provide POSIX atomicity guarantees (as long as the write does not
928  * cross a page boundary).  Other modifications to the data in the folio
929  * do not hold the folio lock and can race with writes, eg DMA and stores
930  * to mapped pages.
931  *
932  * Context: May sleep.  If you need to acquire the locks of two or
933  * more folios, they must be in order of ascending index, if they are
934  * in the same address_space.  If they are in different address_spaces,
935  * acquire the lock of the folio which belongs to the address_space which
936  * has the lowest address in memory first.
937  */
folio_lock(struct folio * folio)938 static inline void folio_lock(struct folio *folio)
939 {
940 	might_sleep();
941 	if (!folio_trylock(folio))
942 		__folio_lock(folio);
943 }
944 
945 /**
946  * lock_page() - Lock the folio containing this page.
947  * @page: The page to lock.
948  *
949  * See folio_lock() for a description of what the lock protects.
950  * This is a legacy function and new code should probably use folio_lock()
951  * instead.
952  *
953  * Context: May sleep.  Pages in the same folio share a lock, so do not
954  * attempt to lock two pages which share a folio.
955  */
lock_page(struct page * page)956 static inline void lock_page(struct page *page)
957 {
958 	struct folio *folio;
959 	might_sleep();
960 
961 	folio = page_folio(page);
962 	if (!folio_trylock(folio))
963 		__folio_lock(folio);
964 }
965 
966 /**
967  * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
968  * @folio: The folio to lock.
969  *
970  * Attempts to lock the folio, like folio_lock(), except that the sleep
971  * to acquire the lock is interruptible by a fatal signal.
972  *
973  * Context: May sleep; see folio_lock().
974  * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
975  */
folio_lock_killable(struct folio * folio)976 static inline int folio_lock_killable(struct folio *folio)
977 {
978 	might_sleep();
979 	if (!folio_trylock(folio))
980 		return __folio_lock_killable(folio);
981 	return 0;
982 }
983 
984 /*
985  * lock_page_killable is like lock_page but can be interrupted by fatal
986  * signals.  It returns 0 if it locked the page and -EINTR if it was
987  * killed while waiting.
988  */
lock_page_killable(struct page * page)989 static inline int lock_page_killable(struct page *page)
990 {
991 	return folio_lock_killable(page_folio(page));
992 }
993 
994 /*
995  * lock_page_or_retry - Lock the page, unless this would block and the
996  * caller indicated that it can handle a retry.
997  *
998  * Return value and mmap_lock implications depend on flags; see
999  * __folio_lock_or_retry().
1000  */
lock_page_or_retry(struct page * page,struct mm_struct * mm,unsigned int flags)1001 static inline bool lock_page_or_retry(struct page *page, struct mm_struct *mm,
1002 				     unsigned int flags)
1003 {
1004 	struct folio *folio;
1005 	might_sleep();
1006 
1007 	folio = page_folio(page);
1008 	return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
1009 }
1010 
1011 /*
1012  * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
1013  * and should not be used directly.
1014  */
1015 void folio_wait_bit(struct folio *folio, int bit_nr);
1016 int folio_wait_bit_killable(struct folio *folio, int bit_nr);
1017 
1018 /*
1019  * Wait for a folio to be unlocked.
1020  *
1021  * This must be called with the caller "holding" the folio,
1022  * ie with increased folio reference count so that the folio won't
1023  * go away during the wait.
1024  */
folio_wait_locked(struct folio * folio)1025 static inline void folio_wait_locked(struct folio *folio)
1026 {
1027 	if (folio_test_locked(folio))
1028 		folio_wait_bit(folio, PG_locked);
1029 }
1030 
folio_wait_locked_killable(struct folio * folio)1031 static inline int folio_wait_locked_killable(struct folio *folio)
1032 {
1033 	if (!folio_test_locked(folio))
1034 		return 0;
1035 	return folio_wait_bit_killable(folio, PG_locked);
1036 }
1037 
wait_on_page_locked(struct page * page)1038 static inline void wait_on_page_locked(struct page *page)
1039 {
1040 	folio_wait_locked(page_folio(page));
1041 }
1042 
wait_on_page_locked_killable(struct page * page)1043 static inline int wait_on_page_locked_killable(struct page *page)
1044 {
1045 	return folio_wait_locked_killable(page_folio(page));
1046 }
1047 
1048 int folio_put_wait_locked(struct folio *folio, int state);
1049 void wait_on_page_writeback(struct page *page);
1050 void folio_wait_writeback(struct folio *folio);
1051 int folio_wait_writeback_killable(struct folio *folio);
1052 void end_page_writeback(struct page *page);
1053 void folio_end_writeback(struct folio *folio);
1054 void wait_for_stable_page(struct page *page);
1055 void folio_wait_stable(struct folio *folio);
1056 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
__set_page_dirty(struct page * page,struct address_space * mapping,int warn)1057 static inline void __set_page_dirty(struct page *page,
1058 		struct address_space *mapping, int warn)
1059 {
1060 	__folio_mark_dirty(page_folio(page), mapping, warn);
1061 }
1062 void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
1063 void __folio_cancel_dirty(struct folio *folio);
folio_cancel_dirty(struct folio * folio)1064 static inline void folio_cancel_dirty(struct folio *folio)
1065 {
1066 	/* Avoid atomic ops, locking, etc. when not actually needed. */
1067 	if (folio_test_dirty(folio))
1068 		__folio_cancel_dirty(folio);
1069 }
1070 bool folio_clear_dirty_for_io(struct folio *folio);
1071 bool clear_page_dirty_for_io(struct page *page);
1072 void folio_invalidate(struct folio *folio, size_t offset, size_t length);
1073 int __must_check folio_write_one(struct folio *folio);
write_one_page(struct page * page)1074 static inline int __must_check write_one_page(struct page *page)
1075 {
1076 	return folio_write_one(page_folio(page));
1077 }
1078 
1079 int __set_page_dirty_nobuffers(struct page *page);
1080 bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
1081 
1082 void page_endio(struct page *page, bool is_write, int err);
1083 
1084 void folio_end_private_2(struct folio *folio);
1085 void folio_wait_private_2(struct folio *folio);
1086 int folio_wait_private_2_killable(struct folio *folio);
1087 
1088 /*
1089  * Add an arbitrary waiter to a page's wait queue
1090  */
1091 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
1092 
1093 /*
1094  * Fault in userspace address range.
1095  */
1096 size_t fault_in_writeable(char __user *uaddr, size_t size);
1097 size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
1098 size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
1099 size_t fault_in_readable(const char __user *uaddr, size_t size);
1100 
1101 int add_to_page_cache_locked(struct page *page, struct address_space *mapping,
1102 		pgoff_t index, gfp_t gfp);
1103 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
1104 		pgoff_t index, gfp_t gfp);
1105 int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1106 		pgoff_t index, gfp_t gfp);
1107 void filemap_remove_folio(struct folio *folio);
1108 void delete_from_page_cache(struct page *page);
1109 void __filemap_remove_folio(struct folio *folio, void *shadow);
__delete_from_page_cache(struct page * page,void * shadow)1110 static inline void __delete_from_page_cache(struct page *page, void *shadow)
1111 {
1112 	__filemap_remove_folio(page_folio(page), shadow);
1113 }
1114 void replace_page_cache_page(struct page *old, struct page *new);
1115 void delete_from_page_cache_batch(struct address_space *mapping,
1116 				  struct folio_batch *fbatch);
1117 int try_to_release_page(struct page *page, gfp_t gfp);
1118 bool filemap_release_folio(struct folio *folio, gfp_t gfp);
1119 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1120 		int whence);
1121 
1122 /*
1123  * Like add_to_page_cache_locked, but used to add newly allocated pages:
1124  * the page is new, so we can just run __SetPageLocked() against it.
1125  */
add_to_page_cache(struct page * page,struct address_space * mapping,pgoff_t offset,gfp_t gfp_mask)1126 static inline int add_to_page_cache(struct page *page,
1127 		struct address_space *mapping, pgoff_t offset, gfp_t gfp_mask)
1128 {
1129 	int error;
1130 
1131 	__SetPageLocked(page);
1132 	error = add_to_page_cache_locked(page, mapping, offset, gfp_mask);
1133 	if (unlikely(error))
1134 		__ClearPageLocked(page);
1135 	return error;
1136 }
1137 
1138 /* Must be non-static for BPF error injection */
1139 int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1140 		pgoff_t index, gfp_t gfp, void **shadowp);
1141 
1142 bool filemap_range_has_writeback(struct address_space *mapping,
1143 				 loff_t start_byte, loff_t end_byte);
1144 
1145 /**
1146  * filemap_range_needs_writeback - check if range potentially needs writeback
1147  * @mapping:           address space within which to check
1148  * @start_byte:        offset in bytes where the range starts
1149  * @end_byte:          offset in bytes where the range ends (inclusive)
1150  *
1151  * Find at least one page in the range supplied, usually used to check if
1152  * direct writing in this range will trigger a writeback. Used by O_DIRECT
1153  * read/write with IOCB_NOWAIT, to see if the caller needs to do
1154  * filemap_write_and_wait_range() before proceeding.
1155  *
1156  * Return: %true if the caller should do filemap_write_and_wait_range() before
1157  * doing O_DIRECT to a page in this range, %false otherwise.
1158  */
filemap_range_needs_writeback(struct address_space * mapping,loff_t start_byte,loff_t end_byte)1159 static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1160 						 loff_t start_byte,
1161 						 loff_t end_byte)
1162 {
1163 	if (!mapping->nrpages)
1164 		return false;
1165 	if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1166 	    !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1167 		return false;
1168 	return filemap_range_has_writeback(mapping, start_byte, end_byte);
1169 }
1170 
1171 /**
1172  * struct readahead_control - Describes a readahead request.
1173  *
1174  * A readahead request is for consecutive pages.  Filesystems which
1175  * implement the ->readahead method should call readahead_page() or
1176  * readahead_page_batch() in a loop and attempt to start I/O against
1177  * each page in the request.
1178  *
1179  * Most of the fields in this struct are private and should be accessed
1180  * by the functions below.
1181  *
1182  * @file: The file, used primarily by network filesystems for authentication.
1183  *	  May be NULL if invoked internally by the filesystem.
1184  * @mapping: Readahead this filesystem object.
1185  * @ra: File readahead state.  May be NULL.
1186  */
1187 struct readahead_control {
1188 	struct file *file;
1189 	struct address_space *mapping;
1190 	struct file_ra_state *ra;
1191 /* private: use the readahead_* accessors instead */
1192 	pgoff_t _index;
1193 	unsigned int _nr_pages;
1194 	unsigned int _batch_count;
1195 };
1196 
1197 #define DEFINE_READAHEAD(ractl, f, r, m, i)				\
1198 	struct readahead_control ractl = {				\
1199 		.file = f,						\
1200 		.mapping = m,						\
1201 		.ra = r,						\
1202 		._index = i,						\
1203 	}
1204 
1205 #define VM_READAHEAD_PAGES	(SZ_128K / PAGE_SIZE)
1206 
1207 void page_cache_ra_unbounded(struct readahead_control *,
1208 		unsigned long nr_to_read, unsigned long lookahead_count);
1209 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
1210 void page_cache_async_ra(struct readahead_control *, struct folio *,
1211 		unsigned long req_count);
1212 void readahead_expand(struct readahead_control *ractl,
1213 		      loff_t new_start, size_t new_len);
1214 
1215 /**
1216  * page_cache_sync_readahead - generic file readahead
1217  * @mapping: address_space which holds the pagecache and I/O vectors
1218  * @ra: file_ra_state which holds the readahead state
1219  * @file: Used by the filesystem for authentication.
1220  * @index: Index of first page to be read.
1221  * @req_count: Total number of pages being read by the caller.
1222  *
1223  * page_cache_sync_readahead() should be called when a cache miss happened:
1224  * it will submit the read.  The readahead logic may decide to piggyback more
1225  * pages onto the read request if access patterns suggest it will improve
1226  * performance.
1227  */
1228 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)1229 void page_cache_sync_readahead(struct address_space *mapping,
1230 		struct file_ra_state *ra, struct file *file, pgoff_t index,
1231 		unsigned long req_count)
1232 {
1233 	DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1234 	page_cache_sync_ra(&ractl, req_count);
1235 }
1236 
1237 /**
1238  * page_cache_async_readahead - file readahead for marked pages
1239  * @mapping: address_space which holds the pagecache and I/O vectors
1240  * @ra: file_ra_state which holds the readahead state
1241  * @file: Used by the filesystem for authentication.
1242  * @folio: The folio at @index which triggered the readahead call.
1243  * @index: Index of first page to be read.
1244  * @req_count: Total number of pages being read by the caller.
1245  *
1246  * page_cache_async_readahead() should be called when a page is used which
1247  * is marked as PageReadahead; this is a marker to suggest that the application
1248  * has used up enough of the readahead window that we should start pulling in
1249  * more pages.
1250  */
1251 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)1252 void page_cache_async_readahead(struct address_space *mapping,
1253 		struct file_ra_state *ra, struct file *file,
1254 		struct folio *folio, pgoff_t index, unsigned long req_count)
1255 {
1256 	DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1257 	page_cache_async_ra(&ractl, folio, req_count);
1258 }
1259 
__readahead_folio(struct readahead_control * ractl)1260 static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1261 {
1262 	struct folio *folio;
1263 
1264 	BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1265 	ractl->_nr_pages -= ractl->_batch_count;
1266 	ractl->_index += ractl->_batch_count;
1267 
1268 	if (!ractl->_nr_pages) {
1269 		ractl->_batch_count = 0;
1270 		return NULL;
1271 	}
1272 
1273 	folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1274 	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1275 	ractl->_batch_count = folio_nr_pages(folio);
1276 
1277 	return folio;
1278 }
1279 
1280 /**
1281  * readahead_page - Get the next page to read.
1282  * @ractl: The current readahead request.
1283  *
1284  * Context: The page is locked and has an elevated refcount.  The caller
1285  * should decreases the refcount once the page has been submitted for I/O
1286  * and unlock the page once all I/O to that page has completed.
1287  * Return: A pointer to the next page, or %NULL if we are done.
1288  */
readahead_page(struct readahead_control * ractl)1289 static inline struct page *readahead_page(struct readahead_control *ractl)
1290 {
1291 	struct folio *folio = __readahead_folio(ractl);
1292 
1293 	return &folio->page;
1294 }
1295 
1296 /**
1297  * readahead_folio - Get the next folio to read.
1298  * @ractl: The current readahead request.
1299  *
1300  * Context: The folio is locked.  The caller should unlock the folio once
1301  * all I/O to that folio has completed.
1302  * Return: A pointer to the next folio, or %NULL if we are done.
1303  */
readahead_folio(struct readahead_control * ractl)1304 static inline struct folio *readahead_folio(struct readahead_control *ractl)
1305 {
1306 	struct folio *folio = __readahead_folio(ractl);
1307 
1308 	if (folio)
1309 		folio_put(folio);
1310 	return folio;
1311 }
1312 
__readahead_batch(struct readahead_control * rac,struct page ** array,unsigned int array_sz)1313 static inline unsigned int __readahead_batch(struct readahead_control *rac,
1314 		struct page **array, unsigned int array_sz)
1315 {
1316 	unsigned int i = 0;
1317 	XA_STATE(xas, &rac->mapping->i_pages, 0);
1318 	struct page *page;
1319 
1320 	BUG_ON(rac->_batch_count > rac->_nr_pages);
1321 	rac->_nr_pages -= rac->_batch_count;
1322 	rac->_index += rac->_batch_count;
1323 	rac->_batch_count = 0;
1324 
1325 	xas_set(&xas, rac->_index);
1326 	rcu_read_lock();
1327 	xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1328 		if (xas_retry(&xas, page))
1329 			continue;
1330 		VM_BUG_ON_PAGE(!PageLocked(page), page);
1331 		VM_BUG_ON_PAGE(PageTail(page), page);
1332 		array[i++] = page;
1333 		rac->_batch_count += thp_nr_pages(page);
1334 		if (i == array_sz)
1335 			break;
1336 	}
1337 	rcu_read_unlock();
1338 
1339 	return i;
1340 }
1341 
1342 /**
1343  * readahead_page_batch - Get a batch of pages to read.
1344  * @rac: The current readahead request.
1345  * @array: An array of pointers to struct page.
1346  *
1347  * Context: The pages are locked and have an elevated refcount.  The caller
1348  * should decreases the refcount once the page has been submitted for I/O
1349  * and unlock the page once all I/O to that page has completed.
1350  * Return: The number of pages placed in the array.  0 indicates the request
1351  * is complete.
1352  */
1353 #define readahead_page_batch(rac, array)				\
1354 	__readahead_batch(rac, array, ARRAY_SIZE(array))
1355 
1356 /**
1357  * readahead_pos - The byte offset into the file of this readahead request.
1358  * @rac: The readahead request.
1359  */
readahead_pos(struct readahead_control * rac)1360 static inline loff_t readahead_pos(struct readahead_control *rac)
1361 {
1362 	return (loff_t)rac->_index * PAGE_SIZE;
1363 }
1364 
1365 /**
1366  * readahead_length - The number of bytes in this readahead request.
1367  * @rac: The readahead request.
1368  */
readahead_length(struct readahead_control * rac)1369 static inline size_t readahead_length(struct readahead_control *rac)
1370 {
1371 	return rac->_nr_pages * PAGE_SIZE;
1372 }
1373 
1374 /**
1375  * readahead_index - The index of the first page in this readahead request.
1376  * @rac: The readahead request.
1377  */
readahead_index(struct readahead_control * rac)1378 static inline pgoff_t readahead_index(struct readahead_control *rac)
1379 {
1380 	return rac->_index;
1381 }
1382 
1383 /**
1384  * readahead_count - The number of pages in this readahead request.
1385  * @rac: The readahead request.
1386  */
readahead_count(struct readahead_control * rac)1387 static inline unsigned int readahead_count(struct readahead_control *rac)
1388 {
1389 	return rac->_nr_pages;
1390 }
1391 
1392 /**
1393  * readahead_batch_length - The number of bytes in the current batch.
1394  * @rac: The readahead request.
1395  */
readahead_batch_length(struct readahead_control * rac)1396 static inline size_t readahead_batch_length(struct readahead_control *rac)
1397 {
1398 	return rac->_batch_count * PAGE_SIZE;
1399 }
1400 
dir_pages(struct inode * inode)1401 static inline unsigned long dir_pages(struct inode *inode)
1402 {
1403 	return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1404 			       PAGE_SHIFT;
1405 }
1406 
1407 /**
1408  * folio_mkwrite_check_truncate - check if folio was truncated
1409  * @folio: the folio to check
1410  * @inode: the inode to check the folio against
1411  *
1412  * Return: the number of bytes in the folio up to EOF,
1413  * or -EFAULT if the folio was truncated.
1414  */
folio_mkwrite_check_truncate(struct folio * folio,struct inode * inode)1415 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1416 					      struct inode *inode)
1417 {
1418 	loff_t size = i_size_read(inode);
1419 	pgoff_t index = size >> PAGE_SHIFT;
1420 	size_t offset = offset_in_folio(folio, size);
1421 
1422 	if (!folio->mapping)
1423 		return -EFAULT;
1424 
1425 	/* folio is wholly inside EOF */
1426 	if (folio_next_index(folio) - 1 < index)
1427 		return folio_size(folio);
1428 	/* folio is wholly past EOF */
1429 	if (folio->index > index || !offset)
1430 		return -EFAULT;
1431 	/* folio is partially inside EOF */
1432 	return offset;
1433 }
1434 
1435 /**
1436  * page_mkwrite_check_truncate - check if page was truncated
1437  * @page: the page to check
1438  * @inode: the inode to check the page against
1439  *
1440  * Returns the number of bytes in the page up to EOF,
1441  * or -EFAULT if the page was truncated.
1442  */
page_mkwrite_check_truncate(struct page * page,struct inode * inode)1443 static inline int page_mkwrite_check_truncate(struct page *page,
1444 					      struct inode *inode)
1445 {
1446 	loff_t size = i_size_read(inode);
1447 	pgoff_t index = size >> PAGE_SHIFT;
1448 	int offset = offset_in_page(size);
1449 
1450 	if (page->mapping != inode->i_mapping)
1451 		return -EFAULT;
1452 
1453 	/* page is wholly inside EOF */
1454 	if (page->index < index)
1455 		return PAGE_SIZE;
1456 	/* page is wholly past EOF */
1457 	if (page->index > index || !offset)
1458 		return -EFAULT;
1459 	/* page is partially inside EOF */
1460 	return offset;
1461 }
1462 
1463 /**
1464  * i_blocks_per_folio - How many blocks fit in this folio.
1465  * @inode: The inode which contains the blocks.
1466  * @folio: The folio.
1467  *
1468  * If the block size is larger than the size of this folio, return zero.
1469  *
1470  * Context: The caller should hold a refcount on the folio to prevent it
1471  * from being split.
1472  * Return: The number of filesystem blocks covered by this folio.
1473  */
1474 static inline
i_blocks_per_folio(struct inode * inode,struct folio * folio)1475 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1476 {
1477 	return folio_size(folio) >> inode->i_blkbits;
1478 }
1479 
1480 static inline
i_blocks_per_page(struct inode * inode,struct page * page)1481 unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1482 {
1483 	return i_blocks_per_folio(inode, page_folio(page));
1484 }
1485 #endif /* _LINUX_PAGEMAP_H */
1486