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