Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | #ifndef _LINUX_PAGEMAP_H |
2 | #define _LINUX_PAGEMAP_H | |
3 | ||
4 | /* | |
5 | * Copyright 1995 Linus Torvalds | |
6 | */ | |
7 | #include <linux/mm.h> | |
8 | #include <linux/fs.h> | |
9 | #include <linux/list.h> | |
10 | #include <linux/highmem.h> | |
11 | #include <linux/compiler.h> | |
7c0f6ba6 | 12 | #include <linux/uaccess.h> |
1da177e4 | 13 | #include <linux/gfp.h> |
3e9f45bd | 14 | #include <linux/bitops.h> |
e286781d | 15 | #include <linux/hardirq.h> /* for in_interrupt() */ |
8edf344c | 16 | #include <linux/hugetlb_inline.h> |
1da177e4 LT |
17 | |
18 | /* | |
9c5d760b | 19 | * Bits in mapping->flags. |
1da177e4 | 20 | */ |
9a896c9a | 21 | enum mapping_flags { |
9c5d760b MH |
22 | AS_EIO = 0, /* IO error on async write */ |
23 | AS_ENOSPC = 1, /* ENOSPC on async write */ | |
24 | AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */ | |
25 | AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */ | |
26 | AS_EXITING = 4, /* final truncate in progress */ | |
371a096e | 27 | /* writeback related tags are not used */ |
9c5d760b | 28 | AS_NO_WRITEBACK_TAGS = 5, |
9a896c9a | 29 | }; |
1da177e4 | 30 | |
8ed1e46a JL |
31 | /** |
32 | * mapping_set_error - record a writeback error in the address_space | |
33 | * @mapping - the mapping in which an error should be set | |
34 | * @error - the error to set in the mapping | |
35 | * | |
36 | * When writeback fails in some way, we must record that error so that | |
37 | * userspace can be informed when fsync and the like are called. We endeavor | |
38 | * to report errors on any file that was open at the time of the error. Some | |
39 | * internal callers also need to know when writeback errors have occurred. | |
40 | * | |
41 | * When a writeback error occurs, most filesystems will want to call | |
42 | * mapping_set_error to record the error in the mapping so that it can be | |
43 | * reported when the application calls fsync(2). | |
44 | */ | |
3e9f45bd GC |
45 | static inline void mapping_set_error(struct address_space *mapping, int error) |
46 | { | |
8ed1e46a JL |
47 | if (likely(!error)) |
48 | return; | |
49 | ||
50 | /* Record in wb_err for checkers using errseq_t based tracking */ | |
51 | filemap_set_wb_err(mapping, error); | |
52 | ||
53 | /* Record it in flags for now, for legacy callers */ | |
54 | if (error == -ENOSPC) | |
55 | set_bit(AS_ENOSPC, &mapping->flags); | |
56 | else | |
57 | set_bit(AS_EIO, &mapping->flags); | |
3e9f45bd GC |
58 | } |
59 | ||
ba9ddf49 LS |
60 | static inline void mapping_set_unevictable(struct address_space *mapping) |
61 | { | |
62 | set_bit(AS_UNEVICTABLE, &mapping->flags); | |
63 | } | |
64 | ||
89e004ea LS |
65 | static inline void mapping_clear_unevictable(struct address_space *mapping) |
66 | { | |
67 | clear_bit(AS_UNEVICTABLE, &mapping->flags); | |
68 | } | |
69 | ||
ba9ddf49 LS |
70 | static inline int mapping_unevictable(struct address_space *mapping) |
71 | { | |
088e5465 | 72 | if (mapping) |
89e004ea LS |
73 | return test_bit(AS_UNEVICTABLE, &mapping->flags); |
74 | return !!mapping; | |
ba9ddf49 | 75 | } |
ba9ddf49 | 76 | |
91b0abe3 JW |
77 | static inline void mapping_set_exiting(struct address_space *mapping) |
78 | { | |
79 | set_bit(AS_EXITING, &mapping->flags); | |
80 | } | |
81 | ||
82 | static inline int mapping_exiting(struct address_space *mapping) | |
83 | { | |
84 | return test_bit(AS_EXITING, &mapping->flags); | |
85 | } | |
86 | ||
371a096e HY |
87 | static inline void mapping_set_no_writeback_tags(struct address_space *mapping) |
88 | { | |
89 | set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags); | |
90 | } | |
91 | ||
92 | static inline int mapping_use_writeback_tags(struct address_space *mapping) | |
93 | { | |
94 | return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags); | |
95 | } | |
96 | ||
dd0fc66f | 97 | static inline gfp_t mapping_gfp_mask(struct address_space * mapping) |
1da177e4 | 98 | { |
9c5d760b | 99 | return mapping->gfp_mask; |
1da177e4 LT |
100 | } |
101 | ||
c62d2555 MH |
102 | /* Restricts the given gfp_mask to what the mapping allows. */ |
103 | static inline gfp_t mapping_gfp_constraint(struct address_space *mapping, | |
104 | gfp_t gfp_mask) | |
105 | { | |
106 | return mapping_gfp_mask(mapping) & gfp_mask; | |
107 | } | |
108 | ||
1da177e4 LT |
109 | /* |
110 | * This is non-atomic. Only to be used before the mapping is activated. | |
111 | * Probably needs a barrier... | |
112 | */ | |
260b2367 | 113 | static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask) |
1da177e4 | 114 | { |
9c5d760b | 115 | m->gfp_mask = mask; |
1da177e4 LT |
116 | } |
117 | ||
b745bc85 | 118 | void release_pages(struct page **pages, int nr, bool cold); |
1da177e4 | 119 | |
e286781d NP |
120 | /* |
121 | * speculatively take a reference to a page. | |
0139aa7b JK |
122 | * If the page is free (_refcount == 0), then _refcount is untouched, and 0 |
123 | * is returned. Otherwise, _refcount is incremented by 1 and 1 is returned. | |
e286781d NP |
124 | * |
125 | * This function must be called inside the same rcu_read_lock() section as has | |
126 | * been used to lookup the page in the pagecache radix-tree (or page table): | |
0139aa7b | 127 | * this allows allocators to use a synchronize_rcu() to stabilize _refcount. |
e286781d NP |
128 | * |
129 | * Unless an RCU grace period has passed, the count of all pages coming out | |
130 | * of the allocator must be considered unstable. page_count may return higher | |
131 | * than expected, and put_page must be able to do the right thing when the | |
132 | * page has been finished with, no matter what it is subsequently allocated | |
133 | * for (because put_page is what is used here to drop an invalid speculative | |
134 | * reference). | |
135 | * | |
136 | * This is the interesting part of the lockless pagecache (and lockless | |
137 | * get_user_pages) locking protocol, where the lookup-side (eg. find_get_page) | |
138 | * has the following pattern: | |
139 | * 1. find page in radix tree | |
140 | * 2. conditionally increment refcount | |
141 | * 3. check the page is still in pagecache (if no, goto 1) | |
142 | * | |
0139aa7b | 143 | * Remove-side that cares about stability of _refcount (eg. reclaim) has the |
e286781d NP |
144 | * following (with tree_lock held for write): |
145 | * A. atomically check refcount is correct and set it to 0 (atomic_cmpxchg) | |
146 | * B. remove page from pagecache | |
147 | * C. free the page | |
148 | * | |
149 | * There are 2 critical interleavings that matter: | |
150 | * - 2 runs before A: in this case, A sees elevated refcount and bails out | |
151 | * - A runs before 2: in this case, 2 sees zero refcount and retries; | |
152 | * subsequently, B will complete and 1 will find no page, causing the | |
153 | * lookup to return NULL. | |
154 | * | |
155 | * It is possible that between 1 and 2, the page is removed then the exact same | |
156 | * page is inserted into the same position in pagecache. That's OK: the | |
157 | * old find_get_page using tree_lock could equally have run before or after | |
158 | * such a re-insertion, depending on order that locks are granted. | |
159 | * | |
160 | * Lookups racing against pagecache insertion isn't a big problem: either 1 | |
161 | * will find the page or it will not. Likewise, the old find_get_page could run | |
162 | * either before the insertion or afterwards, depending on timing. | |
163 | */ | |
164 | static inline int page_cache_get_speculative(struct page *page) | |
165 | { | |
166 | VM_BUG_ON(in_interrupt()); | |
167 | ||
8375ad98 | 168 | #ifdef CONFIG_TINY_RCU |
bdd4e85d | 169 | # ifdef CONFIG_PREEMPT_COUNT |
591a3d7c | 170 | VM_BUG_ON(!in_atomic() && !irqs_disabled()); |
e286781d NP |
171 | # endif |
172 | /* | |
173 | * Preempt must be disabled here - we rely on rcu_read_lock doing | |
174 | * this for us. | |
175 | * | |
176 | * Pagecache won't be truncated from interrupt context, so if we have | |
177 | * found a page in the radix tree here, we have pinned its refcount by | |
178 | * disabling preempt, and hence no need for the "speculative get" that | |
179 | * SMP requires. | |
180 | */ | |
309381fe | 181 | VM_BUG_ON_PAGE(page_count(page) == 0, page); |
fe896d18 | 182 | page_ref_inc(page); |
e286781d NP |
183 | |
184 | #else | |
185 | if (unlikely(!get_page_unless_zero(page))) { | |
186 | /* | |
187 | * Either the page has been freed, or will be freed. | |
188 | * In either case, retry here and the caller should | |
189 | * do the right thing (see comments above). | |
190 | */ | |
191 | return 0; | |
192 | } | |
193 | #endif | |
309381fe | 194 | VM_BUG_ON_PAGE(PageTail(page), page); |
e286781d NP |
195 | |
196 | return 1; | |
197 | } | |
198 | ||
ce0ad7f0 NP |
199 | /* |
200 | * Same as above, but add instead of inc (could just be merged) | |
201 | */ | |
202 | static inline int page_cache_add_speculative(struct page *page, int count) | |
203 | { | |
204 | VM_BUG_ON(in_interrupt()); | |
205 | ||
b560d8ad | 206 | #if !defined(CONFIG_SMP) && defined(CONFIG_TREE_RCU) |
bdd4e85d | 207 | # ifdef CONFIG_PREEMPT_COUNT |
591a3d7c | 208 | VM_BUG_ON(!in_atomic() && !irqs_disabled()); |
ce0ad7f0 | 209 | # endif |
309381fe | 210 | VM_BUG_ON_PAGE(page_count(page) == 0, page); |
fe896d18 | 211 | page_ref_add(page, count); |
ce0ad7f0 NP |
212 | |
213 | #else | |
fe896d18 | 214 | if (unlikely(!page_ref_add_unless(page, count, 0))) |
ce0ad7f0 NP |
215 | return 0; |
216 | #endif | |
309381fe | 217 | VM_BUG_ON_PAGE(PageCompound(page) && page != compound_head(page), page); |
ce0ad7f0 NP |
218 | |
219 | return 1; | |
220 | } | |
221 | ||
44110fe3 | 222 | #ifdef CONFIG_NUMA |
2ae88149 | 223 | extern struct page *__page_cache_alloc(gfp_t gfp); |
44110fe3 | 224 | #else |
2ae88149 NP |
225 | static inline struct page *__page_cache_alloc(gfp_t gfp) |
226 | { | |
227 | return alloc_pages(gfp, 0); | |
228 | } | |
229 | #endif | |
230 | ||
1da177e4 LT |
231 | static inline struct page *page_cache_alloc(struct address_space *x) |
232 | { | |
2ae88149 | 233 | return __page_cache_alloc(mapping_gfp_mask(x)); |
1da177e4 LT |
234 | } |
235 | ||
236 | static inline struct page *page_cache_alloc_cold(struct address_space *x) | |
237 | { | |
2ae88149 | 238 | return __page_cache_alloc(mapping_gfp_mask(x)|__GFP_COLD); |
1da177e4 LT |
239 | } |
240 | ||
8a5c743e | 241 | static inline gfp_t readahead_gfp_mask(struct address_space *x) |
7b1de586 | 242 | { |
8a5c743e MH |
243 | return mapping_gfp_mask(x) | |
244 | __GFP_COLD | __GFP_NORETRY | __GFP_NOWARN; | |
7b1de586 WF |
245 | } |
246 | ||
1da177e4 LT |
247 | typedef int filler_t(void *, struct page *); |
248 | ||
e7b563bb JW |
249 | pgoff_t page_cache_next_hole(struct address_space *mapping, |
250 | pgoff_t index, unsigned long max_scan); | |
251 | pgoff_t page_cache_prev_hole(struct address_space *mapping, | |
252 | pgoff_t index, unsigned long max_scan); | |
253 | ||
2457aec6 MG |
254 | #define FGP_ACCESSED 0x00000001 |
255 | #define FGP_LOCK 0x00000002 | |
256 | #define FGP_CREAT 0x00000004 | |
257 | #define FGP_WRITE 0x00000008 | |
258 | #define FGP_NOFS 0x00000010 | |
259 | #define FGP_NOWAIT 0x00000020 | |
260 | ||
261 | struct page *pagecache_get_page(struct address_space *mapping, pgoff_t offset, | |
45f87de5 | 262 | int fgp_flags, gfp_t cache_gfp_mask); |
2457aec6 MG |
263 | |
264 | /** | |
265 | * find_get_page - find and get a page reference | |
266 | * @mapping: the address_space to search | |
267 | * @offset: the page index | |
268 | * | |
269 | * Looks up the page cache slot at @mapping & @offset. If there is a | |
270 | * page cache page, it is returned with an increased refcount. | |
271 | * | |
272 | * Otherwise, %NULL is returned. | |
273 | */ | |
274 | static inline struct page *find_get_page(struct address_space *mapping, | |
275 | pgoff_t offset) | |
276 | { | |
45f87de5 | 277 | return pagecache_get_page(mapping, offset, 0, 0); |
2457aec6 MG |
278 | } |
279 | ||
280 | static inline struct page *find_get_page_flags(struct address_space *mapping, | |
281 | pgoff_t offset, int fgp_flags) | |
282 | { | |
45f87de5 | 283 | return pagecache_get_page(mapping, offset, fgp_flags, 0); |
2457aec6 MG |
284 | } |
285 | ||
286 | /** | |
287 | * find_lock_page - locate, pin and lock a pagecache page | |
2457aec6 MG |
288 | * @mapping: the address_space to search |
289 | * @offset: the page index | |
290 | * | |
291 | * Looks up the page cache slot at @mapping & @offset. If there is a | |
292 | * page cache page, it is returned locked and with an increased | |
293 | * refcount. | |
294 | * | |
295 | * Otherwise, %NULL is returned. | |
296 | * | |
297 | * find_lock_page() may sleep. | |
298 | */ | |
299 | static inline struct page *find_lock_page(struct address_space *mapping, | |
300 | pgoff_t offset) | |
301 | { | |
45f87de5 | 302 | return pagecache_get_page(mapping, offset, FGP_LOCK, 0); |
2457aec6 MG |
303 | } |
304 | ||
305 | /** | |
306 | * find_or_create_page - locate or add a pagecache page | |
307 | * @mapping: the page's address_space | |
308 | * @index: the page's index into the mapping | |
309 | * @gfp_mask: page allocation mode | |
310 | * | |
311 | * Looks up the page cache slot at @mapping & @offset. If there is a | |
312 | * page cache page, it is returned locked and with an increased | |
313 | * refcount. | |
314 | * | |
315 | * If the page is not present, a new page is allocated using @gfp_mask | |
316 | * and added to the page cache and the VM's LRU list. The page is | |
317 | * returned locked and with an increased refcount. | |
318 | * | |
319 | * On memory exhaustion, %NULL is returned. | |
320 | * | |
321 | * find_or_create_page() may sleep, even if @gfp_flags specifies an | |
322 | * atomic allocation! | |
323 | */ | |
324 | static inline struct page *find_or_create_page(struct address_space *mapping, | |
325 | pgoff_t offset, gfp_t gfp_mask) | |
326 | { | |
327 | return pagecache_get_page(mapping, offset, | |
328 | FGP_LOCK|FGP_ACCESSED|FGP_CREAT, | |
45f87de5 | 329 | gfp_mask); |
2457aec6 MG |
330 | } |
331 | ||
332 | /** | |
333 | * grab_cache_page_nowait - returns locked page at given index in given cache | |
334 | * @mapping: target address_space | |
335 | * @index: the page index | |
336 | * | |
337 | * Same as grab_cache_page(), but do not wait if the page is unavailable. | |
338 | * This is intended for speculative data generators, where the data can | |
339 | * be regenerated if the page couldn't be grabbed. This routine should | |
340 | * be safe to call while holding the lock for another page. | |
341 | * | |
342 | * Clear __GFP_FS when allocating the page to avoid recursion into the fs | |
343 | * and deadlock against the caller's locked page. | |
344 | */ | |
345 | static inline struct page *grab_cache_page_nowait(struct address_space *mapping, | |
346 | pgoff_t index) | |
347 | { | |
348 | return pagecache_get_page(mapping, index, | |
349 | FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT, | |
45f87de5 | 350 | mapping_gfp_mask(mapping)); |
2457aec6 MG |
351 | } |
352 | ||
0cd6144a | 353 | struct page *find_get_entry(struct address_space *mapping, pgoff_t offset); |
0cd6144a | 354 | struct page *find_lock_entry(struct address_space *mapping, pgoff_t offset); |
0cd6144a JW |
355 | unsigned find_get_entries(struct address_space *mapping, pgoff_t start, |
356 | unsigned int nr_entries, struct page **entries, | |
357 | pgoff_t *indices); | |
1da177e4 LT |
358 | unsigned find_get_pages(struct address_space *mapping, pgoff_t start, |
359 | unsigned int nr_pages, struct page **pages); | |
ebf43500 JA |
360 | unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t start, |
361 | unsigned int nr_pages, struct page **pages); | |
1da177e4 LT |
362 | unsigned find_get_pages_tag(struct address_space *mapping, pgoff_t *index, |
363 | int tag, unsigned int nr_pages, struct page **pages); | |
7e7f7749 RZ |
364 | unsigned find_get_entries_tag(struct address_space *mapping, pgoff_t start, |
365 | int tag, unsigned int nr_entries, | |
366 | struct page **entries, pgoff_t *indices); | |
1da177e4 | 367 | |
54566b2c NP |
368 | struct page *grab_cache_page_write_begin(struct address_space *mapping, |
369 | pgoff_t index, unsigned flags); | |
afddba49 | 370 | |
1da177e4 LT |
371 | /* |
372 | * Returns locked page at given index in given cache, creating it if needed. | |
373 | */ | |
57f6b96c FW |
374 | static inline struct page *grab_cache_page(struct address_space *mapping, |
375 | pgoff_t index) | |
1da177e4 LT |
376 | { |
377 | return find_or_create_page(mapping, index, mapping_gfp_mask(mapping)); | |
378 | } | |
379 | ||
1da177e4 | 380 | extern struct page * read_cache_page(struct address_space *mapping, |
5e5358e7 | 381 | pgoff_t index, filler_t *filler, void *data); |
0531b2aa LT |
382 | extern struct page * read_cache_page_gfp(struct address_space *mapping, |
383 | pgoff_t index, gfp_t gfp_mask); | |
1da177e4 LT |
384 | extern int read_cache_pages(struct address_space *mapping, |
385 | struct list_head *pages, filler_t *filler, void *data); | |
386 | ||
090d2b18 | 387 | static inline struct page *read_mapping_page(struct address_space *mapping, |
5e5358e7 | 388 | pgoff_t index, void *data) |
090d2b18 PE |
389 | { |
390 | filler_t *filler = (filler_t *)mapping->a_ops->readpage; | |
391 | return read_cache_page(mapping, index, filler, data); | |
392 | } | |
393 | ||
a0f7a756 | 394 | /* |
5cbc198a KS |
395 | * Get index of the page with in radix-tree |
396 | * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE) | |
a0f7a756 | 397 | */ |
5cbc198a | 398 | static inline pgoff_t page_to_index(struct page *page) |
a0f7a756 | 399 | { |
e9b61f19 KS |
400 | pgoff_t pgoff; |
401 | ||
e9b61f19 | 402 | if (likely(!PageTransTail(page))) |
09cbfeaf | 403 | return page->index; |
e9b61f19 KS |
404 | |
405 | /* | |
406 | * We don't initialize ->index for tail pages: calculate based on | |
407 | * head page | |
408 | */ | |
09cbfeaf | 409 | pgoff = compound_head(page)->index; |
e9b61f19 KS |
410 | pgoff += page - compound_head(page); |
411 | return pgoff; | |
a0f7a756 NH |
412 | } |
413 | ||
5cbc198a KS |
414 | /* |
415 | * Get the offset in PAGE_SIZE. | |
416 | * (TODO: hugepage should have ->index in PAGE_SIZE) | |
417 | */ | |
418 | static inline pgoff_t page_to_pgoff(struct page *page) | |
419 | { | |
420 | if (unlikely(PageHeadHuge(page))) | |
421 | return page->index << compound_order(page); | |
422 | ||
423 | return page_to_index(page); | |
424 | } | |
425 | ||
1da177e4 LT |
426 | /* |
427 | * Return byte-offset into filesystem object for page. | |
428 | */ | |
429 | static inline loff_t page_offset(struct page *page) | |
430 | { | |
09cbfeaf | 431 | return ((loff_t)page->index) << PAGE_SHIFT; |
1da177e4 LT |
432 | } |
433 | ||
f981c595 MG |
434 | static inline loff_t page_file_offset(struct page *page) |
435 | { | |
8cd79788 | 436 | return ((loff_t)page_index(page)) << PAGE_SHIFT; |
f981c595 MG |
437 | } |
438 | ||
0fe6e20b NH |
439 | extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma, |
440 | unsigned long address); | |
441 | ||
1da177e4 LT |
442 | static inline pgoff_t linear_page_index(struct vm_area_struct *vma, |
443 | unsigned long address) | |
444 | { | |
0fe6e20b NH |
445 | pgoff_t pgoff; |
446 | if (unlikely(is_vm_hugetlb_page(vma))) | |
447 | return linear_hugepage_index(vma, address); | |
448 | pgoff = (address - vma->vm_start) >> PAGE_SHIFT; | |
1da177e4 | 449 | pgoff += vma->vm_pgoff; |
09cbfeaf | 450 | return pgoff; |
1da177e4 LT |
451 | } |
452 | ||
b3c97528 HH |
453 | extern void __lock_page(struct page *page); |
454 | extern int __lock_page_killable(struct page *page); | |
d065bd81 ML |
455 | extern int __lock_page_or_retry(struct page *page, struct mm_struct *mm, |
456 | unsigned int flags); | |
b3c97528 | 457 | extern void unlock_page(struct page *page); |
1da177e4 | 458 | |
529ae9aa NP |
459 | static inline int trylock_page(struct page *page) |
460 | { | |
48c935ad | 461 | page = compound_head(page); |
8413ac9d | 462 | return (likely(!test_and_set_bit_lock(PG_locked, &page->flags))); |
529ae9aa NP |
463 | } |
464 | ||
db37648c NP |
465 | /* |
466 | * lock_page may only be called if we have the page's inode pinned. | |
467 | */ | |
1da177e4 LT |
468 | static inline void lock_page(struct page *page) |
469 | { | |
470 | might_sleep(); | |
529ae9aa | 471 | if (!trylock_page(page)) |
1da177e4 LT |
472 | __lock_page(page); |
473 | } | |
db37648c | 474 | |
2687a356 MW |
475 | /* |
476 | * lock_page_killable is like lock_page but can be interrupted by fatal | |
477 | * signals. It returns 0 if it locked the page and -EINTR if it was | |
478 | * killed while waiting. | |
479 | */ | |
480 | static inline int lock_page_killable(struct page *page) | |
481 | { | |
482 | might_sleep(); | |
529ae9aa | 483 | if (!trylock_page(page)) |
2687a356 MW |
484 | return __lock_page_killable(page); |
485 | return 0; | |
486 | } | |
487 | ||
d065bd81 ML |
488 | /* |
489 | * lock_page_or_retry - Lock the page, unless this would block and the | |
490 | * caller indicated that it can handle a retry. | |
9a95f3cf PC |
491 | * |
492 | * Return value and mmap_sem implications depend on flags; see | |
493 | * __lock_page_or_retry(). | |
d065bd81 ML |
494 | */ |
495 | static inline int lock_page_or_retry(struct page *page, struct mm_struct *mm, | |
496 | unsigned int flags) | |
497 | { | |
498 | might_sleep(); | |
499 | return trylock_page(page) || __lock_page_or_retry(page, mm, flags); | |
500 | } | |
501 | ||
1da177e4 | 502 | /* |
74d81bfa NP |
503 | * This is exported only for wait_on_page_locked/wait_on_page_writeback, etc., |
504 | * and should not be used directly. | |
1da177e4 | 505 | */ |
b3c97528 | 506 | extern void wait_on_page_bit(struct page *page, int bit_nr); |
f62e00cc | 507 | extern int wait_on_page_bit_killable(struct page *page, int bit_nr); |
a4796e37 | 508 | |
1da177e4 LT |
509 | /* |
510 | * Wait for a page to be unlocked. | |
511 | * | |
512 | * This must be called with the caller "holding" the page, | |
513 | * ie with increased "page->count" so that the page won't | |
514 | * go away during the wait.. | |
515 | */ | |
516 | static inline void wait_on_page_locked(struct page *page) | |
517 | { | |
518 | if (PageLocked(page)) | |
48c935ad | 519 | wait_on_page_bit(compound_head(page), PG_locked); |
1da177e4 LT |
520 | } |
521 | ||
62906027 NP |
522 | static inline int wait_on_page_locked_killable(struct page *page) |
523 | { | |
524 | if (!PageLocked(page)) | |
525 | return 0; | |
526 | return wait_on_page_bit_killable(compound_head(page), PG_locked); | |
527 | } | |
528 | ||
1da177e4 LT |
529 | /* |
530 | * Wait for a page to complete writeback | |
531 | */ | |
532 | static inline void wait_on_page_writeback(struct page *page) | |
533 | { | |
534 | if (PageWriteback(page)) | |
535 | wait_on_page_bit(page, PG_writeback); | |
536 | } | |
537 | ||
538 | extern void end_page_writeback(struct page *page); | |
1d1d1a76 | 539 | void wait_for_stable_page(struct page *page); |
1da177e4 | 540 | |
c11f0c0b | 541 | void page_endio(struct page *page, bool is_write, int err); |
57d99845 | 542 | |
385e1ca5 DH |
543 | /* |
544 | * Add an arbitrary waiter to a page's wait queue | |
545 | */ | |
ac6424b9 | 546 | extern void add_page_wait_queue(struct page *page, wait_queue_entry_t *waiter); |
385e1ca5 | 547 | |
1da177e4 | 548 | /* |
4bce9f6e | 549 | * Fault everything in given userspace address range in. |
1da177e4 LT |
550 | */ |
551 | static inline int fault_in_pages_writeable(char __user *uaddr, int size) | |
f56f821f | 552 | { |
9923777d | 553 | char __user *end = uaddr + size - 1; |
f56f821f DV |
554 | |
555 | if (unlikely(size == 0)) | |
e23d4159 | 556 | return 0; |
f56f821f | 557 | |
e23d4159 AV |
558 | if (unlikely(uaddr > end)) |
559 | return -EFAULT; | |
f56f821f DV |
560 | /* |
561 | * Writing zeroes into userspace here is OK, because we know that if | |
562 | * the zero gets there, we'll be overwriting it. | |
563 | */ | |
e23d4159 AV |
564 | do { |
565 | if (unlikely(__put_user(0, uaddr) != 0)) | |
566 | return -EFAULT; | |
f56f821f | 567 | uaddr += PAGE_SIZE; |
e23d4159 | 568 | } while (uaddr <= end); |
f56f821f DV |
569 | |
570 | /* Check whether the range spilled into the next page. */ | |
571 | if (((unsigned long)uaddr & PAGE_MASK) == | |
572 | ((unsigned long)end & PAGE_MASK)) | |
e23d4159 | 573 | return __put_user(0, end); |
f56f821f | 574 | |
e23d4159 | 575 | return 0; |
f56f821f DV |
576 | } |
577 | ||
4bce9f6e | 578 | static inline int fault_in_pages_readable(const char __user *uaddr, int size) |
f56f821f DV |
579 | { |
580 | volatile char c; | |
f56f821f DV |
581 | const char __user *end = uaddr + size - 1; |
582 | ||
583 | if (unlikely(size == 0)) | |
e23d4159 | 584 | return 0; |
f56f821f | 585 | |
e23d4159 AV |
586 | if (unlikely(uaddr > end)) |
587 | return -EFAULT; | |
588 | ||
589 | do { | |
590 | if (unlikely(__get_user(c, uaddr) != 0)) | |
591 | return -EFAULT; | |
f56f821f | 592 | uaddr += PAGE_SIZE; |
e23d4159 | 593 | } while (uaddr <= end); |
f56f821f DV |
594 | |
595 | /* Check whether the range spilled into the next page. */ | |
596 | if (((unsigned long)uaddr & PAGE_MASK) == | |
597 | ((unsigned long)end & PAGE_MASK)) { | |
e23d4159 | 598 | return __get_user(c, end); |
f56f821f DV |
599 | } |
600 | ||
90b75db6 | 601 | (void)c; |
e23d4159 | 602 | return 0; |
f56f821f DV |
603 | } |
604 | ||
529ae9aa NP |
605 | int add_to_page_cache_locked(struct page *page, struct address_space *mapping, |
606 | pgoff_t index, gfp_t gfp_mask); | |
607 | int add_to_page_cache_lru(struct page *page, struct address_space *mapping, | |
608 | pgoff_t index, gfp_t gfp_mask); | |
97cecb5a | 609 | extern void delete_from_page_cache(struct page *page); |
62cccb8c | 610 | extern void __delete_from_page_cache(struct page *page, void *shadow); |
ef6a3c63 | 611 | int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask); |
529ae9aa NP |
612 | |
613 | /* | |
614 | * Like add_to_page_cache_locked, but used to add newly allocated pages: | |
48c935ad | 615 | * the page is new, so we can just run __SetPageLocked() against it. |
529ae9aa NP |
616 | */ |
617 | static inline int add_to_page_cache(struct page *page, | |
618 | struct address_space *mapping, pgoff_t offset, gfp_t gfp_mask) | |
619 | { | |
620 | int error; | |
621 | ||
48c935ad | 622 | __SetPageLocked(page); |
529ae9aa NP |
623 | error = add_to_page_cache_locked(page, mapping, offset, gfp_mask); |
624 | if (unlikely(error)) | |
48c935ad | 625 | __ClearPageLocked(page); |
529ae9aa NP |
626 | return error; |
627 | } | |
628 | ||
b57c2cb9 FF |
629 | static inline unsigned long dir_pages(struct inode *inode) |
630 | { | |
09cbfeaf KS |
631 | return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >> |
632 | PAGE_SHIFT; | |
b57c2cb9 FF |
633 | } |
634 | ||
1da177e4 | 635 | #endif /* _LINUX_PAGEMAP_H */ |