xref: /openbmc/linux/include/linux/pagemap.h (revision 1f3988ca)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_PAGEMAP_H
3 #define _LINUX_PAGEMAP_H
4 
5 /*
6  * Copyright 1995 Linus Torvalds
7  */
8 #include <linux/mm.h>
9 #include <linux/fs.h>
10 #include <linux/list.h>
11 #include <linux/highmem.h>
12 #include <linux/compiler.h>
13 #include <linux/uaccess.h>
14 #include <linux/gfp.h>
15 #include <linux/bitops.h>
16 #include <linux/hardirq.h> /* for in_interrupt() */
17 #include <linux/hugetlb_inline.h>
18 
19 struct folio_batch;
20 
21 unsigned long invalidate_mapping_pages(struct address_space *mapping,
22 					pgoff_t start, pgoff_t end);
23 
invalidate_remote_inode(struct inode * inode)24 static inline void invalidate_remote_inode(struct inode *inode)
25 {
26 	if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
27 	    S_ISLNK(inode->i_mode))
28 		invalidate_mapping_pages(inode->i_mapping, 0, -1);
29 }
30 int invalidate_inode_pages2(struct address_space *mapping);
31 int invalidate_inode_pages2_range(struct address_space *mapping,
32 		pgoff_t start, pgoff_t end);
33 int kiocb_invalidate_pages(struct kiocb *iocb, size_t count);
34 void kiocb_invalidate_post_direct_write(struct kiocb *iocb, size_t count);
35 
36 int write_inode_now(struct inode *, int sync);
37 int filemap_fdatawrite(struct address_space *);
38 int filemap_flush(struct address_space *);
39 int filemap_fdatawait_keep_errors(struct address_space *mapping);
40 int filemap_fdatawait_range(struct address_space *, loff_t lstart, loff_t lend);
41 int filemap_fdatawait_range_keep_errors(struct address_space *mapping,
42 		loff_t start_byte, loff_t end_byte);
43 
filemap_fdatawait(struct address_space * mapping)44 static inline int filemap_fdatawait(struct address_space *mapping)
45 {
46 	return filemap_fdatawait_range(mapping, 0, LLONG_MAX);
47 }
48 
49 bool filemap_range_has_page(struct address_space *, loff_t lstart, loff_t lend);
50 int filemap_write_and_wait_range(struct address_space *mapping,
51 		loff_t lstart, loff_t lend);
52 int __filemap_fdatawrite_range(struct address_space *mapping,
53 		loff_t start, loff_t end, int sync_mode);
54 int filemap_fdatawrite_range(struct address_space *mapping,
55 		loff_t start, loff_t end);
56 int filemap_check_errors(struct address_space *mapping);
57 void __filemap_set_wb_err(struct address_space *mapping, int err);
58 int filemap_fdatawrite_wbc(struct address_space *mapping,
59 			   struct writeback_control *wbc);
60 int kiocb_write_and_wait(struct kiocb *iocb, size_t count);
61 
filemap_write_and_wait(struct address_space * mapping)62 static inline int filemap_write_and_wait(struct address_space *mapping)
63 {
64 	return filemap_write_and_wait_range(mapping, 0, LLONG_MAX);
65 }
66 
67 /**
68  * filemap_set_wb_err - set a writeback error on an address_space
69  * @mapping: mapping in which to set writeback error
70  * @err: error to be set in mapping
71  *
72  * When writeback fails in some way, we must record that error so that
73  * userspace can be informed when fsync and the like are called.  We endeavor
74  * to report errors on any file that was open at the time of the error.  Some
75  * internal callers also need to know when writeback errors have occurred.
76  *
77  * When a writeback error occurs, most filesystems will want to call
78  * filemap_set_wb_err to record the error in the mapping so that it will be
79  * automatically reported whenever fsync is called on the file.
80  */
filemap_set_wb_err(struct address_space * mapping,int err)81 static inline void filemap_set_wb_err(struct address_space *mapping, int err)
82 {
83 	/* Fastpath for common case of no error */
84 	if (unlikely(err))
85 		__filemap_set_wb_err(mapping, err);
86 }
87 
88 /**
89  * filemap_check_wb_err - has an error occurred since the mark was sampled?
90  * @mapping: mapping to check for writeback errors
91  * @since: previously-sampled errseq_t
92  *
93  * Grab the errseq_t value from the mapping, and see if it has changed "since"
94  * the given value was sampled.
95  *
96  * If it has then report the latest error set, otherwise return 0.
97  */
filemap_check_wb_err(struct address_space * mapping,errseq_t since)98 static inline int filemap_check_wb_err(struct address_space *mapping,
99 					errseq_t since)
100 {
101 	return errseq_check(&mapping->wb_err, since);
102 }
103 
104 /**
105  * filemap_sample_wb_err - sample the current errseq_t to test for later errors
106  * @mapping: mapping to be sampled
107  *
108  * Writeback errors are always reported relative to a particular sample point
109  * in the past. This function provides those sample points.
110  */
filemap_sample_wb_err(struct address_space * mapping)111 static inline errseq_t filemap_sample_wb_err(struct address_space *mapping)
112 {
113 	return errseq_sample(&mapping->wb_err);
114 }
115 
116 /**
117  * file_sample_sb_err - sample the current errseq_t to test for later errors
118  * @file: file pointer to be sampled
119  *
120  * Grab the most current superblock-level errseq_t value for the given
121  * struct file.
122  */
file_sample_sb_err(struct file * file)123 static inline errseq_t file_sample_sb_err(struct file *file)
124 {
125 	return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err);
126 }
127 
128 /*
129  * Flush file data before changing attributes.  Caller must hold any locks
130  * required to prevent further writes to this file until we're done setting
131  * flags.
132  */
inode_drain_writes(struct inode * inode)133 static inline int inode_drain_writes(struct inode *inode)
134 {
135 	inode_dio_wait(inode);
136 	return filemap_write_and_wait(inode->i_mapping);
137 }
138 
mapping_empty(struct address_space * mapping)139 static inline bool mapping_empty(struct address_space *mapping)
140 {
141 	return xa_empty(&mapping->i_pages);
142 }
143 
144 /*
145  * mapping_shrinkable - test if page cache state allows inode reclaim
146  * @mapping: the page cache mapping
147  *
148  * This checks the mapping's cache state for the pupose of inode
149  * reclaim and LRU management.
150  *
151  * The caller is expected to hold the i_lock, but is not required to
152  * hold the i_pages lock, which usually protects cache state. That's
153  * because the i_lock and the list_lru lock that protect the inode and
154  * its LRU state don't nest inside the irq-safe i_pages lock.
155  *
156  * Cache deletions are performed under the i_lock, which ensures that
157  * when an inode goes empty, it will reliably get queued on the LRU.
158  *
159  * Cache additions do not acquire the i_lock and may race with this
160  * check, in which case we'll report the inode as shrinkable when it
161  * has cache pages. This is okay: the shrinker also checks the
162  * refcount and the referenced bit, which will be elevated or set in
163  * the process of adding new cache pages to an inode.
164  */
mapping_shrinkable(struct address_space * mapping)165 static inline bool mapping_shrinkable(struct address_space *mapping)
166 {
167 	void *head;
168 
169 	/*
170 	 * On highmem systems, there could be lowmem pressure from the
171 	 * inodes before there is highmem pressure from the page
172 	 * cache. Make inodes shrinkable regardless of cache state.
173 	 */
174 	if (IS_ENABLED(CONFIG_HIGHMEM))
175 		return true;
176 
177 	/* Cache completely empty? Shrink away. */
178 	head = rcu_access_pointer(mapping->i_pages.xa_head);
179 	if (!head)
180 		return true;
181 
182 	/*
183 	 * The xarray stores single offset-0 entries directly in the
184 	 * head pointer, which allows non-resident page cache entries
185 	 * to escape the shadow shrinker's list of xarray nodes. The
186 	 * inode shrinker needs to pick them up under memory pressure.
187 	 */
188 	if (!xa_is_node(head) && xa_is_value(head))
189 		return true;
190 
191 	return false;
192 }
193 
194 /*
195  * Bits in mapping->flags.
196  */
197 enum mapping_flags {
198 	AS_EIO		= 0,	/* IO error on async write */
199 	AS_ENOSPC	= 1,	/* ENOSPC on async write */
200 	AS_MM_ALL_LOCKS	= 2,	/* under mm_take_all_locks() */
201 	AS_UNEVICTABLE	= 3,	/* e.g., ramdisk, SHM_LOCK */
202 	AS_EXITING	= 4, 	/* final truncate in progress */
203 	/* writeback related tags are not used */
204 	AS_NO_WRITEBACK_TAGS = 5,
205 	AS_LARGE_FOLIO_SUPPORT = 6,
206 	AS_RELEASE_ALWAYS,	/* Call ->release_folio(), even if no private data */
207 	AS_STABLE_WRITES,	/* must wait for writeback before modifying
208 				   folio contents */
209 };
210 
211 /**
212  * mapping_set_error - record a writeback error in the address_space
213  * @mapping: the mapping in which an error should be set
214  * @error: the error to set in the mapping
215  *
216  * When writeback fails in some way, we must record that error so that
217  * userspace can be informed when fsync and the like are called.  We endeavor
218  * to report errors on any file that was open at the time of the error.  Some
219  * internal callers also need to know when writeback errors have occurred.
220  *
221  * When a writeback error occurs, most filesystems will want to call
222  * mapping_set_error to record the error in the mapping so that it can be
223  * reported when the application calls fsync(2).
224  */
mapping_set_error(struct address_space * mapping,int error)225 static inline void mapping_set_error(struct address_space *mapping, int error)
226 {
227 	if (likely(!error))
228 		return;
229 
230 	/* Record in wb_err for checkers using errseq_t based tracking */
231 	__filemap_set_wb_err(mapping, error);
232 
233 	/* Record it in superblock */
234 	if (mapping->host)
235 		errseq_set(&mapping->host->i_sb->s_wb_err, error);
236 
237 	/* Record it in flags for now, for legacy callers */
238 	if (error == -ENOSPC)
239 		set_bit(AS_ENOSPC, &mapping->flags);
240 	else
241 		set_bit(AS_EIO, &mapping->flags);
242 }
243 
mapping_set_unevictable(struct address_space * mapping)244 static inline void mapping_set_unevictable(struct address_space *mapping)
245 {
246 	set_bit(AS_UNEVICTABLE, &mapping->flags);
247 }
248 
mapping_clear_unevictable(struct address_space * mapping)249 static inline void mapping_clear_unevictable(struct address_space *mapping)
250 {
251 	clear_bit(AS_UNEVICTABLE, &mapping->flags);
252 }
253 
mapping_unevictable(struct address_space * mapping)254 static inline bool mapping_unevictable(struct address_space *mapping)
255 {
256 	return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
257 }
258 
mapping_set_exiting(struct address_space * mapping)259 static inline void mapping_set_exiting(struct address_space *mapping)
260 {
261 	set_bit(AS_EXITING, &mapping->flags);
262 }
263 
mapping_exiting(struct address_space * mapping)264 static inline int mapping_exiting(struct address_space *mapping)
265 {
266 	return test_bit(AS_EXITING, &mapping->flags);
267 }
268 
mapping_set_no_writeback_tags(struct address_space * mapping)269 static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
270 {
271 	set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
272 }
273 
mapping_use_writeback_tags(struct address_space * mapping)274 static inline int mapping_use_writeback_tags(struct address_space *mapping)
275 {
276 	return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
277 }
278 
mapping_release_always(const struct address_space * mapping)279 static inline bool mapping_release_always(const struct address_space *mapping)
280 {
281 	return test_bit(AS_RELEASE_ALWAYS, &mapping->flags);
282 }
283 
mapping_set_release_always(struct address_space * mapping)284 static inline void mapping_set_release_always(struct address_space *mapping)
285 {
286 	set_bit(AS_RELEASE_ALWAYS, &mapping->flags);
287 }
288 
mapping_clear_release_always(struct address_space * mapping)289 static inline void mapping_clear_release_always(struct address_space *mapping)
290 {
291 	clear_bit(AS_RELEASE_ALWAYS, &mapping->flags);
292 }
293 
mapping_stable_writes(const struct address_space * mapping)294 static inline bool mapping_stable_writes(const struct address_space *mapping)
295 {
296 	return test_bit(AS_STABLE_WRITES, &mapping->flags);
297 }
298 
mapping_set_stable_writes(struct address_space * mapping)299 static inline void mapping_set_stable_writes(struct address_space *mapping)
300 {
301 	set_bit(AS_STABLE_WRITES, &mapping->flags);
302 }
303 
mapping_clear_stable_writes(struct address_space * mapping)304 static inline void mapping_clear_stable_writes(struct address_space *mapping)
305 {
306 	clear_bit(AS_STABLE_WRITES, &mapping->flags);
307 }
308 
mapping_gfp_mask(struct address_space * mapping)309 static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
310 {
311 	return mapping->gfp_mask;
312 }
313 
314 /* Restricts the given gfp_mask to what the mapping allows. */
mapping_gfp_constraint(struct address_space * mapping,gfp_t gfp_mask)315 static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
316 		gfp_t gfp_mask)
317 {
318 	return mapping_gfp_mask(mapping) & gfp_mask;
319 }
320 
321 /*
322  * This is non-atomic.  Only to be used before the mapping is activated.
323  * Probably needs a barrier...
324  */
mapping_set_gfp_mask(struct address_space * m,gfp_t mask)325 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
326 {
327 	m->gfp_mask = mask;
328 }
329 
330 /*
331  * There are some parts of the kernel which assume that PMD entries
332  * are exactly HPAGE_PMD_ORDER.  Those should be fixed, but until then,
333  * limit the maximum allocation order to PMD size.  I'm not aware of any
334  * assumptions about maximum order if THP are disabled, but 8 seems like
335  * a good order (that's 1MB if you're using 4kB pages)
336  */
337 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
338 #define MAX_PAGECACHE_ORDER	HPAGE_PMD_ORDER
339 #else
340 #define MAX_PAGECACHE_ORDER	8
341 #endif
342 
343 /**
344  * mapping_set_large_folios() - Indicate the file supports large folios.
345  * @mapping: The file.
346  *
347  * The filesystem should call this function in its inode constructor to
348  * indicate that the VFS can use large folios to cache the contents of
349  * the file.
350  *
351  * Context: This should not be called while the inode is active as it
352  * is non-atomic.
353  */
mapping_set_large_folios(struct address_space * mapping)354 static inline void mapping_set_large_folios(struct address_space *mapping)
355 {
356 	__set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
357 }
358 
359 /*
360  * Large folio support currently depends on THP.  These dependencies are
361  * being worked on but are not yet fixed.
362  */
mapping_large_folio_support(struct address_space * mapping)363 static inline bool mapping_large_folio_support(struct address_space *mapping)
364 {
365 	return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) &&
366 		test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
367 }
368 
369 /* Return the maximum folio size for this pagecache mapping, in bytes. */
mapping_max_folio_size(struct address_space * mapping)370 static inline size_t mapping_max_folio_size(struct address_space *mapping)
371 {
372 	if (mapping_large_folio_support(mapping))
373 		return PAGE_SIZE << MAX_PAGECACHE_ORDER;
374 	return PAGE_SIZE;
375 }
376 
filemap_nr_thps(struct address_space * mapping)377 static inline int filemap_nr_thps(struct address_space *mapping)
378 {
379 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
380 	return atomic_read(&mapping->nr_thps);
381 #else
382 	return 0;
383 #endif
384 }
385 
filemap_nr_thps_inc(struct address_space * mapping)386 static inline void filemap_nr_thps_inc(struct address_space *mapping)
387 {
388 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
389 	if (!mapping_large_folio_support(mapping))
390 		atomic_inc(&mapping->nr_thps);
391 #else
392 	WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
393 #endif
394 }
395 
filemap_nr_thps_dec(struct address_space * mapping)396 static inline void filemap_nr_thps_dec(struct address_space *mapping)
397 {
398 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
399 	if (!mapping_large_folio_support(mapping))
400 		atomic_dec(&mapping->nr_thps);
401 #else
402 	WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
403 #endif
404 }
405 
406 struct address_space *page_mapping(struct page *);
407 struct address_space *folio_mapping(struct folio *);
408 struct address_space *swapcache_mapping(struct folio *);
409 
410 /**
411  * folio_file_mapping - Find the mapping this folio belongs to.
412  * @folio: The folio.
413  *
414  * For folios which are in the page cache, return the mapping that this
415  * page belongs to.  Folios in the swap cache return the mapping of the
416  * swap file or swap device where the data is stored.  This is different
417  * from the mapping returned by folio_mapping().  The only reason to
418  * use it is if, like NFS, you return 0 from ->activate_swapfile.
419  *
420  * Do not call this for folios which aren't in the page cache or swap cache.
421  */
folio_file_mapping(struct folio * folio)422 static inline struct address_space *folio_file_mapping(struct folio *folio)
423 {
424 	if (unlikely(folio_test_swapcache(folio)))
425 		return swapcache_mapping(folio);
426 
427 	return folio->mapping;
428 }
429 
430 /**
431  * folio_flush_mapping - Find the file mapping this folio belongs to.
432  * @folio: The folio.
433  *
434  * For folios which are in the page cache, return the mapping that this
435  * page belongs to.  Anonymous folios return NULL, even if they're in
436  * the swap cache.  Other kinds of folio also return NULL.
437  *
438  * This is ONLY used by architecture cache flushing code.  If you aren't
439  * writing cache flushing code, you want either folio_mapping() or
440  * folio_file_mapping().
441  */
folio_flush_mapping(struct folio * folio)442 static inline struct address_space *folio_flush_mapping(struct folio *folio)
443 {
444 	if (unlikely(folio_test_swapcache(folio)))
445 		return NULL;
446 
447 	return folio_mapping(folio);
448 }
449 
page_file_mapping(struct page * page)450 static inline struct address_space *page_file_mapping(struct page *page)
451 {
452 	return folio_file_mapping(page_folio(page));
453 }
454 
455 /**
456  * folio_inode - Get the host inode for this folio.
457  * @folio: The folio.
458  *
459  * For folios which are in the page cache, return the inode that this folio
460  * belongs to.
461  *
462  * Do not call this for folios which aren't in the page cache.
463  */
folio_inode(struct folio * folio)464 static inline struct inode *folio_inode(struct folio *folio)
465 {
466 	return folio->mapping->host;
467 }
468 
469 /**
470  * folio_attach_private - Attach private data to a folio.
471  * @folio: Folio to attach data to.
472  * @data: Data to attach to folio.
473  *
474  * Attaching private data to a folio increments the page's reference count.
475  * The data must be detached before the folio will be freed.
476  */
folio_attach_private(struct folio * folio,void * data)477 static inline void folio_attach_private(struct folio *folio, void *data)
478 {
479 	folio_get(folio);
480 	folio->private = data;
481 	folio_set_private(folio);
482 }
483 
484 /**
485  * folio_change_private - Change private data on a folio.
486  * @folio: Folio to change the data on.
487  * @data: Data to set on the folio.
488  *
489  * Change the private data attached to a folio and return the old
490  * data.  The page must previously have had data attached and the data
491  * must be detached before the folio will be freed.
492  *
493  * Return: Data that was previously attached to the folio.
494  */
folio_change_private(struct folio * folio,void * data)495 static inline void *folio_change_private(struct folio *folio, void *data)
496 {
497 	void *old = folio_get_private(folio);
498 
499 	folio->private = data;
500 	return old;
501 }
502 
503 /**
504  * folio_detach_private - Detach private data from a folio.
505  * @folio: Folio to detach data from.
506  *
507  * Removes the data that was previously attached to the folio and decrements
508  * the refcount on the page.
509  *
510  * Return: Data that was attached to the folio.
511  */
folio_detach_private(struct folio * folio)512 static inline void *folio_detach_private(struct folio *folio)
513 {
514 	void *data = folio_get_private(folio);
515 
516 	if (!folio_test_private(folio))
517 		return NULL;
518 	folio_clear_private(folio);
519 	folio->private = NULL;
520 	folio_put(folio);
521 
522 	return data;
523 }
524 
attach_page_private(struct page * page,void * data)525 static inline void attach_page_private(struct page *page, void *data)
526 {
527 	folio_attach_private(page_folio(page), data);
528 }
529 
detach_page_private(struct page * page)530 static inline void *detach_page_private(struct page *page)
531 {
532 	return folio_detach_private(page_folio(page));
533 }
534 
535 #ifdef CONFIG_NUMA
536 struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order);
537 #else
filemap_alloc_folio(gfp_t gfp,unsigned int order)538 static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order)
539 {
540 	return folio_alloc(gfp, order);
541 }
542 #endif
543 
__page_cache_alloc(gfp_t gfp)544 static inline struct page *__page_cache_alloc(gfp_t gfp)
545 {
546 	return &filemap_alloc_folio(gfp, 0)->page;
547 }
548 
page_cache_alloc(struct address_space * x)549 static inline struct page *page_cache_alloc(struct address_space *x)
550 {
551 	return __page_cache_alloc(mapping_gfp_mask(x));
552 }
553 
readahead_gfp_mask(struct address_space * x)554 static inline gfp_t readahead_gfp_mask(struct address_space *x)
555 {
556 	return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
557 }
558 
559 typedef int filler_t(struct file *, struct folio *);
560 
561 pgoff_t page_cache_next_miss(struct address_space *mapping,
562 			     pgoff_t index, unsigned long max_scan);
563 pgoff_t page_cache_prev_miss(struct address_space *mapping,
564 			     pgoff_t index, unsigned long max_scan);
565 
566 /**
567  * typedef fgf_t - Flags for getting folios from the page cache.
568  *
569  * Most users of the page cache will not need to use these flags;
570  * there are convenience functions such as filemap_get_folio() and
571  * filemap_lock_folio().  For users which need more control over exactly
572  * what is done with the folios, these flags to __filemap_get_folio()
573  * are available.
574  *
575  * * %FGP_ACCESSED - The folio will be marked accessed.
576  * * %FGP_LOCK - The folio is returned locked.
577  * * %FGP_CREAT - If no folio is present then a new folio is allocated,
578  *   added to the page cache and the VM's LRU list.  The folio is
579  *   returned locked.
580  * * %FGP_FOR_MMAP - The caller wants to do its own locking dance if the
581  *   folio is already in cache.  If the folio was allocated, unlock it
582  *   before returning so the caller can do the same dance.
583  * * %FGP_WRITE - The folio will be written to by the caller.
584  * * %FGP_NOFS - __GFP_FS will get cleared in gfp.
585  * * %FGP_NOWAIT - Don't block on the folio lock.
586  * * %FGP_STABLE - Wait for the folio to be stable (finished writeback)
587  * * %FGP_WRITEBEGIN - The flags to use in a filesystem write_begin()
588  *   implementation.
589  */
590 typedef unsigned int __bitwise fgf_t;
591 
592 #define FGP_ACCESSED		((__force fgf_t)0x00000001)
593 #define FGP_LOCK		((__force fgf_t)0x00000002)
594 #define FGP_CREAT		((__force fgf_t)0x00000004)
595 #define FGP_WRITE		((__force fgf_t)0x00000008)
596 #define FGP_NOFS		((__force fgf_t)0x00000010)
597 #define FGP_NOWAIT		((__force fgf_t)0x00000020)
598 #define FGP_FOR_MMAP		((__force fgf_t)0x00000040)
599 #define FGP_STABLE		((__force fgf_t)0x00000080)
600 #define FGF_GET_ORDER(fgf)	(((__force unsigned)fgf) >> 26)	/* top 6 bits */
601 
602 #define FGP_WRITEBEGIN		(FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE)
603 
604 /**
605  * fgf_set_order - Encode a length in the fgf_t flags.
606  * @size: The suggested size of the folio to create.
607  *
608  * The caller of __filemap_get_folio() can use this to suggest a preferred
609  * size for the folio that is created.  If there is already a folio at
610  * the index, it will be returned, no matter what its size.  If a folio
611  * is freshly created, it may be of a different size than requested
612  * due to alignment constraints, memory pressure, or the presence of
613  * other folios at nearby indices.
614  */
fgf_set_order(size_t size)615 static inline fgf_t fgf_set_order(size_t size)
616 {
617 	unsigned int shift = ilog2(size);
618 
619 	if (shift <= PAGE_SHIFT)
620 		return 0;
621 	return (__force fgf_t)((shift - PAGE_SHIFT) << 26);
622 }
623 
624 void *filemap_get_entry(struct address_space *mapping, pgoff_t index);
625 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
626 		fgf_t fgp_flags, gfp_t gfp);
627 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
628 		fgf_t fgp_flags, gfp_t gfp);
629 
630 /**
631  * filemap_get_folio - Find and get a folio.
632  * @mapping: The address_space to search.
633  * @index: The page index.
634  *
635  * Looks up the page cache entry at @mapping & @index.  If a folio is
636  * present, it is returned with an increased refcount.
637  *
638  * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
639  * this index.  Will not return a shadow, swap or DAX entry.
640  */
filemap_get_folio(struct address_space * mapping,pgoff_t index)641 static inline struct folio *filemap_get_folio(struct address_space *mapping,
642 					pgoff_t index)
643 {
644 	return __filemap_get_folio(mapping, index, 0, 0);
645 }
646 
647 /**
648  * filemap_lock_folio - Find and lock a folio.
649  * @mapping: The address_space to search.
650  * @index: The page index.
651  *
652  * Looks up the page cache entry at @mapping & @index.  If a folio is
653  * present, it is returned locked with an increased refcount.
654  *
655  * Context: May sleep.
656  * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
657  * this index.  Will not return a shadow, swap or DAX entry.
658  */
filemap_lock_folio(struct address_space * mapping,pgoff_t index)659 static inline struct folio *filemap_lock_folio(struct address_space *mapping,
660 					pgoff_t index)
661 {
662 	return __filemap_get_folio(mapping, index, FGP_LOCK, 0);
663 }
664 
665 /**
666  * filemap_grab_folio - grab a folio from the page cache
667  * @mapping: The address space to search
668  * @index: The page index
669  *
670  * Looks up the page cache entry at @mapping & @index. If no folio is found,
671  * a new folio is created. The folio is locked, marked as accessed, and
672  * returned.
673  *
674  * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found
675  * and failed to create a folio.
676  */
filemap_grab_folio(struct address_space * mapping,pgoff_t index)677 static inline struct folio *filemap_grab_folio(struct address_space *mapping,
678 					pgoff_t index)
679 {
680 	return __filemap_get_folio(mapping, index,
681 			FGP_LOCK | FGP_ACCESSED | FGP_CREAT,
682 			mapping_gfp_mask(mapping));
683 }
684 
685 /**
686  * find_get_page - find and get a page reference
687  * @mapping: the address_space to search
688  * @offset: the page index
689  *
690  * Looks up the page cache slot at @mapping & @offset.  If there is a
691  * page cache page, it is returned with an increased refcount.
692  *
693  * Otherwise, %NULL is returned.
694  */
find_get_page(struct address_space * mapping,pgoff_t offset)695 static inline struct page *find_get_page(struct address_space *mapping,
696 					pgoff_t offset)
697 {
698 	return pagecache_get_page(mapping, offset, 0, 0);
699 }
700 
find_get_page_flags(struct address_space * mapping,pgoff_t offset,fgf_t fgp_flags)701 static inline struct page *find_get_page_flags(struct address_space *mapping,
702 					pgoff_t offset, fgf_t fgp_flags)
703 {
704 	return pagecache_get_page(mapping, offset, fgp_flags, 0);
705 }
706 
707 /**
708  * find_lock_page - locate, pin and lock a pagecache page
709  * @mapping: the address_space to search
710  * @index: the page index
711  *
712  * Looks up the page cache entry at @mapping & @index.  If there is a
713  * page cache page, it is returned locked and with an increased
714  * refcount.
715  *
716  * Context: May sleep.
717  * Return: A struct page or %NULL if there is no page in the cache for this
718  * index.
719  */
find_lock_page(struct address_space * mapping,pgoff_t index)720 static inline struct page *find_lock_page(struct address_space *mapping,
721 					pgoff_t index)
722 {
723 	return pagecache_get_page(mapping, index, FGP_LOCK, 0);
724 }
725 
726 /**
727  * find_or_create_page - locate or add a pagecache page
728  * @mapping: the page's address_space
729  * @index: the page's index into the mapping
730  * @gfp_mask: page allocation mode
731  *
732  * Looks up the page cache slot at @mapping & @offset.  If there is a
733  * page cache page, it is returned locked and with an increased
734  * refcount.
735  *
736  * If the page is not present, a new page is allocated using @gfp_mask
737  * and added to the page cache and the VM's LRU list.  The page is
738  * returned locked and with an increased refcount.
739  *
740  * On memory exhaustion, %NULL is returned.
741  *
742  * find_or_create_page() may sleep, even if @gfp_flags specifies an
743  * atomic allocation!
744  */
find_or_create_page(struct address_space * mapping,pgoff_t index,gfp_t gfp_mask)745 static inline struct page *find_or_create_page(struct address_space *mapping,
746 					pgoff_t index, gfp_t gfp_mask)
747 {
748 	return pagecache_get_page(mapping, index,
749 					FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
750 					gfp_mask);
751 }
752 
753 /**
754  * grab_cache_page_nowait - returns locked page at given index in given cache
755  * @mapping: target address_space
756  * @index: the page index
757  *
758  * Same as grab_cache_page(), but do not wait if the page is unavailable.
759  * This is intended for speculative data generators, where the data can
760  * be regenerated if the page couldn't be grabbed.  This routine should
761  * be safe to call while holding the lock for another page.
762  *
763  * Clear __GFP_FS when allocating the page to avoid recursion into the fs
764  * and deadlock against the caller's locked page.
765  */
grab_cache_page_nowait(struct address_space * mapping,pgoff_t index)766 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
767 				pgoff_t index)
768 {
769 	return pagecache_get_page(mapping, index,
770 			FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
771 			mapping_gfp_mask(mapping));
772 }
773 
774 #define swapcache_index(folio)	__page_file_index(&(folio)->page)
775 
776 /**
777  * folio_index - File index of a folio.
778  * @folio: The folio.
779  *
780  * For a folio which is either in the page cache or the swap cache,
781  * return its index within the address_space it belongs to.  If you know
782  * the page is definitely in the page cache, you can look at the folio's
783  * index directly.
784  *
785  * Return: The index (offset in units of pages) of a folio in its file.
786  */
folio_index(struct folio * folio)787 static inline pgoff_t folio_index(struct folio *folio)
788 {
789         if (unlikely(folio_test_swapcache(folio)))
790                 return swapcache_index(folio);
791         return folio->index;
792 }
793 
794 /**
795  * folio_next_index - Get the index of the next folio.
796  * @folio: The current folio.
797  *
798  * Return: The index of the folio which follows this folio in the file.
799  */
folio_next_index(struct folio * folio)800 static inline pgoff_t folio_next_index(struct folio *folio)
801 {
802 	return folio->index + folio_nr_pages(folio);
803 }
804 
805 /**
806  * folio_file_page - The page for a particular index.
807  * @folio: The folio which contains this index.
808  * @index: The index we want to look up.
809  *
810  * Sometimes after looking up a folio in the page cache, we need to
811  * obtain the specific page for an index (eg a page fault).
812  *
813  * Return: The page containing the file data for this index.
814  */
folio_file_page(struct folio * folio,pgoff_t index)815 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
816 {
817 	/* HugeTLBfs indexes the page cache in units of hpage_size */
818 	if (folio_test_hugetlb(folio))
819 		return &folio->page;
820 	return folio_page(folio, index & (folio_nr_pages(folio) - 1));
821 }
822 
823 /**
824  * folio_contains - Does this folio contain this index?
825  * @folio: The folio.
826  * @index: The page index within the file.
827  *
828  * Context: The caller should have the page locked in order to prevent
829  * (eg) shmem from moving the page between the page cache and swap cache
830  * and changing its index in the middle of the operation.
831  * Return: true or false.
832  */
folio_contains(struct folio * folio,pgoff_t index)833 static inline bool folio_contains(struct folio *folio, pgoff_t index)
834 {
835 	/* HugeTLBfs indexes the page cache in units of hpage_size */
836 	if (folio_test_hugetlb(folio))
837 		return folio->index == index;
838 	return index - folio_index(folio) < folio_nr_pages(folio);
839 }
840 
841 /*
842  * Given the page we found in the page cache, return the page corresponding
843  * to this index in the file
844  */
find_subpage(struct page * head,pgoff_t index)845 static inline struct page *find_subpage(struct page *head, pgoff_t index)
846 {
847 	/* HugeTLBfs wants the head page regardless */
848 	if (PageHuge(head))
849 		return head;
850 
851 	return head + (index & (thp_nr_pages(head) - 1));
852 }
853 
854 unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start,
855 		pgoff_t end, struct folio_batch *fbatch);
856 unsigned filemap_get_folios_contig(struct address_space *mapping,
857 		pgoff_t *start, pgoff_t end, struct folio_batch *fbatch);
858 unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start,
859 		pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch);
860 
861 struct page *grab_cache_page_write_begin(struct address_space *mapping,
862 			pgoff_t index);
863 
864 /*
865  * Returns locked page at given index in given cache, creating it if needed.
866  */
grab_cache_page(struct address_space * mapping,pgoff_t index)867 static inline struct page *grab_cache_page(struct address_space *mapping,
868 								pgoff_t index)
869 {
870 	return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
871 }
872 
873 struct folio *read_cache_folio(struct address_space *, pgoff_t index,
874 		filler_t *filler, struct file *file);
875 struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index,
876 		gfp_t flags);
877 struct page *read_cache_page(struct address_space *, pgoff_t index,
878 		filler_t *filler, struct file *file);
879 extern struct page * read_cache_page_gfp(struct address_space *mapping,
880 				pgoff_t index, gfp_t gfp_mask);
881 
read_mapping_page(struct address_space * mapping,pgoff_t index,struct file * file)882 static inline struct page *read_mapping_page(struct address_space *mapping,
883 				pgoff_t index, struct file *file)
884 {
885 	return read_cache_page(mapping, index, NULL, file);
886 }
887 
read_mapping_folio(struct address_space * mapping,pgoff_t index,struct file * file)888 static inline struct folio *read_mapping_folio(struct address_space *mapping,
889 				pgoff_t index, struct file *file)
890 {
891 	return read_cache_folio(mapping, index, NULL, file);
892 }
893 
894 /*
895  * Get index of the page within radix-tree (but not for hugetlb pages).
896  * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
897  */
page_to_index(struct page * page)898 static inline pgoff_t page_to_index(struct page *page)
899 {
900 	struct page *head;
901 
902 	if (likely(!PageTransTail(page)))
903 		return page->index;
904 
905 	head = compound_head(page);
906 	/*
907 	 *  We don't initialize ->index for tail pages: calculate based on
908 	 *  head page
909 	 */
910 	return head->index + page - head;
911 }
912 
913 extern pgoff_t hugetlb_basepage_index(struct page *page);
914 
915 /*
916  * Get the offset in PAGE_SIZE (even for hugetlb pages).
917  * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
918  */
page_to_pgoff(struct page * page)919 static inline pgoff_t page_to_pgoff(struct page *page)
920 {
921 	if (unlikely(PageHuge(page)))
922 		return hugetlb_basepage_index(page);
923 	return page_to_index(page);
924 }
925 
926 /*
927  * Return byte-offset into filesystem object for page.
928  */
page_offset(struct page * page)929 static inline loff_t page_offset(struct page *page)
930 {
931 	return ((loff_t)page->index) << PAGE_SHIFT;
932 }
933 
page_file_offset(struct page * page)934 static inline loff_t page_file_offset(struct page *page)
935 {
936 	return ((loff_t)page_index(page)) << PAGE_SHIFT;
937 }
938 
939 /**
940  * folio_pos - Returns the byte position of this folio in its file.
941  * @folio: The folio.
942  */
folio_pos(struct folio * folio)943 static inline loff_t folio_pos(struct folio *folio)
944 {
945 	return page_offset(&folio->page);
946 }
947 
948 /**
949  * folio_file_pos - Returns the byte position of this folio in its file.
950  * @folio: The folio.
951  *
952  * This differs from folio_pos() for folios which belong to a swap file.
953  * NFS is the only filesystem today which needs to use folio_file_pos().
954  */
folio_file_pos(struct folio * folio)955 static inline loff_t folio_file_pos(struct folio *folio)
956 {
957 	return page_file_offset(&folio->page);
958 }
959 
960 /*
961  * Get the offset in PAGE_SIZE (even for hugetlb folios).
962  * (TODO: hugetlb folios should have ->index in PAGE_SIZE)
963  */
folio_pgoff(struct folio * folio)964 static inline pgoff_t folio_pgoff(struct folio *folio)
965 {
966 	if (unlikely(folio_test_hugetlb(folio)))
967 		return hugetlb_basepage_index(&folio->page);
968 	return folio->index;
969 }
970 
971 extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
972 				     unsigned long address);
973 
linear_page_index(struct vm_area_struct * vma,unsigned long address)974 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
975 					unsigned long address)
976 {
977 	pgoff_t pgoff;
978 	if (unlikely(is_vm_hugetlb_page(vma)))
979 		return linear_hugepage_index(vma, address);
980 	pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
981 	pgoff += vma->vm_pgoff;
982 	return pgoff;
983 }
984 
985 struct wait_page_key {
986 	struct folio *folio;
987 	int bit_nr;
988 	int page_match;
989 };
990 
991 struct wait_page_queue {
992 	struct folio *folio;
993 	int bit_nr;
994 	wait_queue_entry_t wait;
995 };
996 
wake_page_match(struct wait_page_queue * wait_page,struct wait_page_key * key)997 static inline bool wake_page_match(struct wait_page_queue *wait_page,
998 				  struct wait_page_key *key)
999 {
1000 	if (wait_page->folio != key->folio)
1001 	       return false;
1002 	key->page_match = 1;
1003 
1004 	if (wait_page->bit_nr != key->bit_nr)
1005 		return false;
1006 
1007 	return true;
1008 }
1009 
1010 void __folio_lock(struct folio *folio);
1011 int __folio_lock_killable(struct folio *folio);
1012 vm_fault_t __folio_lock_or_retry(struct folio *folio, struct vm_fault *vmf);
1013 void unlock_page(struct page *page);
1014 void folio_unlock(struct folio *folio);
1015 
1016 /**
1017  * folio_trylock() - Attempt to lock a folio.
1018  * @folio: The folio to attempt to lock.
1019  *
1020  * Sometimes it is undesirable to wait for a folio to be unlocked (eg
1021  * when the locks are being taken in the wrong order, or if making
1022  * progress through a batch of folios is more important than processing
1023  * them in order).  Usually folio_lock() is the correct function to call.
1024  *
1025  * Context: Any context.
1026  * Return: Whether the lock was successfully acquired.
1027  */
folio_trylock(struct folio * folio)1028 static inline bool folio_trylock(struct folio *folio)
1029 {
1030 	return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
1031 }
1032 
1033 /*
1034  * Return true if the page was successfully locked
1035  */
trylock_page(struct page * page)1036 static inline int trylock_page(struct page *page)
1037 {
1038 	return folio_trylock(page_folio(page));
1039 }
1040 
1041 /**
1042  * folio_lock() - Lock this folio.
1043  * @folio: The folio to lock.
1044  *
1045  * The folio lock protects against many things, probably more than it
1046  * should.  It is primarily held while a folio is being brought uptodate,
1047  * either from its backing file or from swap.  It is also held while a
1048  * folio is being truncated from its address_space, so holding the lock
1049  * is sufficient to keep folio->mapping stable.
1050  *
1051  * The folio lock is also held while write() is modifying the page to
1052  * provide POSIX atomicity guarantees (as long as the write does not
1053  * cross a page boundary).  Other modifications to the data in the folio
1054  * do not hold the folio lock and can race with writes, eg DMA and stores
1055  * to mapped pages.
1056  *
1057  * Context: May sleep.  If you need to acquire the locks of two or
1058  * more folios, they must be in order of ascending index, if they are
1059  * in the same address_space.  If they are in different address_spaces,
1060  * acquire the lock of the folio which belongs to the address_space which
1061  * has the lowest address in memory first.
1062  */
folio_lock(struct folio * folio)1063 static inline void folio_lock(struct folio *folio)
1064 {
1065 	might_sleep();
1066 	if (!folio_trylock(folio))
1067 		__folio_lock(folio);
1068 }
1069 
1070 /**
1071  * lock_page() - Lock the folio containing this page.
1072  * @page: The page to lock.
1073  *
1074  * See folio_lock() for a description of what the lock protects.
1075  * This is a legacy function and new code should probably use folio_lock()
1076  * instead.
1077  *
1078  * Context: May sleep.  Pages in the same folio share a lock, so do not
1079  * attempt to lock two pages which share a folio.
1080  */
lock_page(struct page * page)1081 static inline void lock_page(struct page *page)
1082 {
1083 	struct folio *folio;
1084 	might_sleep();
1085 
1086 	folio = page_folio(page);
1087 	if (!folio_trylock(folio))
1088 		__folio_lock(folio);
1089 }
1090 
1091 /**
1092  * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
1093  * @folio: The folio to lock.
1094  *
1095  * Attempts to lock the folio, like folio_lock(), except that the sleep
1096  * to acquire the lock is interruptible by a fatal signal.
1097  *
1098  * Context: May sleep; see folio_lock().
1099  * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
1100  */
folio_lock_killable(struct folio * folio)1101 static inline int folio_lock_killable(struct folio *folio)
1102 {
1103 	might_sleep();
1104 	if (!folio_trylock(folio))
1105 		return __folio_lock_killable(folio);
1106 	return 0;
1107 }
1108 
1109 /*
1110  * folio_lock_or_retry - Lock the folio, unless this would block and the
1111  * caller indicated that it can handle a retry.
1112  *
1113  * Return value and mmap_lock implications depend on flags; see
1114  * __folio_lock_or_retry().
1115  */
folio_lock_or_retry(struct folio * folio,struct vm_fault * vmf)1116 static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
1117 					     struct vm_fault *vmf)
1118 {
1119 	might_sleep();
1120 	if (!folio_trylock(folio))
1121 		return __folio_lock_or_retry(folio, vmf);
1122 	return 0;
1123 }
1124 
1125 /*
1126  * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
1127  * and should not be used directly.
1128  */
1129 void folio_wait_bit(struct folio *folio, int bit_nr);
1130 int folio_wait_bit_killable(struct folio *folio, int bit_nr);
1131 
1132 /*
1133  * Wait for a folio to be unlocked.
1134  *
1135  * This must be called with the caller "holding" the folio,
1136  * ie with increased folio reference count so that the folio won't
1137  * go away during the wait.
1138  */
folio_wait_locked(struct folio * folio)1139 static inline void folio_wait_locked(struct folio *folio)
1140 {
1141 	if (folio_test_locked(folio))
1142 		folio_wait_bit(folio, PG_locked);
1143 }
1144 
folio_wait_locked_killable(struct folio * folio)1145 static inline int folio_wait_locked_killable(struct folio *folio)
1146 {
1147 	if (!folio_test_locked(folio))
1148 		return 0;
1149 	return folio_wait_bit_killable(folio, PG_locked);
1150 }
1151 
wait_on_page_locked(struct page * page)1152 static inline void wait_on_page_locked(struct page *page)
1153 {
1154 	folio_wait_locked(page_folio(page));
1155 }
1156 
1157 void wait_on_page_writeback(struct page *page);
1158 void folio_wait_writeback(struct folio *folio);
1159 int folio_wait_writeback_killable(struct folio *folio);
1160 void end_page_writeback(struct page *page);
1161 void folio_end_writeback(struct folio *folio);
1162 void wait_for_stable_page(struct page *page);
1163 void folio_wait_stable(struct folio *folio);
1164 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
__set_page_dirty(struct page * page,struct address_space * mapping,int warn)1165 static inline void __set_page_dirty(struct page *page,
1166 		struct address_space *mapping, int warn)
1167 {
1168 	__folio_mark_dirty(page_folio(page), mapping, warn);
1169 }
1170 void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
1171 void __folio_cancel_dirty(struct folio *folio);
folio_cancel_dirty(struct folio * folio)1172 static inline void folio_cancel_dirty(struct folio *folio)
1173 {
1174 	/* Avoid atomic ops, locking, etc. when not actually needed. */
1175 	if (folio_test_dirty(folio))
1176 		__folio_cancel_dirty(folio);
1177 }
1178 bool folio_clear_dirty_for_io(struct folio *folio);
1179 bool clear_page_dirty_for_io(struct page *page);
1180 void folio_invalidate(struct folio *folio, size_t offset, size_t length);
1181 int __set_page_dirty_nobuffers(struct page *page);
1182 bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
1183 
1184 #ifdef CONFIG_MIGRATION
1185 int filemap_migrate_folio(struct address_space *mapping, struct folio *dst,
1186 		struct folio *src, enum migrate_mode mode);
1187 #else
1188 #define filemap_migrate_folio NULL
1189 #endif
1190 void folio_end_private_2(struct folio *folio);
1191 void folio_wait_private_2(struct folio *folio);
1192 int folio_wait_private_2_killable(struct folio *folio);
1193 
1194 /*
1195  * Add an arbitrary waiter to a page's wait queue
1196  */
1197 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
1198 
1199 /*
1200  * Fault in userspace address range.
1201  */
1202 size_t fault_in_writeable(char __user *uaddr, size_t size);
1203 size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
1204 size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
1205 size_t fault_in_readable(const char __user *uaddr, size_t size);
1206 
1207 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
1208 		pgoff_t index, gfp_t gfp);
1209 int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1210 		pgoff_t index, gfp_t gfp);
1211 void filemap_remove_folio(struct folio *folio);
1212 void __filemap_remove_folio(struct folio *folio, void *shadow);
1213 void replace_page_cache_folio(struct folio *old, struct folio *new);
1214 void delete_from_page_cache_batch(struct address_space *mapping,
1215 				  struct folio_batch *fbatch);
1216 bool filemap_release_folio(struct folio *folio, gfp_t gfp);
1217 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1218 		int whence);
1219 
1220 /* Must be non-static for BPF error injection */
1221 int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1222 		pgoff_t index, gfp_t gfp, void **shadowp);
1223 
1224 bool filemap_range_has_writeback(struct address_space *mapping,
1225 				 loff_t start_byte, loff_t end_byte);
1226 
1227 /**
1228  * filemap_range_needs_writeback - check if range potentially needs writeback
1229  * @mapping:           address space within which to check
1230  * @start_byte:        offset in bytes where the range starts
1231  * @end_byte:          offset in bytes where the range ends (inclusive)
1232  *
1233  * Find at least one page in the range supplied, usually used to check if
1234  * direct writing in this range will trigger a writeback. Used by O_DIRECT
1235  * read/write with IOCB_NOWAIT, to see if the caller needs to do
1236  * filemap_write_and_wait_range() before proceeding.
1237  *
1238  * Return: %true if the caller should do filemap_write_and_wait_range() before
1239  * doing O_DIRECT to a page in this range, %false otherwise.
1240  */
filemap_range_needs_writeback(struct address_space * mapping,loff_t start_byte,loff_t end_byte)1241 static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1242 						 loff_t start_byte,
1243 						 loff_t end_byte)
1244 {
1245 	if (!mapping->nrpages)
1246 		return false;
1247 	if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1248 	    !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1249 		return false;
1250 	return filemap_range_has_writeback(mapping, start_byte, end_byte);
1251 }
1252 
1253 /**
1254  * struct readahead_control - Describes a readahead request.
1255  *
1256  * A readahead request is for consecutive pages.  Filesystems which
1257  * implement the ->readahead method should call readahead_page() or
1258  * readahead_page_batch() in a loop and attempt to start I/O against
1259  * each page in the request.
1260  *
1261  * Most of the fields in this struct are private and should be accessed
1262  * by the functions below.
1263  *
1264  * @file: The file, used primarily by network filesystems for authentication.
1265  *	  May be NULL if invoked internally by the filesystem.
1266  * @mapping: Readahead this filesystem object.
1267  * @ra: File readahead state.  May be NULL.
1268  */
1269 struct readahead_control {
1270 	struct file *file;
1271 	struct address_space *mapping;
1272 	struct file_ra_state *ra;
1273 /* private: use the readahead_* accessors instead */
1274 	pgoff_t _index;
1275 	unsigned int _nr_pages;
1276 	unsigned int _batch_count;
1277 	bool _workingset;
1278 	unsigned long _pflags;
1279 };
1280 
1281 #define DEFINE_READAHEAD(ractl, f, r, m, i)				\
1282 	struct readahead_control ractl = {				\
1283 		.file = f,						\
1284 		.mapping = m,						\
1285 		.ra = r,						\
1286 		._index = i,						\
1287 	}
1288 
1289 #define VM_READAHEAD_PAGES	(SZ_128K / PAGE_SIZE)
1290 
1291 void page_cache_ra_unbounded(struct readahead_control *,
1292 		unsigned long nr_to_read, unsigned long lookahead_count);
1293 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
1294 void page_cache_async_ra(struct readahead_control *, struct folio *,
1295 		unsigned long req_count);
1296 void readahead_expand(struct readahead_control *ractl,
1297 		      loff_t new_start, size_t new_len);
1298 
1299 /**
1300  * page_cache_sync_readahead - generic file readahead
1301  * @mapping: address_space which holds the pagecache and I/O vectors
1302  * @ra: file_ra_state which holds the readahead state
1303  * @file: Used by the filesystem for authentication.
1304  * @index: Index of first page to be read.
1305  * @req_count: Total number of pages being read by the caller.
1306  *
1307  * page_cache_sync_readahead() should be called when a cache miss happened:
1308  * it will submit the read.  The readahead logic may decide to piggyback more
1309  * pages onto the read request if access patterns suggest it will improve
1310  * performance.
1311  */
1312 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)1313 void page_cache_sync_readahead(struct address_space *mapping,
1314 		struct file_ra_state *ra, struct file *file, pgoff_t index,
1315 		unsigned long req_count)
1316 {
1317 	DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1318 	page_cache_sync_ra(&ractl, req_count);
1319 }
1320 
1321 /**
1322  * page_cache_async_readahead - file readahead for marked pages
1323  * @mapping: address_space which holds the pagecache and I/O vectors
1324  * @ra: file_ra_state which holds the readahead state
1325  * @file: Used by the filesystem for authentication.
1326  * @folio: The folio at @index which triggered the readahead call.
1327  * @index: Index of first page to be read.
1328  * @req_count: Total number of pages being read by the caller.
1329  *
1330  * page_cache_async_readahead() should be called when a page is used which
1331  * is marked as PageReadahead; this is a marker to suggest that the application
1332  * has used up enough of the readahead window that we should start pulling in
1333  * more pages.
1334  */
1335 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)1336 void page_cache_async_readahead(struct address_space *mapping,
1337 		struct file_ra_state *ra, struct file *file,
1338 		struct folio *folio, pgoff_t index, unsigned long req_count)
1339 {
1340 	DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1341 	page_cache_async_ra(&ractl, folio, req_count);
1342 }
1343 
__readahead_folio(struct readahead_control * ractl)1344 static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1345 {
1346 	struct folio *folio;
1347 
1348 	BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1349 	ractl->_nr_pages -= ractl->_batch_count;
1350 	ractl->_index += ractl->_batch_count;
1351 
1352 	if (!ractl->_nr_pages) {
1353 		ractl->_batch_count = 0;
1354 		return NULL;
1355 	}
1356 
1357 	folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1358 	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1359 	ractl->_batch_count = folio_nr_pages(folio);
1360 
1361 	return folio;
1362 }
1363 
1364 /**
1365  * readahead_page - Get the next page to read.
1366  * @ractl: The current readahead request.
1367  *
1368  * Context: The page is locked and has an elevated refcount.  The caller
1369  * should decreases the refcount once the page has been submitted for I/O
1370  * and unlock the page once all I/O to that page has completed.
1371  * Return: A pointer to the next page, or %NULL if we are done.
1372  */
readahead_page(struct readahead_control * ractl)1373 static inline struct page *readahead_page(struct readahead_control *ractl)
1374 {
1375 	struct folio *folio = __readahead_folio(ractl);
1376 
1377 	return &folio->page;
1378 }
1379 
1380 /**
1381  * readahead_folio - Get the next folio to read.
1382  * @ractl: The current readahead request.
1383  *
1384  * Context: The folio is locked.  The caller should unlock the folio once
1385  * all I/O to that folio has completed.
1386  * Return: A pointer to the next folio, or %NULL if we are done.
1387  */
readahead_folio(struct readahead_control * ractl)1388 static inline struct folio *readahead_folio(struct readahead_control *ractl)
1389 {
1390 	struct folio *folio = __readahead_folio(ractl);
1391 
1392 	if (folio)
1393 		folio_put(folio);
1394 	return folio;
1395 }
1396 
__readahead_batch(struct readahead_control * rac,struct page ** array,unsigned int array_sz)1397 static inline unsigned int __readahead_batch(struct readahead_control *rac,
1398 		struct page **array, unsigned int array_sz)
1399 {
1400 	unsigned int i = 0;
1401 	XA_STATE(xas, &rac->mapping->i_pages, 0);
1402 	struct page *page;
1403 
1404 	BUG_ON(rac->_batch_count > rac->_nr_pages);
1405 	rac->_nr_pages -= rac->_batch_count;
1406 	rac->_index += rac->_batch_count;
1407 	rac->_batch_count = 0;
1408 
1409 	xas_set(&xas, rac->_index);
1410 	rcu_read_lock();
1411 	xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1412 		if (xas_retry(&xas, page))
1413 			continue;
1414 		VM_BUG_ON_PAGE(!PageLocked(page), page);
1415 		VM_BUG_ON_PAGE(PageTail(page), page);
1416 		array[i++] = page;
1417 		rac->_batch_count += thp_nr_pages(page);
1418 		if (i == array_sz)
1419 			break;
1420 	}
1421 	rcu_read_unlock();
1422 
1423 	return i;
1424 }
1425 
1426 /**
1427  * readahead_page_batch - Get a batch of pages to read.
1428  * @rac: The current readahead request.
1429  * @array: An array of pointers to struct page.
1430  *
1431  * Context: The pages are locked and have an elevated refcount.  The caller
1432  * should decreases the refcount once the page has been submitted for I/O
1433  * and unlock the page once all I/O to that page has completed.
1434  * Return: The number of pages placed in the array.  0 indicates the request
1435  * is complete.
1436  */
1437 #define readahead_page_batch(rac, array)				\
1438 	__readahead_batch(rac, array, ARRAY_SIZE(array))
1439 
1440 /**
1441  * readahead_pos - The byte offset into the file of this readahead request.
1442  * @rac: The readahead request.
1443  */
readahead_pos(struct readahead_control * rac)1444 static inline loff_t readahead_pos(struct readahead_control *rac)
1445 {
1446 	return (loff_t)rac->_index * PAGE_SIZE;
1447 }
1448 
1449 /**
1450  * readahead_length - The number of bytes in this readahead request.
1451  * @rac: The readahead request.
1452  */
readahead_length(struct readahead_control * rac)1453 static inline size_t readahead_length(struct readahead_control *rac)
1454 {
1455 	return rac->_nr_pages * PAGE_SIZE;
1456 }
1457 
1458 /**
1459  * readahead_index - The index of the first page in this readahead request.
1460  * @rac: The readahead request.
1461  */
readahead_index(struct readahead_control * rac)1462 static inline pgoff_t readahead_index(struct readahead_control *rac)
1463 {
1464 	return rac->_index;
1465 }
1466 
1467 /**
1468  * readahead_count - The number of pages in this readahead request.
1469  * @rac: The readahead request.
1470  */
readahead_count(struct readahead_control * rac)1471 static inline unsigned int readahead_count(struct readahead_control *rac)
1472 {
1473 	return rac->_nr_pages;
1474 }
1475 
1476 /**
1477  * readahead_batch_length - The number of bytes in the current batch.
1478  * @rac: The readahead request.
1479  */
readahead_batch_length(struct readahead_control * rac)1480 static inline size_t readahead_batch_length(struct readahead_control *rac)
1481 {
1482 	return rac->_batch_count * PAGE_SIZE;
1483 }
1484 
dir_pages(struct inode * inode)1485 static inline unsigned long dir_pages(struct inode *inode)
1486 {
1487 	return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1488 			       PAGE_SHIFT;
1489 }
1490 
1491 /**
1492  * folio_mkwrite_check_truncate - check if folio was truncated
1493  * @folio: the folio to check
1494  * @inode: the inode to check the folio against
1495  *
1496  * Return: the number of bytes in the folio up to EOF,
1497  * or -EFAULT if the folio was truncated.
1498  */
folio_mkwrite_check_truncate(struct folio * folio,struct inode * inode)1499 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1500 					      struct inode *inode)
1501 {
1502 	loff_t size = i_size_read(inode);
1503 	pgoff_t index = size >> PAGE_SHIFT;
1504 	size_t offset = offset_in_folio(folio, size);
1505 
1506 	if (!folio->mapping)
1507 		return -EFAULT;
1508 
1509 	/* folio is wholly inside EOF */
1510 	if (folio_next_index(folio) - 1 < index)
1511 		return folio_size(folio);
1512 	/* folio is wholly past EOF */
1513 	if (folio->index > index || !offset)
1514 		return -EFAULT;
1515 	/* folio is partially inside EOF */
1516 	return offset;
1517 }
1518 
1519 /**
1520  * page_mkwrite_check_truncate - check if page was truncated
1521  * @page: the page to check
1522  * @inode: the inode to check the page against
1523  *
1524  * Returns the number of bytes in the page up to EOF,
1525  * or -EFAULT if the page was truncated.
1526  */
page_mkwrite_check_truncate(struct page * page,struct inode * inode)1527 static inline int page_mkwrite_check_truncate(struct page *page,
1528 					      struct inode *inode)
1529 {
1530 	loff_t size = i_size_read(inode);
1531 	pgoff_t index = size >> PAGE_SHIFT;
1532 	int offset = offset_in_page(size);
1533 
1534 	if (page->mapping != inode->i_mapping)
1535 		return -EFAULT;
1536 
1537 	/* page is wholly inside EOF */
1538 	if (page->index < index)
1539 		return PAGE_SIZE;
1540 	/* page is wholly past EOF */
1541 	if (page->index > index || !offset)
1542 		return -EFAULT;
1543 	/* page is partially inside EOF */
1544 	return offset;
1545 }
1546 
1547 /**
1548  * i_blocks_per_folio - How many blocks fit in this folio.
1549  * @inode: The inode which contains the blocks.
1550  * @folio: The folio.
1551  *
1552  * If the block size is larger than the size of this folio, return zero.
1553  *
1554  * Context: The caller should hold a refcount on the folio to prevent it
1555  * from being split.
1556  * Return: The number of filesystem blocks covered by this folio.
1557  */
1558 static inline
i_blocks_per_folio(struct inode * inode,struct folio * folio)1559 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1560 {
1561 	return folio_size(folio) >> inode->i_blkbits;
1562 }
1563 
1564 static inline
i_blocks_per_page(struct inode * inode,struct page * page)1565 unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1566 {
1567 	return i_blocks_per_folio(inode, page_folio(page));
1568 }
1569 #endif /* _LINUX_PAGEMAP_H */
1570