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 | |
7716506a MWO |
21 | static inline bool mapping_empty(struct address_space *mapping) |
22 | { | |
23 | return xa_empty(&mapping->i_pages); | |
24 | } | |
25 | ||
51b8c1fe JW |
26 | /* |
27 | * mapping_shrinkable - test if page cache state allows inode reclaim | |
28 | * @mapping: the page cache mapping | |
29 | * | |
30 | * This checks the mapping's cache state for the pupose of inode | |
31 | * reclaim and LRU management. | |
32 | * | |
33 | * The caller is expected to hold the i_lock, but is not required to | |
34 | * hold the i_pages lock, which usually protects cache state. That's | |
35 | * because the i_lock and the list_lru lock that protect the inode and | |
36 | * its LRU state don't nest inside the irq-safe i_pages lock. | |
37 | * | |
38 | * Cache deletions are performed under the i_lock, which ensures that | |
39 | * when an inode goes empty, it will reliably get queued on the LRU. | |
40 | * | |
41 | * Cache additions do not acquire the i_lock and may race with this | |
42 | * check, in which case we'll report the inode as shrinkable when it | |
43 | * has cache pages. This is okay: the shrinker also checks the | |
44 | * refcount and the referenced bit, which will be elevated or set in | |
45 | * the process of adding new cache pages to an inode. | |
46 | */ | |
47 | static inline bool mapping_shrinkable(struct address_space *mapping) | |
48 | { | |
49 | void *head; | |
50 | ||
51 | /* | |
52 | * On highmem systems, there could be lowmem pressure from the | |
53 | * inodes before there is highmem pressure from the page | |
54 | * cache. Make inodes shrinkable regardless of cache state. | |
55 | */ | |
56 | if (IS_ENABLED(CONFIG_HIGHMEM)) | |
57 | return true; | |
58 | ||
59 | /* Cache completely empty? Shrink away. */ | |
60 | head = rcu_access_pointer(mapping->i_pages.xa_head); | |
61 | if (!head) | |
62 | return true; | |
63 | ||
64 | /* | |
65 | * The xarray stores single offset-0 entries directly in the | |
66 | * head pointer, which allows non-resident page cache entries | |
67 | * to escape the shadow shrinker's list of xarray nodes. The | |
68 | * inode shrinker needs to pick them up under memory pressure. | |
69 | */ | |
70 | if (!xa_is_node(head) && xa_is_value(head)) | |
71 | return true; | |
72 | ||
73 | return false; | |
74 | } | |
75 | ||
1da177e4 | 76 | /* |
9c5d760b | 77 | * Bits in mapping->flags. |
1da177e4 | 78 | */ |
9a896c9a | 79 | enum mapping_flags { |
9c5d760b MH |
80 | AS_EIO = 0, /* IO error on async write */ |
81 | AS_ENOSPC = 1, /* ENOSPC on async write */ | |
82 | AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */ | |
83 | AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */ | |
84 | AS_EXITING = 4, /* final truncate in progress */ | |
371a096e | 85 | /* writeback related tags are not used */ |
9c5d760b | 86 | AS_NO_WRITEBACK_TAGS = 5, |
ed2145c4 | 87 | AS_LARGE_FOLIO_SUPPORT = 6, |
9a896c9a | 88 | }; |
1da177e4 | 89 | |
8ed1e46a JL |
90 | /** |
91 | * mapping_set_error - record a writeback error in the address_space | |
767e5ee5 MWO |
92 | * @mapping: the mapping in which an error should be set |
93 | * @error: the error to set in the mapping | |
8ed1e46a JL |
94 | * |
95 | * When writeback fails in some way, we must record that error so that | |
96 | * userspace can be informed when fsync and the like are called. We endeavor | |
97 | * to report errors on any file that was open at the time of the error. Some | |
98 | * internal callers also need to know when writeback errors have occurred. | |
99 | * | |
100 | * When a writeback error occurs, most filesystems will want to call | |
101 | * mapping_set_error to record the error in the mapping so that it can be | |
102 | * reported when the application calls fsync(2). | |
103 | */ | |
3e9f45bd GC |
104 | static inline void mapping_set_error(struct address_space *mapping, int error) |
105 | { | |
8ed1e46a JL |
106 | if (likely(!error)) |
107 | return; | |
108 | ||
109 | /* Record in wb_err for checkers using errseq_t based tracking */ | |
735e4ae5 JL |
110 | __filemap_set_wb_err(mapping, error); |
111 | ||
112 | /* Record it in superblock */ | |
8b7b2eb1 MK |
113 | if (mapping->host) |
114 | errseq_set(&mapping->host->i_sb->s_wb_err, error); | |
8ed1e46a JL |
115 | |
116 | /* Record it in flags for now, for legacy callers */ | |
117 | if (error == -ENOSPC) | |
118 | set_bit(AS_ENOSPC, &mapping->flags); | |
119 | else | |
120 | set_bit(AS_EIO, &mapping->flags); | |
3e9f45bd GC |
121 | } |
122 | ||
ba9ddf49 LS |
123 | static inline void mapping_set_unevictable(struct address_space *mapping) |
124 | { | |
125 | set_bit(AS_UNEVICTABLE, &mapping->flags); | |
126 | } | |
127 | ||
89e004ea LS |
128 | static inline void mapping_clear_unevictable(struct address_space *mapping) |
129 | { | |
130 | clear_bit(AS_UNEVICTABLE, &mapping->flags); | |
131 | } | |
132 | ||
1eb6234e | 133 | static inline bool mapping_unevictable(struct address_space *mapping) |
ba9ddf49 | 134 | { |
1eb6234e | 135 | return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags); |
ba9ddf49 | 136 | } |
ba9ddf49 | 137 | |
91b0abe3 JW |
138 | static inline void mapping_set_exiting(struct address_space *mapping) |
139 | { | |
140 | set_bit(AS_EXITING, &mapping->flags); | |
141 | } | |
142 | ||
143 | static inline int mapping_exiting(struct address_space *mapping) | |
144 | { | |
145 | return test_bit(AS_EXITING, &mapping->flags); | |
146 | } | |
147 | ||
371a096e HY |
148 | static inline void mapping_set_no_writeback_tags(struct address_space *mapping) |
149 | { | |
150 | set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags); | |
151 | } | |
152 | ||
153 | static inline int mapping_use_writeback_tags(struct address_space *mapping) | |
154 | { | |
155 | return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags); | |
156 | } | |
157 | ||
dd0fc66f | 158 | static inline gfp_t mapping_gfp_mask(struct address_space * mapping) |
1da177e4 | 159 | { |
9c5d760b | 160 | return mapping->gfp_mask; |
1da177e4 LT |
161 | } |
162 | ||
c62d2555 MH |
163 | /* Restricts the given gfp_mask to what the mapping allows. */ |
164 | static inline gfp_t mapping_gfp_constraint(struct address_space *mapping, | |
165 | gfp_t gfp_mask) | |
166 | { | |
167 | return mapping_gfp_mask(mapping) & gfp_mask; | |
168 | } | |
169 | ||
1da177e4 LT |
170 | /* |
171 | * This is non-atomic. Only to be used before the mapping is activated. | |
172 | * Probably needs a barrier... | |
173 | */ | |
260b2367 | 174 | static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask) |
1da177e4 | 175 | { |
9c5d760b | 176 | m->gfp_mask = mask; |
1da177e4 LT |
177 | } |
178 | ||
ff36da69 MWO |
179 | /** |
180 | * mapping_set_large_folios() - Indicate the file supports large folios. | |
181 | * @mapping: The file. | |
182 | * | |
183 | * The filesystem should call this function in its inode constructor to | |
184 | * indicate that the VFS can use large folios to cache the contents of | |
185 | * the file. | |
186 | * | |
187 | * Context: This should not be called while the inode is active as it | |
188 | * is non-atomic. | |
189 | */ | |
190 | static inline void mapping_set_large_folios(struct address_space *mapping) | |
191 | { | |
ed2145c4 | 192 | __set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags); |
ff36da69 MWO |
193 | } |
194 | ||
ed2145c4 | 195 | static inline bool mapping_large_folio_support(struct address_space *mapping) |
01c70267 | 196 | { |
ed2145c4 | 197 | return test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags); |
01c70267 MWO |
198 | } |
199 | ||
6f4d2f97 MWO |
200 | static inline int filemap_nr_thps(struct address_space *mapping) |
201 | { | |
202 | #ifdef CONFIG_READ_ONLY_THP_FOR_FS | |
203 | return atomic_read(&mapping->nr_thps); | |
204 | #else | |
205 | return 0; | |
206 | #endif | |
207 | } | |
208 | ||
209 | static inline void filemap_nr_thps_inc(struct address_space *mapping) | |
210 | { | |
211 | #ifdef CONFIG_READ_ONLY_THP_FOR_FS | |
ed2145c4 | 212 | if (!mapping_large_folio_support(mapping)) |
6f4d2f97 MWO |
213 | atomic_inc(&mapping->nr_thps); |
214 | #else | |
215 | WARN_ON_ONCE(1); | |
216 | #endif | |
217 | } | |
218 | ||
219 | static inline void filemap_nr_thps_dec(struct address_space *mapping) | |
220 | { | |
221 | #ifdef CONFIG_READ_ONLY_THP_FOR_FS | |
ed2145c4 | 222 | if (!mapping_large_folio_support(mapping)) |
6f4d2f97 MWO |
223 | atomic_dec(&mapping->nr_thps); |
224 | #else | |
225 | WARN_ON_ONCE(1); | |
226 | #endif | |
227 | } | |
228 | ||
c6f92f9f | 229 | void release_pages(struct page **pages, int nr); |
1da177e4 | 230 | |
2f52578f MWO |
231 | struct address_space *page_mapping(struct page *); |
232 | struct address_space *folio_mapping(struct folio *); | |
233 | struct address_space *swapcache_mapping(struct folio *); | |
234 | ||
235 | /** | |
236 | * folio_file_mapping - Find the mapping this folio belongs to. | |
237 | * @folio: The folio. | |
238 | * | |
239 | * For folios which are in the page cache, return the mapping that this | |
240 | * page belongs to. Folios in the swap cache return the mapping of the | |
241 | * swap file or swap device where the data is stored. This is different | |
242 | * from the mapping returned by folio_mapping(). The only reason to | |
243 | * use it is if, like NFS, you return 0 from ->activate_swapfile. | |
244 | * | |
245 | * Do not call this for folios which aren't in the page cache or swap cache. | |
246 | */ | |
247 | static inline struct address_space *folio_file_mapping(struct folio *folio) | |
248 | { | |
249 | if (unlikely(folio_test_swapcache(folio))) | |
250 | return swapcache_mapping(folio); | |
251 | ||
252 | return folio->mapping; | |
253 | } | |
254 | ||
255 | static inline struct address_space *page_file_mapping(struct page *page) | |
256 | { | |
257 | return folio_file_mapping(page_folio(page)); | |
258 | } | |
259 | ||
842ca547 MWO |
260 | /* |
261 | * For file cache pages, return the address_space, otherwise return NULL | |
262 | */ | |
263 | static inline struct address_space *page_mapping_file(struct page *page) | |
264 | { | |
2f52578f MWO |
265 | struct folio *folio = page_folio(page); |
266 | ||
267 | if (unlikely(folio_test_swapcache(folio))) | |
842ca547 | 268 | return NULL; |
2f52578f | 269 | return folio_mapping(folio); |
842ca547 MWO |
270 | } |
271 | ||
452c472e DH |
272 | /** |
273 | * folio_inode - Get the host inode for this folio. | |
274 | * @folio: The folio. | |
275 | * | |
276 | * For folios which are in the page cache, return the inode that this folio | |
277 | * belongs to. | |
278 | * | |
279 | * Do not call this for folios which aren't in the page cache. | |
280 | */ | |
281 | static inline struct inode *folio_inode(struct folio *folio) | |
282 | { | |
283 | return folio->mapping->host; | |
284 | } | |
285 | ||
020853b6 | 286 | static inline bool page_cache_add_speculative(struct page *page, int count) |
e286781d | 287 | { |
020853b6 | 288 | return folio_ref_try_add_rcu((struct folio *)page, count); |
494eec70 | 289 | } |
ce0ad7f0 | 290 | |
020853b6 | 291 | static inline bool page_cache_get_speculative(struct page *page) |
494eec70 | 292 | { |
020853b6 | 293 | return page_cache_add_speculative(page, 1); |
ce0ad7f0 NP |
294 | } |
295 | ||
b03143ac | 296 | /** |
85d0a2ed MWO |
297 | * folio_attach_private - Attach private data to a folio. |
298 | * @folio: Folio to attach data to. | |
299 | * @data: Data to attach to folio. | |
b03143ac | 300 | * |
85d0a2ed MWO |
301 | * Attaching private data to a folio increments the page's reference count. |
302 | * The data must be detached before the folio will be freed. | |
b03143ac | 303 | */ |
85d0a2ed | 304 | static inline void folio_attach_private(struct folio *folio, void *data) |
b03143ac | 305 | { |
85d0a2ed MWO |
306 | folio_get(folio); |
307 | folio->private = data; | |
308 | folio_set_private(folio); | |
b03143ac GJ |
309 | } |
310 | ||
a19672f6 DH |
311 | /** |
312 | * folio_change_private - Change private data on a folio. | |
313 | * @folio: Folio to change the data on. | |
314 | * @data: Data to set on the folio. | |
315 | * | |
316 | * Change the private data attached to a folio and return the old | |
317 | * data. The page must previously have had data attached and the data | |
318 | * must be detached before the folio will be freed. | |
319 | * | |
320 | * Return: Data that was previously attached to the folio. | |
321 | */ | |
322 | static inline void *folio_change_private(struct folio *folio, void *data) | |
323 | { | |
324 | void *old = folio_get_private(folio); | |
325 | ||
326 | folio->private = data; | |
327 | return old; | |
328 | } | |
329 | ||
b03143ac | 330 | /** |
85d0a2ed MWO |
331 | * folio_detach_private - Detach private data from a folio. |
332 | * @folio: Folio to detach data from. | |
b03143ac | 333 | * |
85d0a2ed | 334 | * Removes the data that was previously attached to the folio and decrements |
b03143ac GJ |
335 | * the refcount on the page. |
336 | * | |
85d0a2ed | 337 | * Return: Data that was attached to the folio. |
b03143ac | 338 | */ |
85d0a2ed | 339 | static inline void *folio_detach_private(struct folio *folio) |
b03143ac | 340 | { |
85d0a2ed | 341 | void *data = folio_get_private(folio); |
b03143ac | 342 | |
85d0a2ed | 343 | if (!folio_test_private(folio)) |
b03143ac | 344 | return NULL; |
85d0a2ed MWO |
345 | folio_clear_private(folio); |
346 | folio->private = NULL; | |
347 | folio_put(folio); | |
b03143ac GJ |
348 | |
349 | return data; | |
350 | } | |
351 | ||
85d0a2ed MWO |
352 | static inline void attach_page_private(struct page *page, void *data) |
353 | { | |
354 | folio_attach_private(page_folio(page), data); | |
355 | } | |
356 | ||
357 | static inline void *detach_page_private(struct page *page) | |
358 | { | |
359 | return folio_detach_private(page_folio(page)); | |
360 | } | |
361 | ||
44110fe3 | 362 | #ifdef CONFIG_NUMA |
bb3c579e | 363 | struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order); |
44110fe3 | 364 | #else |
bb3c579e | 365 | static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order) |
2ae88149 | 366 | { |
bb3c579e | 367 | return folio_alloc(gfp, order); |
2ae88149 NP |
368 | } |
369 | #endif | |
370 | ||
bb3c579e MWO |
371 | static inline struct page *__page_cache_alloc(gfp_t gfp) |
372 | { | |
373 | return &filemap_alloc_folio(gfp, 0)->page; | |
374 | } | |
375 | ||
1da177e4 LT |
376 | static inline struct page *page_cache_alloc(struct address_space *x) |
377 | { | |
2ae88149 | 378 | return __page_cache_alloc(mapping_gfp_mask(x)); |
1da177e4 LT |
379 | } |
380 | ||
8a5c743e | 381 | static inline gfp_t readahead_gfp_mask(struct address_space *x) |
7b1de586 | 382 | { |
453f85d4 | 383 | return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN; |
7b1de586 WF |
384 | } |
385 | ||
1da177e4 LT |
386 | typedef int filler_t(void *, struct page *); |
387 | ||
0d3f9296 | 388 | pgoff_t page_cache_next_miss(struct address_space *mapping, |
e7b563bb | 389 | pgoff_t index, unsigned long max_scan); |
0d3f9296 | 390 | pgoff_t page_cache_prev_miss(struct address_space *mapping, |
e7b563bb JW |
391 | pgoff_t index, unsigned long max_scan); |
392 | ||
2457aec6 MG |
393 | #define FGP_ACCESSED 0x00000001 |
394 | #define FGP_LOCK 0x00000002 | |
395 | #define FGP_CREAT 0x00000004 | |
396 | #define FGP_WRITE 0x00000008 | |
397 | #define FGP_NOFS 0x00000010 | |
398 | #define FGP_NOWAIT 0x00000020 | |
a75d4c33 | 399 | #define FGP_FOR_MMAP 0x00000040 |
a8cf7f27 | 400 | #define FGP_HEAD 0x00000080 |
44835d20 | 401 | #define FGP_ENTRY 0x00000100 |
b27652d9 | 402 | #define FGP_STABLE 0x00000200 |
2457aec6 | 403 | |
3f0c6a07 MWO |
404 | struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index, |
405 | int fgp_flags, gfp_t gfp); | |
406 | struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index, | |
407 | int fgp_flags, gfp_t gfp); | |
408 | ||
409 | /** | |
410 | * filemap_get_folio - Find and get a folio. | |
411 | * @mapping: The address_space to search. | |
412 | * @index: The page index. | |
413 | * | |
414 | * Looks up the page cache entry at @mapping & @index. If a folio is | |
415 | * present, it is returned with an increased refcount. | |
416 | * | |
417 | * Otherwise, %NULL is returned. | |
418 | */ | |
419 | static inline struct folio *filemap_get_folio(struct address_space *mapping, | |
420 | pgoff_t index) | |
421 | { | |
422 | return __filemap_get_folio(mapping, index, 0, 0); | |
423 | } | |
2457aec6 MG |
424 | |
425 | /** | |
426 | * find_get_page - find and get a page reference | |
427 | * @mapping: the address_space to search | |
428 | * @offset: the page index | |
429 | * | |
430 | * Looks up the page cache slot at @mapping & @offset. If there is a | |
431 | * page cache page, it is returned with an increased refcount. | |
432 | * | |
433 | * Otherwise, %NULL is returned. | |
434 | */ | |
435 | static inline struct page *find_get_page(struct address_space *mapping, | |
436 | pgoff_t offset) | |
437 | { | |
45f87de5 | 438 | return pagecache_get_page(mapping, offset, 0, 0); |
2457aec6 MG |
439 | } |
440 | ||
441 | static inline struct page *find_get_page_flags(struct address_space *mapping, | |
442 | pgoff_t offset, int fgp_flags) | |
443 | { | |
45f87de5 | 444 | return pagecache_get_page(mapping, offset, fgp_flags, 0); |
2457aec6 MG |
445 | } |
446 | ||
447 | /** | |
448 | * find_lock_page - locate, pin and lock a pagecache page | |
2457aec6 | 449 | * @mapping: the address_space to search |
89b42235 | 450 | * @index: the page index |
2457aec6 | 451 | * |
89b42235 | 452 | * Looks up the page cache entry at @mapping & @index. If there is a |
2457aec6 MG |
453 | * page cache page, it is returned locked and with an increased |
454 | * refcount. | |
455 | * | |
a8cf7f27 MWO |
456 | * Context: May sleep. |
457 | * Return: A struct page or %NULL if there is no page in the cache for this | |
458 | * index. | |
2457aec6 MG |
459 | */ |
460 | static inline struct page *find_lock_page(struct address_space *mapping, | |
a8cf7f27 MWO |
461 | pgoff_t index) |
462 | { | |
463 | return pagecache_get_page(mapping, index, FGP_LOCK, 0); | |
464 | } | |
465 | ||
2457aec6 MG |
466 | /** |
467 | * find_or_create_page - locate or add a pagecache page | |
468 | * @mapping: the page's address_space | |
469 | * @index: the page's index into the mapping | |
470 | * @gfp_mask: page allocation mode | |
471 | * | |
472 | * Looks up the page cache slot at @mapping & @offset. If there is a | |
473 | * page cache page, it is returned locked and with an increased | |
474 | * refcount. | |
475 | * | |
476 | * If the page is not present, a new page is allocated using @gfp_mask | |
477 | * and added to the page cache and the VM's LRU list. The page is | |
478 | * returned locked and with an increased refcount. | |
479 | * | |
480 | * On memory exhaustion, %NULL is returned. | |
481 | * | |
482 | * find_or_create_page() may sleep, even if @gfp_flags specifies an | |
483 | * atomic allocation! | |
484 | */ | |
485 | static inline struct page *find_or_create_page(struct address_space *mapping, | |
767e5ee5 | 486 | pgoff_t index, gfp_t gfp_mask) |
2457aec6 | 487 | { |
767e5ee5 | 488 | return pagecache_get_page(mapping, index, |
2457aec6 | 489 | FGP_LOCK|FGP_ACCESSED|FGP_CREAT, |
45f87de5 | 490 | gfp_mask); |
2457aec6 MG |
491 | } |
492 | ||
493 | /** | |
494 | * grab_cache_page_nowait - returns locked page at given index in given cache | |
495 | * @mapping: target address_space | |
496 | * @index: the page index | |
497 | * | |
498 | * Same as grab_cache_page(), but do not wait if the page is unavailable. | |
499 | * This is intended for speculative data generators, where the data can | |
500 | * be regenerated if the page couldn't be grabbed. This routine should | |
501 | * be safe to call while holding the lock for another page. | |
502 | * | |
503 | * Clear __GFP_FS when allocating the page to avoid recursion into the fs | |
504 | * and deadlock against the caller's locked page. | |
505 | */ | |
506 | static inline struct page *grab_cache_page_nowait(struct address_space *mapping, | |
507 | pgoff_t index) | |
508 | { | |
509 | return pagecache_get_page(mapping, index, | |
510 | FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT, | |
45f87de5 | 511 | mapping_gfp_mask(mapping)); |
2457aec6 MG |
512 | } |
513 | ||
9257e156 MWO |
514 | #define swapcache_index(folio) __page_file_index(&(folio)->page) |
515 | ||
516 | /** | |
517 | * folio_index - File index of a folio. | |
518 | * @folio: The folio. | |
519 | * | |
520 | * For a folio which is either in the page cache or the swap cache, | |
521 | * return its index within the address_space it belongs to. If you know | |
522 | * the page is definitely in the page cache, you can look at the folio's | |
523 | * index directly. | |
524 | * | |
525 | * Return: The index (offset in units of pages) of a folio in its file. | |
526 | */ | |
527 | static inline pgoff_t folio_index(struct folio *folio) | |
528 | { | |
529 | if (unlikely(folio_test_swapcache(folio))) | |
530 | return swapcache_index(folio); | |
531 | return folio->index; | |
532 | } | |
533 | ||
f94b18f6 MWO |
534 | /** |
535 | * folio_next_index - Get the index of the next folio. | |
536 | * @folio: The current folio. | |
537 | * | |
538 | * Return: The index of the folio which follows this folio in the file. | |
539 | */ | |
540 | static inline pgoff_t folio_next_index(struct folio *folio) | |
541 | { | |
542 | return folio->index + folio_nr_pages(folio); | |
543 | } | |
544 | ||
9257e156 MWO |
545 | /** |
546 | * folio_file_page - The page for a particular index. | |
547 | * @folio: The folio which contains this index. | |
548 | * @index: The index we want to look up. | |
549 | * | |
550 | * Sometimes after looking up a folio in the page cache, we need to | |
551 | * obtain the specific page for an index (eg a page fault). | |
552 | * | |
553 | * Return: The page containing the file data for this index. | |
554 | */ | |
555 | static inline struct page *folio_file_page(struct folio *folio, pgoff_t index) | |
556 | { | |
557 | /* HugeTLBfs indexes the page cache in units of hpage_size */ | |
558 | if (folio_test_hugetlb(folio)) | |
559 | return &folio->page; | |
560 | return folio_page(folio, index & (folio_nr_pages(folio) - 1)); | |
561 | } | |
562 | ||
563 | /** | |
564 | * folio_contains - Does this folio contain this index? | |
565 | * @folio: The folio. | |
566 | * @index: The page index within the file. | |
567 | * | |
568 | * Context: The caller should have the page locked in order to prevent | |
569 | * (eg) shmem from moving the page between the page cache and swap cache | |
570 | * and changing its index in the middle of the operation. | |
571 | * Return: true or false. | |
572 | */ | |
573 | static inline bool folio_contains(struct folio *folio, pgoff_t index) | |
574 | { | |
575 | /* HugeTLBfs indexes the page cache in units of hpage_size */ | |
576 | if (folio_test_hugetlb(folio)) | |
577 | return folio->index == index; | |
578 | return index - folio_index(folio) < folio_nr_pages(folio); | |
579 | } | |
580 | ||
ec848215 MWO |
581 | /* |
582 | * Given the page we found in the page cache, return the page corresponding | |
583 | * to this index in the file | |
584 | */ | |
585 | static inline struct page *find_subpage(struct page *head, pgoff_t index) | |
4101196b | 586 | { |
ec848215 MWO |
587 | /* HugeTLBfs wants the head page regardless */ |
588 | if (PageHuge(head)) | |
589 | return head; | |
4101196b | 590 | |
6c357848 | 591 | return head + (index & (thp_nr_pages(head) - 1)); |
4101196b MWO |
592 | } |
593 | ||
b947cee4 JK |
594 | unsigned find_get_pages_range(struct address_space *mapping, pgoff_t *start, |
595 | pgoff_t end, unsigned int nr_pages, | |
596 | struct page **pages); | |
597 | static inline unsigned find_get_pages(struct address_space *mapping, | |
598 | pgoff_t *start, unsigned int nr_pages, | |
599 | struct page **pages) | |
600 | { | |
601 | return find_get_pages_range(mapping, start, (pgoff_t)-1, nr_pages, | |
602 | pages); | |
603 | } | |
ebf43500 JA |
604 | unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t start, |
605 | unsigned int nr_pages, struct page **pages); | |
72b045ae | 606 | unsigned find_get_pages_range_tag(struct address_space *mapping, pgoff_t *index, |
a6906972 | 607 | pgoff_t end, xa_mark_t tag, unsigned int nr_pages, |
72b045ae JK |
608 | struct page **pages); |
609 | static inline unsigned find_get_pages_tag(struct address_space *mapping, | |
a6906972 | 610 | pgoff_t *index, xa_mark_t tag, unsigned int nr_pages, |
72b045ae JK |
611 | struct page **pages) |
612 | { | |
613 | return find_get_pages_range_tag(mapping, index, (pgoff_t)-1, tag, | |
614 | nr_pages, pages); | |
615 | } | |
1da177e4 | 616 | |
54566b2c NP |
617 | struct page *grab_cache_page_write_begin(struct address_space *mapping, |
618 | pgoff_t index, unsigned flags); | |
afddba49 | 619 | |
1da177e4 LT |
620 | /* |
621 | * Returns locked page at given index in given cache, creating it if needed. | |
622 | */ | |
57f6b96c FW |
623 | static inline struct page *grab_cache_page(struct address_space *mapping, |
624 | pgoff_t index) | |
1da177e4 LT |
625 | { |
626 | return find_or_create_page(mapping, index, mapping_gfp_mask(mapping)); | |
627 | } | |
628 | ||
539a3322 MWO |
629 | struct folio *read_cache_folio(struct address_space *, pgoff_t index, |
630 | filler_t *filler, void *data); | |
631 | struct page *read_cache_page(struct address_space *, pgoff_t index, | |
632 | filler_t *filler, void *data); | |
0531b2aa LT |
633 | extern struct page * read_cache_page_gfp(struct address_space *mapping, |
634 | pgoff_t index, gfp_t gfp_mask); | |
1da177e4 LT |
635 | extern int read_cache_pages(struct address_space *mapping, |
636 | struct list_head *pages, filler_t *filler, void *data); | |
637 | ||
090d2b18 | 638 | static inline struct page *read_mapping_page(struct address_space *mapping, |
5e5358e7 | 639 | pgoff_t index, void *data) |
090d2b18 | 640 | { |
6c45b454 | 641 | return read_cache_page(mapping, index, NULL, data); |
090d2b18 PE |
642 | } |
643 | ||
539a3322 MWO |
644 | static inline struct folio *read_mapping_folio(struct address_space *mapping, |
645 | pgoff_t index, void *data) | |
646 | { | |
647 | return read_cache_folio(mapping, index, NULL, data); | |
648 | } | |
649 | ||
a0f7a756 | 650 | /* |
fe19bd3d | 651 | * Get index of the page within radix-tree (but not for hugetlb pages). |
5cbc198a | 652 | * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE) |
a0f7a756 | 653 | */ |
5cbc198a | 654 | static inline pgoff_t page_to_index(struct page *page) |
a0f7a756 | 655 | { |
fe3df441 | 656 | struct page *head; |
e9b61f19 | 657 | |
e9b61f19 | 658 | if (likely(!PageTransTail(page))) |
09cbfeaf | 659 | return page->index; |
e9b61f19 | 660 | |
fe3df441 | 661 | head = compound_head(page); |
e9b61f19 KS |
662 | /* |
663 | * We don't initialize ->index for tail pages: calculate based on | |
664 | * head page | |
665 | */ | |
fe3df441 | 666 | return head->index + page - head; |
a0f7a756 NH |
667 | } |
668 | ||
fe19bd3d HD |
669 | extern pgoff_t hugetlb_basepage_index(struct page *page); |
670 | ||
5cbc198a | 671 | /* |
fe19bd3d HD |
672 | * Get the offset in PAGE_SIZE (even for hugetlb pages). |
673 | * (TODO: hugetlb pages should have ->index in PAGE_SIZE) | |
5cbc198a KS |
674 | */ |
675 | static inline pgoff_t page_to_pgoff(struct page *page) | |
676 | { | |
fe19bd3d HD |
677 | if (unlikely(PageHuge(page))) |
678 | return hugetlb_basepage_index(page); | |
5cbc198a KS |
679 | return page_to_index(page); |
680 | } | |
681 | ||
1da177e4 LT |
682 | /* |
683 | * Return byte-offset into filesystem object for page. | |
684 | */ | |
685 | static inline loff_t page_offset(struct page *page) | |
686 | { | |
09cbfeaf | 687 | return ((loff_t)page->index) << PAGE_SHIFT; |
1da177e4 LT |
688 | } |
689 | ||
f981c595 MG |
690 | static inline loff_t page_file_offset(struct page *page) |
691 | { | |
8cd79788 | 692 | return ((loff_t)page_index(page)) << PAGE_SHIFT; |
f981c595 MG |
693 | } |
694 | ||
352b47a6 MWO |
695 | /** |
696 | * folio_pos - Returns the byte position of this folio in its file. | |
697 | * @folio: The folio. | |
698 | */ | |
699 | static inline loff_t folio_pos(struct folio *folio) | |
700 | { | |
701 | return page_offset(&folio->page); | |
702 | } | |
703 | ||
704 | /** | |
705 | * folio_file_pos - Returns the byte position of this folio in its file. | |
706 | * @folio: The folio. | |
707 | * | |
708 | * This differs from folio_pos() for folios which belong to a swap file. | |
709 | * NFS is the only filesystem today which needs to use folio_file_pos(). | |
710 | */ | |
711 | static inline loff_t folio_file_pos(struct folio *folio) | |
712 | { | |
713 | return page_file_offset(&folio->page); | |
714 | } | |
715 | ||
0fe6e20b NH |
716 | extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma, |
717 | unsigned long address); | |
718 | ||
1da177e4 LT |
719 | static inline pgoff_t linear_page_index(struct vm_area_struct *vma, |
720 | unsigned long address) | |
721 | { | |
0fe6e20b NH |
722 | pgoff_t pgoff; |
723 | if (unlikely(is_vm_hugetlb_page(vma))) | |
724 | return linear_hugepage_index(vma, address); | |
725 | pgoff = (address - vma->vm_start) >> PAGE_SHIFT; | |
1da177e4 | 726 | pgoff += vma->vm_pgoff; |
09cbfeaf | 727 | return pgoff; |
1da177e4 LT |
728 | } |
729 | ||
c7510ab2 | 730 | struct wait_page_key { |
df4d4f12 | 731 | struct folio *folio; |
c7510ab2 JA |
732 | int bit_nr; |
733 | int page_match; | |
734 | }; | |
735 | ||
736 | struct wait_page_queue { | |
df4d4f12 | 737 | struct folio *folio; |
c7510ab2 JA |
738 | int bit_nr; |
739 | wait_queue_entry_t wait; | |
740 | }; | |
741 | ||
cdc8fcb4 | 742 | static inline bool wake_page_match(struct wait_page_queue *wait_page, |
c7510ab2 JA |
743 | struct wait_page_key *key) |
744 | { | |
df4d4f12 | 745 | if (wait_page->folio != key->folio) |
cdc8fcb4 | 746 | return false; |
c7510ab2 JA |
747 | key->page_match = 1; |
748 | ||
749 | if (wait_page->bit_nr != key->bit_nr) | |
cdc8fcb4 | 750 | return false; |
d1932dc3 | 751 | |
cdc8fcb4 | 752 | return true; |
d1932dc3 JA |
753 | } |
754 | ||
7c23c782 | 755 | void __folio_lock(struct folio *folio); |
af7f29d9 | 756 | int __folio_lock_killable(struct folio *folio); |
9138e47e | 757 | bool __folio_lock_or_retry(struct folio *folio, struct mm_struct *mm, |
d065bd81 | 758 | unsigned int flags); |
4e136428 MWO |
759 | void unlock_page(struct page *page); |
760 | void folio_unlock(struct folio *folio); | |
1da177e4 | 761 | |
7c23c782 MWO |
762 | static inline bool folio_trylock(struct folio *folio) |
763 | { | |
764 | return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0))); | |
765 | } | |
766 | ||
f4458845 AM |
767 | /* |
768 | * Return true if the page was successfully locked | |
769 | */ | |
529ae9aa NP |
770 | static inline int trylock_page(struct page *page) |
771 | { | |
7c23c782 MWO |
772 | return folio_trylock(page_folio(page)); |
773 | } | |
774 | ||
775 | static inline void folio_lock(struct folio *folio) | |
776 | { | |
777 | might_sleep(); | |
778 | if (!folio_trylock(folio)) | |
779 | __folio_lock(folio); | |
529ae9aa NP |
780 | } |
781 | ||
db37648c NP |
782 | /* |
783 | * lock_page may only be called if we have the page's inode pinned. | |
784 | */ | |
1da177e4 LT |
785 | static inline void lock_page(struct page *page) |
786 | { | |
7c23c782 | 787 | struct folio *folio; |
1da177e4 | 788 | might_sleep(); |
7c23c782 MWO |
789 | |
790 | folio = page_folio(page); | |
791 | if (!folio_trylock(folio)) | |
792 | __folio_lock(folio); | |
1da177e4 | 793 | } |
db37648c | 794 | |
af7f29d9 MWO |
795 | static inline int folio_lock_killable(struct folio *folio) |
796 | { | |
797 | might_sleep(); | |
798 | if (!folio_trylock(folio)) | |
799 | return __folio_lock_killable(folio); | |
800 | return 0; | |
801 | } | |
802 | ||
2687a356 MW |
803 | /* |
804 | * lock_page_killable is like lock_page but can be interrupted by fatal | |
805 | * signals. It returns 0 if it locked the page and -EINTR if it was | |
806 | * killed while waiting. | |
807 | */ | |
808 | static inline int lock_page_killable(struct page *page) | |
809 | { | |
af7f29d9 | 810 | return folio_lock_killable(page_folio(page)); |
2687a356 MW |
811 | } |
812 | ||
d065bd81 ML |
813 | /* |
814 | * lock_page_or_retry - Lock the page, unless this would block and the | |
815 | * caller indicated that it can handle a retry. | |
9a95f3cf | 816 | * |
c1e8d7c6 | 817 | * Return value and mmap_lock implications depend on flags; see |
9138e47e | 818 | * __folio_lock_or_retry(). |
d065bd81 | 819 | */ |
9138e47e | 820 | static inline bool lock_page_or_retry(struct page *page, struct mm_struct *mm, |
d065bd81 ML |
821 | unsigned int flags) |
822 | { | |
9138e47e | 823 | struct folio *folio; |
d065bd81 | 824 | might_sleep(); |
9138e47e MWO |
825 | |
826 | folio = page_folio(page); | |
827 | return folio_trylock(folio) || __folio_lock_or_retry(folio, mm, flags); | |
d065bd81 ML |
828 | } |
829 | ||
1da177e4 | 830 | /* |
101c0bf6 | 831 | * This is exported only for folio_wait_locked/folio_wait_writeback, etc., |
74d81bfa | 832 | * and should not be used directly. |
1da177e4 | 833 | */ |
101c0bf6 MWO |
834 | void folio_wait_bit(struct folio *folio, int bit_nr); |
835 | int folio_wait_bit_killable(struct folio *folio, int bit_nr); | |
a4796e37 | 836 | |
1da177e4 | 837 | /* |
6baa8d60 | 838 | * Wait for a folio to be unlocked. |
1da177e4 | 839 | * |
6baa8d60 MWO |
840 | * This must be called with the caller "holding" the folio, |
841 | * ie with increased "page->count" so that the folio won't | |
1da177e4 LT |
842 | * go away during the wait.. |
843 | */ | |
6baa8d60 MWO |
844 | static inline void folio_wait_locked(struct folio *folio) |
845 | { | |
846 | if (folio_test_locked(folio)) | |
101c0bf6 | 847 | folio_wait_bit(folio, PG_locked); |
6baa8d60 MWO |
848 | } |
849 | ||
850 | static inline int folio_wait_locked_killable(struct folio *folio) | |
851 | { | |
852 | if (!folio_test_locked(folio)) | |
853 | return 0; | |
101c0bf6 | 854 | return folio_wait_bit_killable(folio, PG_locked); |
6baa8d60 MWO |
855 | } |
856 | ||
1da177e4 LT |
857 | static inline void wait_on_page_locked(struct page *page) |
858 | { | |
6baa8d60 | 859 | folio_wait_locked(page_folio(page)); |
1da177e4 LT |
860 | } |
861 | ||
62906027 NP |
862 | static inline int wait_on_page_locked_killable(struct page *page) |
863 | { | |
6baa8d60 | 864 | return folio_wait_locked_killable(page_folio(page)); |
62906027 NP |
865 | } |
866 | ||
9f2b04a2 | 867 | int folio_put_wait_locked(struct folio *folio, int state); |
19343b5b | 868 | void wait_on_page_writeback(struct page *page); |
490e016f MWO |
869 | void folio_wait_writeback(struct folio *folio); |
870 | int folio_wait_writeback_killable(struct folio *folio); | |
4268b480 MWO |
871 | void end_page_writeback(struct page *page); |
872 | void folio_end_writeback(struct folio *folio); | |
1d1d1a76 | 873 | void wait_for_stable_page(struct page *page); |
a49d0c50 | 874 | void folio_wait_stable(struct folio *folio); |
203a3151 MWO |
875 | void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn); |
876 | static inline void __set_page_dirty(struct page *page, | |
877 | struct address_space *mapping, int warn) | |
878 | { | |
879 | __folio_mark_dirty(page_folio(page), mapping, warn); | |
880 | } | |
fc9b6a53 MWO |
881 | void folio_account_cleaned(struct folio *folio, struct address_space *mapping, |
882 | struct bdi_writeback *wb); | |
fdaf532a MWO |
883 | void __folio_cancel_dirty(struct folio *folio); |
884 | static inline void folio_cancel_dirty(struct folio *folio) | |
885 | { | |
886 | /* Avoid atomic ops, locking, etc. when not actually needed. */ | |
887 | if (folio_test_dirty(folio)) | |
888 | __folio_cancel_dirty(folio); | |
889 | } | |
890 | static inline void cancel_dirty_page(struct page *page) | |
891 | { | |
892 | folio_cancel_dirty(page_folio(page)); | |
893 | } | |
9350f20a MWO |
894 | bool folio_clear_dirty_for_io(struct folio *folio); |
895 | bool clear_page_dirty_for_io(struct page *page); | |
121703c1 MWO |
896 | int __must_check folio_write_one(struct folio *folio); |
897 | static inline int __must_check write_one_page(struct page *page) | |
898 | { | |
899 | return folio_write_one(page_folio(page)); | |
900 | } | |
1da177e4 | 901 | |
3a6b2162 MWO |
902 | int __set_page_dirty_nobuffers(struct page *page); |
903 | int __set_page_dirty_no_writeback(struct page *page); | |
904 | ||
c11f0c0b | 905 | void page_endio(struct page *page, bool is_write, int err); |
57d99845 | 906 | |
b47393f8 MWO |
907 | void folio_end_private_2(struct folio *folio); |
908 | void folio_wait_private_2(struct folio *folio); | |
909 | int folio_wait_private_2_killable(struct folio *folio); | |
73e10ded | 910 | |
385e1ca5 DH |
911 | /* |
912 | * Add an arbitrary waiter to a page's wait queue | |
913 | */ | |
df4d4f12 | 914 | void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter); |
385e1ca5 | 915 | |
1da177e4 | 916 | /* |
bb523b40 | 917 | * Fault in userspace address range. |
1da177e4 | 918 | */ |
bb523b40 | 919 | size_t fault_in_writeable(char __user *uaddr, size_t size); |
cdd591fc | 920 | size_t fault_in_safe_writeable(const char __user *uaddr, size_t size); |
bb523b40 | 921 | size_t fault_in_readable(const char __user *uaddr, size_t size); |
f56f821f | 922 | |
529ae9aa | 923 | int add_to_page_cache_locked(struct page *page, struct address_space *mapping, |
9dd3d069 | 924 | pgoff_t index, gfp_t gfp); |
529ae9aa | 925 | int add_to_page_cache_lru(struct page *page, struct address_space *mapping, |
9dd3d069 MWO |
926 | pgoff_t index, gfp_t gfp); |
927 | int filemap_add_folio(struct address_space *mapping, struct folio *folio, | |
928 | pgoff_t index, gfp_t gfp); | |
452e9e69 MWO |
929 | void filemap_remove_folio(struct folio *folio); |
930 | void delete_from_page_cache(struct page *page); | |
931 | void __filemap_remove_folio(struct folio *folio, void *shadow); | |
932 | static inline void __delete_from_page_cache(struct page *page, void *shadow) | |
933 | { | |
934 | __filemap_remove_folio(page_folio(page), shadow); | |
935 | } | |
1f7ef657 | 936 | void replace_page_cache_page(struct page *old, struct page *new); |
aa65c29c | 937 | void delete_from_page_cache_batch(struct address_space *mapping, |
51dcbdac | 938 | struct folio_batch *fbatch); |
82c50f8b MWO |
939 | int try_to_release_page(struct page *page, gfp_t gfp); |
940 | bool filemap_release_folio(struct folio *folio, gfp_t gfp); | |
41139aa4 MWO |
941 | loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end, |
942 | int whence); | |
529ae9aa NP |
943 | |
944 | /* | |
945 | * Like add_to_page_cache_locked, but used to add newly allocated pages: | |
48c935ad | 946 | * the page is new, so we can just run __SetPageLocked() against it. |
529ae9aa NP |
947 | */ |
948 | static inline int add_to_page_cache(struct page *page, | |
949 | struct address_space *mapping, pgoff_t offset, gfp_t gfp_mask) | |
950 | { | |
951 | int error; | |
952 | ||
48c935ad | 953 | __SetPageLocked(page); |
529ae9aa NP |
954 | error = add_to_page_cache_locked(page, mapping, offset, gfp_mask); |
955 | if (unlikely(error)) | |
48c935ad | 956 | __ClearPageLocked(page); |
529ae9aa NP |
957 | return error; |
958 | } | |
959 | ||
9dd3d069 MWO |
960 | /* Must be non-static for BPF error injection */ |
961 | int __filemap_add_folio(struct address_space *mapping, struct folio *folio, | |
962 | pgoff_t index, gfp_t gfp, void **shadowp); | |
963 | ||
4bdcd1dd JA |
964 | bool filemap_range_has_writeback(struct address_space *mapping, |
965 | loff_t start_byte, loff_t end_byte); | |
966 | ||
967 | /** | |
968 | * filemap_range_needs_writeback - check if range potentially needs writeback | |
969 | * @mapping: address space within which to check | |
970 | * @start_byte: offset in bytes where the range starts | |
971 | * @end_byte: offset in bytes where the range ends (inclusive) | |
972 | * | |
973 | * Find at least one page in the range supplied, usually used to check if | |
974 | * direct writing in this range will trigger a writeback. Used by O_DIRECT | |
975 | * read/write with IOCB_NOWAIT, to see if the caller needs to do | |
976 | * filemap_write_and_wait_range() before proceeding. | |
977 | * | |
978 | * Return: %true if the caller should do filemap_write_and_wait_range() before | |
979 | * doing O_DIRECT to a page in this range, %false otherwise. | |
980 | */ | |
981 | static inline bool filemap_range_needs_writeback(struct address_space *mapping, | |
982 | loff_t start_byte, | |
983 | loff_t end_byte) | |
984 | { | |
985 | if (!mapping->nrpages) | |
986 | return false; | |
987 | if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) && | |
988 | !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK)) | |
989 | return false; | |
990 | return filemap_range_has_writeback(mapping, start_byte, end_byte); | |
991 | } | |
992 | ||
042124cc MWO |
993 | /** |
994 | * struct readahead_control - Describes a readahead request. | |
995 | * | |
996 | * A readahead request is for consecutive pages. Filesystems which | |
997 | * implement the ->readahead method should call readahead_page() or | |
998 | * readahead_page_batch() in a loop and attempt to start I/O against | |
999 | * each page in the request. | |
1000 | * | |
1001 | * Most of the fields in this struct are private and should be accessed | |
1002 | * by the functions below. | |
1003 | * | |
1004 | * @file: The file, used primarily by network filesystems for authentication. | |
1005 | * May be NULL if invoked internally by the filesystem. | |
1006 | * @mapping: Readahead this filesystem object. | |
fcd9ae4f | 1007 | * @ra: File readahead state. May be NULL. |
042124cc MWO |
1008 | */ |
1009 | struct readahead_control { | |
1010 | struct file *file; | |
1011 | struct address_space *mapping; | |
fcd9ae4f | 1012 | struct file_ra_state *ra; |
042124cc MWO |
1013 | /* private: use the readahead_* accessors instead */ |
1014 | pgoff_t _index; | |
1015 | unsigned int _nr_pages; | |
1016 | unsigned int _batch_count; | |
1017 | }; | |
1018 | ||
fcd9ae4f MWO |
1019 | #define DEFINE_READAHEAD(ractl, f, r, m, i) \ |
1020 | struct readahead_control ractl = { \ | |
1aa83cfa MWO |
1021 | .file = f, \ |
1022 | .mapping = m, \ | |
fcd9ae4f | 1023 | .ra = r, \ |
1aa83cfa MWO |
1024 | ._index = i, \ |
1025 | } | |
1026 | ||
fefa7c47 MWO |
1027 | #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE) |
1028 | ||
1029 | void page_cache_ra_unbounded(struct readahead_control *, | |
1030 | unsigned long nr_to_read, unsigned long lookahead_count); | |
fcd9ae4f | 1031 | void page_cache_sync_ra(struct readahead_control *, unsigned long req_count); |
7836d999 | 1032 | void page_cache_async_ra(struct readahead_control *, struct folio *, |
fefa7c47 | 1033 | unsigned long req_count); |
3ca23644 DH |
1034 | void readahead_expand(struct readahead_control *ractl, |
1035 | loff_t new_start, size_t new_len); | |
fefa7c47 MWO |
1036 | |
1037 | /** | |
1038 | * page_cache_sync_readahead - generic file readahead | |
1039 | * @mapping: address_space which holds the pagecache and I/O vectors | |
1040 | * @ra: file_ra_state which holds the readahead state | |
1041 | * @file: Used by the filesystem for authentication. | |
1042 | * @index: Index of first page to be read. | |
1043 | * @req_count: Total number of pages being read by the caller. | |
1044 | * | |
1045 | * page_cache_sync_readahead() should be called when a cache miss happened: | |
1046 | * it will submit the read. The readahead logic may decide to piggyback more | |
1047 | * pages onto the read request if access patterns suggest it will improve | |
1048 | * performance. | |
1049 | */ | |
1050 | static inline | |
1051 | void page_cache_sync_readahead(struct address_space *mapping, | |
1052 | struct file_ra_state *ra, struct file *file, pgoff_t index, | |
1053 | unsigned long req_count) | |
1054 | { | |
fcd9ae4f MWO |
1055 | DEFINE_READAHEAD(ractl, file, ra, mapping, index); |
1056 | page_cache_sync_ra(&ractl, req_count); | |
fefa7c47 MWO |
1057 | } |
1058 | ||
1059 | /** | |
1060 | * page_cache_async_readahead - file readahead for marked pages | |
1061 | * @mapping: address_space which holds the pagecache and I/O vectors | |
1062 | * @ra: file_ra_state which holds the readahead state | |
1063 | * @file: Used by the filesystem for authentication. | |
1064 | * @page: The page at @index which triggered the readahead call. | |
1065 | * @index: Index of first page to be read. | |
1066 | * @req_count: Total number of pages being read by the caller. | |
1067 | * | |
1068 | * page_cache_async_readahead() should be called when a page is used which | |
1069 | * is marked as PageReadahead; this is a marker to suggest that the application | |
1070 | * has used up enough of the readahead window that we should start pulling in | |
1071 | * more pages. | |
1072 | */ | |
1073 | static inline | |
1074 | void page_cache_async_readahead(struct address_space *mapping, | |
1075 | struct file_ra_state *ra, struct file *file, | |
1076 | struct page *page, pgoff_t index, unsigned long req_count) | |
1077 | { | |
fcd9ae4f | 1078 | DEFINE_READAHEAD(ractl, file, ra, mapping, index); |
7836d999 | 1079 | page_cache_async_ra(&ractl, page_folio(page), req_count); |
fefa7c47 MWO |
1080 | } |
1081 | ||
9bf70167 MWO |
1082 | static inline struct folio *__readahead_folio(struct readahead_control *ractl) |
1083 | { | |
1084 | struct folio *folio; | |
1085 | ||
1086 | BUG_ON(ractl->_batch_count > ractl->_nr_pages); | |
1087 | ractl->_nr_pages -= ractl->_batch_count; | |
1088 | ractl->_index += ractl->_batch_count; | |
1089 | ||
1090 | if (!ractl->_nr_pages) { | |
1091 | ractl->_batch_count = 0; | |
1092 | return NULL; | |
1093 | } | |
1094 | ||
1095 | folio = xa_load(&ractl->mapping->i_pages, ractl->_index); | |
1096 | VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio); | |
1097 | ractl->_batch_count = folio_nr_pages(folio); | |
1098 | ||
1099 | return folio; | |
1100 | } | |
1101 | ||
042124cc MWO |
1102 | /** |
1103 | * readahead_page - Get the next page to read. | |
9bf70167 | 1104 | * @ractl: The current readahead request. |
042124cc MWO |
1105 | * |
1106 | * Context: The page is locked and has an elevated refcount. The caller | |
1107 | * should decreases the refcount once the page has been submitted for I/O | |
1108 | * and unlock the page once all I/O to that page has completed. | |
1109 | * Return: A pointer to the next page, or %NULL if we are done. | |
1110 | */ | |
9bf70167 | 1111 | static inline struct page *readahead_page(struct readahead_control *ractl) |
042124cc | 1112 | { |
9bf70167 | 1113 | struct folio *folio = __readahead_folio(ractl); |
042124cc | 1114 | |
9bf70167 MWO |
1115 | return &folio->page; |
1116 | } | |
042124cc | 1117 | |
9bf70167 MWO |
1118 | /** |
1119 | * readahead_folio - Get the next folio to read. | |
1120 | * @ractl: The current readahead request. | |
1121 | * | |
1122 | * Context: The folio is locked. The caller should unlock the folio once | |
1123 | * all I/O to that folio has completed. | |
1124 | * Return: A pointer to the next folio, or %NULL if we are done. | |
1125 | */ | |
1126 | static inline struct folio *readahead_folio(struct readahead_control *ractl) | |
1127 | { | |
1128 | struct folio *folio = __readahead_folio(ractl); | |
042124cc | 1129 | |
9bf70167 MWO |
1130 | if (folio) |
1131 | folio_put(folio); | |
1132 | return folio; | |
042124cc MWO |
1133 | } |
1134 | ||
1135 | static inline unsigned int __readahead_batch(struct readahead_control *rac, | |
1136 | struct page **array, unsigned int array_sz) | |
1137 | { | |
1138 | unsigned int i = 0; | |
1139 | XA_STATE(xas, &rac->mapping->i_pages, 0); | |
1140 | struct page *page; | |
1141 | ||
1142 | BUG_ON(rac->_batch_count > rac->_nr_pages); | |
1143 | rac->_nr_pages -= rac->_batch_count; | |
1144 | rac->_index += rac->_batch_count; | |
1145 | rac->_batch_count = 0; | |
1146 | ||
1147 | xas_set(&xas, rac->_index); | |
1148 | rcu_read_lock(); | |
1149 | xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) { | |
4349a83a MWO |
1150 | if (xas_retry(&xas, page)) |
1151 | continue; | |
042124cc MWO |
1152 | VM_BUG_ON_PAGE(!PageLocked(page), page); |
1153 | VM_BUG_ON_PAGE(PageTail(page), page); | |
1154 | array[i++] = page; | |
6c357848 | 1155 | rac->_batch_count += thp_nr_pages(page); |
042124cc MWO |
1156 | if (i == array_sz) |
1157 | break; | |
1158 | } | |
1159 | rcu_read_unlock(); | |
1160 | ||
1161 | return i; | |
1162 | } | |
1163 | ||
1164 | /** | |
1165 | * readahead_page_batch - Get a batch of pages to read. | |
1166 | * @rac: The current readahead request. | |
1167 | * @array: An array of pointers to struct page. | |
1168 | * | |
1169 | * Context: The pages are locked and have an elevated refcount. The caller | |
1170 | * should decreases the refcount once the page has been submitted for I/O | |
1171 | * and unlock the page once all I/O to that page has completed. | |
1172 | * Return: The number of pages placed in the array. 0 indicates the request | |
1173 | * is complete. | |
1174 | */ | |
1175 | #define readahead_page_batch(rac, array) \ | |
1176 | __readahead_batch(rac, array, ARRAY_SIZE(array)) | |
1177 | ||
1178 | /** | |
1179 | * readahead_pos - The byte offset into the file of this readahead request. | |
1180 | * @rac: The readahead request. | |
1181 | */ | |
1182 | static inline loff_t readahead_pos(struct readahead_control *rac) | |
1183 | { | |
1184 | return (loff_t)rac->_index * PAGE_SIZE; | |
1185 | } | |
1186 | ||
1187 | /** | |
1188 | * readahead_length - The number of bytes in this readahead request. | |
1189 | * @rac: The readahead request. | |
1190 | */ | |
076171a6 | 1191 | static inline size_t readahead_length(struct readahead_control *rac) |
042124cc | 1192 | { |
076171a6 | 1193 | return rac->_nr_pages * PAGE_SIZE; |
042124cc MWO |
1194 | } |
1195 | ||
1196 | /** | |
1197 | * readahead_index - The index of the first page in this readahead request. | |
1198 | * @rac: The readahead request. | |
1199 | */ | |
1200 | static inline pgoff_t readahead_index(struct readahead_control *rac) | |
1201 | { | |
1202 | return rac->_index; | |
1203 | } | |
1204 | ||
1205 | /** | |
1206 | * readahead_count - The number of pages in this readahead request. | |
1207 | * @rac: The readahead request. | |
1208 | */ | |
1209 | static inline unsigned int readahead_count(struct readahead_control *rac) | |
1210 | { | |
1211 | return rac->_nr_pages; | |
1212 | } | |
1213 | ||
32c0a6bc MWO |
1214 | /** |
1215 | * readahead_batch_length - The number of bytes in the current batch. | |
1216 | * @rac: The readahead request. | |
1217 | */ | |
076171a6 | 1218 | static inline size_t readahead_batch_length(struct readahead_control *rac) |
32c0a6bc MWO |
1219 | { |
1220 | return rac->_batch_count * PAGE_SIZE; | |
1221 | } | |
1222 | ||
b57c2cb9 FF |
1223 | static inline unsigned long dir_pages(struct inode *inode) |
1224 | { | |
09cbfeaf KS |
1225 | return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >> |
1226 | PAGE_SHIFT; | |
b57c2cb9 FF |
1227 | } |
1228 | ||
f705bf84 MWO |
1229 | /** |
1230 | * folio_mkwrite_check_truncate - check if folio was truncated | |
1231 | * @folio: the folio to check | |
1232 | * @inode: the inode to check the folio against | |
1233 | * | |
1234 | * Return: the number of bytes in the folio up to EOF, | |
1235 | * or -EFAULT if the folio was truncated. | |
1236 | */ | |
1237 | static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio, | |
1238 | struct inode *inode) | |
1239 | { | |
1240 | loff_t size = i_size_read(inode); | |
1241 | pgoff_t index = size >> PAGE_SHIFT; | |
1242 | size_t offset = offset_in_folio(folio, size); | |
1243 | ||
1244 | if (!folio->mapping) | |
1245 | return -EFAULT; | |
1246 | ||
1247 | /* folio is wholly inside EOF */ | |
1248 | if (folio_next_index(folio) - 1 < index) | |
1249 | return folio_size(folio); | |
1250 | /* folio is wholly past EOF */ | |
1251 | if (folio->index > index || !offset) | |
1252 | return -EFAULT; | |
1253 | /* folio is partially inside EOF */ | |
1254 | return offset; | |
1255 | } | |
1256 | ||
243145bc AG |
1257 | /** |
1258 | * page_mkwrite_check_truncate - check if page was truncated | |
1259 | * @page: the page to check | |
1260 | * @inode: the inode to check the page against | |
1261 | * | |
1262 | * Returns the number of bytes in the page up to EOF, | |
1263 | * or -EFAULT if the page was truncated. | |
1264 | */ | |
1265 | static inline int page_mkwrite_check_truncate(struct page *page, | |
1266 | struct inode *inode) | |
1267 | { | |
1268 | loff_t size = i_size_read(inode); | |
1269 | pgoff_t index = size >> PAGE_SHIFT; | |
1270 | int offset = offset_in_page(size); | |
1271 | ||
1272 | if (page->mapping != inode->i_mapping) | |
1273 | return -EFAULT; | |
1274 | ||
1275 | /* page is wholly inside EOF */ | |
1276 | if (page->index < index) | |
1277 | return PAGE_SIZE; | |
1278 | /* page is wholly past EOF */ | |
1279 | if (page->index > index || !offset) | |
1280 | return -EFAULT; | |
1281 | /* page is partially inside EOF */ | |
1282 | return offset; | |
1283 | } | |
1284 | ||
24addd84 | 1285 | /** |
9eb7c76d | 1286 | * i_blocks_per_folio - How many blocks fit in this folio. |
24addd84 | 1287 | * @inode: The inode which contains the blocks. |
9eb7c76d | 1288 | * @folio: The folio. |
24addd84 | 1289 | * |
9eb7c76d | 1290 | * If the block size is larger than the size of this folio, return zero. |
24addd84 | 1291 | * |
9eb7c76d | 1292 | * Context: The caller should hold a refcount on the folio to prevent it |
24addd84 | 1293 | * from being split. |
9eb7c76d | 1294 | * Return: The number of filesystem blocks covered by this folio. |
24addd84 | 1295 | */ |
9eb7c76d MWO |
1296 | static inline |
1297 | unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio) | |
1298 | { | |
1299 | return folio_size(folio) >> inode->i_blkbits; | |
1300 | } | |
1301 | ||
24addd84 MWO |
1302 | static inline |
1303 | unsigned int i_blocks_per_page(struct inode *inode, struct page *page) | |
1304 | { | |
9eb7c76d | 1305 | return i_blocks_per_folio(inode, page_folio(page)); |
24addd84 | 1306 | } |
1da177e4 | 1307 | #endif /* _LINUX_PAGEMAP_H */ |