1 #ifndef _LINUX_PAGEMAP_H
2 #define _LINUX_PAGEMAP_H
5 * Copyright 1995 Linus Torvalds
9 #include <linux/list.h>
10 #include <linux/highmem.h>
11 #include <linux/compiler.h>
12 #include <asm/uaccess.h>
13 #include <linux/gfp.h>
14 #include <linux/bitops.h>
15 #include <linux/hardirq.h> /* for in_interrupt() */
18 * Bits in mapping->flags. The lower __GFP_BITS_SHIFT bits are the page
19 * allocation mode flags.
21 #define AS_EIO (__GFP_BITS_SHIFT + 0) /* IO error on async write */
22 #define AS_ENOSPC (__GFP_BITS_SHIFT + 1) /* ENOSPC on async write */
23 #define AS_MM_ALL_LOCKS (__GFP_BITS_SHIFT + 2) /* under mm_take_all_locks() */
25 static inline void mapping_set_error(struct address_space
*mapping
, int error
)
27 if (unlikely(error
)) {
29 set_bit(AS_ENOSPC
, &mapping
->flags
);
31 set_bit(AS_EIO
, &mapping
->flags
);
35 #ifdef CONFIG_UNEVICTABLE_LRU
36 #define AS_UNEVICTABLE (__GFP_BITS_SHIFT + 2) /* e.g., ramdisk, SHM_LOCK */
38 static inline void mapping_set_unevictable(struct address_space
*mapping
)
40 set_bit(AS_UNEVICTABLE
, &mapping
->flags
);
43 static inline int mapping_unevictable(struct address_space
*mapping
)
45 if (mapping
&& (mapping
->flags
& AS_UNEVICTABLE
))
50 static inline void mapping_set_unevictable(struct address_space
*mapping
) { }
51 static inline int mapping_unevictable(struct address_space
*mapping
)
57 static inline gfp_t
mapping_gfp_mask(struct address_space
* mapping
)
59 return (__force gfp_t
)mapping
->flags
& __GFP_BITS_MASK
;
63 * This is non-atomic. Only to be used before the mapping is activated.
64 * Probably needs a barrier...
66 static inline void mapping_set_gfp_mask(struct address_space
*m
, gfp_t mask
)
68 m
->flags
= (m
->flags
& ~(__force
unsigned long)__GFP_BITS_MASK
) |
69 (__force
unsigned long)mask
;
73 * The page cache can done in larger chunks than
74 * one page, because it allows for more efficient
75 * throughput (it can then be mapped into user
76 * space in smaller chunks for same flexibility).
78 * Or rather, it _will_ be done in larger chunks.
80 #define PAGE_CACHE_SHIFT PAGE_SHIFT
81 #define PAGE_CACHE_SIZE PAGE_SIZE
82 #define PAGE_CACHE_MASK PAGE_MASK
83 #define PAGE_CACHE_ALIGN(addr) (((addr)+PAGE_CACHE_SIZE-1)&PAGE_CACHE_MASK)
85 #define page_cache_get(page) get_page(page)
86 #define page_cache_release(page) put_page(page)
87 void release_pages(struct page
**pages
, int nr
, int cold
);
90 * speculatively take a reference to a page.
91 * If the page is free (_count == 0), then _count is untouched, and 0
92 * is returned. Otherwise, _count is incremented by 1 and 1 is returned.
94 * This function must be called inside the same rcu_read_lock() section as has
95 * been used to lookup the page in the pagecache radix-tree (or page table):
96 * this allows allocators to use a synchronize_rcu() to stabilize _count.
98 * Unless an RCU grace period has passed, the count of all pages coming out
99 * of the allocator must be considered unstable. page_count may return higher
100 * than expected, and put_page must be able to do the right thing when the
101 * page has been finished with, no matter what it is subsequently allocated
102 * for (because put_page is what is used here to drop an invalid speculative
105 * This is the interesting part of the lockless pagecache (and lockless
106 * get_user_pages) locking protocol, where the lookup-side (eg. find_get_page)
107 * has the following pattern:
108 * 1. find page in radix tree
109 * 2. conditionally increment refcount
110 * 3. check the page is still in pagecache (if no, goto 1)
112 * Remove-side that cares about stability of _count (eg. reclaim) has the
113 * following (with tree_lock held for write):
114 * A. atomically check refcount is correct and set it to 0 (atomic_cmpxchg)
115 * B. remove page from pagecache
118 * There are 2 critical interleavings that matter:
119 * - 2 runs before A: in this case, A sees elevated refcount and bails out
120 * - A runs before 2: in this case, 2 sees zero refcount and retries;
121 * subsequently, B will complete and 1 will find no page, causing the
122 * lookup to return NULL.
124 * It is possible that between 1 and 2, the page is removed then the exact same
125 * page is inserted into the same position in pagecache. That's OK: the
126 * old find_get_page using tree_lock could equally have run before or after
127 * such a re-insertion, depending on order that locks are granted.
129 * Lookups racing against pagecache insertion isn't a big problem: either 1
130 * will find the page or it will not. Likewise, the old find_get_page could run
131 * either before the insertion or afterwards, depending on timing.
133 static inline int page_cache_get_speculative(struct page
*page
)
135 VM_BUG_ON(in_interrupt());
137 #if !defined(CONFIG_SMP) && defined(CONFIG_CLASSIC_RCU)
138 # ifdef CONFIG_PREEMPT
139 VM_BUG_ON(!in_atomic());
142 * Preempt must be disabled here - we rely on rcu_read_lock doing
145 * Pagecache won't be truncated from interrupt context, so if we have
146 * found a page in the radix tree here, we have pinned its refcount by
147 * disabling preempt, and hence no need for the "speculative get" that
150 VM_BUG_ON(page_count(page
) == 0);
151 atomic_inc(&page
->_count
);
154 if (unlikely(!get_page_unless_zero(page
))) {
156 * Either the page has been freed, or will be freed.
157 * In either case, retry here and the caller should
158 * do the right thing (see comments above).
163 VM_BUG_ON(PageTail(page
));
169 * Same as above, but add instead of inc (could just be merged)
171 static inline int page_cache_add_speculative(struct page
*page
, int count
)
173 VM_BUG_ON(in_interrupt());
175 #if !defined(CONFIG_SMP) && defined(CONFIG_CLASSIC_RCU)
176 # ifdef CONFIG_PREEMPT
177 VM_BUG_ON(!in_atomic());
179 VM_BUG_ON(page_count(page
) == 0);
180 atomic_add(count
, &page
->_count
);
183 if (unlikely(!atomic_add_unless(&page
->_count
, count
, 0)))
186 VM_BUG_ON(PageCompound(page
) && page
!= compound_head(page
));
191 static inline int page_freeze_refs(struct page
*page
, int count
)
193 return likely(atomic_cmpxchg(&page
->_count
, count
, 0) == count
);
196 static inline void page_unfreeze_refs(struct page
*page
, int count
)
198 VM_BUG_ON(page_count(page
) != 0);
199 VM_BUG_ON(count
== 0);
201 atomic_set(&page
->_count
, count
);
205 extern struct page
*__page_cache_alloc(gfp_t gfp
);
207 static inline struct page
*__page_cache_alloc(gfp_t gfp
)
209 return alloc_pages(gfp
, 0);
213 static inline struct page
*page_cache_alloc(struct address_space
*x
)
215 return __page_cache_alloc(mapping_gfp_mask(x
));
218 static inline struct page
*page_cache_alloc_cold(struct address_space
*x
)
220 return __page_cache_alloc(mapping_gfp_mask(x
)|__GFP_COLD
);
223 typedef int filler_t(void *, struct page
*);
225 extern struct page
* find_get_page(struct address_space
*mapping
,
227 extern struct page
* find_lock_page(struct address_space
*mapping
,
229 extern struct page
* find_or_create_page(struct address_space
*mapping
,
230 pgoff_t index
, gfp_t gfp_mask
);
231 unsigned find_get_pages(struct address_space
*mapping
, pgoff_t start
,
232 unsigned int nr_pages
, struct page
**pages
);
233 unsigned find_get_pages_contig(struct address_space
*mapping
, pgoff_t start
,
234 unsigned int nr_pages
, struct page
**pages
);
235 unsigned find_get_pages_tag(struct address_space
*mapping
, pgoff_t
*index
,
236 int tag
, unsigned int nr_pages
, struct page
**pages
);
238 struct page
*__grab_cache_page(struct address_space
*mapping
, pgoff_t index
);
241 * Returns locked page at given index in given cache, creating it if needed.
243 static inline struct page
*grab_cache_page(struct address_space
*mapping
,
246 return find_or_create_page(mapping
, index
, mapping_gfp_mask(mapping
));
249 extern struct page
* grab_cache_page_nowait(struct address_space
*mapping
,
251 extern struct page
* read_cache_page_async(struct address_space
*mapping
,
252 pgoff_t index
, filler_t
*filler
,
254 extern struct page
* read_cache_page(struct address_space
*mapping
,
255 pgoff_t index
, filler_t
*filler
,
257 extern int read_cache_pages(struct address_space
*mapping
,
258 struct list_head
*pages
, filler_t
*filler
, void *data
);
260 static inline struct page
*read_mapping_page_async(
261 struct address_space
*mapping
,
262 pgoff_t index
, void *data
)
264 filler_t
*filler
= (filler_t
*)mapping
->a_ops
->readpage
;
265 return read_cache_page_async(mapping
, index
, filler
, data
);
268 static inline struct page
*read_mapping_page(struct address_space
*mapping
,
269 pgoff_t index
, void *data
)
271 filler_t
*filler
= (filler_t
*)mapping
->a_ops
->readpage
;
272 return read_cache_page(mapping
, index
, filler
, data
);
276 * Return byte-offset into filesystem object for page.
278 static inline loff_t
page_offset(struct page
*page
)
280 return ((loff_t
)page
->index
) << PAGE_CACHE_SHIFT
;
283 static inline pgoff_t
linear_page_index(struct vm_area_struct
*vma
,
284 unsigned long address
)
286 pgoff_t pgoff
= (address
- vma
->vm_start
) >> PAGE_SHIFT
;
287 pgoff
+= vma
->vm_pgoff
;
288 return pgoff
>> (PAGE_CACHE_SHIFT
- PAGE_SHIFT
);
291 extern void __lock_page(struct page
*page
);
292 extern int __lock_page_killable(struct page
*page
);
293 extern void __lock_page_nosync(struct page
*page
);
294 extern void unlock_page(struct page
*page
);
296 static inline void set_page_locked(struct page
*page
)
298 set_bit(PG_locked
, &page
->flags
);
301 static inline void clear_page_locked(struct page
*page
)
303 clear_bit(PG_locked
, &page
->flags
);
306 static inline int trylock_page(struct page
*page
)
308 return !test_and_set_bit(PG_locked
, &page
->flags
);
312 * lock_page may only be called if we have the page's inode pinned.
314 static inline void lock_page(struct page
*page
)
317 if (!trylock_page(page
))
322 * lock_page_killable is like lock_page but can be interrupted by fatal
323 * signals. It returns 0 if it locked the page and -EINTR if it was
324 * killed while waiting.
326 static inline int lock_page_killable(struct page
*page
)
329 if (!trylock_page(page
))
330 return __lock_page_killable(page
);
335 * lock_page_nosync should only be used if we can't pin the page's inode.
336 * Doesn't play quite so well with block device plugging.
338 static inline void lock_page_nosync(struct page
*page
)
341 if (!trylock_page(page
))
342 __lock_page_nosync(page
);
346 * This is exported only for wait_on_page_locked/wait_on_page_writeback.
347 * Never use this directly!
349 extern void wait_on_page_bit(struct page
*page
, int bit_nr
);
352 * Wait for a page to be unlocked.
354 * This must be called with the caller "holding" the page,
355 * ie with increased "page->count" so that the page won't
356 * go away during the wait..
358 static inline void wait_on_page_locked(struct page
*page
)
360 if (PageLocked(page
))
361 wait_on_page_bit(page
, PG_locked
);
365 * Wait for a page to complete writeback
367 static inline void wait_on_page_writeback(struct page
*page
)
369 if (PageWriteback(page
))
370 wait_on_page_bit(page
, PG_writeback
);
373 extern void end_page_writeback(struct page
*page
);
376 * Fault a userspace page into pagetables. Return non-zero on a fault.
378 * This assumes that two userspace pages are always sufficient. That's
379 * not true if PAGE_CACHE_SIZE > PAGE_SIZE.
381 static inline int fault_in_pages_writeable(char __user
*uaddr
, int size
)
385 if (unlikely(size
== 0))
389 * Writing zeroes into userspace here is OK, because we know that if
390 * the zero gets there, we'll be overwriting it.
392 ret
= __put_user(0, uaddr
);
394 char __user
*end
= uaddr
+ size
- 1;
397 * If the page was already mapped, this will get a cache miss
398 * for sure, so try to avoid doing it.
400 if (((unsigned long)uaddr
& PAGE_MASK
) !=
401 ((unsigned long)end
& PAGE_MASK
))
402 ret
= __put_user(0, end
);
407 static inline int fault_in_pages_readable(const char __user
*uaddr
, int size
)
412 if (unlikely(size
== 0))
415 ret
= __get_user(c
, uaddr
);
417 const char __user
*end
= uaddr
+ size
- 1;
419 if (((unsigned long)uaddr
& PAGE_MASK
) !=
420 ((unsigned long)end
& PAGE_MASK
))
421 ret
= __get_user(c
, end
);
426 int add_to_page_cache_locked(struct page
*page
, struct address_space
*mapping
,
427 pgoff_t index
, gfp_t gfp_mask
);
428 int add_to_page_cache_lru(struct page
*page
, struct address_space
*mapping
,
429 pgoff_t index
, gfp_t gfp_mask
);
430 extern void remove_from_page_cache(struct page
*page
);
431 extern void __remove_from_page_cache(struct page
*page
);
434 * Like add_to_page_cache_locked, but used to add newly allocated pages:
435 * the page is new, so we can just run set_page_locked() against it.
437 static inline int add_to_page_cache(struct page
*page
,
438 struct address_space
*mapping
, pgoff_t offset
, gfp_t gfp_mask
)
442 set_page_locked(page
);
443 error
= add_to_page_cache_locked(page
, mapping
, offset
, gfp_mask
);
445 clear_page_locked(page
);
449 #endif /* _LINUX_PAGEMAP_H */
This page took 0.061764 seconds and 6 git commands to generate.