Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
1da177e4 LT |
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> | |
7c0f6ba6 | 13 | #include <linux/uaccess.h> |
1da177e4 | 14 | #include <linux/gfp.h> |
3e9f45bd | 15 | #include <linux/bitops.h> |
e286781d | 16 | #include <linux/hardirq.h> /* for in_interrupt() */ |
8edf344c | 17 | #include <linux/hugetlb_inline.h> |
1da177e4 | 18 | |
51dcbdac | 19 | struct folio_batch; |
aa65c29c | 20 | |
cbcc268b MWO |
21 | unsigned long invalidate_mapping_pages(struct address_space *mapping, |
22 | pgoff_t start, pgoff_t end); | |
23 | ||
24 | static inline void invalidate_remote_inode(struct inode *inode) | |
25 | { | |
26 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
27 | S_ISLNK(inode->i_mode)) | |
28 | invalidate_mapping_pages(inode->i_mapping, 0, -1); | |
29 | } | |
30 | int invalidate_inode_pages2(struct address_space *mapping); | |
31 | int invalidate_inode_pages2_range(struct address_space *mapping, | |
32 | pgoff_t start, pgoff_t end); | |
33 | int write_inode_now(struct inode *, int sync); | |
34 | int filemap_fdatawrite(struct address_space *); | |
35 | int filemap_flush(struct address_space *); | |
36 | int filemap_fdatawait_keep_errors(struct address_space *mapping); | |
37 | int filemap_fdatawait_range(struct address_space *, loff_t lstart, loff_t lend); | |
38 | int filemap_fdatawait_range_keep_errors(struct address_space *mapping, | |
39 | loff_t start_byte, loff_t end_byte); | |
40 | ||
41 | static inline int filemap_fdatawait(struct address_space *mapping) | |
42 | { | |
43 | return filemap_fdatawait_range(mapping, 0, LLONG_MAX); | |
44 | } | |
45 | ||
46 | bool filemap_range_has_page(struct address_space *, loff_t lstart, loff_t lend); | |
47 | int filemap_write_and_wait_range(struct address_space *mapping, | |
48 | loff_t lstart, loff_t lend); | |
49 | int __filemap_fdatawrite_range(struct address_space *mapping, | |
50 | loff_t start, loff_t end, int sync_mode); | |
51 | int filemap_fdatawrite_range(struct address_space *mapping, | |
52 | loff_t start, loff_t end); | |
53 | int filemap_check_errors(struct address_space *mapping); | |
54 | void __filemap_set_wb_err(struct address_space *mapping, int err); | |
55 | int filemap_fdatawrite_wbc(struct address_space *mapping, | |
56 | struct writeback_control *wbc); | |
57 | ||
58 | static inline int filemap_write_and_wait(struct address_space *mapping) | |
59 | { | |
60 | return filemap_write_and_wait_range(mapping, 0, LLONG_MAX); | |
61 | } | |
62 | ||
63 | /** | |
64 | * filemap_set_wb_err - set a writeback error on an address_space | |
65 | * @mapping: mapping in which to set writeback error | |
66 | * @err: error to be set in mapping | |
67 | * | |
68 | * When writeback fails in some way, we must record that error so that | |
69 | * userspace can be informed when fsync and the like are called. We endeavor | |
70 | * to report errors on any file that was open at the time of the error. Some | |
71 | * internal callers also need to know when writeback errors have occurred. | |
72 | * | |
73 | * When a writeback error occurs, most filesystems will want to call | |
74 | * filemap_set_wb_err to record the error in the mapping so that it will be | |
75 | * automatically reported whenever fsync is called on the file. | |
76 | */ | |
77 | static inline void filemap_set_wb_err(struct address_space *mapping, int err) | |
78 | { | |
79 | /* Fastpath for common case of no error */ | |
80 | if (unlikely(err)) | |
81 | __filemap_set_wb_err(mapping, err); | |
82 | } | |
83 | ||
84 | /** | |
85 | * filemap_check_wb_err - has an error occurred since the mark was sampled? | |
86 | * @mapping: mapping to check for writeback errors | |
87 | * @since: previously-sampled errseq_t | |
88 | * | |
89 | * Grab the errseq_t value from the mapping, and see if it has changed "since" | |
90 | * the given value was sampled. | |
91 | * | |
92 | * If it has then report the latest error set, otherwise return 0. | |
93 | */ | |
94 | static inline int filemap_check_wb_err(struct address_space *mapping, | |
95 | errseq_t since) | |
96 | { | |
97 | return errseq_check(&mapping->wb_err, since); | |
98 | } | |
99 | ||
100 | /** | |
101 | * filemap_sample_wb_err - sample the current errseq_t to test for later errors | |
102 | * @mapping: mapping to be sampled | |
103 | * | |
104 | * Writeback errors are always reported relative to a particular sample point | |
105 | * in the past. This function provides those sample points. | |
106 | */ | |
107 | static inline errseq_t filemap_sample_wb_err(struct address_space *mapping) | |
108 | { | |
109 | return errseq_sample(&mapping->wb_err); | |
110 | } | |
111 | ||
112 | /** | |
113 | * file_sample_sb_err - sample the current errseq_t to test for later errors | |
114 | * @file: file pointer to be sampled | |
115 | * | |
116 | * Grab the most current superblock-level errseq_t value for the given | |
117 | * struct file. | |
118 | */ | |
119 | static inline errseq_t file_sample_sb_err(struct file *file) | |
120 | { | |
121 | return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err); | |
122 | } | |
123 | ||
124 | /* | |
125 | * Flush file data before changing attributes. Caller must hold any locks | |
126 | * required to prevent further writes to this file until we're done setting | |
127 | * flags. | |
128 | */ | |
129 | static inline int inode_drain_writes(struct inode *inode) | |
130 | { | |
131 | inode_dio_wait(inode); | |
132 | return filemap_write_and_wait(inode->i_mapping); | |
133 | } | |
134 | ||
7716506a MWO |
135 | static inline bool mapping_empty(struct address_space *mapping) |
136 | { | |
137 | return xa_empty(&mapping->i_pages); | |
138 | } | |
139 | ||
51b8c1fe JW |
140 | /* |
141 | * mapping_shrinkable - test if page cache state allows inode reclaim | |
142 | * @mapping: the page cache mapping | |
143 | * | |
144 | * This checks the mapping's cache state for the pupose of inode | |
145 | * reclaim and LRU management. | |
146 | * | |
147 | * The caller is expected to hold the i_lock, but is not required to | |
148 | * hold the i_pages lock, which usually protects cache state. That's | |
149 | * because the i_lock and the list_lru lock that protect the inode and | |
150 | * its LRU state don't nest inside the irq-safe i_pages lock. | |
151 | * | |
152 | * Cache deletions are performed under the i_lock, which ensures that | |
153 | * when an inode goes empty, it will reliably get queued on the LRU. | |
154 | * | |
155 | * Cache additions do not acquire the i_lock and may race with this | |
156 | * check, in which case we'll report the inode as shrinkable when it | |
157 | * has cache pages. This is okay: the shrinker also checks the | |
158 | * refcount and the referenced bit, which will be elevated or set in | |
159 | * the process of adding new cache pages to an inode. | |
160 | */ | |
161 | static inline bool mapping_shrinkable(struct address_space *mapping) | |
162 | { | |
163 | void *head; | |
164 | ||
165 | /* | |
166 | * On highmem systems, there could be lowmem pressure from the | |
167 | * inodes before there is highmem pressure from the page | |
168 | * cache. Make inodes shrinkable regardless of cache state. | |
169 | */ | |
170 | if (IS_ENABLED(CONFIG_HIGHMEM)) | |
171 | return true; | |
172 | ||
173 | /* Cache completely empty? Shrink away. */ | |
174 | head = rcu_access_pointer(mapping->i_pages.xa_head); | |
175 | if (!head) | |
176 | return true; | |
177 | ||
178 | /* | |
179 | * The xarray stores single offset-0 entries directly in the | |
180 | * head pointer, which allows non-resident page cache entries | |
181 | * to escape the shadow shrinker's list of xarray nodes. The | |
182 | * inode shrinker needs to pick them up under memory pressure. | |
183 | */ | |
184 | if (!xa_is_node(head) && xa_is_value(head)) | |
185 | return true; | |
186 | ||
187 | return false; | |
188 | } | |
189 | ||
1da177e4 | 190 | /* |
9c5d760b | 191 | * Bits in mapping->flags. |
1da177e4 | 192 | */ |
9a896c9a | 193 | enum mapping_flags { |
9c5d760b MH |
194 | AS_EIO = 0, /* IO error on async write */ |
195 | AS_ENOSPC = 1, /* ENOSPC on async write */ | |
196 | AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */ | |
197 | AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */ | |
198 | AS_EXITING = 4, /* final truncate in progress */ | |
371a096e | 199 | /* writeback related tags are not used */ |
9c5d760b | 200 | AS_NO_WRITEBACK_TAGS = 5, |
ed2145c4 | 201 | AS_LARGE_FOLIO_SUPPORT = 6, |
9a896c9a | 202 | }; |
1da177e4 | 203 | |
8ed1e46a JL |
204 | /** |
205 | * mapping_set_error - record a writeback error in the address_space | |
767e5ee5 MWO |
206 | * @mapping: the mapping in which an error should be set |
207 | * @error: the error to set in the mapping | |
8ed1e46a JL |
208 | * |
209 | * When writeback fails in some way, we must record that error so that | |
210 | * userspace can be informed when fsync and the like are called. We endeavor | |
211 | * to report errors on any file that was open at the time of the error. Some | |
212 | * internal callers also need to know when writeback errors have occurred. | |
213 | * | |
214 | * When a writeback error occurs, most filesystems will want to call | |
215 | * mapping_set_error to record the error in the mapping so that it can be | |
216 | * reported when the application calls fsync(2). | |
217 | */ | |
3e9f45bd GC |
218 | static inline void mapping_set_error(struct address_space *mapping, int error) |
219 | { | |
8ed1e46a JL |
220 | if (likely(!error)) |
221 | return; | |
222 | ||
223 | /* Record in wb_err for checkers using errseq_t based tracking */ | |
735e4ae5 JL |
224 | __filemap_set_wb_err(mapping, error); |
225 | ||
226 | /* Record it in superblock */ | |
8b7b2eb1 MK |
227 | if (mapping->host) |
228 | errseq_set(&mapping->host->i_sb->s_wb_err, error); | |
8ed1e46a JL |
229 | |
230 | /* Record it in flags for now, for legacy callers */ | |
231 | if (error == -ENOSPC) | |
232 | set_bit(AS_ENOSPC, &mapping->flags); | |
233 | else | |
234 | set_bit(AS_EIO, &mapping->flags); | |
3e9f45bd GC |
235 | } |
236 | ||
ba9ddf49 LS |
237 | static inline void mapping_set_unevictable(struct address_space *mapping) |
238 | { | |
239 | set_bit(AS_UNEVICTABLE, &mapping->flags); | |
240 | } | |
241 | ||
89e004ea LS |
242 | static inline void mapping_clear_unevictable(struct address_space *mapping) |
243 | { | |
244 | clear_bit(AS_UNEVICTABLE, &mapping->flags); | |
245 | } | |
246 | ||
1eb6234e | 247 | static inline bool mapping_unevictable(struct address_space *mapping) |
ba9ddf49 | 248 | { |
1eb6234e | 249 | return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags); |
ba9ddf49 | 250 | } |
ba9ddf49 | 251 | |
91b0abe3 JW |
252 | static inline void mapping_set_exiting(struct address_space *mapping) |
253 | { | |
254 | set_bit(AS_EXITING, &mapping->flags); | |
255 | } | |
256 | ||
257 | static inline int mapping_exiting(struct address_space *mapping) | |
258 | { | |
259 | return test_bit(AS_EXITING, &mapping->flags); | |
260 | } | |
261 | ||
371a096e HY |
262 | static inline void mapping_set_no_writeback_tags(struct address_space *mapping) |
263 | { | |
264 | set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags); | |
265 | } | |
266 | ||
267 | static inline int mapping_use_writeback_tags(struct address_space *mapping) | |
268 | { | |
269 | return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags); | |
270 | } | |
271 | ||
dd0fc66f | 272 | static inline gfp_t mapping_gfp_mask(struct address_space * mapping) |
1da177e4 | 273 | { |
9c5d760b | 274 | return mapping->gfp_mask; |
1da177e4 LT |
275 | } |
276 | ||
c62d2555 MH |
277 | /* Restricts the given gfp_mask to what the mapping allows. */ |
278 | static inline gfp_t mapping_gfp_constraint(struct address_space *mapping, | |
279 | gfp_t gfp_mask) | |
280 | { | |
281 | return mapping_gfp_mask(mapping) & gfp_mask; | |
282 | } | |
283 | ||
1da177e4 LT |
284 | /* |
285 | * This is non-atomic. Only to be used before the mapping is activated. | |
286 | * Probably needs a barrier... | |
287 | */ | |
260b2367 | 288 | static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask) |
1da177e4 | 289 | { |
9c5d760b | 290 | m->gfp_mask = mask; |
1da177e4 LT |
291 | } |
292 | ||
ff36da69 MWO |
293 | /** |
294 | * mapping_set_large_folios() - Indicate the file supports large folios. | |
295 | * @mapping: The file. | |
296 | * | |
297 | * The filesystem should call this function in its inode constructor to | |
298 | * indicate that the VFS can use large folios to cache the contents of | |
299 | * the file. | |
300 | * | |
301 | * Context: This should not be called while the inode is active as it | |
302 | * is non-atomic. | |
303 | */ | |
304 | static inline void mapping_set_large_folios(struct address_space *mapping) | |
305 | { | |
ed2145c4 | 306 | __set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags); |
ff36da69 MWO |
307 | } |
308 | ||
421f1ab4 MWO |
309 | /* |
310 | * Large folio support currently depends on THP. These dependencies are | |
311 | * being worked on but are not yet fixed. | |
312 | */ | |
ed2145c4 | 313 | static inline bool mapping_large_folio_support(struct address_space *mapping) |
01c70267 | 314 | { |
421f1ab4 MWO |
315 | return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) && |
316 | test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags); | |
01c70267 MWO |
317 | } |
318 | ||
6f4d2f97 MWO |
319 | static inline int filemap_nr_thps(struct address_space *mapping) |
320 | { | |
321 | #ifdef CONFIG_READ_ONLY_THP_FOR_FS | |
322 | return atomic_read(&mapping->nr_thps); | |
323 | #else | |
324 | return 0; | |
325 | #endif | |
326 | } | |
327 | ||
328 | static inline void filemap_nr_thps_inc(struct address_space *mapping) | |
329 | { | |
330 | #ifdef CONFIG_READ_ONLY_THP_FOR_FS | |
ed2145c4 | 331 | if (!mapping_large_folio_support(mapping)) |
6f4d2f97 MWO |
332 | atomic_inc(&mapping->nr_thps); |
333 | #else | |
06d44142 | 334 | WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0); |
6f4d2f97 MWO |
335 | #endif |
336 | } | |
337 | ||
338 | static inline void filemap_nr_thps_dec(struct address_space *mapping) | |
339 | { | |
340 | #ifdef CONFIG_READ_ONLY_THP_FOR_FS | |
ed2145c4 | 341 | if (!mapping_large_folio_support(mapping)) |
6f4d2f97 MWO |
342 | atomic_dec(&mapping->nr_thps); |
343 | #else | |
06d44142 | 344 | WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0); |
6f4d2f97 MWO |
345 | #endif |
346 | } | |
347 | ||
2f52578f MWO |
348 | struct address_space *page_mapping(struct page *); |
349 | struct address_space *folio_mapping(struct folio *); | |
350 | struct address_space *swapcache_mapping(struct folio *); | |
351 | ||
352 | /** | |
353 | * folio_file_mapping - Find the mapping this folio belongs to. | |
354 | * @folio: The folio. | |
355 | * | |
356 | * For folios which are in the page cache, return the mapping that this | |
357 | * page belongs to. Folios in the swap cache return the mapping of the | |
358 | * swap file or swap device where the data is stored. This is different | |
359 | * from the mapping returned by folio_mapping(). The only reason to | |
360 | * use it is if, like NFS, you return 0 from ->activate_swapfile. | |
361 | * | |
362 | * Do not call this for folios which aren't in the page cache or swap cache. | |
363 | */ | |
364 | static inline struct address_space *folio_file_mapping(struct folio *folio) | |
365 | { | |
366 | if (unlikely(folio_test_swapcache(folio))) | |
367 | return swapcache_mapping(folio); | |
368 | ||
369 | return folio->mapping; | |
370 | } | |
371 | ||
372 | static inline struct address_space *page_file_mapping(struct page *page) | |
373 | { | |
374 | return folio_file_mapping(page_folio(page)); | |
375 | } | |
376 | ||
842ca547 MWO |
377 | /* |
378 | * For file cache pages, return the address_space, otherwise return NULL | |
379 | */ | |
380 | static inline struct address_space *page_mapping_file(struct page *page) | |
381 | { | |
2f52578f MWO |
382 | struct folio *folio = page_folio(page); |
383 | ||
384 | if (unlikely(folio_test_swapcache(folio))) | |
842ca547 | 385 | return NULL; |
2f52578f | 386 | return folio_mapping(folio); |
842ca547 MWO |
387 | } |
388 | ||
452c472e DH |
389 | /** |
390 | * folio_inode - Get the host inode for this folio. | |
391 | * @folio: The folio. | |
392 | * | |
393 | * For folios which are in the page cache, return the inode that this folio | |
394 | * belongs to. | |
395 | * | |
396 | * Do not call this for folios which aren't in the page cache. | |
397 | */ | |
398 | static inline struct inode *folio_inode(struct folio *folio) | |
399 | { | |
400 | return folio->mapping->host; | |
401 | } | |
402 | ||
b03143ac | 403 | /** |
85d0a2ed MWO |
404 | * folio_attach_private - Attach private data to a folio. |
405 | * @folio: Folio to attach data to. | |
406 | * @data: Data to attach to folio. | |
b03143ac | 407 | * |
85d0a2ed MWO |
408 | * Attaching private data to a folio increments the page's reference count. |
409 | * The data must be detached before the folio will be freed. | |
b03143ac | 410 | */ |
85d0a2ed | 411 | static inline void folio_attach_private(struct folio *folio, void *data) |
b03143ac | 412 | { |
85d0a2ed MWO |
413 | folio_get(folio); |
414 | folio->private = data; | |
415 | folio_set_private(folio); | |
b03143ac GJ |
416 | } |
417 | ||
a19672f6 DH |
418 | /** |
419 | * folio_change_private - Change private data on a folio. | |
420 | * @folio: Folio to change the data on. | |
421 | * @data: Data to set on the folio. | |
422 | * | |
423 | * Change the private data attached to a folio and return the old | |
424 | * data. The page must previously have had data attached and the data | |
425 | * must be detached before the folio will be freed. | |
426 | * | |
427 | * Return: Data that was previously attached to the folio. | |
428 | */ | |
429 | static inline void *folio_change_private(struct folio *folio, void *data) | |
430 | { | |
431 | void *old = folio_get_private(folio); | |
432 | ||
433 | folio->private = data; | |
434 | return old; | |
435 | } | |
436 | ||
b03143ac | 437 | /** |
85d0a2ed MWO |
438 | * folio_detach_private - Detach private data from a folio. |
439 | * @folio: Folio to detach data from. | |
b03143ac | 440 | * |
85d0a2ed | 441 | * Removes the data that was previously attached to the folio and decrements |
b03143ac GJ |
442 | * the refcount on the page. |
443 | * | |
85d0a2ed | 444 | * Return: Data that was attached to the folio. |
b03143ac | 445 | */ |
85d0a2ed | 446 | static inline void *folio_detach_private(struct folio *folio) |
b03143ac | 447 | { |
85d0a2ed | 448 | void *data = folio_get_private(folio); |
b03143ac | 449 | |
85d0a2ed | 450 | if (!folio_test_private(folio)) |
b03143ac | 451 | return NULL; |
85d0a2ed MWO |
452 | folio_clear_private(folio); |
453 | folio->private = NULL; | |
454 | folio_put(folio); | |
b03143ac GJ |
455 | |
456 | return data; | |
457 | } | |
458 | ||
85d0a2ed MWO |
459 | static inline void attach_page_private(struct page *page, void *data) |
460 | { | |
461 | folio_attach_private(page_folio(page), data); | |
462 | } | |
463 | ||
464 | static inline void *detach_page_private(struct page *page) | |
465 | { | |
466 | return folio_detach_private(page_folio(page)); | |
467 | } | |
468 | ||
44110fe3 | 469 | #ifdef CONFIG_NUMA |
bb3c579e | 470 | struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order); |
44110fe3 | 471 | #else |
bb3c579e | 472 | static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order) |
2ae88149 | 473 | { |
bb3c579e | 474 | return folio_alloc(gfp, order); |
2ae88149 NP |
475 | } |
476 | #endif | |
477 | ||
bb3c579e MWO |
478 | static inline struct page *__page_cache_alloc(gfp_t gfp) |
479 | { | |
480 | return &filemap_alloc_folio(gfp, 0)->page; | |
481 | } | |
482 | ||
1da177e4 LT |
483 | static inline struct page *page_cache_alloc(struct address_space *x) |
484 | { | |
2ae88149 | 485 | return __page_cache_alloc(mapping_gfp_mask(x)); |
1da177e4 LT |
486 | } |
487 | ||
8a5c743e | 488 | static inline gfp_t readahead_gfp_mask(struct address_space *x) |
7b1de586 | 489 | { |
453f85d4 | 490 | return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN; |
7b1de586 WF |
491 | } |
492 | ||
e9b5b23e | 493 | typedef int filler_t(struct file *, struct folio *); |
1da177e4 | 494 | |
0d3f9296 | 495 | pgoff_t page_cache_next_miss(struct address_space *mapping, |
e7b563bb | 496 | pgoff_t index, unsigned long max_scan); |
0d3f9296 | 497 | pgoff_t page_cache_prev_miss(struct address_space *mapping, |
e7b563bb JW |
498 | pgoff_t index, unsigned long max_scan); |
499 | ||
2457aec6 MG |
500 | #define FGP_ACCESSED 0x00000001 |
501 | #define FGP_LOCK 0x00000002 | |
502 | #define FGP_CREAT 0x00000004 | |
503 | #define FGP_WRITE 0x00000008 | |
504 | #define FGP_NOFS 0x00000010 | |
505 | #define FGP_NOWAIT 0x00000020 | |
a75d4c33 | 506 | #define FGP_FOR_MMAP 0x00000040 |
48c9d113 | 507 | #define FGP_STABLE 0x00000080 |
2457aec6 | 508 | |
e999a5c5 MW |
509 | #define FGP_WRITEBEGIN (FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE) |
510 | ||
263e721e | 511 | void *filemap_get_entry(struct address_space *mapping, pgoff_t index); |
3f0c6a07 MWO |
512 | struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index, |
513 | int fgp_flags, gfp_t gfp); | |
514 | struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index, | |
515 | int fgp_flags, gfp_t gfp); | |
516 | ||
517 | /** | |
518 | * filemap_get_folio - Find and get a folio. | |
519 | * @mapping: The address_space to search. | |
520 | * @index: The page index. | |
521 | * | |
522 | * Looks up the page cache entry at @mapping & @index. If a folio is | |
523 | * present, it is returned with an increased refcount. | |
524 | * | |
66dabbb6 CH |
525 | * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for |
526 | * this index. Will not return a shadow, swap or DAX entry. | |
3f0c6a07 MWO |
527 | */ |
528 | static inline struct folio *filemap_get_folio(struct address_space *mapping, | |
529 | pgoff_t index) | |
530 | { | |
531 | return __filemap_get_folio(mapping, index, 0, 0); | |
532 | } | |
2457aec6 | 533 | |
ccd16945 MWO |
534 | /** |
535 | * filemap_lock_folio - Find and lock a folio. | |
536 | * @mapping: The address_space to search. | |
537 | * @index: The page index. | |
538 | * | |
539 | * Looks up the page cache entry at @mapping & @index. If a folio is | |
540 | * present, it is returned locked with an increased refcount. | |
541 | * | |
542 | * Context: May sleep. | |
66dabbb6 CH |
543 | * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for |
544 | * this index. Will not return a shadow, swap or DAX entry. | |
ccd16945 MWO |
545 | */ |
546 | static inline struct folio *filemap_lock_folio(struct address_space *mapping, | |
547 | pgoff_t index) | |
548 | { | |
549 | return __filemap_get_folio(mapping, index, FGP_LOCK, 0); | |
550 | } | |
551 | ||
ee7a5906 VMO |
552 | /** |
553 | * filemap_grab_folio - grab a folio from the page cache | |
554 | * @mapping: The address space to search | |
555 | * @index: The page index | |
556 | * | |
557 | * Looks up the page cache entry at @mapping & @index. If no folio is found, | |
558 | * a new folio is created. The folio is locked, marked as accessed, and | |
559 | * returned. | |
560 | * | |
66dabbb6 CH |
561 | * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found |
562 | * and failed to create a folio. | |
ee7a5906 VMO |
563 | */ |
564 | static inline struct folio *filemap_grab_folio(struct address_space *mapping, | |
565 | pgoff_t index) | |
566 | { | |
567 | return __filemap_get_folio(mapping, index, | |
568 | FGP_LOCK | FGP_ACCESSED | FGP_CREAT, | |
569 | mapping_gfp_mask(mapping)); | |
570 | } | |
571 | ||
2457aec6 MG |
572 | /** |
573 | * find_get_page - find and get a page reference | |
574 | * @mapping: the address_space to search | |
575 | * @offset: the page index | |
576 | * | |
577 | * Looks up the page cache slot at @mapping & @offset. If there is a | |
578 | * page cache page, it is returned with an increased refcount. | |
579 | * | |
580 | * Otherwise, %NULL is returned. | |
581 | */ | |
582 | static inline struct page *find_get_page(struct address_space *mapping, | |
583 | pgoff_t offset) | |
584 | { | |
45f87de5 | 585 | return pagecache_get_page(mapping, offset, 0, 0); |
2457aec6 MG |
586 | } |
587 | ||
588 | static inline struct page *find_get_page_flags(struct address_space *mapping, | |
589 | pgoff_t offset, int fgp_flags) | |
590 | { | |
45f87de5 | 591 | return pagecache_get_page(mapping, offset, fgp_flags, 0); |
2457aec6 MG |
592 | } |
593 | ||
594 | /** | |
595 | * find_lock_page - locate, pin and lock a pagecache page | |
2457aec6 | 596 | * @mapping: the address_space to search |
89b42235 | 597 | * @index: the page index |
2457aec6 | 598 | * |
89b42235 | 599 | * Looks up the page cache entry at @mapping & @index. If there is a |
2457aec6 MG |
600 | * page cache page, it is returned locked and with an increased |
601 | * refcount. | |
602 | * | |
a8cf7f27 MWO |
603 | * Context: May sleep. |
604 | * Return: A struct page or %NULL if there is no page in the cache for this | |
605 | * index. | |
2457aec6 MG |
606 | */ |
607 | static inline struct page *find_lock_page(struct address_space *mapping, | |
a8cf7f27 MWO |
608 | pgoff_t index) |
609 | { | |
610 | return pagecache_get_page(mapping, index, FGP_LOCK, 0); | |
611 | } | |
612 | ||
2457aec6 MG |
613 | /** |
614 | * find_or_create_page - locate or add a pagecache page | |
615 | * @mapping: the page's address_space | |
616 | * @index: the page's index into the mapping | |
617 | * @gfp_mask: page allocation mode | |
618 | * | |
619 | * Looks up the page cache slot at @mapping & @offset. If there is a | |
620 | * page cache page, it is returned locked and with an increased | |
621 | * refcount. | |
622 | * | |
623 | * If the page is not present, a new page is allocated using @gfp_mask | |
624 | * and added to the page cache and the VM's LRU list. The page is | |
625 | * returned locked and with an increased refcount. | |
626 | * | |
627 | * On memory exhaustion, %NULL is returned. | |
628 | * | |
629 | * find_or_create_page() may sleep, even if @gfp_flags specifies an | |
630 | * atomic allocation! | |
631 | */ | |
632 | static inline struct page *find_or_create_page(struct address_space *mapping, | |
767e5ee5 | 633 | pgoff_t index, gfp_t gfp_mask) |
2457aec6 | 634 | { |
767e5ee5 | 635 | return pagecache_get_page(mapping, index, |
2457aec6 | 636 | FGP_LOCK|FGP_ACCESSED|FGP_CREAT, |
45f87de5 | 637 | gfp_mask); |
2457aec6 MG |
638 | } |
639 | ||
640 | /** | |
641 | * grab_cache_page_nowait - returns locked page at given index in given cache | |
642 | * @mapping: target address_space | |
643 | * @index: the page index | |
644 | * | |
645 | * Same as grab_cache_page(), but do not wait if the page is unavailable. | |
646 | * This is intended for speculative data generators, where the data can | |
647 | * be regenerated if the page couldn't be grabbed. This routine should | |
648 | * be safe to call while holding the lock for another page. | |
649 | * | |
650 | * Clear __GFP_FS when allocating the page to avoid recursion into the fs | |
651 | * and deadlock against the caller's locked page. | |
652 | */ | |
653 | static inline struct page *grab_cache_page_nowait(struct address_space *mapping, | |
654 | pgoff_t index) | |
655 | { | |
656 | return pagecache_get_page(mapping, index, | |
657 | FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT, | |
45f87de5 | 658 | mapping_gfp_mask(mapping)); |
2457aec6 MG |
659 | } |
660 | ||
9257e156 MWO |
661 | #define swapcache_index(folio) __page_file_index(&(folio)->page) |
662 | ||
663 | /** | |
664 | * folio_index - File index of a folio. | |
665 | * @folio: The folio. | |
666 | * | |
667 | * For a folio which is either in the page cache or the swap cache, | |
668 | * return its index within the address_space it belongs to. If you know | |
669 | * the page is definitely in the page cache, you can look at the folio's | |
670 | * index directly. | |
671 | * | |
672 | * Return: The index (offset in units of pages) of a folio in its file. | |
673 | */ | |
674 | static inline pgoff_t folio_index(struct folio *folio) | |
675 | { | |
676 | if (unlikely(folio_test_swapcache(folio))) | |
677 | return swapcache_index(folio); | |
678 | return folio->index; | |
679 | } | |
680 | ||
f94b18f6 MWO |
681 | /** |
682 | * folio_next_index - Get the index of the next folio. | |
683 | * @folio: The current folio. | |
684 | * | |
685 | * Return: The index of the folio which follows this folio in the file. | |
686 | */ | |
687 | static inline pgoff_t folio_next_index(struct folio *folio) | |
688 | { | |
689 | return folio->index + folio_nr_pages(folio); | |
690 | } | |
691 | ||
9257e156 MWO |
692 | /** |
693 | * folio_file_page - The page for a particular index. | |
694 | * @folio: The folio which contains this index. | |
695 | * @index: The index we want to look up. | |
696 | * | |
697 | * Sometimes after looking up a folio in the page cache, we need to | |
698 | * obtain the specific page for an index (eg a page fault). | |
699 | * | |
700 | * Return: The page containing the file data for this index. | |
701 | */ | |
702 | static inline struct page *folio_file_page(struct folio *folio, pgoff_t index) | |
703 | { | |
704 | /* HugeTLBfs indexes the page cache in units of hpage_size */ | |
705 | if (folio_test_hugetlb(folio)) | |
706 | return &folio->page; | |
707 | return folio_page(folio, index & (folio_nr_pages(folio) - 1)); | |
708 | } | |
709 | ||
710 | /** | |
711 | * folio_contains - Does this folio contain this index? | |
712 | * @folio: The folio. | |
713 | * @index: The page index within the file. | |
714 | * | |
715 | * Context: The caller should have the page locked in order to prevent | |
716 | * (eg) shmem from moving the page between the page cache and swap cache | |
717 | * and changing its index in the middle of the operation. | |
718 | * Return: true or false. | |
719 | */ | |
720 | static inline bool folio_contains(struct folio *folio, pgoff_t index) | |
721 | { | |
722 | /* HugeTLBfs indexes the page cache in units of hpage_size */ | |
723 | if (folio_test_hugetlb(folio)) | |
724 | return folio->index == index; | |
725 | return index - folio_index(folio) < folio_nr_pages(folio); | |
726 | } | |
727 | ||
ec848215 MWO |
728 | /* |
729 | * Given the page we found in the page cache, return the page corresponding | |
730 | * to this index in the file | |
731 | */ | |
732 | static inline struct page *find_subpage(struct page *head, pgoff_t index) | |
4101196b | 733 | { |
ec848215 MWO |
734 | /* HugeTLBfs wants the head page regardless */ |
735 | if (PageHuge(head)) | |
736 | return head; | |
4101196b | 737 | |
6c357848 | 738 | return head + (index & (thp_nr_pages(head) - 1)); |
4101196b MWO |
739 | } |
740 | ||
be0ced5e MWO |
741 | unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start, |
742 | pgoff_t end, struct folio_batch *fbatch); | |
35b47146 VMO |
743 | unsigned filemap_get_folios_contig(struct address_space *mapping, |
744 | pgoff_t *start, pgoff_t end, struct folio_batch *fbatch); | |
247f9e1f VMO |
745 | unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start, |
746 | pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch); | |
1da177e4 | 747 | |
54566b2c | 748 | struct page *grab_cache_page_write_begin(struct address_space *mapping, |
b7446e7c | 749 | pgoff_t index); |
afddba49 | 750 | |
1da177e4 LT |
751 | /* |
752 | * Returns locked page at given index in given cache, creating it if needed. | |
753 | */ | |
57f6b96c FW |
754 | static inline struct page *grab_cache_page(struct address_space *mapping, |
755 | pgoff_t index) | |
1da177e4 LT |
756 | { |
757 | return find_or_create_page(mapping, index, mapping_gfp_mask(mapping)); | |
758 | } | |
759 | ||
539a3322 | 760 | struct folio *read_cache_folio(struct address_space *, pgoff_t index, |
e9b5b23e | 761 | filler_t *filler, struct file *file); |
3e629597 MWO |
762 | struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index, |
763 | gfp_t flags); | |
539a3322 | 764 | struct page *read_cache_page(struct address_space *, pgoff_t index, |
e9b5b23e | 765 | filler_t *filler, struct file *file); |
0531b2aa LT |
766 | extern struct page * read_cache_page_gfp(struct address_space *mapping, |
767 | pgoff_t index, gfp_t gfp_mask); | |
1da177e4 | 768 | |
090d2b18 | 769 | static inline struct page *read_mapping_page(struct address_space *mapping, |
c14c6843 | 770 | pgoff_t index, struct file *file) |
090d2b18 | 771 | { |
c14c6843 | 772 | return read_cache_page(mapping, index, NULL, file); |
090d2b18 PE |
773 | } |
774 | ||
539a3322 | 775 | static inline struct folio *read_mapping_folio(struct address_space *mapping, |
c14c6843 | 776 | pgoff_t index, struct file *file) |
539a3322 | 777 | { |
c14c6843 | 778 | return read_cache_folio(mapping, index, NULL, file); |
539a3322 MWO |
779 | } |
780 | ||
a0f7a756 | 781 | /* |
fe19bd3d | 782 | * Get index of the page within radix-tree (but not for hugetlb pages). |
5cbc198a | 783 | * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE) |
a0f7a756 | 784 | */ |
5cbc198a | 785 | static inline pgoff_t page_to_index(struct page *page) |
a0f7a756 | 786 | { |
fe3df441 | 787 | struct page *head; |
e9b61f19 | 788 | |
e9b61f19 | 789 | if (likely(!PageTransTail(page))) |
09cbfeaf | 790 | return page->index; |
e9b61f19 | 791 | |
fe3df441 | 792 | head = compound_head(page); |
e9b61f19 KS |
793 | /* |
794 | * We don't initialize ->index for tail pages: calculate based on | |
795 | * head page | |
796 | */ | |
fe3df441 | 797 | return head->index + page - head; |
a0f7a756 NH |
798 | } |
799 | ||
fe19bd3d HD |
800 | extern pgoff_t hugetlb_basepage_index(struct page *page); |
801 | ||
5cbc198a | 802 | /* |
fe19bd3d HD |
803 | * Get the offset in PAGE_SIZE (even for hugetlb pages). |
804 | * (TODO: hugetlb pages should have ->index in PAGE_SIZE) | |
5cbc198a KS |
805 | */ |
806 | static inline pgoff_t page_to_pgoff(struct page *page) | |
807 | { | |
fe19bd3d HD |
808 | if (unlikely(PageHuge(page))) |
809 | return hugetlb_basepage_index(page); | |
5cbc198a KS |
810 | return page_to_index(page); |
811 | } | |
812 | ||
1da177e4 LT |
813 | /* |
814 | * Return byte-offset into filesystem object for page. | |
815 | */ | |
816 | static inline loff_t page_offset(struct page *page) | |
817 | { | |
09cbfeaf | 818 | return ((loff_t)page->index) << PAGE_SHIFT; |
1da177e4 LT |
819 | } |
820 | ||
f981c595 MG |
821 | static inline loff_t page_file_offset(struct page *page) |
822 | { | |
8cd79788 | 823 | return ((loff_t)page_index(page)) << PAGE_SHIFT; |
f981c595 MG |
824 | } |
825 | ||
352b47a6 MWO |
826 | /** |
827 | * folio_pos - Returns the byte position of this folio in its file. | |
828 | * @folio: The folio. | |
829 | */ | |
830 | static inline loff_t folio_pos(struct folio *folio) | |
831 | { | |
832 | return page_offset(&folio->page); | |
833 | } | |
834 | ||
835 | /** | |
836 | * folio_file_pos - Returns the byte position of this folio in its file. | |
837 | * @folio: The folio. | |
838 | * | |
839 | * This differs from folio_pos() for folios which belong to a swap file. | |
840 | * NFS is the only filesystem today which needs to use folio_file_pos(). | |
841 | */ | |
842 | static inline loff_t folio_file_pos(struct folio *folio) | |
843 | { | |
844 | return page_file_offset(&folio->page); | |
845 | } | |
846 | ||
f087b903 MWO |
847 | /* |
848 | * Get the offset in PAGE_SIZE (even for hugetlb folios). | |
849 | * (TODO: hugetlb folios should have ->index in PAGE_SIZE) | |
850 | */ | |
851 | static inline pgoff_t folio_pgoff(struct folio *folio) | |
852 | { | |
853 | if (unlikely(folio_test_hugetlb(folio))) | |
854 | return hugetlb_basepage_index(&folio->page); | |
855 | return folio->index; | |
856 | } | |
857 | ||
0fe6e20b NH |
858 | extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma, |
859 | unsigned long address); | |
860 | ||
1da177e4 LT |
861 | static inline pgoff_t linear_page_index(struct vm_area_struct *vma, |
862 | unsigned long address) | |
863 | { | |
0fe6e20b NH |
864 | pgoff_t pgoff; |
865 | if (unlikely(is_vm_hugetlb_page(vma))) | |
866 | return linear_hugepage_index(vma, address); | |
867 | pgoff = (address - vma->vm_start) >> PAGE_SHIFT; | |
1da177e4 | 868 | pgoff += vma->vm_pgoff; |
09cbfeaf | 869 | return pgoff; |
1da177e4 LT |
870 | } |
871 | ||
c7510ab2 | 872 | struct wait_page_key { |
df4d4f12 | 873 | struct folio *folio; |
c7510ab2 JA |
874 | int bit_nr; |
875 | int page_match; | |
876 | }; | |
877 | ||
878 | struct wait_page_queue { | |
df4d4f12 | 879 | struct folio *folio; |
c7510ab2 JA |
880 | int bit_nr; |
881 | wait_queue_entry_t wait; | |
882 | }; | |
883 | ||
cdc8fcb4 | 884 | static inline bool wake_page_match(struct wait_page_queue *wait_page, |
c7510ab2 JA |
885 | struct wait_page_key *key) |
886 | { | |
df4d4f12 | 887 | if (wait_page->folio != key->folio) |
cdc8fcb4 | 888 | return false; |
c7510ab2 JA |
889 | key->page_match = 1; |
890 | ||
891 | if (wait_page->bit_nr != key->bit_nr) | |
cdc8fcb4 | 892 | return false; |
d1932dc3 | 893 | |
cdc8fcb4 | 894 | return true; |
d1932dc3 JA |
895 | } |
896 | ||
7c23c782 | 897 | void __folio_lock(struct folio *folio); |
af7f29d9 | 898 | int __folio_lock_killable(struct folio *folio); |
9138e47e | 899 | bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm, |
d065bd81 | 900 | unsigned int flags); |
4e136428 MWO |
901 | void unlock_page(struct page *page); |
902 | void folio_unlock(struct folio *folio); | |
1da177e4 | 903 | |
cd125eea MWO |
904 | /** |
905 | * folio_trylock() - Attempt to lock a folio. | |
906 | * @folio: The folio to attempt to lock. | |
907 | * | |
908 | * Sometimes it is undesirable to wait for a folio to be unlocked (eg | |
909 | * when the locks are being taken in the wrong order, or if making | |
910 | * progress through a batch of folios is more important than processing | |
911 | * them in order). Usually folio_lock() is the correct function to call. | |
912 | * | |
913 | * Context: Any context. | |
914 | * Return: Whether the lock was successfully acquired. | |
915 | */ | |
7c23c782 MWO |
916 | static inline bool folio_trylock(struct folio *folio) |
917 | { | |
918 | return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0))); | |
919 | } | |
920 | ||
f4458845 AM |
921 | /* |
922 | * Return true if the page was successfully locked | |
923 | */ | |
529ae9aa NP |
924 | static inline int trylock_page(struct page *page) |
925 | { | |
7c23c782 MWO |
926 | return folio_trylock(page_folio(page)); |
927 | } | |
928 | ||
cd125eea MWO |
929 | /** |
930 | * folio_lock() - Lock this folio. | |
931 | * @folio: The folio to lock. | |
932 | * | |
933 | * The folio lock protects against many things, probably more than it | |
934 | * should. It is primarily held while a folio is being brought uptodate, | |
935 | * either from its backing file or from swap. It is also held while a | |
936 | * folio is being truncated from its address_space, so holding the lock | |
937 | * is sufficient to keep folio->mapping stable. | |
938 | * | |
939 | * The folio lock is also held while write() is modifying the page to | |
940 | * provide POSIX atomicity guarantees (as long as the write does not | |
941 | * cross a page boundary). Other modifications to the data in the folio | |
942 | * do not hold the folio lock and can race with writes, eg DMA and stores | |
943 | * to mapped pages. | |
944 | * | |
945 | * Context: May sleep. If you need to acquire the locks of two or | |
946 | * more folios, they must be in order of ascending index, if they are | |
947 | * in the same address_space. If they are in different address_spaces, | |
948 | * acquire the lock of the folio which belongs to the address_space which | |
949 | * has the lowest address in memory first. | |
950 | */ | |
7c23c782 MWO |
951 | static inline void folio_lock(struct folio *folio) |
952 | { | |
953 | might_sleep(); | |
954 | if (!folio_trylock(folio)) | |
955 | __folio_lock(folio); | |
529ae9aa NP |
956 | } |
957 | ||
cd125eea MWO |
958 | /** |
959 | * lock_page() - Lock the folio containing this page. | |
960 | * @page: The page to lock. | |
961 | * | |
962 | * See folio_lock() for a description of what the lock protects. | |
963 | * This is a legacy function and new code should probably use folio_lock() | |
964 | * instead. | |
965 | * | |
966 | * Context: May sleep. Pages in the same folio share a lock, so do not | |
967 | * attempt to lock two pages which share a folio. | |
db37648c | 968 | */ |
1da177e4 LT |
969 | static inline void lock_page(struct page *page) |
970 | { | |
7c23c782 | 971 | struct folio *folio; |
1da177e4 | 972 | might_sleep(); |
7c23c782 MWO |
973 | |
974 | folio = page_folio(page); | |
975 | if (!folio_trylock(folio)) | |
976 | __folio_lock(folio); | |
1da177e4 | 977 | } |
db37648c | 978 | |
cd125eea MWO |
979 | /** |
980 | * folio_lock_killable() - Lock this folio, interruptible by a fatal signal. | |
981 | * @folio: The folio to lock. | |
982 | * | |
983 | * Attempts to lock the folio, like folio_lock(), except that the sleep | |
984 | * to acquire the lock is interruptible by a fatal signal. | |
985 | * | |
986 | * Context: May sleep; see folio_lock(). | |
987 | * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received. | |
988 | */ | |
af7f29d9 MWO |
989 | static inline int folio_lock_killable(struct folio *folio) |
990 | { | |
991 | might_sleep(); | |
992 | if (!folio_trylock(folio)) | |
993 | return __folio_lock_killable(folio); | |
994 | return 0; | |
995 | } | |
996 | ||
d065bd81 | 997 | /* |
19672a9e | 998 | * folio_lock_or_retry - Lock the folio, unless this would block and the |
d065bd81 | 999 | * caller indicated that it can handle a retry. |
9a95f3cf | 1000 | * |
c1e8d7c6 | 1001 | * Return value and mmap_lock implications depend on flags; see |
9138e47e | 1002 | * __folio_lock_or_retry(). |
d065bd81 | 1003 | */ |
19672a9e MWO |
1004 | static inline bool folio_lock_or_retry(struct folio *folio, |
1005 | struct mm_struct *mm, unsigned int flags) | |
d065bd81 ML |
1006 | { |
1007 | might_sleep(); | |
9138e47e | 1008 | return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags); |
d065bd81 ML |
1009 | } |
1010 | ||
1da177e4 | 1011 | /* |
101c0bf6 | 1012 | * This is exported only for folio_wait_locked/folio_wait_writeback, etc., |
74d81bfa | 1013 | * and should not be used directly. |
1da177e4 | 1014 | */ |
101c0bf6 MWO |
1015 | void folio_wait_bit(struct folio *folio, int bit_nr); |
1016 | int folio_wait_bit_killable(struct folio *folio, int bit_nr); | |
a4796e37 | 1017 | |
1da177e4 | 1018 | /* |
6baa8d60 | 1019 | * Wait for a folio to be unlocked. |
1da177e4 | 1020 | * |
6baa8d60 | 1021 | * This must be called with the caller "holding" the folio, |
cd125eea MWO |
1022 | * ie with increased folio reference count so that the folio won't |
1023 | * go away during the wait. | |
1da177e4 | 1024 | */ |
6baa8d60 MWO |
1025 | static inline void folio_wait_locked(struct folio *folio) |
1026 | { | |
1027 | if (folio_test_locked(folio)) | |
101c0bf6 | 1028 | folio_wait_bit(folio, PG_locked); |
6baa8d60 MWO |
1029 | } |
1030 | ||
1031 | static inline int folio_wait_locked_killable(struct folio *folio) | |
1032 | { | |
1033 | if (!folio_test_locked(folio)) | |
1034 | return 0; | |
101c0bf6 | 1035 | return folio_wait_bit_killable(folio, PG_locked); |
6baa8d60 MWO |
1036 | } |
1037 | ||
1da177e4 LT |
1038 | static inline void wait_on_page_locked(struct page *page) |
1039 | { | |
6baa8d60 | 1040 | folio_wait_locked(page_folio(page)); |
1da177e4 LT |
1041 | } |
1042 | ||
62906027 NP |
1043 | static inline int wait_on_page_locked_killable(struct page *page) |
1044 | { | |
6baa8d60 | 1045 | return folio_wait_locked_killable(page_folio(page)); |
62906027 NP |
1046 | } |
1047 | ||
19343b5b | 1048 | void wait_on_page_writeback(struct page *page); |
490e016f MWO |
1049 | void folio_wait_writeback(struct folio *folio); |
1050 | int folio_wait_writeback_killable(struct folio *folio); | |
4268b480 MWO |
1051 | void end_page_writeback(struct page *page); |
1052 | void folio_end_writeback(struct folio *folio); | |
1d1d1a76 | 1053 | void wait_for_stable_page(struct page *page); |
a49d0c50 | 1054 | void folio_wait_stable(struct folio *folio); |
203a3151 MWO |
1055 | void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn); |
1056 | static inline void __set_page_dirty(struct page *page, | |
1057 | struct address_space *mapping, int warn) | |
1058 | { | |
1059 | __folio_mark_dirty(page_folio(page), mapping, warn); | |
1060 | } | |
566d3362 | 1061 | void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb); |
fdaf532a MWO |
1062 | void __folio_cancel_dirty(struct folio *folio); |
1063 | static inline void folio_cancel_dirty(struct folio *folio) | |
1064 | { | |
1065 | /* Avoid atomic ops, locking, etc. when not actually needed. */ | |
1066 | if (folio_test_dirty(folio)) | |
1067 | __folio_cancel_dirty(folio); | |
1068 | } | |
9350f20a MWO |
1069 | bool folio_clear_dirty_for_io(struct folio *folio); |
1070 | bool clear_page_dirty_for_io(struct page *page); | |
5ad6b2bd | 1071 | void folio_invalidate(struct folio *folio, size_t offset, size_t length); |
3a6b2162 | 1072 | int __set_page_dirty_nobuffers(struct page *page); |
46de8b97 | 1073 | bool noop_dirty_folio(struct address_space *mapping, struct folio *folio); |
3a6b2162 | 1074 | |
2ec810d5 MWO |
1075 | #ifdef CONFIG_MIGRATION |
1076 | int filemap_migrate_folio(struct address_space *mapping, struct folio *dst, | |
1077 | struct folio *src, enum migrate_mode mode); | |
1078 | #else | |
1079 | #define filemap_migrate_folio NULL | |
1080 | #endif | |
b47393f8 MWO |
1081 | void folio_end_private_2(struct folio *folio); |
1082 | void folio_wait_private_2(struct folio *folio); | |
1083 | int folio_wait_private_2_killable(struct folio *folio); | |
73e10ded | 1084 | |
385e1ca5 DH |
1085 | /* |
1086 | * Add an arbitrary waiter to a page's wait queue | |
1087 | */ | |
df4d4f12 | 1088 | void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter); |
385e1ca5 | 1089 | |
1da177e4 | 1090 | /* |
bb523b40 | 1091 | * Fault in userspace address range. |
1da177e4 | 1092 | */ |
bb523b40 | 1093 | size_t fault_in_writeable(char __user *uaddr, size_t size); |
da32b581 | 1094 | size_t fault_in_subpage_writeable(char __user *uaddr, size_t size); |
cdd591fc | 1095 | size_t fault_in_safe_writeable(const char __user *uaddr, size_t size); |
bb523b40 | 1096 | size_t fault_in_readable(const char __user *uaddr, size_t size); |
f56f821f | 1097 | |
529ae9aa | 1098 | int add_to_page_cache_lru(struct page *page, struct address_space *mapping, |
9dd3d069 MWO |
1099 | pgoff_t index, gfp_t gfp); |
1100 | int filemap_add_folio(struct address_space *mapping, struct folio *folio, | |
1101 | pgoff_t index, gfp_t gfp); | |
452e9e69 | 1102 | void filemap_remove_folio(struct folio *folio); |
452e9e69 | 1103 | void __filemap_remove_folio(struct folio *folio, void *shadow); |
3720dd6d | 1104 | void replace_page_cache_folio(struct folio *old, struct folio *new); |
aa65c29c | 1105 | void delete_from_page_cache_batch(struct address_space *mapping, |
51dcbdac | 1106 | struct folio_batch *fbatch); |
82c50f8b | 1107 | bool filemap_release_folio(struct folio *folio, gfp_t gfp); |
41139aa4 MWO |
1108 | loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end, |
1109 | int whence); | |
529ae9aa | 1110 | |
9dd3d069 MWO |
1111 | /* Must be non-static for BPF error injection */ |
1112 | int __filemap_add_folio(struct address_space *mapping, struct folio *folio, | |
1113 | pgoff_t index, gfp_t gfp, void **shadowp); | |
1114 | ||
4bdcd1dd JA |
1115 | bool filemap_range_has_writeback(struct address_space *mapping, |
1116 | loff_t start_byte, loff_t end_byte); | |
1117 | ||
1118 | /** | |
1119 | * filemap_range_needs_writeback - check if range potentially needs writeback | |
1120 | * @mapping: address space within which to check | |
1121 | * @start_byte: offset in bytes where the range starts | |
1122 | * @end_byte: offset in bytes where the range ends (inclusive) | |
1123 | * | |
1124 | * Find at least one page in the range supplied, usually used to check if | |
1125 | * direct writing in this range will trigger a writeback. Used by O_DIRECT | |
1126 | * read/write with IOCB_NOWAIT, to see if the caller needs to do | |
1127 | * filemap_write_and_wait_range() before proceeding. | |
1128 | * | |
1129 | * Return: %true if the caller should do filemap_write_and_wait_range() before | |
1130 | * doing O_DIRECT to a page in this range, %false otherwise. | |
1131 | */ | |
1132 | static inline bool filemap_range_needs_writeback(struct address_space *mapping, | |
1133 | loff_t start_byte, | |
1134 | loff_t end_byte) | |
1135 | { | |
1136 | if (!mapping->nrpages) | |
1137 | return false; | |
1138 | if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) && | |
1139 | !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK)) | |
1140 | return false; | |
1141 | return filemap_range_has_writeback(mapping, start_byte, end_byte); | |
1142 | } | |
1143 | ||
042124cc MWO |
1144 | /** |
1145 | * struct readahead_control - Describes a readahead request. | |
1146 | * | |
1147 | * A readahead request is for consecutive pages. Filesystems which | |
1148 | * implement the ->readahead method should call readahead_page() or | |
1149 | * readahead_page_batch() in a loop and attempt to start I/O against | |
1150 | * each page in the request. | |
1151 | * | |
1152 | * Most of the fields in this struct are private and should be accessed | |
1153 | * by the functions below. | |
1154 | * | |
1155 | * @file: The file, used primarily by network filesystems for authentication. | |
1156 | * May be NULL if invoked internally by the filesystem. | |
1157 | * @mapping: Readahead this filesystem object. | |
fcd9ae4f | 1158 | * @ra: File readahead state. May be NULL. |
042124cc MWO |
1159 | */ |
1160 | struct readahead_control { | |
1161 | struct file *file; | |
1162 | struct address_space *mapping; | |
fcd9ae4f | 1163 | struct file_ra_state *ra; |
042124cc MWO |
1164 | /* private: use the readahead_* accessors instead */ |
1165 | pgoff_t _index; | |
1166 | unsigned int _nr_pages; | |
1167 | unsigned int _batch_count; | |
17604240 CH |
1168 | bool _workingset; |
1169 | unsigned long _pflags; | |
042124cc MWO |
1170 | }; |
1171 | ||
fcd9ae4f MWO |
1172 | #define DEFINE_READAHEAD(ractl, f, r, m, i) \ |
1173 | struct readahead_control ractl = { \ | |
1aa83cfa MWO |
1174 | .file = f, \ |
1175 | .mapping = m, \ | |
fcd9ae4f | 1176 | .ra = r, \ |
1aa83cfa MWO |
1177 | ._index = i, \ |
1178 | } | |
1179 | ||
fefa7c47 MWO |
1180 | #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE) |
1181 | ||
1182 | void page_cache_ra_unbounded(struct readahead_control *, | |
1183 | unsigned long nr_to_read, unsigned long lookahead_count); | |
fcd9ae4f | 1184 | void page_cache_sync_ra(struct readahead_control *, unsigned long req_count); |
7836d999 | 1185 | void page_cache_async_ra(struct readahead_control *, struct folio *, |
fefa7c47 | 1186 | unsigned long req_count); |
3ca23644 DH |
1187 | void readahead_expand(struct readahead_control *ractl, |
1188 | loff_t new_start, size_t new_len); | |
fefa7c47 MWO |
1189 | |
1190 | /** | |
1191 | * page_cache_sync_readahead - generic file readahead | |
1192 | * @mapping: address_space which holds the pagecache and I/O vectors | |
1193 | * @ra: file_ra_state which holds the readahead state | |
1194 | * @file: Used by the filesystem for authentication. | |
1195 | * @index: Index of first page to be read. | |
1196 | * @req_count: Total number of pages being read by the caller. | |
1197 | * | |
1198 | * page_cache_sync_readahead() should be called when a cache miss happened: | |
1199 | * it will submit the read. The readahead logic may decide to piggyback more | |
1200 | * pages onto the read request if access patterns suggest it will improve | |
1201 | * performance. | |
1202 | */ | |
1203 | static inline | |
1204 | void page_cache_sync_readahead(struct address_space *mapping, | |
1205 | struct file_ra_state *ra, struct file *file, pgoff_t index, | |
1206 | unsigned long req_count) | |
1207 | { | |
fcd9ae4f MWO |
1208 | DEFINE_READAHEAD(ractl, file, ra, mapping, index); |
1209 | page_cache_sync_ra(&ractl, req_count); | |
fefa7c47 MWO |
1210 | } |
1211 | ||
1212 | /** | |
1213 | * page_cache_async_readahead - file readahead for marked pages | |
1214 | * @mapping: address_space which holds the pagecache and I/O vectors | |
1215 | * @ra: file_ra_state which holds the readahead state | |
1216 | * @file: Used by the filesystem for authentication. | |
2ebdd1df | 1217 | * @folio: The folio at @index which triggered the readahead call. |
fefa7c47 MWO |
1218 | * @index: Index of first page to be read. |
1219 | * @req_count: Total number of pages being read by the caller. | |
1220 | * | |
1221 | * page_cache_async_readahead() should be called when a page is used which | |
1222 | * is marked as PageReadahead; this is a marker to suggest that the application | |
1223 | * has used up enough of the readahead window that we should start pulling in | |
1224 | * more pages. | |
1225 | */ | |
1226 | static inline | |
1227 | void page_cache_async_readahead(struct address_space *mapping, | |
1228 | struct file_ra_state *ra, struct file *file, | |
2ebdd1df | 1229 | struct folio *folio, pgoff_t index, unsigned long req_count) |
fefa7c47 | 1230 | { |
fcd9ae4f | 1231 | DEFINE_READAHEAD(ractl, file, ra, mapping, index); |
2ebdd1df | 1232 | page_cache_async_ra(&ractl, folio, req_count); |
fefa7c47 MWO |
1233 | } |
1234 | ||
9bf70167 MWO |
1235 | static inline struct folio *__readahead_folio(struct readahead_control *ractl) |
1236 | { | |
1237 | struct folio *folio; | |
1238 | ||
1239 | BUG_ON(ractl->_batch_count > ractl->_nr_pages); | |
1240 | ractl->_nr_pages -= ractl->_batch_count; | |
1241 | ractl->_index += ractl->_batch_count; | |
1242 | ||
1243 | if (!ractl->_nr_pages) { | |
1244 | ractl->_batch_count = 0; | |
1245 | return NULL; | |
1246 | } | |
1247 | ||
1248 | folio = xa_load(&ractl->mapping->i_pages, ractl->_index); | |
1249 | VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio); | |
1250 | ractl->_batch_count = folio_nr_pages(folio); | |
1251 | ||
1252 | return folio; | |
1253 | } | |
1254 | ||
042124cc MWO |
1255 | /** |
1256 | * readahead_page - Get the next page to read. | |
9bf70167 | 1257 | * @ractl: The current readahead request. |
042124cc MWO |
1258 | * |
1259 | * Context: The page is locked and has an elevated refcount. The caller | |
1260 | * should decreases the refcount once the page has been submitted for I/O | |
1261 | * and unlock the page once all I/O to that page has completed. | |
1262 | * Return: A pointer to the next page, or %NULL if we are done. | |
1263 | */ | |
9bf70167 | 1264 | static inline struct page *readahead_page(struct readahead_control *ractl) |
042124cc | 1265 | { |
9bf70167 | 1266 | struct folio *folio = __readahead_folio(ractl); |
042124cc | 1267 | |
9bf70167 MWO |
1268 | return &folio->page; |
1269 | } | |
042124cc | 1270 | |
9bf70167 MWO |
1271 | /** |
1272 | * readahead_folio - Get the next folio to read. | |
1273 | * @ractl: The current readahead request. | |
1274 | * | |
1275 | * Context: The folio is locked. The caller should unlock the folio once | |
1276 | * all I/O to that folio has completed. | |
1277 | * Return: A pointer to the next folio, or %NULL if we are done. | |
1278 | */ | |
1279 | static inline struct folio *readahead_folio(struct readahead_control *ractl) | |
1280 | { | |
1281 | struct folio *folio = __readahead_folio(ractl); | |
042124cc | 1282 | |
9bf70167 MWO |
1283 | if (folio) |
1284 | folio_put(folio); | |
1285 | return folio; | |
042124cc MWO |
1286 | } |
1287 | ||
1288 | static inline unsigned int __readahead_batch(struct readahead_control *rac, | |
1289 | struct page **array, unsigned int array_sz) | |
1290 | { | |
1291 | unsigned int i = 0; | |
1292 | XA_STATE(xas, &rac->mapping->i_pages, 0); | |
1293 | struct page *page; | |
1294 | ||
1295 | BUG_ON(rac->_batch_count > rac->_nr_pages); | |
1296 | rac->_nr_pages -= rac->_batch_count; | |
1297 | rac->_index += rac->_batch_count; | |
1298 | rac->_batch_count = 0; | |
1299 | ||
1300 | xas_set(&xas, rac->_index); | |
1301 | rcu_read_lock(); | |
1302 | xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) { | |
4349a83a MWO |
1303 | if (xas_retry(&xas, page)) |
1304 | continue; | |
042124cc MWO |
1305 | VM_BUG_ON_PAGE(!PageLocked(page), page); |
1306 | VM_BUG_ON_PAGE(PageTail(page), page); | |
1307 | array[i++] = page; | |
6c357848 | 1308 | rac->_batch_count += thp_nr_pages(page); |
042124cc MWO |
1309 | if (i == array_sz) |
1310 | break; | |
1311 | } | |
1312 | rcu_read_unlock(); | |
1313 | ||
1314 | return i; | |
1315 | } | |
1316 | ||
1317 | /** | |
1318 | * readahead_page_batch - Get a batch of pages to read. | |
1319 | * @rac: The current readahead request. | |
1320 | * @array: An array of pointers to struct page. | |
1321 | * | |
1322 | * Context: The pages are locked and have an elevated refcount. The caller | |
1323 | * should decreases the refcount once the page has been submitted for I/O | |
1324 | * and unlock the page once all I/O to that page has completed. | |
1325 | * Return: The number of pages placed in the array. 0 indicates the request | |
1326 | * is complete. | |
1327 | */ | |
1328 | #define readahead_page_batch(rac, array) \ | |
1329 | __readahead_batch(rac, array, ARRAY_SIZE(array)) | |
1330 | ||
1331 | /** | |
1332 | * readahead_pos - The byte offset into the file of this readahead request. | |
1333 | * @rac: The readahead request. | |
1334 | */ | |
1335 | static inline loff_t readahead_pos(struct readahead_control *rac) | |
1336 | { | |
1337 | return (loff_t)rac->_index * PAGE_SIZE; | |
1338 | } | |
1339 | ||
1340 | /** | |
1341 | * readahead_length - The number of bytes in this readahead request. | |
1342 | * @rac: The readahead request. | |
1343 | */ | |
076171a6 | 1344 | static inline size_t readahead_length(struct readahead_control *rac) |
042124cc | 1345 | { |
076171a6 | 1346 | return rac->_nr_pages * PAGE_SIZE; |
042124cc MWO |
1347 | } |
1348 | ||
1349 | /** | |
1350 | * readahead_index - The index of the first page in this readahead request. | |
1351 | * @rac: The readahead request. | |
1352 | */ | |
1353 | static inline pgoff_t readahead_index(struct readahead_control *rac) | |
1354 | { | |
1355 | return rac->_index; | |
1356 | } | |
1357 | ||
1358 | /** | |
1359 | * readahead_count - The number of pages in this readahead request. | |
1360 | * @rac: The readahead request. | |
1361 | */ | |
1362 | static inline unsigned int readahead_count(struct readahead_control *rac) | |
1363 | { | |
1364 | return rac->_nr_pages; | |
1365 | } | |
1366 | ||
32c0a6bc MWO |
1367 | /** |
1368 | * readahead_batch_length - The number of bytes in the current batch. | |
1369 | * @rac: The readahead request. | |
1370 | */ | |
076171a6 | 1371 | static inline size_t readahead_batch_length(struct readahead_control *rac) |
32c0a6bc MWO |
1372 | { |
1373 | return rac->_batch_count * PAGE_SIZE; | |
1374 | } | |
1375 | ||
b57c2cb9 FF |
1376 | static inline unsigned long dir_pages(struct inode *inode) |
1377 | { | |
09cbfeaf KS |
1378 | return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >> |
1379 | PAGE_SHIFT; | |
b57c2cb9 FF |
1380 | } |
1381 | ||
f705bf84 MWO |
1382 | /** |
1383 | * folio_mkwrite_check_truncate - check if folio was truncated | |
1384 | * @folio: the folio to check | |
1385 | * @inode: the inode to check the folio against | |
1386 | * | |
1387 | * Return: the number of bytes in the folio up to EOF, | |
1388 | * or -EFAULT if the folio was truncated. | |
1389 | */ | |
1390 | static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio, | |
1391 | struct inode *inode) | |
1392 | { | |
1393 | loff_t size = i_size_read(inode); | |
1394 | pgoff_t index = size >> PAGE_SHIFT; | |
1395 | size_t offset = offset_in_folio(folio, size); | |
1396 | ||
1397 | if (!folio->mapping) | |
1398 | return -EFAULT; | |
1399 | ||
1400 | /* folio is wholly inside EOF */ | |
1401 | if (folio_next_index(folio) - 1 < index) | |
1402 | return folio_size(folio); | |
1403 | /* folio is wholly past EOF */ | |
1404 | if (folio->index > index || !offset) | |
1405 | return -EFAULT; | |
1406 | /* folio is partially inside EOF */ | |
1407 | return offset; | |
1408 | } | |
1409 | ||
243145bc AG |
1410 | /** |
1411 | * page_mkwrite_check_truncate - check if page was truncated | |
1412 | * @page: the page to check | |
1413 | * @inode: the inode to check the page against | |
1414 | * | |
1415 | * Returns the number of bytes in the page up to EOF, | |
1416 | * or -EFAULT if the page was truncated. | |
1417 | */ | |
1418 | static inline int page_mkwrite_check_truncate(struct page *page, | |
1419 | struct inode *inode) | |
1420 | { | |
1421 | loff_t size = i_size_read(inode); | |
1422 | pgoff_t index = size >> PAGE_SHIFT; | |
1423 | int offset = offset_in_page(size); | |
1424 | ||
1425 | if (page->mapping != inode->i_mapping) | |
1426 | return -EFAULT; | |
1427 | ||
1428 | /* page is wholly inside EOF */ | |
1429 | if (page->index < index) | |
1430 | return PAGE_SIZE; | |
1431 | /* page is wholly past EOF */ | |
1432 | if (page->index > index || !offset) | |
1433 | return -EFAULT; | |
1434 | /* page is partially inside EOF */ | |
1435 | return offset; | |
1436 | } | |
1437 | ||
24addd84 | 1438 | /** |
9eb7c76d | 1439 | * i_blocks_per_folio - How many blocks fit in this folio. |
24addd84 | 1440 | * @inode: The inode which contains the blocks. |
9eb7c76d | 1441 | * @folio: The folio. |
24addd84 | 1442 | * |
9eb7c76d | 1443 | * If the block size is larger than the size of this folio, return zero. |
24addd84 | 1444 | * |
9eb7c76d | 1445 | * Context: The caller should hold a refcount on the folio to prevent it |
24addd84 | 1446 | * from being split. |
9eb7c76d | 1447 | * Return: The number of filesystem blocks covered by this folio. |
24addd84 | 1448 | */ |
9eb7c76d MWO |
1449 | static inline |
1450 | unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio) | |
1451 | { | |
1452 | return folio_size(folio) >> inode->i_blkbits; | |
1453 | } | |
1454 | ||
24addd84 MWO |
1455 | static inline |
1456 | unsigned int i_blocks_per_page(struct inode *inode, struct page *page) | |
1457 | { | |
9eb7c76d | 1458 | return i_blocks_per_folio(inode, page_folio(page)); |
24addd84 | 1459 | } |
1da177e4 | 1460 | #endif /* _LINUX_PAGEMAP_H */ |