Merge tag 'kvmgt-vfio-mdev-for-v4.10-rc1' of git://github.com/01org/gvt-linux
[linux-2.6-block.git] / include / linux / buffer_head.h
1 /*
2  * include/linux/buffer_head.h
3  *
4  * Everything to do with buffer_heads.
5  */
6
7 #ifndef _LINUX_BUFFER_HEAD_H
8 #define _LINUX_BUFFER_HEAD_H
9
10 #include <linux/types.h>
11 #include <linux/fs.h>
12 #include <linux/linkage.h>
13 #include <linux/pagemap.h>
14 #include <linux/wait.h>
15 #include <linux/atomic.h>
16
17 #ifdef CONFIG_BLOCK
18
19 enum bh_state_bits {
20         BH_Uptodate,    /* Contains valid data */
21         BH_Dirty,       /* Is dirty */
22         BH_Lock,        /* Is locked */
23         BH_Req,         /* Has been submitted for I/O */
24         BH_Uptodate_Lock,/* Used by the first bh in a page, to serialise
25                           * IO completion of other buffers in the page
26                           */
27
28         BH_Mapped,      /* Has a disk mapping */
29         BH_New,         /* Disk mapping was newly created by get_block */
30         BH_Async_Read,  /* Is under end_buffer_async_read I/O */
31         BH_Async_Write, /* Is under end_buffer_async_write I/O */
32         BH_Delay,       /* Buffer is not yet allocated on disk */
33         BH_Boundary,    /* Block is followed by a discontiguity */
34         BH_Write_EIO,   /* I/O error on write */
35         BH_Unwritten,   /* Buffer is allocated on disk but not written */
36         BH_Quiet,       /* Buffer Error Prinks to be quiet */
37         BH_Meta,        /* Buffer contains metadata */
38         BH_Prio,        /* Buffer should be submitted with REQ_PRIO */
39         BH_Defer_Completion, /* Defer AIO completion to workqueue */
40
41         BH_PrivateStart,/* not a state bit, but the first bit available
42                          * for private allocation by other entities
43                          */
44 };
45
46 #define MAX_BUF_PER_PAGE (PAGE_SIZE / 512)
47
48 struct page;
49 struct buffer_head;
50 struct address_space;
51 typedef void (bh_end_io_t)(struct buffer_head *bh, int uptodate);
52
53 /*
54  * Historically, a buffer_head was used to map a single block
55  * within a page, and of course as the unit of I/O through the
56  * filesystem and block layers.  Nowadays the basic I/O unit
57  * is the bio, and buffer_heads are used for extracting block
58  * mappings (via a get_block_t call), for tracking state within
59  * a page (via a page_mapping) and for wrapping bio submission
60  * for backward compatibility reasons (e.g. submit_bh).
61  */
62 struct buffer_head {
63         unsigned long b_state;          /* buffer state bitmap (see above) */
64         struct buffer_head *b_this_page;/* circular list of page's buffers */
65         struct page *b_page;            /* the page this bh is mapped to */
66
67         sector_t b_blocknr;             /* start block number */
68         size_t b_size;                  /* size of mapping */
69         char *b_data;                   /* pointer to data within the page */
70
71         struct block_device *b_bdev;
72         bh_end_io_t *b_end_io;          /* I/O completion */
73         void *b_private;                /* reserved for b_end_io */
74         struct list_head b_assoc_buffers; /* associated with another mapping */
75         struct address_space *b_assoc_map;      /* mapping this buffer is
76                                                    associated with */
77         atomic_t b_count;               /* users using this buffer_head */
78 };
79
80 /*
81  * macro tricks to expand the set_buffer_foo(), clear_buffer_foo()
82  * and buffer_foo() functions.
83  */
84 #define BUFFER_FNS(bit, name)                                           \
85 static __always_inline void set_buffer_##name(struct buffer_head *bh)   \
86 {                                                                       \
87         set_bit(BH_##bit, &(bh)->b_state);                              \
88 }                                                                       \
89 static __always_inline void clear_buffer_##name(struct buffer_head *bh) \
90 {                                                                       \
91         clear_bit(BH_##bit, &(bh)->b_state);                            \
92 }                                                                       \
93 static __always_inline int buffer_##name(const struct buffer_head *bh)  \
94 {                                                                       \
95         return test_bit(BH_##bit, &(bh)->b_state);                      \
96 }
97
98 /*
99  * test_set_buffer_foo() and test_clear_buffer_foo()
100  */
101 #define TAS_BUFFER_FNS(bit, name)                                       \
102 static __always_inline int test_set_buffer_##name(struct buffer_head *bh) \
103 {                                                                       \
104         return test_and_set_bit(BH_##bit, &(bh)->b_state);              \
105 }                                                                       \
106 static __always_inline int test_clear_buffer_##name(struct buffer_head *bh) \
107 {                                                                       \
108         return test_and_clear_bit(BH_##bit, &(bh)->b_state);            \
109 }                                                                       \
110
111 /*
112  * Emit the buffer bitops functions.   Note that there are also functions
113  * of the form "mark_buffer_foo()".  These are higher-level functions which
114  * do something in addition to setting a b_state bit.
115  */
116 BUFFER_FNS(Uptodate, uptodate)
117 BUFFER_FNS(Dirty, dirty)
118 TAS_BUFFER_FNS(Dirty, dirty)
119 BUFFER_FNS(Lock, locked)
120 BUFFER_FNS(Req, req)
121 TAS_BUFFER_FNS(Req, req)
122 BUFFER_FNS(Mapped, mapped)
123 BUFFER_FNS(New, new)
124 BUFFER_FNS(Async_Read, async_read)
125 BUFFER_FNS(Async_Write, async_write)
126 BUFFER_FNS(Delay, delay)
127 BUFFER_FNS(Boundary, boundary)
128 BUFFER_FNS(Write_EIO, write_io_error)
129 BUFFER_FNS(Unwritten, unwritten)
130 BUFFER_FNS(Meta, meta)
131 BUFFER_FNS(Prio, prio)
132 BUFFER_FNS(Defer_Completion, defer_completion)
133
134 #define bh_offset(bh)           ((unsigned long)(bh)->b_data & ~PAGE_MASK)
135
136 /* If we *know* page->private refers to buffer_heads */
137 #define page_buffers(page)                                      \
138         ({                                                      \
139                 BUG_ON(!PagePrivate(page));                     \
140                 ((struct buffer_head *)page_private(page));     \
141         })
142 #define page_has_buffers(page)  PagePrivate(page)
143
144 void buffer_check_dirty_writeback(struct page *page,
145                                      bool *dirty, bool *writeback);
146
147 /*
148  * Declarations
149  */
150
151 void mark_buffer_dirty(struct buffer_head *bh);
152 void init_buffer(struct buffer_head *, bh_end_io_t *, void *);
153 void touch_buffer(struct buffer_head *bh);
154 void set_bh_page(struct buffer_head *bh,
155                 struct page *page, unsigned long offset);
156 int try_to_free_buffers(struct page *);
157 struct buffer_head *alloc_page_buffers(struct page *page, unsigned long size,
158                 int retry);
159 void create_empty_buffers(struct page *, unsigned long,
160                         unsigned long b_state);
161 void end_buffer_read_sync(struct buffer_head *bh, int uptodate);
162 void end_buffer_write_sync(struct buffer_head *bh, int uptodate);
163 void end_buffer_async_write(struct buffer_head *bh, int uptodate);
164
165 /* Things to do with buffers at mapping->private_list */
166 void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode);
167 int inode_has_buffers(struct inode *);
168 void invalidate_inode_buffers(struct inode *);
169 int remove_inode_buffers(struct inode *inode);
170 int sync_mapping_buffers(struct address_space *mapping);
171 void clean_bdev_aliases(struct block_device *bdev, sector_t block,
172                         sector_t len);
173 static inline void clean_bdev_bh_alias(struct buffer_head *bh)
174 {
175         clean_bdev_aliases(bh->b_bdev, bh->b_blocknr, 1);
176 }
177
178 void mark_buffer_async_write(struct buffer_head *bh);
179 void __wait_on_buffer(struct buffer_head *);
180 wait_queue_head_t *bh_waitq_head(struct buffer_head *bh);
181 struct buffer_head *__find_get_block(struct block_device *bdev, sector_t block,
182                         unsigned size);
183 struct buffer_head *__getblk_gfp(struct block_device *bdev, sector_t block,
184                                   unsigned size, gfp_t gfp);
185 void __brelse(struct buffer_head *);
186 void __bforget(struct buffer_head *);
187 void __breadahead(struct block_device *, sector_t block, unsigned int size);
188 struct buffer_head *__bread_gfp(struct block_device *,
189                                 sector_t block, unsigned size, gfp_t gfp);
190 void invalidate_bh_lrus(void);
191 struct buffer_head *alloc_buffer_head(gfp_t gfp_flags);
192 void free_buffer_head(struct buffer_head * bh);
193 void unlock_buffer(struct buffer_head *bh);
194 void __lock_buffer(struct buffer_head *bh);
195 void ll_rw_block(int, int, int, struct buffer_head * bh[]);
196 int sync_dirty_buffer(struct buffer_head *bh);
197 int __sync_dirty_buffer(struct buffer_head *bh, int op_flags);
198 void write_dirty_buffer(struct buffer_head *bh, int op_flags);
199 int _submit_bh(int op, int op_flags, struct buffer_head *bh,
200                unsigned long bio_flags);
201 int submit_bh(int, int, struct buffer_head *);
202 void write_boundary_block(struct block_device *bdev,
203                         sector_t bblock, unsigned blocksize);
204 int bh_uptodate_or_lock(struct buffer_head *bh);
205 int bh_submit_read(struct buffer_head *bh);
206
207 extern int buffer_heads_over_limit;
208
209 /*
210  * Generic address_space_operations implementations for buffer_head-backed
211  * address_spaces.
212  */
213 void block_invalidatepage(struct page *page, unsigned int offset,
214                           unsigned int length);
215 int block_write_full_page(struct page *page, get_block_t *get_block,
216                                 struct writeback_control *wbc);
217 int __block_write_full_page(struct inode *inode, struct page *page,
218                         get_block_t *get_block, struct writeback_control *wbc,
219                         bh_end_io_t *handler);
220 int block_read_full_page(struct page*, get_block_t*);
221 int block_is_partially_uptodate(struct page *page, unsigned long from,
222                                 unsigned long count);
223 int block_write_begin(struct address_space *mapping, loff_t pos, unsigned len,
224                 unsigned flags, struct page **pagep, get_block_t *get_block);
225 int __block_write_begin(struct page *page, loff_t pos, unsigned len,
226                 get_block_t *get_block);
227 int block_write_end(struct file *, struct address_space *,
228                                 loff_t, unsigned, unsigned,
229                                 struct page *, void *);
230 int generic_write_end(struct file *, struct address_space *,
231                                 loff_t, unsigned, unsigned,
232                                 struct page *, void *);
233 void page_zero_new_buffers(struct page *page, unsigned from, unsigned to);
234 int cont_write_begin(struct file *, struct address_space *, loff_t,
235                         unsigned, unsigned, struct page **, void **,
236                         get_block_t *, loff_t *);
237 int generic_cont_expand_simple(struct inode *inode, loff_t size);
238 int block_commit_write(struct page *page, unsigned from, unsigned to);
239 int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
240                                 get_block_t get_block);
241 /* Convert errno to return value from ->page_mkwrite() call */
242 static inline int block_page_mkwrite_return(int err)
243 {
244         if (err == 0)
245                 return VM_FAULT_LOCKED;
246         if (err == -EFAULT)
247                 return VM_FAULT_NOPAGE;
248         if (err == -ENOMEM)
249                 return VM_FAULT_OOM;
250         if (err == -EAGAIN)
251                 return VM_FAULT_RETRY;
252         /* -ENOSPC, -EDQUOT, -EIO ... */
253         return VM_FAULT_SIGBUS;
254 }
255 sector_t generic_block_bmap(struct address_space *, sector_t, get_block_t *);
256 int block_truncate_page(struct address_space *, loff_t, get_block_t *);
257 int nobh_write_begin(struct address_space *, loff_t, unsigned, unsigned,
258                                 struct page **, void **, get_block_t*);
259 int nobh_write_end(struct file *, struct address_space *,
260                                 loff_t, unsigned, unsigned,
261                                 struct page *, void *);
262 int nobh_truncate_page(struct address_space *, loff_t, get_block_t *);
263 int nobh_writepage(struct page *page, get_block_t *get_block,
264                         struct writeback_control *wbc);
265
266 void buffer_init(void);
267
268 /*
269  * inline definitions
270  */
271
272 static inline void attach_page_buffers(struct page *page,
273                 struct buffer_head *head)
274 {
275         get_page(page);
276         SetPagePrivate(page);
277         set_page_private(page, (unsigned long)head);
278 }
279
280 static inline void get_bh(struct buffer_head *bh)
281 {
282         atomic_inc(&bh->b_count);
283 }
284
285 static inline void put_bh(struct buffer_head *bh)
286 {
287         smp_mb__before_atomic();
288         atomic_dec(&bh->b_count);
289 }
290
291 static inline void brelse(struct buffer_head *bh)
292 {
293         if (bh)
294                 __brelse(bh);
295 }
296
297 static inline void bforget(struct buffer_head *bh)
298 {
299         if (bh)
300                 __bforget(bh);
301 }
302
303 static inline struct buffer_head *
304 sb_bread(struct super_block *sb, sector_t block)
305 {
306         return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE);
307 }
308
309 static inline struct buffer_head *
310 sb_bread_unmovable(struct super_block *sb, sector_t block)
311 {
312         return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, 0);
313 }
314
315 static inline void
316 sb_breadahead(struct super_block *sb, sector_t block)
317 {
318         __breadahead(sb->s_bdev, block, sb->s_blocksize);
319 }
320
321 static inline struct buffer_head *
322 sb_getblk(struct super_block *sb, sector_t block)
323 {
324         return __getblk_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE);
325 }
326
327
328 static inline struct buffer_head *
329 sb_getblk_gfp(struct super_block *sb, sector_t block, gfp_t gfp)
330 {
331         return __getblk_gfp(sb->s_bdev, block, sb->s_blocksize, gfp);
332 }
333
334 static inline struct buffer_head *
335 sb_find_get_block(struct super_block *sb, sector_t block)
336 {
337         return __find_get_block(sb->s_bdev, block, sb->s_blocksize);
338 }
339
340 static inline void
341 map_bh(struct buffer_head *bh, struct super_block *sb, sector_t block)
342 {
343         set_buffer_mapped(bh);
344         bh->b_bdev = sb->s_bdev;
345         bh->b_blocknr = block;
346         bh->b_size = sb->s_blocksize;
347 }
348
349 static inline void wait_on_buffer(struct buffer_head *bh)
350 {
351         might_sleep();
352         if (buffer_locked(bh))
353                 __wait_on_buffer(bh);
354 }
355
356 static inline int trylock_buffer(struct buffer_head *bh)
357 {
358         return likely(!test_and_set_bit_lock(BH_Lock, &bh->b_state));
359 }
360
361 static inline void lock_buffer(struct buffer_head *bh)
362 {
363         might_sleep();
364         if (!trylock_buffer(bh))
365                 __lock_buffer(bh);
366 }
367
368 static inline struct buffer_head *getblk_unmovable(struct block_device *bdev,
369                                                    sector_t block,
370                                                    unsigned size)
371 {
372         return __getblk_gfp(bdev, block, size, 0);
373 }
374
375 static inline struct buffer_head *__getblk(struct block_device *bdev,
376                                            sector_t block,
377                                            unsigned size)
378 {
379         return __getblk_gfp(bdev, block, size, __GFP_MOVABLE);
380 }
381
382 /**
383  *  __bread() - reads a specified block and returns the bh
384  *  @bdev: the block_device to read from
385  *  @block: number of block
386  *  @size: size (in bytes) to read
387  *
388  *  Reads a specified block, and returns buffer head that contains it.
389  *  The page cache is allocated from movable area so that it can be migrated.
390  *  It returns NULL if the block was unreadable.
391  */
392 static inline struct buffer_head *
393 __bread(struct block_device *bdev, sector_t block, unsigned size)
394 {
395         return __bread_gfp(bdev, block, size, __GFP_MOVABLE);
396 }
397
398 extern int __set_page_dirty_buffers(struct page *page);
399
400 #else /* CONFIG_BLOCK */
401
402 static inline void buffer_init(void) {}
403 static inline int try_to_free_buffers(struct page *page) { return 1; }
404 static inline int inode_has_buffers(struct inode *inode) { return 0; }
405 static inline void invalidate_inode_buffers(struct inode *inode) {}
406 static inline int remove_inode_buffers(struct inode *inode) { return 1; }
407 static inline int sync_mapping_buffers(struct address_space *mapping) { return 0; }
408
409 #endif /* CONFIG_BLOCK */
410 #endif /* _LINUX_BUFFER_HEAD_H */