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 struct address_space *page_mapping(struct page *);
349 struct address_space *folio_mapping(struct folio *);
350 struct address_space *swapcache_mapping(struct folio *);
351 
352 /**
353  * folio_file_mapping - Find the mapping this folio belongs to.
354  * @folio: The folio.
355  *
356  * For folios which are in the page cache, return the mapping that this
357  * page belongs to.  Folios in the swap cache return the mapping of the
358  * swap file or swap device where the data is stored.  This is different
359  * from the mapping returned by folio_mapping().  The only reason to
360  * use it is if, like NFS, you return 0 from ->activate_swapfile.
361  *
362  * Do not call this for folios which aren't in the page cache or swap cache.
363  */
folio_file_mapping(struct folio * folio)364 static inline struct address_space *folio_file_mapping(struct folio *folio)
365 {
366 	if (unlikely(folio_test_swapcache(folio)))
367 		return swapcache_mapping(folio);
368 
369 	return folio->mapping;
370 }
371 
page_file_mapping(struct page * page)372 static inline struct address_space *page_file_mapping(struct page *page)
373 {
374 	return folio_file_mapping(page_folio(page));
375 }
376 
377 /*
378  * For file cache pages, return the address_space, otherwise return NULL
379  */
page_mapping_file(struct page * page)380 static inline struct address_space *page_mapping_file(struct page *page)
381 {
382 	struct folio *folio = page_folio(page);
383 
384 	if (unlikely(folio_test_swapcache(folio)))
385 		return NULL;
386 	return folio_mapping(folio);
387 }
388 
389 /**
390  * folio_inode - Get the host inode for this folio.
391  * @folio: The folio.
392  *
393  * For folios which are in the page cache, return the inode that this folio
394  * belongs to.
395  *
396  * Do not call this for folios which aren't in the page cache.
397  */
folio_inode(struct folio * folio)398 static inline struct inode *folio_inode(struct folio *folio)
399 {
400 	return folio->mapping->host;
401 }
402 
403 /**
404  * folio_attach_private - Attach private data to a folio.
405  * @folio: Folio to attach data to.
406  * @data: Data to attach to folio.
407  *
408  * Attaching private data to a folio increments the page's reference count.
409  * The data must be detached before the folio will be freed.
410  */
folio_attach_private(struct folio * folio,void * data)411 static inline void folio_attach_private(struct folio *folio, void *data)
412 {
413 	folio_get(folio);
414 	folio->private = data;
415 	folio_set_private(folio);
416 }
417 
418 /**
419  * folio_change_private - Change private data on a folio.
420  * @folio: Folio to change the data on.
421  * @data: Data to set on the folio.
422  *
423  * Change the private data attached to a folio and return the old
424  * data.  The page must previously have had data attached and the data
425  * must be detached before the folio will be freed.
426  *
427  * Return: Data that was previously attached to the folio.
428  */
folio_change_private(struct folio * folio,void * data)429 static inline void *folio_change_private(struct folio *folio, void *data)
430 {
431 	void *old = folio_get_private(folio);
432 
433 	folio->private = data;
434 	return old;
435 }
436 
437 /**
438  * folio_detach_private - Detach private data from a folio.
439  * @folio: Folio to detach data from.
440  *
441  * Removes the data that was previously attached to the folio and decrements
442  * the refcount on the page.
443  *
444  * Return: Data that was attached to the folio.
445  */
folio_detach_private(struct folio * folio)446 static inline void *folio_detach_private(struct folio *folio)
447 {
448 	void *data = folio_get_private(folio);
449 
450 	if (!folio_test_private(folio))
451 		return NULL;
452 	folio_clear_private(folio);
453 	folio->private = NULL;
454 	folio_put(folio);
455 
456 	return data;
457 }
458 
attach_page_private(struct page * page,void * data)459 static inline void attach_page_private(struct page *page, void *data)
460 {
461 	folio_attach_private(page_folio(page), data);
462 }
463 
detach_page_private(struct page * page)464 static inline void *detach_page_private(struct page *page)
465 {
466 	return folio_detach_private(page_folio(page));
467 }
468 
469 #ifdef CONFIG_NUMA
470 struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order);
471 #else
filemap_alloc_folio(gfp_t gfp,unsigned int order)472 static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order)
473 {
474 	return folio_alloc(gfp, order);
475 }
476 #endif
477 
__page_cache_alloc(gfp_t gfp)478 static inline struct page *__page_cache_alloc(gfp_t gfp)
479 {
480 	return &filemap_alloc_folio(gfp, 0)->page;
481 }
482 
page_cache_alloc(struct address_space * x)483 static inline struct page *page_cache_alloc(struct address_space *x)
484 {
485 	return __page_cache_alloc(mapping_gfp_mask(x));
486 }
487 
readahead_gfp_mask(struct address_space * x)488 static inline gfp_t readahead_gfp_mask(struct address_space *x)
489 {
490 	return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
491 }
492 
493 typedef int filler_t(struct file *, struct folio *);
494 
495 pgoff_t page_cache_next_miss(struct address_space *mapping,
496 			     pgoff_t index, unsigned long max_scan);
497 pgoff_t page_cache_prev_miss(struct address_space *mapping,
498 			     pgoff_t index, unsigned long max_scan);
499 
500 #define FGP_ACCESSED		0x00000001
501 #define FGP_LOCK		0x00000002
502 #define FGP_CREAT		0x00000004
503 #define FGP_WRITE		0x00000008
504 #define FGP_NOFS		0x00000010
505 #define FGP_NOWAIT		0x00000020
506 #define FGP_FOR_MMAP		0x00000040
507 #define FGP_HEAD		0x00000080
508 #define FGP_ENTRY		0x00000100
509 #define FGP_STABLE		0x00000200
510 
511 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
512 		int fgp_flags, gfp_t gfp);
513 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
514 		int fgp_flags, gfp_t gfp);
515 
516 /**
517  * filemap_get_folio - Find and get a folio.
518  * @mapping: The address_space to search.
519  * @index: The page index.
520  *
521  * Looks up the page cache entry at @mapping & @index.  If a folio is
522  * present, it is returned with an increased refcount.
523  *
524  * Otherwise, %NULL is returned.
525  */
filemap_get_folio(struct address_space * mapping,pgoff_t index)526 static inline struct folio *filemap_get_folio(struct address_space *mapping,
527 					pgoff_t index)
528 {
529 	return __filemap_get_folio(mapping, index, 0, 0);
530 }
531 
532 /**
533  * filemap_lock_folio - Find and lock a folio.
534  * @mapping: The address_space to search.
535  * @index: The page index.
536  *
537  * Looks up the page cache entry at @mapping & @index.  If a folio is
538  * present, it is returned locked with an increased refcount.
539  *
540  * Context: May sleep.
541  * Return: A folio or %NULL if there is no folio in the cache for this
542  * index.  Will not return a shadow, swap or DAX entry.
543  */
filemap_lock_folio(struct address_space * mapping,pgoff_t index)544 static inline struct folio *filemap_lock_folio(struct address_space *mapping,
545 					pgoff_t index)
546 {
547 	return __filemap_get_folio(mapping, index, FGP_LOCK, 0);
548 }
549 
550 /**
551  * find_get_page - find and get a page reference
552  * @mapping: the address_space to search
553  * @offset: the page index
554  *
555  * Looks up the page cache slot at @mapping & @offset.  If there is a
556  * page cache page, it is returned with an increased refcount.
557  *
558  * Otherwise, %NULL is returned.
559  */
find_get_page(struct address_space * mapping,pgoff_t offset)560 static inline struct page *find_get_page(struct address_space *mapping,
561 					pgoff_t offset)
562 {
563 	return pagecache_get_page(mapping, offset, 0, 0);
564 }
565 
find_get_page_flags(struct address_space * mapping,pgoff_t offset,int fgp_flags)566 static inline struct page *find_get_page_flags(struct address_space *mapping,
567 					pgoff_t offset, int fgp_flags)
568 {
569 	return pagecache_get_page(mapping, offset, fgp_flags, 0);
570 }
571 
572 /**
573  * find_lock_page - locate, pin and lock a pagecache page
574  * @mapping: the address_space to search
575  * @index: the page index
576  *
577  * Looks up the page cache entry at @mapping & @index.  If there is a
578  * page cache page, it is returned locked and with an increased
579  * refcount.
580  *
581  * Context: May sleep.
582  * Return: A struct page or %NULL if there is no page in the cache for this
583  * index.
584  */
find_lock_page(struct address_space * mapping,pgoff_t index)585 static inline struct page *find_lock_page(struct address_space *mapping,
586 					pgoff_t index)
587 {
588 	return pagecache_get_page(mapping, index, FGP_LOCK, 0);
589 }
590 
591 /**
592  * find_or_create_page - locate or add a pagecache page
593  * @mapping: the page's address_space
594  * @index: the page's index into the mapping
595  * @gfp_mask: page allocation mode
596  *
597  * Looks up the page cache slot at @mapping & @offset.  If there is a
598  * page cache page, it is returned locked and with an increased
599  * refcount.
600  *
601  * If the page is not present, a new page is allocated using @gfp_mask
602  * and added to the page cache and the VM's LRU list.  The page is
603  * returned locked and with an increased refcount.
604  *
605  * On memory exhaustion, %NULL is returned.
606  *
607  * find_or_create_page() may sleep, even if @gfp_flags specifies an
608  * atomic allocation!
609  */
find_or_create_page(struct address_space * mapping,pgoff_t index,gfp_t gfp_mask)610 static inline struct page *find_or_create_page(struct address_space *mapping,
611 					pgoff_t index, gfp_t gfp_mask)
612 {
613 	return pagecache_get_page(mapping, index,
614 					FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
615 					gfp_mask);
616 }
617 
618 /**
619  * grab_cache_page_nowait - returns locked page at given index in given cache
620  * @mapping: target address_space
621  * @index: the page index
622  *
623  * Same as grab_cache_page(), but do not wait if the page is unavailable.
624  * This is intended for speculative data generators, where the data can
625  * be regenerated if the page couldn't be grabbed.  This routine should
626  * be safe to call while holding the lock for another page.
627  *
628  * Clear __GFP_FS when allocating the page to avoid recursion into the fs
629  * and deadlock against the caller's locked page.
630  */
grab_cache_page_nowait(struct address_space * mapping,pgoff_t index)631 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
632 				pgoff_t index)
633 {
634 	return pagecache_get_page(mapping, index,
635 			FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
636 			mapping_gfp_mask(mapping));
637 }
638 
639 #define swapcache_index(folio)	__page_file_index(&(folio)->page)
640 
641 /**
642  * folio_index - File index of a folio.
643  * @folio: The folio.
644  *
645  * For a folio which is either in the page cache or the swap cache,
646  * return its index within the address_space it belongs to.  If you know
647  * the page is definitely in the page cache, you can look at the folio's
648  * index directly.
649  *
650  * Return: The index (offset in units of pages) of a folio in its file.
651  */
folio_index(struct folio * folio)652 static inline pgoff_t folio_index(struct folio *folio)
653 {
654         if (unlikely(folio_test_swapcache(folio)))
655                 return swapcache_index(folio);
656         return folio->index;
657 }
658 
659 /**
660  * folio_next_index - Get the index of the next folio.
661  * @folio: The current folio.
662  *
663  * Return: The index of the folio which follows this folio in the file.
664  */
folio_next_index(struct folio * folio)665 static inline pgoff_t folio_next_index(struct folio *folio)
666 {
667 	return folio->index + folio_nr_pages(folio);
668 }
669 
670 /**
671  * folio_file_page - The page for a particular index.
672  * @folio: The folio which contains this index.
673  * @index: The index we want to look up.
674  *
675  * Sometimes after looking up a folio in the page cache, we need to
676  * obtain the specific page for an index (eg a page fault).
677  *
678  * Return: The page containing the file data for this index.
679  */
folio_file_page(struct folio * folio,pgoff_t index)680 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
681 {
682 	/* HugeTLBfs indexes the page cache in units of hpage_size */
683 	if (folio_test_hugetlb(folio))
684 		return &folio->page;
685 	return folio_page(folio, index & (folio_nr_pages(folio) - 1));
686 }
687 
688 /**
689  * folio_contains - Does this folio contain this index?
690  * @folio: The folio.
691  * @index: The page index within the file.
692  *
693  * Context: The caller should have the page locked in order to prevent
694  * (eg) shmem from moving the page between the page cache and swap cache
695  * and changing its index in the middle of the operation.
696  * Return: true or false.
697  */
folio_contains(struct folio * folio,pgoff_t index)698 static inline bool folio_contains(struct folio *folio, pgoff_t index)
699 {
700 	/* HugeTLBfs indexes the page cache in units of hpage_size */
701 	if (folio_test_hugetlb(folio))
702 		return folio->index == index;
703 	return index - folio_index(folio) < folio_nr_pages(folio);
704 }
705 
706 /*
707  * Given the page we found in the page cache, return the page corresponding
708  * to this index in the file
709  */
find_subpage(struct page * head,pgoff_t index)710 static inline struct page *find_subpage(struct page *head, pgoff_t index)
711 {
712 	/* HugeTLBfs wants the head page regardless */
713 	if (PageHuge(head))
714 		return head;
715 
716 	return head + (index & (thp_nr_pages(head) - 1));
717 }
718 
719 unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start,
720 		pgoff_t end, struct folio_batch *fbatch);
721 unsigned filemap_get_folios_contig(struct address_space *mapping,
722 		pgoff_t *start, pgoff_t end, struct folio_batch *fbatch);
723 unsigned find_get_pages_range_tag(struct address_space *mapping, pgoff_t *index,
724 			pgoff_t end, xa_mark_t tag, unsigned int nr_pages,
725 			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)726 static inline unsigned find_get_pages_tag(struct address_space *mapping,
727 			pgoff_t *index, xa_mark_t tag, unsigned int nr_pages,
728 			struct page **pages)
729 {
730 	return find_get_pages_range_tag(mapping, index, (pgoff_t)-1, tag,
731 					nr_pages, pages);
732 }
733 
734 struct page *grab_cache_page_write_begin(struct address_space *mapping,
735 			pgoff_t index);
736 
737 /*
738  * Returns locked page at given index in given cache, creating it if needed.
739  */
grab_cache_page(struct address_space * mapping,pgoff_t index)740 static inline struct page *grab_cache_page(struct address_space *mapping,
741 								pgoff_t index)
742 {
743 	return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
744 }
745 
746 struct folio *read_cache_folio(struct address_space *, pgoff_t index,
747 		filler_t *filler, struct file *file);
748 struct page *read_cache_page(struct address_space *, pgoff_t index,
749 		filler_t *filler, struct file *file);
750 extern struct page * read_cache_page_gfp(struct address_space *mapping,
751 				pgoff_t index, gfp_t gfp_mask);
752 
read_mapping_page(struct address_space * mapping,pgoff_t index,struct file * file)753 static inline struct page *read_mapping_page(struct address_space *mapping,
754 				pgoff_t index, struct file *file)
755 {
756 	return read_cache_page(mapping, index, NULL, file);
757 }
758 
read_mapping_folio(struct address_space * mapping,pgoff_t index,struct file * file)759 static inline struct folio *read_mapping_folio(struct address_space *mapping,
760 				pgoff_t index, struct file *file)
761 {
762 	return read_cache_folio(mapping, index, NULL, file);
763 }
764 
765 /*
766  * Get index of the page within radix-tree (but not for hugetlb pages).
767  * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
768  */
page_to_index(struct page * page)769 static inline pgoff_t page_to_index(struct page *page)
770 {
771 	struct page *head;
772 
773 	if (likely(!PageTransTail(page)))
774 		return page->index;
775 
776 	head = compound_head(page);
777 	/*
778 	 *  We don't initialize ->index for tail pages: calculate based on
779 	 *  head page
780 	 */
781 	return head->index + page - head;
782 }
783 
784 extern pgoff_t hugetlb_basepage_index(struct page *page);
785 
786 /*
787  * Get the offset in PAGE_SIZE (even for hugetlb pages).
788  * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
789  */
page_to_pgoff(struct page * page)790 static inline pgoff_t page_to_pgoff(struct page *page)
791 {
792 	if (unlikely(PageHuge(page)))
793 		return hugetlb_basepage_index(page);
794 	return page_to_index(page);
795 }
796 
797 /*
798  * Return byte-offset into filesystem object for page.
799  */
page_offset(struct page * page)800 static inline loff_t page_offset(struct page *page)
801 {
802 	return ((loff_t)page->index) << PAGE_SHIFT;
803 }
804 
page_file_offset(struct page * page)805 static inline loff_t page_file_offset(struct page *page)
806 {
807 	return ((loff_t)page_index(page)) << PAGE_SHIFT;
808 }
809 
810 /**
811  * folio_pos - Returns the byte position of this folio in its file.
812  * @folio: The folio.
813  */
folio_pos(struct folio * folio)814 static inline loff_t folio_pos(struct folio *folio)
815 {
816 	return page_offset(&folio->page);
817 }
818 
819 /**
820  * folio_file_pos - Returns the byte position of this folio in its file.
821  * @folio: The folio.
822  *
823  * This differs from folio_pos() for folios which belong to a swap file.
824  * NFS is the only filesystem today which needs to use folio_file_pos().
825  */
folio_file_pos(struct folio * folio)826 static inline loff_t folio_file_pos(struct folio *folio)
827 {
828 	return page_file_offset(&folio->page);
829 }
830 
831 /*
832  * Get the offset in PAGE_SIZE (even for hugetlb folios).
833  * (TODO: hugetlb folios should have ->index in PAGE_SIZE)
834  */
folio_pgoff(struct folio * folio)835 static inline pgoff_t folio_pgoff(struct folio *folio)
836 {
837 	if (unlikely(folio_test_hugetlb(folio)))
838 		return hugetlb_basepage_index(&folio->page);
839 	return folio->index;
840 }
841 
842 extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
843 				     unsigned long address);
844 
linear_page_index(struct vm_area_struct * vma,unsigned long address)845 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
846 					unsigned long address)
847 {
848 	pgoff_t pgoff;
849 	if (unlikely(is_vm_hugetlb_page(vma)))
850 		return linear_hugepage_index(vma, address);
851 	pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
852 	pgoff += vma->vm_pgoff;
853 	return pgoff;
854 }
855 
856 struct wait_page_key {
857 	struct folio *folio;
858 	int bit_nr;
859 	int page_match;
860 };
861 
862 struct wait_page_queue {
863 	struct folio *folio;
864 	int bit_nr;
865 	wait_queue_entry_t wait;
866 };
867 
wake_page_match(struct wait_page_queue * wait_page,struct wait_page_key * key)868 static inline bool wake_page_match(struct wait_page_queue *wait_page,
869 				  struct wait_page_key *key)
870 {
871 	if (wait_page->folio != key->folio)
872 	       return false;
873 	key->page_match = 1;
874 
875 	if (wait_page->bit_nr != key->bit_nr)
876 		return false;
877 
878 	return true;
879 }
880 
881 void __folio_lock(struct folio *folio);
882 int __folio_lock_killable(struct folio *folio);
883 bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm,
884 				unsigned int flags);
885 void unlock_page(struct page *page);
886 void folio_unlock(struct folio *folio);
887 
888 /**
889  * folio_trylock() - Attempt to lock a folio.
890  * @folio: The folio to attempt to lock.
891  *
892  * Sometimes it is undesirable to wait for a folio to be unlocked (eg
893  * when the locks are being taken in the wrong order, or if making
894  * progress through a batch of folios is more important than processing
895  * them in order).  Usually folio_lock() is the correct function to call.
896  *
897  * Context: Any context.
898  * Return: Whether the lock was successfully acquired.
899  */
folio_trylock(struct folio * folio)900 static inline bool folio_trylock(struct folio *folio)
901 {
902 	return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
903 }
904 
905 /*
906  * Return true if the page was successfully locked
907  */
trylock_page(struct page * page)908 static inline int trylock_page(struct page *page)
909 {
910 	return folio_trylock(page_folio(page));
911 }
912 
913 /**
914  * folio_lock() - Lock this folio.
915  * @folio: The folio to lock.
916  *
917  * The folio lock protects against many things, probably more than it
918  * should.  It is primarily held while a folio is being brought uptodate,
919  * either from its backing file or from swap.  It is also held while a
920  * folio is being truncated from its address_space, so holding the lock
921  * is sufficient to keep folio->mapping stable.
922  *
923  * The folio lock is also held while write() is modifying the page to
924  * provide POSIX atomicity guarantees (as long as the write does not
925  * cross a page boundary).  Other modifications to the data in the folio
926  * do not hold the folio lock and can race with writes, eg DMA and stores
927  * to mapped pages.
928  *
929  * Context: May sleep.  If you need to acquire the locks of two or
930  * more folios, they must be in order of ascending index, if they are
931  * in the same address_space.  If they are in different address_spaces,
932  * acquire the lock of the folio which belongs to the address_space which
933  * has the lowest address in memory first.
934  */
folio_lock(struct folio * folio)935 static inline void folio_lock(struct folio *folio)
936 {
937 	might_sleep();
938 	if (!folio_trylock(folio))
939 		__folio_lock(folio);
940 }
941 
942 /**
943  * lock_page() - Lock the folio containing this page.
944  * @page: The page to lock.
945  *
946  * See folio_lock() for a description of what the lock protects.
947  * This is a legacy function and new code should probably use folio_lock()
948  * instead.
949  *
950  * Context: May sleep.  Pages in the same folio share a lock, so do not
951  * attempt to lock two pages which share a folio.
952  */
lock_page(struct page * page)953 static inline void lock_page(struct page *page)
954 {
955 	struct folio *folio;
956 	might_sleep();
957 
958 	folio = page_folio(page);
959 	if (!folio_trylock(folio))
960 		__folio_lock(folio);
961 }
962 
963 /**
964  * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
965  * @folio: The folio to lock.
966  *
967  * Attempts to lock the folio, like folio_lock(), except that the sleep
968  * to acquire the lock is interruptible by a fatal signal.
969  *
970  * Context: May sleep; see folio_lock().
971  * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
972  */
folio_lock_killable(struct folio * folio)973 static inline int folio_lock_killable(struct folio *folio)
974 {
975 	might_sleep();
976 	if (!folio_trylock(folio))
977 		return __folio_lock_killable(folio);
978 	return 0;
979 }
980 
981 /*
982  * lock_page_killable is like lock_page but can be interrupted by fatal
983  * signals.  It returns 0 if it locked the page and -EINTR if it was
984  * killed while waiting.
985  */
lock_page_killable(struct page * page)986 static inline int lock_page_killable(struct page *page)
987 {
988 	return folio_lock_killable(page_folio(page));
989 }
990 
991 /*
992  * folio_lock_or_retry - Lock the folio, unless this would block and the
993  * caller indicated that it can handle a retry.
994  *
995  * Return value and mmap_lock implications depend on flags; see
996  * __folio_lock_or_retry().
997  */
folio_lock_or_retry(struct folio * folio,struct mm_struct * mm,unsigned int flags)998 static inline bool folio_lock_or_retry(struct folio *folio,
999 		struct mm_struct *mm, unsigned int flags)
1000 {
1001 	might_sleep();
1002 	return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags);
1003 }
1004 
1005 /*
1006  * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
1007  * and should not be used directly.
1008  */
1009 void folio_wait_bit(struct folio *folio, int bit_nr);
1010 int folio_wait_bit_killable(struct folio *folio, int bit_nr);
1011 
1012 /*
1013  * Wait for a folio to be unlocked.
1014  *
1015  * This must be called with the caller "holding" the folio,
1016  * ie with increased folio reference count so that the folio won't
1017  * go away during the wait.
1018  */
folio_wait_locked(struct folio * folio)1019 static inline void folio_wait_locked(struct folio *folio)
1020 {
1021 	if (folio_test_locked(folio))
1022 		folio_wait_bit(folio, PG_locked);
1023 }
1024 
folio_wait_locked_killable(struct folio * folio)1025 static inline int folio_wait_locked_killable(struct folio *folio)
1026 {
1027 	if (!folio_test_locked(folio))
1028 		return 0;
1029 	return folio_wait_bit_killable(folio, PG_locked);
1030 }
1031 
wait_on_page_locked(struct page * page)1032 static inline void wait_on_page_locked(struct page *page)
1033 {
1034 	folio_wait_locked(page_folio(page));
1035 }
1036 
wait_on_page_locked_killable(struct page * page)1037 static inline int wait_on_page_locked_killable(struct page *page)
1038 {
1039 	return folio_wait_locked_killable(page_folio(page));
1040 }
1041 
1042 void wait_on_page_writeback(struct page *page);
1043 void folio_wait_writeback(struct folio *folio);
1044 int folio_wait_writeback_killable(struct folio *folio);
1045 void end_page_writeback(struct page *page);
1046 void folio_end_writeback(struct folio *folio);
1047 void wait_for_stable_page(struct page *page);
1048 void folio_wait_stable(struct folio *folio);
1049 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
__set_page_dirty(struct page * page,struct address_space * mapping,int warn)1050 static inline void __set_page_dirty(struct page *page,
1051 		struct address_space *mapping, int warn)
1052 {
1053 	__folio_mark_dirty(page_folio(page), mapping, warn);
1054 }
1055 void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
1056 void __folio_cancel_dirty(struct folio *folio);
folio_cancel_dirty(struct folio * folio)1057 static inline void folio_cancel_dirty(struct folio *folio)
1058 {
1059 	/* Avoid atomic ops, locking, etc. when not actually needed. */
1060 	if (folio_test_dirty(folio))
1061 		__folio_cancel_dirty(folio);
1062 }
1063 bool folio_clear_dirty_for_io(struct folio *folio);
1064 bool clear_page_dirty_for_io(struct page *page);
1065 void folio_invalidate(struct folio *folio, size_t offset, size_t length);
1066 int __must_check folio_write_one(struct folio *folio);
write_one_page(struct page * page)1067 static inline int __must_check write_one_page(struct page *page)
1068 {
1069 	return folio_write_one(page_folio(page));
1070 }
1071 
1072 int __set_page_dirty_nobuffers(struct page *page);
1073 bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
1074 
1075 #ifdef CONFIG_MIGRATION
1076 int filemap_migrate_folio(struct address_space *mapping, struct folio *dst,
1077 		struct folio *src, enum migrate_mode mode);
1078 #else
1079 #define filemap_migrate_folio NULL
1080 #endif
1081 void page_endio(struct page *page, bool is_write, int err);
1082 
1083 void folio_end_private_2(struct folio *folio);
1084 void folio_wait_private_2(struct folio *folio);
1085 int folio_wait_private_2_killable(struct folio *folio);
1086 
1087 /*
1088  * Add an arbitrary waiter to a page's wait queue
1089  */
1090 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
1091 
1092 /*
1093  * Fault in userspace address range.
1094  */
1095 size_t fault_in_writeable(char __user *uaddr, size_t size);
1096 size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
1097 size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
1098 size_t fault_in_readable(const char __user *uaddr, size_t size);
1099 
1100 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
1101 		pgoff_t index, gfp_t gfp);
1102 int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1103 		pgoff_t index, gfp_t gfp);
1104 void filemap_remove_folio(struct folio *folio);
1105 void delete_from_page_cache(struct page *page);
1106 void __filemap_remove_folio(struct folio *folio, void *shadow);
1107 void replace_page_cache_page(struct page *old, struct page *new);
1108 void delete_from_page_cache_batch(struct address_space *mapping,
1109 				  struct folio_batch *fbatch);
1110 int try_to_release_page(struct page *page, gfp_t gfp);
1111 bool filemap_release_folio(struct folio *folio, gfp_t gfp);
1112 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1113 		int whence);
1114 
1115 /* Must be non-static for BPF error injection */
1116 int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1117 		pgoff_t index, gfp_t gfp, void **shadowp);
1118 
1119 bool filemap_range_has_writeback(struct address_space *mapping,
1120 				 loff_t start_byte, loff_t end_byte);
1121 
1122 /**
1123  * filemap_range_needs_writeback - check if range potentially needs writeback
1124  * @mapping:           address space within which to check
1125  * @start_byte:        offset in bytes where the range starts
1126  * @end_byte:          offset in bytes where the range ends (inclusive)
1127  *
1128  * Find at least one page in the range supplied, usually used to check if
1129  * direct writing in this range will trigger a writeback. Used by O_DIRECT
1130  * read/write with IOCB_NOWAIT, to see if the caller needs to do
1131  * filemap_write_and_wait_range() before proceeding.
1132  *
1133  * Return: %true if the caller should do filemap_write_and_wait_range() before
1134  * doing O_DIRECT to a page in this range, %false otherwise.
1135  */
filemap_range_needs_writeback(struct address_space * mapping,loff_t start_byte,loff_t end_byte)1136 static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1137 						 loff_t start_byte,
1138 						 loff_t end_byte)
1139 {
1140 	if (!mapping->nrpages)
1141 		return false;
1142 	if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1143 	    !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1144 		return false;
1145 	return filemap_range_has_writeback(mapping, start_byte, end_byte);
1146 }
1147 
1148 /**
1149  * struct readahead_control - Describes a readahead request.
1150  *
1151  * A readahead request is for consecutive pages.  Filesystems which
1152  * implement the ->readahead method should call readahead_page() or
1153  * readahead_page_batch() in a loop and attempt to start I/O against
1154  * each page in the request.
1155  *
1156  * Most of the fields in this struct are private and should be accessed
1157  * by the functions below.
1158  *
1159  * @file: The file, used primarily by network filesystems for authentication.
1160  *	  May be NULL if invoked internally by the filesystem.
1161  * @mapping: Readahead this filesystem object.
1162  * @ra: File readahead state.  May be NULL.
1163  */
1164 struct readahead_control {
1165 	struct file *file;
1166 	struct address_space *mapping;
1167 	struct file_ra_state *ra;
1168 /* private: use the readahead_* accessors instead */
1169 	pgoff_t _index;
1170 	unsigned int _nr_pages;
1171 	unsigned int _batch_count;
1172 	bool _workingset;
1173 	unsigned long _pflags;
1174 };
1175 
1176 #define DEFINE_READAHEAD(ractl, f, r, m, i)				\
1177 	struct readahead_control ractl = {				\
1178 		.file = f,						\
1179 		.mapping = m,						\
1180 		.ra = r,						\
1181 		._index = i,						\
1182 	}
1183 
1184 #define VM_READAHEAD_PAGES	(SZ_128K / PAGE_SIZE)
1185 
1186 void page_cache_ra_unbounded(struct readahead_control *,
1187 		unsigned long nr_to_read, unsigned long lookahead_count);
1188 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
1189 void page_cache_async_ra(struct readahead_control *, struct folio *,
1190 		unsigned long req_count);
1191 void readahead_expand(struct readahead_control *ractl,
1192 		      loff_t new_start, size_t new_len);
1193 
1194 /**
1195  * page_cache_sync_readahead - generic file readahead
1196  * @mapping: address_space which holds the pagecache and I/O vectors
1197  * @ra: file_ra_state which holds the readahead state
1198  * @file: Used by the filesystem for authentication.
1199  * @index: Index of first page to be read.
1200  * @req_count: Total number of pages being read by the caller.
1201  *
1202  * page_cache_sync_readahead() should be called when a cache miss happened:
1203  * it will submit the read.  The readahead logic may decide to piggyback more
1204  * pages onto the read request if access patterns suggest it will improve
1205  * performance.
1206  */
1207 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)1208 void page_cache_sync_readahead(struct address_space *mapping,
1209 		struct file_ra_state *ra, struct file *file, pgoff_t index,
1210 		unsigned long req_count)
1211 {
1212 	DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1213 	page_cache_sync_ra(&ractl, req_count);
1214 }
1215 
1216 /**
1217  * page_cache_async_readahead - file readahead for marked pages
1218  * @mapping: address_space which holds the pagecache and I/O vectors
1219  * @ra: file_ra_state which holds the readahead state
1220  * @file: Used by the filesystem for authentication.
1221  * @folio: The folio at @index which triggered the readahead call.
1222  * @index: Index of first page to be read.
1223  * @req_count: Total number of pages being read by the caller.
1224  *
1225  * page_cache_async_readahead() should be called when a page is used which
1226  * is marked as PageReadahead; this is a marker to suggest that the application
1227  * has used up enough of the readahead window that we should start pulling in
1228  * more pages.
1229  */
1230 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)1231 void page_cache_async_readahead(struct address_space *mapping,
1232 		struct file_ra_state *ra, struct file *file,
1233 		struct folio *folio, pgoff_t index, unsigned long req_count)
1234 {
1235 	DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1236 	page_cache_async_ra(&ractl, folio, req_count);
1237 }
1238 
__readahead_folio(struct readahead_control * ractl)1239 static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1240 {
1241 	struct folio *folio;
1242 
1243 	BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1244 	ractl->_nr_pages -= ractl->_batch_count;
1245 	ractl->_index += ractl->_batch_count;
1246 
1247 	if (!ractl->_nr_pages) {
1248 		ractl->_batch_count = 0;
1249 		return NULL;
1250 	}
1251 
1252 	folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1253 	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1254 	ractl->_batch_count = folio_nr_pages(folio);
1255 
1256 	return folio;
1257 }
1258 
1259 /**
1260  * readahead_page - Get the next page to read.
1261  * @ractl: The current readahead request.
1262  *
1263  * Context: The page is locked and has an elevated refcount.  The caller
1264  * should decreases the refcount once the page has been submitted for I/O
1265  * and unlock the page once all I/O to that page has completed.
1266  * Return: A pointer to the next page, or %NULL if we are done.
1267  */
readahead_page(struct readahead_control * ractl)1268 static inline struct page *readahead_page(struct readahead_control *ractl)
1269 {
1270 	struct folio *folio = __readahead_folio(ractl);
1271 
1272 	return &folio->page;
1273 }
1274 
1275 /**
1276  * readahead_folio - Get the next folio to read.
1277  * @ractl: The current readahead request.
1278  *
1279  * Context: The folio is locked.  The caller should unlock the folio once
1280  * all I/O to that folio has completed.
1281  * Return: A pointer to the next folio, or %NULL if we are done.
1282  */
readahead_folio(struct readahead_control * ractl)1283 static inline struct folio *readahead_folio(struct readahead_control *ractl)
1284 {
1285 	struct folio *folio = __readahead_folio(ractl);
1286 
1287 	if (folio)
1288 		folio_put(folio);
1289 	return folio;
1290 }
1291 
__readahead_batch(struct readahead_control * rac,struct page ** array,unsigned int array_sz)1292 static inline unsigned int __readahead_batch(struct readahead_control *rac,
1293 		struct page **array, unsigned int array_sz)
1294 {
1295 	unsigned int i = 0;
1296 	XA_STATE(xas, &rac->mapping->i_pages, 0);
1297 	struct page *page;
1298 
1299 	BUG_ON(rac->_batch_count > rac->_nr_pages);
1300 	rac->_nr_pages -= rac->_batch_count;
1301 	rac->_index += rac->_batch_count;
1302 	rac->_batch_count = 0;
1303 
1304 	xas_set(&xas, rac->_index);
1305 	rcu_read_lock();
1306 	xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1307 		if (xas_retry(&xas, page))
1308 			continue;
1309 		VM_BUG_ON_PAGE(!PageLocked(page), page);
1310 		VM_BUG_ON_PAGE(PageTail(page), page);
1311 		array[i++] = page;
1312 		rac->_batch_count += thp_nr_pages(page);
1313 		if (i == array_sz)
1314 			break;
1315 	}
1316 	rcu_read_unlock();
1317 
1318 	return i;
1319 }
1320 
1321 /**
1322  * readahead_page_batch - Get a batch of pages to read.
1323  * @rac: The current readahead request.
1324  * @array: An array of pointers to struct page.
1325  *
1326  * Context: The pages are locked and have an elevated refcount.  The caller
1327  * should decreases the refcount once the page has been submitted for I/O
1328  * and unlock the page once all I/O to that page has completed.
1329  * Return: The number of pages placed in the array.  0 indicates the request
1330  * is complete.
1331  */
1332 #define readahead_page_batch(rac, array)				\
1333 	__readahead_batch(rac, array, ARRAY_SIZE(array))
1334 
1335 /**
1336  * readahead_pos - The byte offset into the file of this readahead request.
1337  * @rac: The readahead request.
1338  */
readahead_pos(struct readahead_control * rac)1339 static inline loff_t readahead_pos(struct readahead_control *rac)
1340 {
1341 	return (loff_t)rac->_index * PAGE_SIZE;
1342 }
1343 
1344 /**
1345  * readahead_length - The number of bytes in this readahead request.
1346  * @rac: The readahead request.
1347  */
readahead_length(struct readahead_control * rac)1348 static inline size_t readahead_length(struct readahead_control *rac)
1349 {
1350 	return rac->_nr_pages * PAGE_SIZE;
1351 }
1352 
1353 /**
1354  * readahead_index - The index of the first page in this readahead request.
1355  * @rac: The readahead request.
1356  */
readahead_index(struct readahead_control * rac)1357 static inline pgoff_t readahead_index(struct readahead_control *rac)
1358 {
1359 	return rac->_index;
1360 }
1361 
1362 /**
1363  * readahead_count - The number of pages in this readahead request.
1364  * @rac: The readahead request.
1365  */
readahead_count(struct readahead_control * rac)1366 static inline unsigned int readahead_count(struct readahead_control *rac)
1367 {
1368 	return rac->_nr_pages;
1369 }
1370 
1371 /**
1372  * readahead_batch_length - The number of bytes in the current batch.
1373  * @rac: The readahead request.
1374  */
readahead_batch_length(struct readahead_control * rac)1375 static inline size_t readahead_batch_length(struct readahead_control *rac)
1376 {
1377 	return rac->_batch_count * PAGE_SIZE;
1378 }
1379 
dir_pages(struct inode * inode)1380 static inline unsigned long dir_pages(struct inode *inode)
1381 {
1382 	return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1383 			       PAGE_SHIFT;
1384 }
1385 
1386 /**
1387  * folio_mkwrite_check_truncate - check if folio was truncated
1388  * @folio: the folio to check
1389  * @inode: the inode to check the folio against
1390  *
1391  * Return: the number of bytes in the folio up to EOF,
1392  * or -EFAULT if the folio was truncated.
1393  */
folio_mkwrite_check_truncate(struct folio * folio,struct inode * inode)1394 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1395 					      struct inode *inode)
1396 {
1397 	loff_t size = i_size_read(inode);
1398 	pgoff_t index = size >> PAGE_SHIFT;
1399 	size_t offset = offset_in_folio(folio, size);
1400 
1401 	if (!folio->mapping)
1402 		return -EFAULT;
1403 
1404 	/* folio is wholly inside EOF */
1405 	if (folio_next_index(folio) - 1 < index)
1406 		return folio_size(folio);
1407 	/* folio is wholly past EOF */
1408 	if (folio->index > index || !offset)
1409 		return -EFAULT;
1410 	/* folio is partially inside EOF */
1411 	return offset;
1412 }
1413 
1414 /**
1415  * page_mkwrite_check_truncate - check if page was truncated
1416  * @page: the page to check
1417  * @inode: the inode to check the page against
1418  *
1419  * Returns the number of bytes in the page up to EOF,
1420  * or -EFAULT if the page was truncated.
1421  */
page_mkwrite_check_truncate(struct page * page,struct inode * inode)1422 static inline int page_mkwrite_check_truncate(struct page *page,
1423 					      struct inode *inode)
1424 {
1425 	loff_t size = i_size_read(inode);
1426 	pgoff_t index = size >> PAGE_SHIFT;
1427 	int offset = offset_in_page(size);
1428 
1429 	if (page->mapping != inode->i_mapping)
1430 		return -EFAULT;
1431 
1432 	/* page is wholly inside EOF */
1433 	if (page->index < index)
1434 		return PAGE_SIZE;
1435 	/* page is wholly past EOF */
1436 	if (page->index > index || !offset)
1437 		return -EFAULT;
1438 	/* page is partially inside EOF */
1439 	return offset;
1440 }
1441 
1442 /**
1443  * i_blocks_per_folio - How many blocks fit in this folio.
1444  * @inode: The inode which contains the blocks.
1445  * @folio: The folio.
1446  *
1447  * If the block size is larger than the size of this folio, return zero.
1448  *
1449  * Context: The caller should hold a refcount on the folio to prevent it
1450  * from being split.
1451  * Return: The number of filesystem blocks covered by this folio.
1452  */
1453 static inline
i_blocks_per_folio(struct inode * inode,struct folio * folio)1454 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1455 {
1456 	return folio_size(folio) >> inode->i_blkbits;
1457 }
1458 
1459 static inline
i_blocks_per_page(struct inode * inode,struct page * page)1460 unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1461 {
1462 	return i_blocks_per_folio(inode, page_folio(page));
1463 }
1464 #endif /* _LINUX_PAGEMAP_H */
1465