xref: /openbmc/linux/include/linux/pagemap.h (revision 3461e3bf)
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  * mapping_set_large_folios() - Indicate the file supports large folios.
332  * @mapping: The file.
333  *
334  * The filesystem should call this function in its inode constructor to
335  * indicate that the VFS can use large folios to cache the contents of
336  * the file.
337  *
338  * Context: This should not be called while the inode is active as it
339  * is non-atomic.
340  */
mapping_set_large_folios(struct address_space * mapping)341 static inline void mapping_set_large_folios(struct address_space *mapping)
342 {
343 	__set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
344 }
345 
346 /*
347  * Large folio support currently depends on THP.  These dependencies are
348  * being worked on but are not yet fixed.
349  */
mapping_large_folio_support(struct address_space * mapping)350 static inline bool mapping_large_folio_support(struct address_space *mapping)
351 {
352 	return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) &&
353 		test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags);
354 }
355 
filemap_nr_thps(struct address_space * mapping)356 static inline int filemap_nr_thps(struct address_space *mapping)
357 {
358 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
359 	return atomic_read(&mapping->nr_thps);
360 #else
361 	return 0;
362 #endif
363 }
364 
filemap_nr_thps_inc(struct address_space * mapping)365 static inline void filemap_nr_thps_inc(struct address_space *mapping)
366 {
367 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
368 	if (!mapping_large_folio_support(mapping))
369 		atomic_inc(&mapping->nr_thps);
370 #else
371 	WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
372 #endif
373 }
374 
filemap_nr_thps_dec(struct address_space * mapping)375 static inline void filemap_nr_thps_dec(struct address_space *mapping)
376 {
377 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
378 	if (!mapping_large_folio_support(mapping))
379 		atomic_dec(&mapping->nr_thps);
380 #else
381 	WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0);
382 #endif
383 }
384 
385 struct address_space *page_mapping(struct page *);
386 struct address_space *folio_mapping(struct folio *);
387 struct address_space *swapcache_mapping(struct folio *);
388 
389 /**
390  * folio_file_mapping - Find the mapping this folio belongs to.
391  * @folio: The folio.
392  *
393  * For folios which are in the page cache, return the mapping that this
394  * page belongs to.  Folios in the swap cache return the mapping of the
395  * swap file or swap device where the data is stored.  This is different
396  * from the mapping returned by folio_mapping().  The only reason to
397  * use it is if, like NFS, you return 0 from ->activate_swapfile.
398  *
399  * Do not call this for folios which aren't in the page cache or swap cache.
400  */
folio_file_mapping(struct folio * folio)401 static inline struct address_space *folio_file_mapping(struct folio *folio)
402 {
403 	if (unlikely(folio_test_swapcache(folio)))
404 		return swapcache_mapping(folio);
405 
406 	return folio->mapping;
407 }
408 
409 /**
410  * folio_flush_mapping - Find the file mapping this folio belongs to.
411  * @folio: The folio.
412  *
413  * For folios which are in the page cache, return the mapping that this
414  * page belongs to.  Anonymous folios return NULL, even if they're in
415  * the swap cache.  Other kinds of folio also return NULL.
416  *
417  * This is ONLY used by architecture cache flushing code.  If you aren't
418  * writing cache flushing code, you want either folio_mapping() or
419  * folio_file_mapping().
420  */
folio_flush_mapping(struct folio * folio)421 static inline struct address_space *folio_flush_mapping(struct folio *folio)
422 {
423 	if (unlikely(folio_test_swapcache(folio)))
424 		return NULL;
425 
426 	return folio_mapping(folio);
427 }
428 
page_file_mapping(struct page * page)429 static inline struct address_space *page_file_mapping(struct page *page)
430 {
431 	return folio_file_mapping(page_folio(page));
432 }
433 
434 /**
435  * folio_inode - Get the host inode for this folio.
436  * @folio: The folio.
437  *
438  * For folios which are in the page cache, return the inode that this folio
439  * belongs to.
440  *
441  * Do not call this for folios which aren't in the page cache.
442  */
folio_inode(struct folio * folio)443 static inline struct inode *folio_inode(struct folio *folio)
444 {
445 	return folio->mapping->host;
446 }
447 
448 /**
449  * folio_attach_private - Attach private data to a folio.
450  * @folio: Folio to attach data to.
451  * @data: Data to attach to folio.
452  *
453  * Attaching private data to a folio increments the page's reference count.
454  * The data must be detached before the folio will be freed.
455  */
folio_attach_private(struct folio * folio,void * data)456 static inline void folio_attach_private(struct folio *folio, void *data)
457 {
458 	folio_get(folio);
459 	folio->private = data;
460 	folio_set_private(folio);
461 }
462 
463 /**
464  * folio_change_private - Change private data on a folio.
465  * @folio: Folio to change the data on.
466  * @data: Data to set on the folio.
467  *
468  * Change the private data attached to a folio and return the old
469  * data.  The page must previously have had data attached and the data
470  * must be detached before the folio will be freed.
471  *
472  * Return: Data that was previously attached to the folio.
473  */
folio_change_private(struct folio * folio,void * data)474 static inline void *folio_change_private(struct folio *folio, void *data)
475 {
476 	void *old = folio_get_private(folio);
477 
478 	folio->private = data;
479 	return old;
480 }
481 
482 /**
483  * folio_detach_private - Detach private data from a folio.
484  * @folio: Folio to detach data from.
485  *
486  * Removes the data that was previously attached to the folio and decrements
487  * the refcount on the page.
488  *
489  * Return: Data that was attached to the folio.
490  */
folio_detach_private(struct folio * folio)491 static inline void *folio_detach_private(struct folio *folio)
492 {
493 	void *data = folio_get_private(folio);
494 
495 	if (!folio_test_private(folio))
496 		return NULL;
497 	folio_clear_private(folio);
498 	folio->private = NULL;
499 	folio_put(folio);
500 
501 	return data;
502 }
503 
attach_page_private(struct page * page,void * data)504 static inline void attach_page_private(struct page *page, void *data)
505 {
506 	folio_attach_private(page_folio(page), data);
507 }
508 
detach_page_private(struct page * page)509 static inline void *detach_page_private(struct page *page)
510 {
511 	return folio_detach_private(page_folio(page));
512 }
513 
514 /*
515  * There are some parts of the kernel which assume that PMD entries
516  * are exactly HPAGE_PMD_ORDER.  Those should be fixed, but until then,
517  * limit the maximum allocation order to PMD size.  I'm not aware of any
518  * assumptions about maximum order if THP are disabled, but 8 seems like
519  * a good order (that's 1MB if you're using 4kB pages)
520  */
521 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
522 #define MAX_PAGECACHE_ORDER	HPAGE_PMD_ORDER
523 #else
524 #define MAX_PAGECACHE_ORDER	8
525 #endif
526 
527 #ifdef CONFIG_NUMA
528 struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order);
529 #else
filemap_alloc_folio(gfp_t gfp,unsigned int order)530 static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order)
531 {
532 	return folio_alloc(gfp, order);
533 }
534 #endif
535 
__page_cache_alloc(gfp_t gfp)536 static inline struct page *__page_cache_alloc(gfp_t gfp)
537 {
538 	return &filemap_alloc_folio(gfp, 0)->page;
539 }
540 
page_cache_alloc(struct address_space * x)541 static inline struct page *page_cache_alloc(struct address_space *x)
542 {
543 	return __page_cache_alloc(mapping_gfp_mask(x));
544 }
545 
readahead_gfp_mask(struct address_space * x)546 static inline gfp_t readahead_gfp_mask(struct address_space *x)
547 {
548 	return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
549 }
550 
551 typedef int filler_t(struct file *, struct folio *);
552 
553 pgoff_t page_cache_next_miss(struct address_space *mapping,
554 			     pgoff_t index, unsigned long max_scan);
555 pgoff_t page_cache_prev_miss(struct address_space *mapping,
556 			     pgoff_t index, unsigned long max_scan);
557 
558 /**
559  * typedef fgf_t - Flags for getting folios from the page cache.
560  *
561  * Most users of the page cache will not need to use these flags;
562  * there are convenience functions such as filemap_get_folio() and
563  * filemap_lock_folio().  For users which need more control over exactly
564  * what is done with the folios, these flags to __filemap_get_folio()
565  * are available.
566  *
567  * * %FGP_ACCESSED - The folio will be marked accessed.
568  * * %FGP_LOCK - The folio is returned locked.
569  * * %FGP_CREAT - If no folio is present then a new folio is allocated,
570  *   added to the page cache and the VM's LRU list.  The folio is
571  *   returned locked.
572  * * %FGP_FOR_MMAP - The caller wants to do its own locking dance if the
573  *   folio is already in cache.  If the folio was allocated, unlock it
574  *   before returning so the caller can do the same dance.
575  * * %FGP_WRITE - The folio will be written to by the caller.
576  * * %FGP_NOFS - __GFP_FS will get cleared in gfp.
577  * * %FGP_NOWAIT - Don't block on the folio lock.
578  * * %FGP_STABLE - Wait for the folio to be stable (finished writeback)
579  * * %FGP_WRITEBEGIN - The flags to use in a filesystem write_begin()
580  *   implementation.
581  */
582 typedef unsigned int __bitwise fgf_t;
583 
584 #define FGP_ACCESSED		((__force fgf_t)0x00000001)
585 #define FGP_LOCK		((__force fgf_t)0x00000002)
586 #define FGP_CREAT		((__force fgf_t)0x00000004)
587 #define FGP_WRITE		((__force fgf_t)0x00000008)
588 #define FGP_NOFS		((__force fgf_t)0x00000010)
589 #define FGP_NOWAIT		((__force fgf_t)0x00000020)
590 #define FGP_FOR_MMAP		((__force fgf_t)0x00000040)
591 #define FGP_STABLE		((__force fgf_t)0x00000080)
592 #define FGF_GET_ORDER(fgf)	(((__force unsigned)fgf) >> 26)	/* top 6 bits */
593 
594 #define FGP_WRITEBEGIN		(FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE)
595 
596 /**
597  * fgf_set_order - Encode a length in the fgf_t flags.
598  * @size: The suggested size of the folio to create.
599  *
600  * The caller of __filemap_get_folio() can use this to suggest a preferred
601  * size for the folio that is created.  If there is already a folio at
602  * the index, it will be returned, no matter what its size.  If a folio
603  * is freshly created, it may be of a different size than requested
604  * due to alignment constraints, memory pressure, or the presence of
605  * other folios at nearby indices.
606  */
fgf_set_order(size_t size)607 static inline fgf_t fgf_set_order(size_t size)
608 {
609 	unsigned int shift = ilog2(size);
610 
611 	if (shift <= PAGE_SHIFT)
612 		return 0;
613 	return (__force fgf_t)((shift - PAGE_SHIFT) << 26);
614 }
615 
616 void *filemap_get_entry(struct address_space *mapping, pgoff_t index);
617 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index,
618 		fgf_t fgp_flags, gfp_t gfp);
619 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index,
620 		fgf_t fgp_flags, gfp_t gfp);
621 
622 /**
623  * filemap_get_folio - Find and get a folio.
624  * @mapping: The address_space to search.
625  * @index: The page index.
626  *
627  * Looks up the page cache entry at @mapping & @index.  If a folio is
628  * present, it is returned with an increased refcount.
629  *
630  * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
631  * this index.  Will not return a shadow, swap or DAX entry.
632  */
filemap_get_folio(struct address_space * mapping,pgoff_t index)633 static inline struct folio *filemap_get_folio(struct address_space *mapping,
634 					pgoff_t index)
635 {
636 	return __filemap_get_folio(mapping, index, 0, 0);
637 }
638 
639 /**
640  * filemap_lock_folio - Find and lock a folio.
641  * @mapping: The address_space to search.
642  * @index: The page index.
643  *
644  * Looks up the page cache entry at @mapping & @index.  If a folio is
645  * present, it is returned locked with an increased refcount.
646  *
647  * Context: May sleep.
648  * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for
649  * this index.  Will not return a shadow, swap or DAX entry.
650  */
filemap_lock_folio(struct address_space * mapping,pgoff_t index)651 static inline struct folio *filemap_lock_folio(struct address_space *mapping,
652 					pgoff_t index)
653 {
654 	return __filemap_get_folio(mapping, index, FGP_LOCK, 0);
655 }
656 
657 /**
658  * filemap_grab_folio - grab a folio from the page cache
659  * @mapping: The address space to search
660  * @index: The page index
661  *
662  * Looks up the page cache entry at @mapping & @index. If no folio is found,
663  * a new folio is created. The folio is locked, marked as accessed, and
664  * returned.
665  *
666  * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found
667  * and failed to create a folio.
668  */
filemap_grab_folio(struct address_space * mapping,pgoff_t index)669 static inline struct folio *filemap_grab_folio(struct address_space *mapping,
670 					pgoff_t index)
671 {
672 	return __filemap_get_folio(mapping, index,
673 			FGP_LOCK | FGP_ACCESSED | FGP_CREAT,
674 			mapping_gfp_mask(mapping));
675 }
676 
677 /**
678  * find_get_page - find and get a page reference
679  * @mapping: the address_space to search
680  * @offset: the page index
681  *
682  * Looks up the page cache slot at @mapping & @offset.  If there is a
683  * page cache page, it is returned with an increased refcount.
684  *
685  * Otherwise, %NULL is returned.
686  */
find_get_page(struct address_space * mapping,pgoff_t offset)687 static inline struct page *find_get_page(struct address_space *mapping,
688 					pgoff_t offset)
689 {
690 	return pagecache_get_page(mapping, offset, 0, 0);
691 }
692 
find_get_page_flags(struct address_space * mapping,pgoff_t offset,fgf_t fgp_flags)693 static inline struct page *find_get_page_flags(struct address_space *mapping,
694 					pgoff_t offset, fgf_t fgp_flags)
695 {
696 	return pagecache_get_page(mapping, offset, fgp_flags, 0);
697 }
698 
699 /**
700  * find_lock_page - locate, pin and lock a pagecache page
701  * @mapping: the address_space to search
702  * @index: the page index
703  *
704  * Looks up the page cache entry at @mapping & @index.  If there is a
705  * page cache page, it is returned locked and with an increased
706  * refcount.
707  *
708  * Context: May sleep.
709  * Return: A struct page or %NULL if there is no page in the cache for this
710  * index.
711  */
find_lock_page(struct address_space * mapping,pgoff_t index)712 static inline struct page *find_lock_page(struct address_space *mapping,
713 					pgoff_t index)
714 {
715 	return pagecache_get_page(mapping, index, FGP_LOCK, 0);
716 }
717 
718 /**
719  * find_or_create_page - locate or add a pagecache page
720  * @mapping: the page's address_space
721  * @index: the page's index into the mapping
722  * @gfp_mask: page allocation mode
723  *
724  * Looks up the page cache slot at @mapping & @offset.  If there is a
725  * page cache page, it is returned locked and with an increased
726  * refcount.
727  *
728  * If the page is not present, a new page is allocated using @gfp_mask
729  * and added to the page cache and the VM's LRU list.  The page is
730  * returned locked and with an increased refcount.
731  *
732  * On memory exhaustion, %NULL is returned.
733  *
734  * find_or_create_page() may sleep, even if @gfp_flags specifies an
735  * atomic allocation!
736  */
find_or_create_page(struct address_space * mapping,pgoff_t index,gfp_t gfp_mask)737 static inline struct page *find_or_create_page(struct address_space *mapping,
738 					pgoff_t index, gfp_t gfp_mask)
739 {
740 	return pagecache_get_page(mapping, index,
741 					FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
742 					gfp_mask);
743 }
744 
745 /**
746  * grab_cache_page_nowait - returns locked page at given index in given cache
747  * @mapping: target address_space
748  * @index: the page index
749  *
750  * Same as grab_cache_page(), but do not wait if the page is unavailable.
751  * This is intended for speculative data generators, where the data can
752  * be regenerated if the page couldn't be grabbed.  This routine should
753  * be safe to call while holding the lock for another page.
754  *
755  * Clear __GFP_FS when allocating the page to avoid recursion into the fs
756  * and deadlock against the caller's locked page.
757  */
grab_cache_page_nowait(struct address_space * mapping,pgoff_t index)758 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
759 				pgoff_t index)
760 {
761 	return pagecache_get_page(mapping, index,
762 			FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
763 			mapping_gfp_mask(mapping));
764 }
765 
766 #define swapcache_index(folio)	__page_file_index(&(folio)->page)
767 
768 /**
769  * folio_index - File index of a folio.
770  * @folio: The folio.
771  *
772  * For a folio which is either in the page cache or the swap cache,
773  * return its index within the address_space it belongs to.  If you know
774  * the page is definitely in the page cache, you can look at the folio's
775  * index directly.
776  *
777  * Return: The index (offset in units of pages) of a folio in its file.
778  */
folio_index(struct folio * folio)779 static inline pgoff_t folio_index(struct folio *folio)
780 {
781         if (unlikely(folio_test_swapcache(folio)))
782                 return swapcache_index(folio);
783         return folio->index;
784 }
785 
786 /**
787  * folio_next_index - Get the index of the next folio.
788  * @folio: The current folio.
789  *
790  * Return: The index of the folio which follows this folio in the file.
791  */
folio_next_index(struct folio * folio)792 static inline pgoff_t folio_next_index(struct folio *folio)
793 {
794 	return folio->index + folio_nr_pages(folio);
795 }
796 
797 /**
798  * folio_file_page - The page for a particular index.
799  * @folio: The folio which contains this index.
800  * @index: The index we want to look up.
801  *
802  * Sometimes after looking up a folio in the page cache, we need to
803  * obtain the specific page for an index (eg a page fault).
804  *
805  * Return: The page containing the file data for this index.
806  */
folio_file_page(struct folio * folio,pgoff_t index)807 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index)
808 {
809 	/* HugeTLBfs indexes the page cache in units of hpage_size */
810 	if (folio_test_hugetlb(folio))
811 		return &folio->page;
812 	return folio_page(folio, index & (folio_nr_pages(folio) - 1));
813 }
814 
815 /**
816  * folio_contains - Does this folio contain this index?
817  * @folio: The folio.
818  * @index: The page index within the file.
819  *
820  * Context: The caller should have the page locked in order to prevent
821  * (eg) shmem from moving the page between the page cache and swap cache
822  * and changing its index in the middle of the operation.
823  * Return: true or false.
824  */
folio_contains(struct folio * folio,pgoff_t index)825 static inline bool folio_contains(struct folio *folio, pgoff_t index)
826 {
827 	/* HugeTLBfs indexes the page cache in units of hpage_size */
828 	if (folio_test_hugetlb(folio))
829 		return folio->index == index;
830 	return index - folio_index(folio) < folio_nr_pages(folio);
831 }
832 
833 /*
834  * Given the page we found in the page cache, return the page corresponding
835  * to this index in the file
836  */
find_subpage(struct page * head,pgoff_t index)837 static inline struct page *find_subpage(struct page *head, pgoff_t index)
838 {
839 	/* HugeTLBfs wants the head page regardless */
840 	if (PageHuge(head))
841 		return head;
842 
843 	return head + (index & (thp_nr_pages(head) - 1));
844 }
845 
846 unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start,
847 		pgoff_t end, struct folio_batch *fbatch);
848 unsigned filemap_get_folios_contig(struct address_space *mapping,
849 		pgoff_t *start, pgoff_t end, struct folio_batch *fbatch);
850 unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start,
851 		pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch);
852 
853 struct page *grab_cache_page_write_begin(struct address_space *mapping,
854 			pgoff_t index);
855 
856 /*
857  * Returns locked page at given index in given cache, creating it if needed.
858  */
grab_cache_page(struct address_space * mapping,pgoff_t index)859 static inline struct page *grab_cache_page(struct address_space *mapping,
860 								pgoff_t index)
861 {
862 	return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
863 }
864 
865 struct folio *read_cache_folio(struct address_space *, pgoff_t index,
866 		filler_t *filler, struct file *file);
867 struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index,
868 		gfp_t flags);
869 struct page *read_cache_page(struct address_space *, pgoff_t index,
870 		filler_t *filler, struct file *file);
871 extern struct page * read_cache_page_gfp(struct address_space *mapping,
872 				pgoff_t index, gfp_t gfp_mask);
873 
read_mapping_page(struct address_space * mapping,pgoff_t index,struct file * file)874 static inline struct page *read_mapping_page(struct address_space *mapping,
875 				pgoff_t index, struct file *file)
876 {
877 	return read_cache_page(mapping, index, NULL, file);
878 }
879 
read_mapping_folio(struct address_space * mapping,pgoff_t index,struct file * file)880 static inline struct folio *read_mapping_folio(struct address_space *mapping,
881 				pgoff_t index, struct file *file)
882 {
883 	return read_cache_folio(mapping, index, NULL, file);
884 }
885 
886 /*
887  * Get index of the page within radix-tree (but not for hugetlb pages).
888  * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
889  */
page_to_index(struct page * page)890 static inline pgoff_t page_to_index(struct page *page)
891 {
892 	struct page *head;
893 
894 	if (likely(!PageTransTail(page)))
895 		return page->index;
896 
897 	head = compound_head(page);
898 	/*
899 	 *  We don't initialize ->index for tail pages: calculate based on
900 	 *  head page
901 	 */
902 	return head->index + page - head;
903 }
904 
905 extern pgoff_t hugetlb_basepage_index(struct page *page);
906 
907 /*
908  * Get the offset in PAGE_SIZE (even for hugetlb pages).
909  * (TODO: hugetlb pages should have ->index in PAGE_SIZE)
910  */
page_to_pgoff(struct page * page)911 static inline pgoff_t page_to_pgoff(struct page *page)
912 {
913 	if (unlikely(PageHuge(page)))
914 		return hugetlb_basepage_index(page);
915 	return page_to_index(page);
916 }
917 
918 /*
919  * Return byte-offset into filesystem object for page.
920  */
page_offset(struct page * page)921 static inline loff_t page_offset(struct page *page)
922 {
923 	return ((loff_t)page->index) << PAGE_SHIFT;
924 }
925 
page_file_offset(struct page * page)926 static inline loff_t page_file_offset(struct page *page)
927 {
928 	return ((loff_t)page_index(page)) << PAGE_SHIFT;
929 }
930 
931 /**
932  * folio_pos - Returns the byte position of this folio in its file.
933  * @folio: The folio.
934  */
folio_pos(struct folio * folio)935 static inline loff_t folio_pos(struct folio *folio)
936 {
937 	return page_offset(&folio->page);
938 }
939 
940 /**
941  * folio_file_pos - Returns the byte position of this folio in its file.
942  * @folio: The folio.
943  *
944  * This differs from folio_pos() for folios which belong to a swap file.
945  * NFS is the only filesystem today which needs to use folio_file_pos().
946  */
folio_file_pos(struct folio * folio)947 static inline loff_t folio_file_pos(struct folio *folio)
948 {
949 	return page_file_offset(&folio->page);
950 }
951 
952 /*
953  * Get the offset in PAGE_SIZE (even for hugetlb folios).
954  * (TODO: hugetlb folios should have ->index in PAGE_SIZE)
955  */
folio_pgoff(struct folio * folio)956 static inline pgoff_t folio_pgoff(struct folio *folio)
957 {
958 	if (unlikely(folio_test_hugetlb(folio)))
959 		return hugetlb_basepage_index(&folio->page);
960 	return folio->index;
961 }
962 
963 extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
964 				     unsigned long address);
965 
linear_page_index(struct vm_area_struct * vma,unsigned long address)966 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
967 					unsigned long address)
968 {
969 	pgoff_t pgoff;
970 	if (unlikely(is_vm_hugetlb_page(vma)))
971 		return linear_hugepage_index(vma, address);
972 	pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
973 	pgoff += vma->vm_pgoff;
974 	return pgoff;
975 }
976 
977 struct wait_page_key {
978 	struct folio *folio;
979 	int bit_nr;
980 	int page_match;
981 };
982 
983 struct wait_page_queue {
984 	struct folio *folio;
985 	int bit_nr;
986 	wait_queue_entry_t wait;
987 };
988 
wake_page_match(struct wait_page_queue * wait_page,struct wait_page_key * key)989 static inline bool wake_page_match(struct wait_page_queue *wait_page,
990 				  struct wait_page_key *key)
991 {
992 	if (wait_page->folio != key->folio)
993 	       return false;
994 	key->page_match = 1;
995 
996 	if (wait_page->bit_nr != key->bit_nr)
997 		return false;
998 
999 	return true;
1000 }
1001 
1002 void __folio_lock(struct folio *folio);
1003 int __folio_lock_killable(struct folio *folio);
1004 vm_fault_t __folio_lock_or_retry(struct folio *folio, struct vm_fault *vmf);
1005 void unlock_page(struct page *page);
1006 void folio_unlock(struct folio *folio);
1007 
1008 /**
1009  * folio_trylock() - Attempt to lock a folio.
1010  * @folio: The folio to attempt to lock.
1011  *
1012  * Sometimes it is undesirable to wait for a folio to be unlocked (eg
1013  * when the locks are being taken in the wrong order, or if making
1014  * progress through a batch of folios is more important than processing
1015  * them in order).  Usually folio_lock() is the correct function to call.
1016  *
1017  * Context: Any context.
1018  * Return: Whether the lock was successfully acquired.
1019  */
folio_trylock(struct folio * folio)1020 static inline bool folio_trylock(struct folio *folio)
1021 {
1022 	return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0)));
1023 }
1024 
1025 /*
1026  * Return true if the page was successfully locked
1027  */
trylock_page(struct page * page)1028 static inline int trylock_page(struct page *page)
1029 {
1030 	return folio_trylock(page_folio(page));
1031 }
1032 
1033 /**
1034  * folio_lock() - Lock this folio.
1035  * @folio: The folio to lock.
1036  *
1037  * The folio lock protects against many things, probably more than it
1038  * should.  It is primarily held while a folio is being brought uptodate,
1039  * either from its backing file or from swap.  It is also held while a
1040  * folio is being truncated from its address_space, so holding the lock
1041  * is sufficient to keep folio->mapping stable.
1042  *
1043  * The folio lock is also held while write() is modifying the page to
1044  * provide POSIX atomicity guarantees (as long as the write does not
1045  * cross a page boundary).  Other modifications to the data in the folio
1046  * do not hold the folio lock and can race with writes, eg DMA and stores
1047  * to mapped pages.
1048  *
1049  * Context: May sleep.  If you need to acquire the locks of two or
1050  * more folios, they must be in order of ascending index, if they are
1051  * in the same address_space.  If they are in different address_spaces,
1052  * acquire the lock of the folio which belongs to the address_space which
1053  * has the lowest address in memory first.
1054  */
folio_lock(struct folio * folio)1055 static inline void folio_lock(struct folio *folio)
1056 {
1057 	might_sleep();
1058 	if (!folio_trylock(folio))
1059 		__folio_lock(folio);
1060 }
1061 
1062 /**
1063  * lock_page() - Lock the folio containing this page.
1064  * @page: The page to lock.
1065  *
1066  * See folio_lock() for a description of what the lock protects.
1067  * This is a legacy function and new code should probably use folio_lock()
1068  * instead.
1069  *
1070  * Context: May sleep.  Pages in the same folio share a lock, so do not
1071  * attempt to lock two pages which share a folio.
1072  */
lock_page(struct page * page)1073 static inline void lock_page(struct page *page)
1074 {
1075 	struct folio *folio;
1076 	might_sleep();
1077 
1078 	folio = page_folio(page);
1079 	if (!folio_trylock(folio))
1080 		__folio_lock(folio);
1081 }
1082 
1083 /**
1084  * folio_lock_killable() - Lock this folio, interruptible by a fatal signal.
1085  * @folio: The folio to lock.
1086  *
1087  * Attempts to lock the folio, like folio_lock(), except that the sleep
1088  * to acquire the lock is interruptible by a fatal signal.
1089  *
1090  * Context: May sleep; see folio_lock().
1091  * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received.
1092  */
folio_lock_killable(struct folio * folio)1093 static inline int folio_lock_killable(struct folio *folio)
1094 {
1095 	might_sleep();
1096 	if (!folio_trylock(folio))
1097 		return __folio_lock_killable(folio);
1098 	return 0;
1099 }
1100 
1101 /*
1102  * folio_lock_or_retry - Lock the folio, unless this would block and the
1103  * caller indicated that it can handle a retry.
1104  *
1105  * Return value and mmap_lock implications depend on flags; see
1106  * __folio_lock_or_retry().
1107  */
folio_lock_or_retry(struct folio * folio,struct vm_fault * vmf)1108 static inline vm_fault_t folio_lock_or_retry(struct folio *folio,
1109 					     struct vm_fault *vmf)
1110 {
1111 	might_sleep();
1112 	if (!folio_trylock(folio))
1113 		return __folio_lock_or_retry(folio, vmf);
1114 	return 0;
1115 }
1116 
1117 /*
1118  * This is exported only for folio_wait_locked/folio_wait_writeback, etc.,
1119  * and should not be used directly.
1120  */
1121 void folio_wait_bit(struct folio *folio, int bit_nr);
1122 int folio_wait_bit_killable(struct folio *folio, int bit_nr);
1123 
1124 /*
1125  * Wait for a folio to be unlocked.
1126  *
1127  * This must be called with the caller "holding" the folio,
1128  * ie with increased folio reference count so that the folio won't
1129  * go away during the wait.
1130  */
folio_wait_locked(struct folio * folio)1131 static inline void folio_wait_locked(struct folio *folio)
1132 {
1133 	if (folio_test_locked(folio))
1134 		folio_wait_bit(folio, PG_locked);
1135 }
1136 
folio_wait_locked_killable(struct folio * folio)1137 static inline int folio_wait_locked_killable(struct folio *folio)
1138 {
1139 	if (!folio_test_locked(folio))
1140 		return 0;
1141 	return folio_wait_bit_killable(folio, PG_locked);
1142 }
1143 
wait_on_page_locked(struct page * page)1144 static inline void wait_on_page_locked(struct page *page)
1145 {
1146 	folio_wait_locked(page_folio(page));
1147 }
1148 
1149 void wait_on_page_writeback(struct page *page);
1150 void folio_wait_writeback(struct folio *folio);
1151 int folio_wait_writeback_killable(struct folio *folio);
1152 void end_page_writeback(struct page *page);
1153 void folio_end_writeback(struct folio *folio);
1154 void wait_for_stable_page(struct page *page);
1155 void folio_wait_stable(struct folio *folio);
1156 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn);
__set_page_dirty(struct page * page,struct address_space * mapping,int warn)1157 static inline void __set_page_dirty(struct page *page,
1158 		struct address_space *mapping, int warn)
1159 {
1160 	__folio_mark_dirty(page_folio(page), mapping, warn);
1161 }
1162 void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb);
1163 void __folio_cancel_dirty(struct folio *folio);
folio_cancel_dirty(struct folio * folio)1164 static inline void folio_cancel_dirty(struct folio *folio)
1165 {
1166 	/* Avoid atomic ops, locking, etc. when not actually needed. */
1167 	if (folio_test_dirty(folio))
1168 		__folio_cancel_dirty(folio);
1169 }
1170 bool folio_clear_dirty_for_io(struct folio *folio);
1171 bool clear_page_dirty_for_io(struct page *page);
1172 void folio_invalidate(struct folio *folio, size_t offset, size_t length);
1173 int __set_page_dirty_nobuffers(struct page *page);
1174 bool noop_dirty_folio(struct address_space *mapping, struct folio *folio);
1175 
1176 #ifdef CONFIG_MIGRATION
1177 int filemap_migrate_folio(struct address_space *mapping, struct folio *dst,
1178 		struct folio *src, enum migrate_mode mode);
1179 #else
1180 #define filemap_migrate_folio NULL
1181 #endif
1182 void folio_end_private_2(struct folio *folio);
1183 void folio_wait_private_2(struct folio *folio);
1184 int folio_wait_private_2_killable(struct folio *folio);
1185 
1186 /*
1187  * Add an arbitrary waiter to a page's wait queue
1188  */
1189 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter);
1190 
1191 /*
1192  * Fault in userspace address range.
1193  */
1194 size_t fault_in_writeable(char __user *uaddr, size_t size);
1195 size_t fault_in_subpage_writeable(char __user *uaddr, size_t size);
1196 size_t fault_in_safe_writeable(const char __user *uaddr, size_t size);
1197 size_t fault_in_readable(const char __user *uaddr, size_t size);
1198 
1199 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
1200 		pgoff_t index, gfp_t gfp);
1201 int filemap_add_folio(struct address_space *mapping, struct folio *folio,
1202 		pgoff_t index, gfp_t gfp);
1203 void filemap_remove_folio(struct folio *folio);
1204 void __filemap_remove_folio(struct folio *folio, void *shadow);
1205 void replace_page_cache_folio(struct folio *old, struct folio *new);
1206 void delete_from_page_cache_batch(struct address_space *mapping,
1207 				  struct folio_batch *fbatch);
1208 bool filemap_release_folio(struct folio *folio, gfp_t gfp);
1209 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end,
1210 		int whence);
1211 
1212 /* Must be non-static for BPF error injection */
1213 int __filemap_add_folio(struct address_space *mapping, struct folio *folio,
1214 		pgoff_t index, gfp_t gfp, void **shadowp);
1215 
1216 bool filemap_range_has_writeback(struct address_space *mapping,
1217 				 loff_t start_byte, loff_t end_byte);
1218 
1219 /**
1220  * filemap_range_needs_writeback - check if range potentially needs writeback
1221  * @mapping:           address space within which to check
1222  * @start_byte:        offset in bytes where the range starts
1223  * @end_byte:          offset in bytes where the range ends (inclusive)
1224  *
1225  * Find at least one page in the range supplied, usually used to check if
1226  * direct writing in this range will trigger a writeback. Used by O_DIRECT
1227  * read/write with IOCB_NOWAIT, to see if the caller needs to do
1228  * filemap_write_and_wait_range() before proceeding.
1229  *
1230  * Return: %true if the caller should do filemap_write_and_wait_range() before
1231  * doing O_DIRECT to a page in this range, %false otherwise.
1232  */
filemap_range_needs_writeback(struct address_space * mapping,loff_t start_byte,loff_t end_byte)1233 static inline bool filemap_range_needs_writeback(struct address_space *mapping,
1234 						 loff_t start_byte,
1235 						 loff_t end_byte)
1236 {
1237 	if (!mapping->nrpages)
1238 		return false;
1239 	if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
1240 	    !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
1241 		return false;
1242 	return filemap_range_has_writeback(mapping, start_byte, end_byte);
1243 }
1244 
1245 /**
1246  * struct readahead_control - Describes a readahead request.
1247  *
1248  * A readahead request is for consecutive pages.  Filesystems which
1249  * implement the ->readahead method should call readahead_page() or
1250  * readahead_page_batch() in a loop and attempt to start I/O against
1251  * each page in the request.
1252  *
1253  * Most of the fields in this struct are private and should be accessed
1254  * by the functions below.
1255  *
1256  * @file: The file, used primarily by network filesystems for authentication.
1257  *	  May be NULL if invoked internally by the filesystem.
1258  * @mapping: Readahead this filesystem object.
1259  * @ra: File readahead state.  May be NULL.
1260  */
1261 struct readahead_control {
1262 	struct file *file;
1263 	struct address_space *mapping;
1264 	struct file_ra_state *ra;
1265 /* private: use the readahead_* accessors instead */
1266 	pgoff_t _index;
1267 	unsigned int _nr_pages;
1268 	unsigned int _batch_count;
1269 	bool _workingset;
1270 	unsigned long _pflags;
1271 };
1272 
1273 #define DEFINE_READAHEAD(ractl, f, r, m, i)				\
1274 	struct readahead_control ractl = {				\
1275 		.file = f,						\
1276 		.mapping = m,						\
1277 		.ra = r,						\
1278 		._index = i,						\
1279 	}
1280 
1281 #define VM_READAHEAD_PAGES	(SZ_128K / PAGE_SIZE)
1282 
1283 void page_cache_ra_unbounded(struct readahead_control *,
1284 		unsigned long nr_to_read, unsigned long lookahead_count);
1285 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count);
1286 void page_cache_async_ra(struct readahead_control *, struct folio *,
1287 		unsigned long req_count);
1288 void readahead_expand(struct readahead_control *ractl,
1289 		      loff_t new_start, size_t new_len);
1290 
1291 /**
1292  * page_cache_sync_readahead - generic file readahead
1293  * @mapping: address_space which holds the pagecache and I/O vectors
1294  * @ra: file_ra_state which holds the readahead state
1295  * @file: Used by the filesystem for authentication.
1296  * @index: Index of first page to be read.
1297  * @req_count: Total number of pages being read by the caller.
1298  *
1299  * page_cache_sync_readahead() should be called when a cache miss happened:
1300  * it will submit the read.  The readahead logic may decide to piggyback more
1301  * pages onto the read request if access patterns suggest it will improve
1302  * performance.
1303  */
1304 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)1305 void page_cache_sync_readahead(struct address_space *mapping,
1306 		struct file_ra_state *ra, struct file *file, pgoff_t index,
1307 		unsigned long req_count)
1308 {
1309 	DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1310 	page_cache_sync_ra(&ractl, req_count);
1311 }
1312 
1313 /**
1314  * page_cache_async_readahead - file readahead for marked pages
1315  * @mapping: address_space which holds the pagecache and I/O vectors
1316  * @ra: file_ra_state which holds the readahead state
1317  * @file: Used by the filesystem for authentication.
1318  * @folio: The folio at @index which triggered the readahead call.
1319  * @index: Index of first page to be read.
1320  * @req_count: Total number of pages being read by the caller.
1321  *
1322  * page_cache_async_readahead() should be called when a page is used which
1323  * is marked as PageReadahead; this is a marker to suggest that the application
1324  * has used up enough of the readahead window that we should start pulling in
1325  * more pages.
1326  */
1327 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)1328 void page_cache_async_readahead(struct address_space *mapping,
1329 		struct file_ra_state *ra, struct file *file,
1330 		struct folio *folio, pgoff_t index, unsigned long req_count)
1331 {
1332 	DEFINE_READAHEAD(ractl, file, ra, mapping, index);
1333 	page_cache_async_ra(&ractl, folio, req_count);
1334 }
1335 
__readahead_folio(struct readahead_control * ractl)1336 static inline struct folio *__readahead_folio(struct readahead_control *ractl)
1337 {
1338 	struct folio *folio;
1339 
1340 	BUG_ON(ractl->_batch_count > ractl->_nr_pages);
1341 	ractl->_nr_pages -= ractl->_batch_count;
1342 	ractl->_index += ractl->_batch_count;
1343 
1344 	if (!ractl->_nr_pages) {
1345 		ractl->_batch_count = 0;
1346 		return NULL;
1347 	}
1348 
1349 	folio = xa_load(&ractl->mapping->i_pages, ractl->_index);
1350 	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
1351 	ractl->_batch_count = folio_nr_pages(folio);
1352 
1353 	return folio;
1354 }
1355 
1356 /**
1357  * readahead_page - Get the next page to read.
1358  * @ractl: The current readahead request.
1359  *
1360  * Context: The page is locked and has an elevated refcount.  The caller
1361  * should decreases the refcount once the page has been submitted for I/O
1362  * and unlock the page once all I/O to that page has completed.
1363  * Return: A pointer to the next page, or %NULL if we are done.
1364  */
readahead_page(struct readahead_control * ractl)1365 static inline struct page *readahead_page(struct readahead_control *ractl)
1366 {
1367 	struct folio *folio = __readahead_folio(ractl);
1368 
1369 	return &folio->page;
1370 }
1371 
1372 /**
1373  * readahead_folio - Get the next folio to read.
1374  * @ractl: The current readahead request.
1375  *
1376  * Context: The folio is locked.  The caller should unlock the folio once
1377  * all I/O to that folio has completed.
1378  * Return: A pointer to the next folio, or %NULL if we are done.
1379  */
readahead_folio(struct readahead_control * ractl)1380 static inline struct folio *readahead_folio(struct readahead_control *ractl)
1381 {
1382 	struct folio *folio = __readahead_folio(ractl);
1383 
1384 	if (folio)
1385 		folio_put(folio);
1386 	return folio;
1387 }
1388 
__readahead_batch(struct readahead_control * rac,struct page ** array,unsigned int array_sz)1389 static inline unsigned int __readahead_batch(struct readahead_control *rac,
1390 		struct page **array, unsigned int array_sz)
1391 {
1392 	unsigned int i = 0;
1393 	XA_STATE(xas, &rac->mapping->i_pages, 0);
1394 	struct page *page;
1395 
1396 	BUG_ON(rac->_batch_count > rac->_nr_pages);
1397 	rac->_nr_pages -= rac->_batch_count;
1398 	rac->_index += rac->_batch_count;
1399 	rac->_batch_count = 0;
1400 
1401 	xas_set(&xas, rac->_index);
1402 	rcu_read_lock();
1403 	xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
1404 		if (xas_retry(&xas, page))
1405 			continue;
1406 		VM_BUG_ON_PAGE(!PageLocked(page), page);
1407 		VM_BUG_ON_PAGE(PageTail(page), page);
1408 		array[i++] = page;
1409 		rac->_batch_count += thp_nr_pages(page);
1410 		if (i == array_sz)
1411 			break;
1412 	}
1413 	rcu_read_unlock();
1414 
1415 	return i;
1416 }
1417 
1418 /**
1419  * readahead_page_batch - Get a batch of pages to read.
1420  * @rac: The current readahead request.
1421  * @array: An array of pointers to struct page.
1422  *
1423  * Context: The pages are locked and have an elevated refcount.  The caller
1424  * should decreases the refcount once the page has been submitted for I/O
1425  * and unlock the page once all I/O to that page has completed.
1426  * Return: The number of pages placed in the array.  0 indicates the request
1427  * is complete.
1428  */
1429 #define readahead_page_batch(rac, array)				\
1430 	__readahead_batch(rac, array, ARRAY_SIZE(array))
1431 
1432 /**
1433  * readahead_pos - The byte offset into the file of this readahead request.
1434  * @rac: The readahead request.
1435  */
readahead_pos(struct readahead_control * rac)1436 static inline loff_t readahead_pos(struct readahead_control *rac)
1437 {
1438 	return (loff_t)rac->_index * PAGE_SIZE;
1439 }
1440 
1441 /**
1442  * readahead_length - The number of bytes in this readahead request.
1443  * @rac: The readahead request.
1444  */
readahead_length(struct readahead_control * rac)1445 static inline size_t readahead_length(struct readahead_control *rac)
1446 {
1447 	return rac->_nr_pages * PAGE_SIZE;
1448 }
1449 
1450 /**
1451  * readahead_index - The index of the first page in this readahead request.
1452  * @rac: The readahead request.
1453  */
readahead_index(struct readahead_control * rac)1454 static inline pgoff_t readahead_index(struct readahead_control *rac)
1455 {
1456 	return rac->_index;
1457 }
1458 
1459 /**
1460  * readahead_count - The number of pages in this readahead request.
1461  * @rac: The readahead request.
1462  */
readahead_count(struct readahead_control * rac)1463 static inline unsigned int readahead_count(struct readahead_control *rac)
1464 {
1465 	return rac->_nr_pages;
1466 }
1467 
1468 /**
1469  * readahead_batch_length - The number of bytes in the current batch.
1470  * @rac: The readahead request.
1471  */
readahead_batch_length(struct readahead_control * rac)1472 static inline size_t readahead_batch_length(struct readahead_control *rac)
1473 {
1474 	return rac->_batch_count * PAGE_SIZE;
1475 }
1476 
dir_pages(struct inode * inode)1477 static inline unsigned long dir_pages(struct inode *inode)
1478 {
1479 	return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
1480 			       PAGE_SHIFT;
1481 }
1482 
1483 /**
1484  * folio_mkwrite_check_truncate - check if folio was truncated
1485  * @folio: the folio to check
1486  * @inode: the inode to check the folio against
1487  *
1488  * Return: the number of bytes in the folio up to EOF,
1489  * or -EFAULT if the folio was truncated.
1490  */
folio_mkwrite_check_truncate(struct folio * folio,struct inode * inode)1491 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio,
1492 					      struct inode *inode)
1493 {
1494 	loff_t size = i_size_read(inode);
1495 	pgoff_t index = size >> PAGE_SHIFT;
1496 	size_t offset = offset_in_folio(folio, size);
1497 
1498 	if (!folio->mapping)
1499 		return -EFAULT;
1500 
1501 	/* folio is wholly inside EOF */
1502 	if (folio_next_index(folio) - 1 < index)
1503 		return folio_size(folio);
1504 	/* folio is wholly past EOF */
1505 	if (folio->index > index || !offset)
1506 		return -EFAULT;
1507 	/* folio is partially inside EOF */
1508 	return offset;
1509 }
1510 
1511 /**
1512  * page_mkwrite_check_truncate - check if page was truncated
1513  * @page: the page to check
1514  * @inode: the inode to check the page against
1515  *
1516  * Returns the number of bytes in the page up to EOF,
1517  * or -EFAULT if the page was truncated.
1518  */
page_mkwrite_check_truncate(struct page * page,struct inode * inode)1519 static inline int page_mkwrite_check_truncate(struct page *page,
1520 					      struct inode *inode)
1521 {
1522 	loff_t size = i_size_read(inode);
1523 	pgoff_t index = size >> PAGE_SHIFT;
1524 	int offset = offset_in_page(size);
1525 
1526 	if (page->mapping != inode->i_mapping)
1527 		return -EFAULT;
1528 
1529 	/* page is wholly inside EOF */
1530 	if (page->index < index)
1531 		return PAGE_SIZE;
1532 	/* page is wholly past EOF */
1533 	if (page->index > index || !offset)
1534 		return -EFAULT;
1535 	/* page is partially inside EOF */
1536 	return offset;
1537 }
1538 
1539 /**
1540  * i_blocks_per_folio - How many blocks fit in this folio.
1541  * @inode: The inode which contains the blocks.
1542  * @folio: The folio.
1543  *
1544  * If the block size is larger than the size of this folio, return zero.
1545  *
1546  * Context: The caller should hold a refcount on the folio to prevent it
1547  * from being split.
1548  * Return: The number of filesystem blocks covered by this folio.
1549  */
1550 static inline
i_blocks_per_folio(struct inode * inode,struct folio * folio)1551 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio)
1552 {
1553 	return folio_size(folio) >> inode->i_blkbits;
1554 }
1555 
1556 static inline
i_blocks_per_page(struct inode * inode,struct page * page)1557 unsigned int i_blocks_per_page(struct inode *inode, struct page *page)
1558 {
1559 	return i_blocks_per_folio(inode, page_folio(page));
1560 }
1561 #endif /* _LINUX_PAGEMAP_H */
1562