]>
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 | |
aa65c29c JK |
19 | struct pagevec; |
20 | ||
1da177e4 | 21 | /* |
9c5d760b | 22 | * Bits in mapping->flags. |
1da177e4 | 23 | */ |
9a896c9a | 24 | enum mapping_flags { |
9c5d760b MH |
25 | AS_EIO = 0, /* IO error on async write */ |
26 | AS_ENOSPC = 1, /* ENOSPC on async write */ | |
27 | AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */ | |
28 | AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */ | |
29 | AS_EXITING = 4, /* final truncate in progress */ | |
371a096e | 30 | /* writeback related tags are not used */ |
9c5d760b | 31 | AS_NO_WRITEBACK_TAGS = 5, |
9a896c9a | 32 | }; |
1da177e4 | 33 | |
8ed1e46a JL |
34 | /** |
35 | * mapping_set_error - record a writeback error in the address_space | |
767e5ee5 MWO |
36 | * @mapping: the mapping in which an error should be set |
37 | * @error: the error to set in the mapping | |
8ed1e46a JL |
38 | * |
39 | * When writeback fails in some way, we must record that error so that | |
40 | * userspace can be informed when fsync and the like are called. We endeavor | |
41 | * to report errors on any file that was open at the time of the error. Some | |
42 | * internal callers also need to know when writeback errors have occurred. | |
43 | * | |
44 | * When a writeback error occurs, most filesystems will want to call | |
45 | * mapping_set_error to record the error in the mapping so that it can be | |
46 | * reported when the application calls fsync(2). | |
47 | */ | |
3e9f45bd GC |
48 | static inline void mapping_set_error(struct address_space *mapping, int error) |
49 | { | |
8ed1e46a JL |
50 | if (likely(!error)) |
51 | return; | |
52 | ||
53 | /* Record in wb_err for checkers using errseq_t based tracking */ | |
735e4ae5 JL |
54 | __filemap_set_wb_err(mapping, error); |
55 | ||
56 | /* Record it in superblock */ | |
57 | errseq_set(&mapping->host->i_sb->s_wb_err, error); | |
8ed1e46a JL |
58 | |
59 | /* Record it in flags for now, for legacy callers */ | |
60 | if (error == -ENOSPC) | |
61 | set_bit(AS_ENOSPC, &mapping->flags); | |
62 | else | |
63 | set_bit(AS_EIO, &mapping->flags); | |
3e9f45bd GC |
64 | } |
65 | ||
ba9ddf49 LS |
66 | static inline void mapping_set_unevictable(struct address_space *mapping) |
67 | { | |
68 | set_bit(AS_UNEVICTABLE, &mapping->flags); | |
69 | } | |
70 | ||
89e004ea LS |
71 | static inline void mapping_clear_unevictable(struct address_space *mapping) |
72 | { | |
73 | clear_bit(AS_UNEVICTABLE, &mapping->flags); | |
74 | } | |
75 | ||
1eb6234e | 76 | static inline bool mapping_unevictable(struct address_space *mapping) |
ba9ddf49 | 77 | { |
1eb6234e | 78 | return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags); |
ba9ddf49 | 79 | } |
ba9ddf49 | 80 | |
91b0abe3 JW |
81 | static inline void mapping_set_exiting(struct address_space *mapping) |
82 | { | |
83 | set_bit(AS_EXITING, &mapping->flags); | |
84 | } | |
85 | ||
86 | static inline int mapping_exiting(struct address_space *mapping) | |
87 | { | |
88 | return test_bit(AS_EXITING, &mapping->flags); | |
89 | } | |
90 | ||
371a096e HY |
91 | static inline void mapping_set_no_writeback_tags(struct address_space *mapping) |
92 | { | |
93 | set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags); | |
94 | } | |
95 | ||
96 | static inline int mapping_use_writeback_tags(struct address_space *mapping) | |
97 | { | |
98 | return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags); | |
99 | } | |
100 | ||
dd0fc66f | 101 | static inline gfp_t mapping_gfp_mask(struct address_space * mapping) |
1da177e4 | 102 | { |
9c5d760b | 103 | return mapping->gfp_mask; |
1da177e4 LT |
104 | } |
105 | ||
c62d2555 MH |
106 | /* Restricts the given gfp_mask to what the mapping allows. */ |
107 | static inline gfp_t mapping_gfp_constraint(struct address_space *mapping, | |
108 | gfp_t gfp_mask) | |
109 | { | |
110 | return mapping_gfp_mask(mapping) & gfp_mask; | |
111 | } | |
112 | ||
1da177e4 LT |
113 | /* |
114 | * This is non-atomic. Only to be used before the mapping is activated. | |
115 | * Probably needs a barrier... | |
116 | */ | |
260b2367 | 117 | static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask) |
1da177e4 | 118 | { |
9c5d760b | 119 | m->gfp_mask = mask; |
1da177e4 LT |
120 | } |
121 | ||
c6f92f9f | 122 | void release_pages(struct page **pages, int nr); |
1da177e4 | 123 | |
e286781d NP |
124 | /* |
125 | * speculatively take a reference to a page. | |
0139aa7b JK |
126 | * If the page is free (_refcount == 0), then _refcount is untouched, and 0 |
127 | * is returned. Otherwise, _refcount is incremented by 1 and 1 is returned. | |
e286781d NP |
128 | * |
129 | * This function must be called inside the same rcu_read_lock() section as has | |
130 | * been used to lookup the page in the pagecache radix-tree (or page table): | |
0139aa7b | 131 | * this allows allocators to use a synchronize_rcu() to stabilize _refcount. |
e286781d NP |
132 | * |
133 | * Unless an RCU grace period has passed, the count of all pages coming out | |
134 | * of the allocator must be considered unstable. page_count may return higher | |
135 | * than expected, and put_page must be able to do the right thing when the | |
136 | * page has been finished with, no matter what it is subsequently allocated | |
137 | * for (because put_page is what is used here to drop an invalid speculative | |
138 | * reference). | |
139 | * | |
140 | * This is the interesting part of the lockless pagecache (and lockless | |
141 | * get_user_pages) locking protocol, where the lookup-side (eg. find_get_page) | |
142 | * has the following pattern: | |
143 | * 1. find page in radix tree | |
144 | * 2. conditionally increment refcount | |
145 | * 3. check the page is still in pagecache (if no, goto 1) | |
146 | * | |
0139aa7b | 147 | * Remove-side that cares about stability of _refcount (eg. reclaim) has the |
b93b0163 | 148 | * following (with the i_pages lock held): |
e286781d NP |
149 | * A. atomically check refcount is correct and set it to 0 (atomic_cmpxchg) |
150 | * B. remove page from pagecache | |
151 | * C. free the page | |
152 | * | |
153 | * There are 2 critical interleavings that matter: | |
154 | * - 2 runs before A: in this case, A sees elevated refcount and bails out | |
155 | * - A runs before 2: in this case, 2 sees zero refcount and retries; | |
156 | * subsequently, B will complete and 1 will find no page, causing the | |
157 | * lookup to return NULL. | |
158 | * | |
159 | * It is possible that between 1 and 2, the page is removed then the exact same | |
160 | * page is inserted into the same position in pagecache. That's OK: the | |
b93b0163 | 161 | * old find_get_page using a lock could equally have run before or after |
e286781d NP |
162 | * such a re-insertion, depending on order that locks are granted. |
163 | * | |
164 | * Lookups racing against pagecache insertion isn't a big problem: either 1 | |
165 | * will find the page or it will not. Likewise, the old find_get_page could run | |
166 | * either before the insertion or afterwards, depending on timing. | |
167 | */ | |
494eec70 | 168 | static inline int __page_cache_add_speculative(struct page *page, int count) |
e286781d | 169 | { |
8375ad98 | 170 | #ifdef CONFIG_TINY_RCU |
bdd4e85d | 171 | # ifdef CONFIG_PREEMPT_COUNT |
591a3d7c | 172 | VM_BUG_ON(!in_atomic() && !irqs_disabled()); |
e286781d NP |
173 | # endif |
174 | /* | |
175 | * Preempt must be disabled here - we rely on rcu_read_lock doing | |
176 | * this for us. | |
177 | * | |
178 | * Pagecache won't be truncated from interrupt context, so if we have | |
179 | * found a page in the radix tree here, we have pinned its refcount by | |
180 | * disabling preempt, and hence no need for the "speculative get" that | |
181 | * SMP requires. | |
182 | */ | |
309381fe | 183 | VM_BUG_ON_PAGE(page_count(page) == 0, page); |
494eec70 | 184 | page_ref_add(page, count); |
e286781d NP |
185 | |
186 | #else | |
494eec70 | 187 | if (unlikely(!page_ref_add_unless(page, count, 0))) { |
e286781d NP |
188 | /* |
189 | * Either the page has been freed, or will be freed. | |
190 | * In either case, retry here and the caller should | |
191 | * do the right thing (see comments above). | |
192 | */ | |
193 | return 0; | |
194 | } | |
195 | #endif | |
309381fe | 196 | VM_BUG_ON_PAGE(PageTail(page), page); |
e286781d NP |
197 | |
198 | return 1; | |
199 | } | |
200 | ||
494eec70 | 201 | static inline int page_cache_get_speculative(struct page *page) |
ce0ad7f0 | 202 | { |
494eec70 | 203 | return __page_cache_add_speculative(page, 1); |
204 | } | |
ce0ad7f0 | 205 | |
494eec70 | 206 | static inline int page_cache_add_speculative(struct page *page, int count) |
207 | { | |
208 | return __page_cache_add_speculative(page, count); | |
ce0ad7f0 NP |
209 | } |
210 | ||
44110fe3 | 211 | #ifdef CONFIG_NUMA |
2ae88149 | 212 | extern struct page *__page_cache_alloc(gfp_t gfp); |
44110fe3 | 213 | #else |
2ae88149 NP |
214 | static inline struct page *__page_cache_alloc(gfp_t gfp) |
215 | { | |
216 | return alloc_pages(gfp, 0); | |
217 | } | |
218 | #endif | |
219 | ||
1da177e4 LT |
220 | static inline struct page *page_cache_alloc(struct address_space *x) |
221 | { | |
2ae88149 | 222 | return __page_cache_alloc(mapping_gfp_mask(x)); |
1da177e4 LT |
223 | } |
224 | ||
8a5c743e | 225 | static inline gfp_t readahead_gfp_mask(struct address_space *x) |
7b1de586 | 226 | { |
453f85d4 | 227 | return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN; |
7b1de586 WF |
228 | } |
229 | ||
1da177e4 LT |
230 | typedef int filler_t(void *, struct page *); |
231 | ||
0d3f9296 | 232 | pgoff_t page_cache_next_miss(struct address_space *mapping, |
e7b563bb | 233 | pgoff_t index, unsigned long max_scan); |
0d3f9296 | 234 | pgoff_t page_cache_prev_miss(struct address_space *mapping, |
e7b563bb JW |
235 | pgoff_t index, unsigned long max_scan); |
236 | ||
2457aec6 MG |
237 | #define FGP_ACCESSED 0x00000001 |
238 | #define FGP_LOCK 0x00000002 | |
239 | #define FGP_CREAT 0x00000004 | |
240 | #define FGP_WRITE 0x00000008 | |
241 | #define FGP_NOFS 0x00000010 | |
242 | #define FGP_NOWAIT 0x00000020 | |
a75d4c33 | 243 | #define FGP_FOR_MMAP 0x00000040 |
2457aec6 MG |
244 | |
245 | struct page *pagecache_get_page(struct address_space *mapping, pgoff_t offset, | |
45f87de5 | 246 | int fgp_flags, gfp_t cache_gfp_mask); |
2457aec6 MG |
247 | |
248 | /** | |
249 | * find_get_page - find and get a page reference | |
250 | * @mapping: the address_space to search | |
251 | * @offset: the page index | |
252 | * | |
253 | * Looks up the page cache slot at @mapping & @offset. If there is a | |
254 | * page cache page, it is returned with an increased refcount. | |
255 | * | |
256 | * Otherwise, %NULL is returned. | |
257 | */ | |
258 | static inline struct page *find_get_page(struct address_space *mapping, | |
259 | pgoff_t offset) | |
260 | { | |
45f87de5 | 261 | return pagecache_get_page(mapping, offset, 0, 0); |
2457aec6 MG |
262 | } |
263 | ||
264 | static inline struct page *find_get_page_flags(struct address_space *mapping, | |
265 | pgoff_t offset, int fgp_flags) | |
266 | { | |
45f87de5 | 267 | return pagecache_get_page(mapping, offset, fgp_flags, 0); |
2457aec6 MG |
268 | } |
269 | ||
270 | /** | |
271 | * find_lock_page - locate, pin and lock a pagecache page | |
2457aec6 MG |
272 | * @mapping: the address_space to search |
273 | * @offset: the page index | |
274 | * | |
275 | * Looks up the page cache slot at @mapping & @offset. If there is a | |
276 | * page cache page, it is returned locked and with an increased | |
277 | * refcount. | |
278 | * | |
279 | * Otherwise, %NULL is returned. | |
280 | * | |
281 | * find_lock_page() may sleep. | |
282 | */ | |
283 | static inline struct page *find_lock_page(struct address_space *mapping, | |
284 | pgoff_t offset) | |
285 | { | |
45f87de5 | 286 | return pagecache_get_page(mapping, offset, FGP_LOCK, 0); |
2457aec6 MG |
287 | } |
288 | ||
289 | /** | |
290 | * find_or_create_page - locate or add a pagecache page | |
291 | * @mapping: the page's address_space | |
292 | * @index: the page's index into the mapping | |
293 | * @gfp_mask: page allocation mode | |
294 | * | |
295 | * Looks up the page cache slot at @mapping & @offset. If there is a | |
296 | * page cache page, it is returned locked and with an increased | |
297 | * refcount. | |
298 | * | |
299 | * If the page is not present, a new page is allocated using @gfp_mask | |
300 | * and added to the page cache and the VM's LRU list. The page is | |
301 | * returned locked and with an increased refcount. | |
302 | * | |
303 | * On memory exhaustion, %NULL is returned. | |
304 | * | |
305 | * find_or_create_page() may sleep, even if @gfp_flags specifies an | |
306 | * atomic allocation! | |
307 | */ | |
308 | static inline struct page *find_or_create_page(struct address_space *mapping, | |
767e5ee5 | 309 | pgoff_t index, gfp_t gfp_mask) |
2457aec6 | 310 | { |
767e5ee5 | 311 | return pagecache_get_page(mapping, index, |
2457aec6 | 312 | FGP_LOCK|FGP_ACCESSED|FGP_CREAT, |
45f87de5 | 313 | gfp_mask); |
2457aec6 MG |
314 | } |
315 | ||
316 | /** | |
317 | * grab_cache_page_nowait - returns locked page at given index in given cache | |
318 | * @mapping: target address_space | |
319 | * @index: the page index | |
320 | * | |
321 | * Same as grab_cache_page(), but do not wait if the page is unavailable. | |
322 | * This is intended for speculative data generators, where the data can | |
323 | * be regenerated if the page couldn't be grabbed. This routine should | |
324 | * be safe to call while holding the lock for another page. | |
325 | * | |
326 | * Clear __GFP_FS when allocating the page to avoid recursion into the fs | |
327 | * and deadlock against the caller's locked page. | |
328 | */ | |
329 | static inline struct page *grab_cache_page_nowait(struct address_space *mapping, | |
330 | pgoff_t index) | |
331 | { | |
332 | return pagecache_get_page(mapping, index, | |
333 | FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT, | |
45f87de5 | 334 | mapping_gfp_mask(mapping)); |
2457aec6 MG |
335 | } |
336 | ||
ec848215 MWO |
337 | /* |
338 | * Given the page we found in the page cache, return the page corresponding | |
339 | * to this index in the file | |
340 | */ | |
341 | static inline struct page *find_subpage(struct page *head, pgoff_t index) | |
4101196b | 342 | { |
ec848215 MWO |
343 | /* HugeTLBfs wants the head page regardless */ |
344 | if (PageHuge(head)) | |
345 | return head; | |
4101196b | 346 | |
a0650604 | 347 | return head + (index & (hpage_nr_pages(head) - 1)); |
4101196b MWO |
348 | } |
349 | ||
0cd6144a | 350 | struct page *find_get_entry(struct address_space *mapping, pgoff_t offset); |
0cd6144a | 351 | struct page *find_lock_entry(struct address_space *mapping, pgoff_t offset); |
0cd6144a JW |
352 | unsigned find_get_entries(struct address_space *mapping, pgoff_t start, |
353 | unsigned int nr_entries, struct page **entries, | |
354 | pgoff_t *indices); | |
b947cee4 JK |
355 | unsigned find_get_pages_range(struct address_space *mapping, pgoff_t *start, |
356 | pgoff_t end, unsigned int nr_pages, | |
357 | struct page **pages); | |
358 | static inline unsigned find_get_pages(struct address_space *mapping, | |
359 | pgoff_t *start, unsigned int nr_pages, | |
360 | struct page **pages) | |
361 | { | |
362 | return find_get_pages_range(mapping, start, (pgoff_t)-1, nr_pages, | |
363 | pages); | |
364 | } | |
ebf43500 JA |
365 | unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t start, |
366 | unsigned int nr_pages, struct page **pages); | |
72b045ae | 367 | unsigned find_get_pages_range_tag(struct address_space *mapping, pgoff_t *index, |
a6906972 | 368 | pgoff_t end, xa_mark_t tag, unsigned int nr_pages, |
72b045ae JK |
369 | struct page **pages); |
370 | static inline unsigned find_get_pages_tag(struct address_space *mapping, | |
a6906972 | 371 | pgoff_t *index, xa_mark_t tag, unsigned int nr_pages, |
72b045ae JK |
372 | struct page **pages) |
373 | { | |
374 | return find_get_pages_range_tag(mapping, index, (pgoff_t)-1, tag, | |
375 | nr_pages, pages); | |
376 | } | |
1da177e4 | 377 | |
54566b2c NP |
378 | struct page *grab_cache_page_write_begin(struct address_space *mapping, |
379 | pgoff_t index, unsigned flags); | |
afddba49 | 380 | |
1da177e4 LT |
381 | /* |
382 | * Returns locked page at given index in given cache, creating it if needed. | |
383 | */ | |
57f6b96c FW |
384 | static inline struct page *grab_cache_page(struct address_space *mapping, |
385 | pgoff_t index) | |
1da177e4 LT |
386 | { |
387 | return find_or_create_page(mapping, index, mapping_gfp_mask(mapping)); | |
388 | } | |
389 | ||
1da177e4 | 390 | extern struct page * read_cache_page(struct address_space *mapping, |
5e5358e7 | 391 | pgoff_t index, filler_t *filler, void *data); |
0531b2aa LT |
392 | extern struct page * read_cache_page_gfp(struct address_space *mapping, |
393 | pgoff_t index, gfp_t gfp_mask); | |
1da177e4 LT |
394 | extern int read_cache_pages(struct address_space *mapping, |
395 | struct list_head *pages, filler_t *filler, void *data); | |
396 | ||
090d2b18 | 397 | static inline struct page *read_mapping_page(struct address_space *mapping, |
5e5358e7 | 398 | pgoff_t index, void *data) |
090d2b18 | 399 | { |
6c45b454 | 400 | return read_cache_page(mapping, index, NULL, data); |
090d2b18 PE |
401 | } |
402 | ||
a0f7a756 | 403 | /* |
5cbc198a KS |
404 | * Get index of the page with in radix-tree |
405 | * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE) | |
a0f7a756 | 406 | */ |
5cbc198a | 407 | static inline pgoff_t page_to_index(struct page *page) |
a0f7a756 | 408 | { |
e9b61f19 KS |
409 | pgoff_t pgoff; |
410 | ||
e9b61f19 | 411 | if (likely(!PageTransTail(page))) |
09cbfeaf | 412 | return page->index; |
e9b61f19 KS |
413 | |
414 | /* | |
415 | * We don't initialize ->index for tail pages: calculate based on | |
416 | * head page | |
417 | */ | |
09cbfeaf | 418 | pgoff = compound_head(page)->index; |
e9b61f19 KS |
419 | pgoff += page - compound_head(page); |
420 | return pgoff; | |
a0f7a756 NH |
421 | } |
422 | ||
5cbc198a KS |
423 | /* |
424 | * Get the offset in PAGE_SIZE. | |
425 | * (TODO: hugepage should have ->index in PAGE_SIZE) | |
426 | */ | |
427 | static inline pgoff_t page_to_pgoff(struct page *page) | |
428 | { | |
429 | if (unlikely(PageHeadHuge(page))) | |
430 | return page->index << compound_order(page); | |
431 | ||
432 | return page_to_index(page); | |
433 | } | |
434 | ||
1da177e4 LT |
435 | /* |
436 | * Return byte-offset into filesystem object for page. | |
437 | */ | |
438 | static inline loff_t page_offset(struct page *page) | |
439 | { | |
09cbfeaf | 440 | return ((loff_t)page->index) << PAGE_SHIFT; |
1da177e4 LT |
441 | } |
442 | ||
f981c595 MG |
443 | static inline loff_t page_file_offset(struct page *page) |
444 | { | |
8cd79788 | 445 | return ((loff_t)page_index(page)) << PAGE_SHIFT; |
f981c595 MG |
446 | } |
447 | ||
0fe6e20b NH |
448 | extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma, |
449 | unsigned long address); | |
450 | ||
1da177e4 LT |
451 | static inline pgoff_t linear_page_index(struct vm_area_struct *vma, |
452 | unsigned long address) | |
453 | { | |
0fe6e20b NH |
454 | pgoff_t pgoff; |
455 | if (unlikely(is_vm_hugetlb_page(vma))) | |
456 | return linear_hugepage_index(vma, address); | |
457 | pgoff = (address - vma->vm_start) >> PAGE_SHIFT; | |
1da177e4 | 458 | pgoff += vma->vm_pgoff; |
09cbfeaf | 459 | return pgoff; |
1da177e4 LT |
460 | } |
461 | ||
b3c97528 HH |
462 | extern void __lock_page(struct page *page); |
463 | extern int __lock_page_killable(struct page *page); | |
d065bd81 ML |
464 | extern int __lock_page_or_retry(struct page *page, struct mm_struct *mm, |
465 | unsigned int flags); | |
b3c97528 | 466 | extern void unlock_page(struct page *page); |
1da177e4 | 467 | |
f4458845 AM |
468 | /* |
469 | * Return true if the page was successfully locked | |
470 | */ | |
529ae9aa NP |
471 | static inline int trylock_page(struct page *page) |
472 | { | |
48c935ad | 473 | page = compound_head(page); |
8413ac9d | 474 | return (likely(!test_and_set_bit_lock(PG_locked, &page->flags))); |
529ae9aa NP |
475 | } |
476 | ||
db37648c NP |
477 | /* |
478 | * lock_page may only be called if we have the page's inode pinned. | |
479 | */ | |
1da177e4 LT |
480 | static inline void lock_page(struct page *page) |
481 | { | |
482 | might_sleep(); | |
529ae9aa | 483 | if (!trylock_page(page)) |
1da177e4 LT |
484 | __lock_page(page); |
485 | } | |
db37648c | 486 | |
2687a356 MW |
487 | /* |
488 | * lock_page_killable is like lock_page but can be interrupted by fatal | |
489 | * signals. It returns 0 if it locked the page and -EINTR if it was | |
490 | * killed while waiting. | |
491 | */ | |
492 | static inline int lock_page_killable(struct page *page) | |
493 | { | |
494 | might_sleep(); | |
529ae9aa | 495 | if (!trylock_page(page)) |
2687a356 MW |
496 | return __lock_page_killable(page); |
497 | return 0; | |
498 | } | |
499 | ||
d065bd81 ML |
500 | /* |
501 | * lock_page_or_retry - Lock the page, unless this would block and the | |
502 | * caller indicated that it can handle a retry. | |
9a95f3cf PC |
503 | * |
504 | * Return value and mmap_sem implications depend on flags; see | |
505 | * __lock_page_or_retry(). | |
d065bd81 ML |
506 | */ |
507 | static inline int lock_page_or_retry(struct page *page, struct mm_struct *mm, | |
508 | unsigned int flags) | |
509 | { | |
510 | might_sleep(); | |
511 | return trylock_page(page) || __lock_page_or_retry(page, mm, flags); | |
512 | } | |
513 | ||
1da177e4 | 514 | /* |
74d81bfa NP |
515 | * This is exported only for wait_on_page_locked/wait_on_page_writeback, etc., |
516 | * and should not be used directly. | |
1da177e4 | 517 | */ |
b3c97528 | 518 | extern void wait_on_page_bit(struct page *page, int bit_nr); |
f62e00cc | 519 | extern int wait_on_page_bit_killable(struct page *page, int bit_nr); |
a4796e37 | 520 | |
1da177e4 LT |
521 | /* |
522 | * Wait for a page to be unlocked. | |
523 | * | |
524 | * This must be called with the caller "holding" the page, | |
525 | * ie with increased "page->count" so that the page won't | |
526 | * go away during the wait.. | |
527 | */ | |
528 | static inline void wait_on_page_locked(struct page *page) | |
529 | { | |
530 | if (PageLocked(page)) | |
48c935ad | 531 | wait_on_page_bit(compound_head(page), PG_locked); |
1da177e4 LT |
532 | } |
533 | ||
62906027 NP |
534 | static inline int wait_on_page_locked_killable(struct page *page) |
535 | { | |
536 | if (!PageLocked(page)) | |
537 | return 0; | |
538 | return wait_on_page_bit_killable(compound_head(page), PG_locked); | |
539 | } | |
540 | ||
9a1ea439 HD |
541 | extern void put_and_wait_on_page_locked(struct page *page); |
542 | ||
19343b5b | 543 | void wait_on_page_writeback(struct page *page); |
1da177e4 | 544 | extern void end_page_writeback(struct page *page); |
1d1d1a76 | 545 | void wait_for_stable_page(struct page *page); |
1da177e4 | 546 | |
c11f0c0b | 547 | void page_endio(struct page *page, bool is_write, int err); |
57d99845 | 548 | |
385e1ca5 DH |
549 | /* |
550 | * Add an arbitrary waiter to a page's wait queue | |
551 | */ | |
ac6424b9 | 552 | extern void add_page_wait_queue(struct page *page, wait_queue_entry_t *waiter); |
385e1ca5 | 553 | |
1da177e4 | 554 | /* |
4bce9f6e | 555 | * Fault everything in given userspace address range in. |
1da177e4 LT |
556 | */ |
557 | static inline int fault_in_pages_writeable(char __user *uaddr, int size) | |
f56f821f | 558 | { |
9923777d | 559 | char __user *end = uaddr + size - 1; |
f56f821f DV |
560 | |
561 | if (unlikely(size == 0)) | |
e23d4159 | 562 | return 0; |
f56f821f | 563 | |
e23d4159 AV |
564 | if (unlikely(uaddr > end)) |
565 | return -EFAULT; | |
f56f821f DV |
566 | /* |
567 | * Writing zeroes into userspace here is OK, because we know that if | |
568 | * the zero gets there, we'll be overwriting it. | |
569 | */ | |
e23d4159 AV |
570 | do { |
571 | if (unlikely(__put_user(0, uaddr) != 0)) | |
572 | return -EFAULT; | |
f56f821f | 573 | uaddr += PAGE_SIZE; |
e23d4159 | 574 | } while (uaddr <= end); |
f56f821f DV |
575 | |
576 | /* Check whether the range spilled into the next page. */ | |
577 | if (((unsigned long)uaddr & PAGE_MASK) == | |
578 | ((unsigned long)end & PAGE_MASK)) | |
e23d4159 | 579 | return __put_user(0, end); |
f56f821f | 580 | |
e23d4159 | 581 | return 0; |
f56f821f DV |
582 | } |
583 | ||
4bce9f6e | 584 | static inline int fault_in_pages_readable(const char __user *uaddr, int size) |
f56f821f DV |
585 | { |
586 | volatile char c; | |
f56f821f DV |
587 | const char __user *end = uaddr + size - 1; |
588 | ||
589 | if (unlikely(size == 0)) | |
e23d4159 | 590 | return 0; |
f56f821f | 591 | |
e23d4159 AV |
592 | if (unlikely(uaddr > end)) |
593 | return -EFAULT; | |
594 | ||
595 | do { | |
596 | if (unlikely(__get_user(c, uaddr) != 0)) | |
597 | return -EFAULT; | |
f56f821f | 598 | uaddr += PAGE_SIZE; |
e23d4159 | 599 | } while (uaddr <= end); |
f56f821f DV |
600 | |
601 | /* Check whether the range spilled into the next page. */ | |
602 | if (((unsigned long)uaddr & PAGE_MASK) == | |
603 | ((unsigned long)end & PAGE_MASK)) { | |
e23d4159 | 604 | return __get_user(c, end); |
f56f821f DV |
605 | } |
606 | ||
90b75db6 | 607 | (void)c; |
e23d4159 | 608 | return 0; |
f56f821f DV |
609 | } |
610 | ||
529ae9aa NP |
611 | int add_to_page_cache_locked(struct page *page, struct address_space *mapping, |
612 | pgoff_t index, gfp_t gfp_mask); | |
613 | int add_to_page_cache_lru(struct page *page, struct address_space *mapping, | |
614 | pgoff_t index, gfp_t gfp_mask); | |
97cecb5a | 615 | extern void delete_from_page_cache(struct page *page); |
62cccb8c | 616 | extern void __delete_from_page_cache(struct page *page, void *shadow); |
ef6a3c63 | 617 | int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask); |
aa65c29c JK |
618 | void delete_from_page_cache_batch(struct address_space *mapping, |
619 | struct pagevec *pvec); | |
529ae9aa | 620 | |
cee9a0c4 MWO |
621 | #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE) |
622 | ||
623 | void page_cache_sync_readahead(struct address_space *, struct file_ra_state *, | |
624 | struct file *, pgoff_t index, unsigned long req_count); | |
625 | void page_cache_async_readahead(struct address_space *, struct file_ra_state *, | |
626 | struct file *, struct page *, pgoff_t index, | |
627 | unsigned long req_count); | |
628 | ||
529ae9aa NP |
629 | /* |
630 | * Like add_to_page_cache_locked, but used to add newly allocated pages: | |
48c935ad | 631 | * the page is new, so we can just run __SetPageLocked() against it. |
529ae9aa NP |
632 | */ |
633 | static inline int add_to_page_cache(struct page *page, | |
634 | struct address_space *mapping, pgoff_t offset, gfp_t gfp_mask) | |
635 | { | |
636 | int error; | |
637 | ||
48c935ad | 638 | __SetPageLocked(page); |
529ae9aa NP |
639 | error = add_to_page_cache_locked(page, mapping, offset, gfp_mask); |
640 | if (unlikely(error)) | |
48c935ad | 641 | __ClearPageLocked(page); |
529ae9aa NP |
642 | return error; |
643 | } | |
644 | ||
b57c2cb9 FF |
645 | static inline unsigned long dir_pages(struct inode *inode) |
646 | { | |
09cbfeaf KS |
647 | return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >> |
648 | PAGE_SHIFT; | |
b57c2cb9 FF |
649 | } |
650 | ||
243145bc AG |
651 | /** |
652 | * page_mkwrite_check_truncate - check if page was truncated | |
653 | * @page: the page to check | |
654 | * @inode: the inode to check the page against | |
655 | * | |
656 | * Returns the number of bytes in the page up to EOF, | |
657 | * or -EFAULT if the page was truncated. | |
658 | */ | |
659 | static inline int page_mkwrite_check_truncate(struct page *page, | |
660 | struct inode *inode) | |
661 | { | |
662 | loff_t size = i_size_read(inode); | |
663 | pgoff_t index = size >> PAGE_SHIFT; | |
664 | int offset = offset_in_page(size); | |
665 | ||
666 | if (page->mapping != inode->i_mapping) | |
667 | return -EFAULT; | |
668 | ||
669 | /* page is wholly inside EOF */ | |
670 | if (page->index < index) | |
671 | return PAGE_SIZE; | |
672 | /* page is wholly past EOF */ | |
673 | if (page->index > index || !offset) | |
674 | return -EFAULT; | |
675 | /* page is partially inside EOF */ | |
676 | return offset; | |
677 | } | |
678 | ||
1da177e4 | 679 | #endif /* _LINUX_PAGEMAP_H */ |