1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3 * include/linux/buffer_head.h
4 *
5 * Everything to do with buffer_heads.
6 */
7
8 #ifndef _LINUX_BUFFER_HEAD_H
9 #define _LINUX_BUFFER_HEAD_H
10
11 #include <linux/types.h>
12 #include <linux/blk_types.h>
13 #include <linux/fs.h>
14 #include <linux/linkage.h>
15 #include <linux/pagemap.h>
16 #include <linux/wait.h>
17 #include <linux/atomic.h>
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
25 BH_Mapped, /* Has a disk mapping */
26 BH_New, /* Disk mapping was newly created by get_block */
27 BH_Async_Read, /* Is under end_buffer_async_read I/O */
28 BH_Async_Write, /* Is under end_buffer_async_write I/O */
29 BH_Delay, /* Buffer is not yet allocated on disk */
30 BH_Boundary, /* Block is followed by a discontiguity */
31 BH_Write_EIO, /* I/O error on write */
32 BH_Unwritten, /* Buffer is allocated on disk but not written */
33 BH_Quiet, /* Buffer Error Prinks to be quiet */
34 BH_Meta, /* Buffer contains metadata */
35 BH_Prio, /* Buffer should be submitted with REQ_PRIO */
36 BH_Defer_Completion, /* Defer AIO completion to workqueue */
37
38 BH_PrivateStart,/* not a state bit, but the first bit available
39 * for private allocation by other entities
40 */
41 };
42
43 #define MAX_BUF_PER_PAGE (PAGE_SIZE / 512)
44
45 struct page;
46 struct buffer_head;
47 struct address_space;
48 typedef void (bh_end_io_t)(struct buffer_head *bh, int uptodate);
49
50 /*
51 * Historically, a buffer_head was used to map a single block
52 * within a page, and of course as the unit of I/O through the
53 * filesystem and block layers. Nowadays the basic I/O unit
54 * is the bio, and buffer_heads are used for extracting block
55 * mappings (via a get_block_t call), for tracking state within
56 * a folio (via a folio_mapping) and for wrapping bio submission
57 * for backward compatibility reasons (e.g. submit_bh).
58 */
59 struct buffer_head {
60 unsigned long b_state; /* buffer state bitmap (see above) */
61 struct buffer_head *b_this_page;/* circular list of page's buffers */
62 union {
63 struct page *b_page; /* the page this bh is mapped to */
64 struct folio *b_folio; /* the folio this bh is mapped to */
65 };
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 spinlock_t b_uptodate_lock; /* Used by the first bh in a page, to
79 * serialise IO completion of other
80 * buffers in the page */
81 };
82
83 /*
84 * macro tricks to expand the set_buffer_foo(), clear_buffer_foo()
85 * and buffer_foo() functions.
86 * To avoid reset buffer flags that are already set, because that causes
87 * a costly cache line transition, check the flag first.
88 */
89 #define BUFFER_FNS(bit, name) \
90 static __always_inline void set_buffer_##name(struct buffer_head *bh) \
91 { \
92 if (!test_bit(BH_##bit, &(bh)->b_state)) \
93 set_bit(BH_##bit, &(bh)->b_state); \
94 } \
95 static __always_inline void clear_buffer_##name(struct buffer_head *bh) \
96 { \
97 clear_bit(BH_##bit, &(bh)->b_state); \
98 } \
99 static __always_inline int buffer_##name(const struct buffer_head *bh) \
100 { \
101 return test_bit(BH_##bit, &(bh)->b_state); \
102 }
103
104 /*
105 * test_set_buffer_foo() and test_clear_buffer_foo()
106 */
107 #define TAS_BUFFER_FNS(bit, name) \
108 static __always_inline int test_set_buffer_##name(struct buffer_head *bh) \
109 { \
110 return test_and_set_bit(BH_##bit, &(bh)->b_state); \
111 } \
112 static __always_inline int test_clear_buffer_##name(struct buffer_head *bh) \
113 { \
114 return test_and_clear_bit(BH_##bit, &(bh)->b_state); \
115 } \
116
117 /*
118 * Emit the buffer bitops functions. Note that there are also functions
119 * of the form "mark_buffer_foo()". These are higher-level functions which
120 * do something in addition to setting a b_state bit.
121 */
122 BUFFER_FNS(Dirty, dirty)
123 TAS_BUFFER_FNS(Dirty, dirty)
124 BUFFER_FNS(Lock, locked)
125 BUFFER_FNS(Req, req)
126 TAS_BUFFER_FNS(Req, req)
127 BUFFER_FNS(Mapped, mapped)
128 BUFFER_FNS(New, new)
129 BUFFER_FNS(Async_Read, async_read)
130 BUFFER_FNS(Async_Write, async_write)
131 BUFFER_FNS(Delay, delay)
132 BUFFER_FNS(Boundary, boundary)
133 BUFFER_FNS(Write_EIO, write_io_error)
134 BUFFER_FNS(Unwritten, unwritten)
135 BUFFER_FNS(Meta, meta)
136 BUFFER_FNS(Prio, prio)
137 BUFFER_FNS(Defer_Completion, defer_completion)
138
139 static __always_inline void set_buffer_uptodate(struct buffer_head *bh)
140 {
141 /*
142 * If somebody else already set this uptodate, they will
143 * have done the memory barrier, and a reader will thus
144 * see *some* valid buffer state.
145 *
146 * Any other serialization (with IO errors or whatever that
147 * might clear the bit) has to come from other state (eg BH_Lock).
148 */
149 if (test_bit(BH_Uptodate, &bh->b_state))
150 return;
151
152 /*
153 * make it consistent with folio_mark_uptodate
154 * pairs with smp_load_acquire in buffer_uptodate
155 */
156 smp_mb__before_atomic();
157 set_bit(BH_Uptodate, &bh->b_state);
158 }
159
160 static __always_inline void clear_buffer_uptodate(struct buffer_head *bh)
161 {
162 clear_bit(BH_Uptodate, &bh->b_state);
163 }
164
165 static __always_inline int buffer_uptodate(const struct buffer_head *bh)
166 {
167 /*
168 * make it consistent with folio_test_uptodate
169 * pairs with smp_mb__before_atomic in set_buffer_uptodate
170 */
171 return test_bit_acquire(BH_Uptodate, &bh->b_state);
172 }
173
174 static inline unsigned long bh_offset(const struct buffer_head *bh)
175 {
176 return (unsigned long)(bh)->b_data & (page_size(bh->b_page) - 1);
177 }
178
179 /* If we *know* page->private refers to buffer_heads */
180 #define page_buffers(page) \
181 ({ \
182 BUG_ON(!PagePrivate(page)); \
183 ((struct buffer_head *)page_private(page)); \
184 })
185 #define page_has_buffers(page) PagePrivate(page)
186 #define folio_buffers(folio) folio_get_private(folio)
187
188 void buffer_check_dirty_writeback(struct folio *folio,
189 bool *dirty, bool *writeback);
190
191 /*
192 * Declarations
193 */
194
195 void mark_buffer_dirty(struct buffer_head *bh);
196 void mark_buffer_write_io_error(struct buffer_head *bh);
197 void touch_buffer(struct buffer_head *bh);
198 void folio_set_bh(struct buffer_head *bh, struct folio *folio,
199 unsigned long offset);
200 struct buffer_head *folio_alloc_buffers(struct folio *folio, unsigned long size,
201 gfp_t gfp);
202 struct buffer_head *alloc_page_buffers(struct page *page, unsigned long size,
203 bool retry);
204 struct buffer_head *create_empty_buffers(struct folio *folio,
205 unsigned long blocksize, unsigned long b_state);
206 void end_buffer_read_sync(struct buffer_head *bh, int uptodate);
207 void end_buffer_write_sync(struct buffer_head *bh, int uptodate);
208
209 /* Things to do with buffers at mapping->private_list */
210 void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode);
211 int generic_buffers_fsync_noflush(struct file *file, loff_t start, loff_t end,
212 bool datasync);
213 int generic_buffers_fsync(struct file *file, loff_t start, loff_t end,
214 bool datasync);
215 void clean_bdev_aliases(struct block_device *bdev, sector_t block,
216 sector_t len);
217 static inline void clean_bdev_bh_alias(struct buffer_head *bh)
218 {
219 clean_bdev_aliases(bh->b_bdev, bh->b_blocknr, 1);
220 }
221
222 void mark_buffer_async_write(struct buffer_head *bh);
223 void __wait_on_buffer(struct buffer_head *);
224 wait_queue_head_t *bh_waitq_head(struct buffer_head *bh);
225 struct buffer_head *__find_get_block(struct block_device *bdev, sector_t block,
226 unsigned size);
227 struct buffer_head *bdev_getblk(struct block_device *bdev, sector_t block,
228 unsigned size, gfp_t gfp);
229 void __brelse(struct buffer_head *);
230 void __bforget(struct buffer_head *);
231 void __breadahead(struct block_device *, sector_t block, unsigned int size);
232 struct buffer_head *__bread_gfp(struct block_device *,
233 sector_t block, unsigned size, gfp_t gfp);
234 struct buffer_head *alloc_buffer_head(gfp_t gfp_flags);
235 void free_buffer_head(struct buffer_head * bh);
236 void unlock_buffer(struct buffer_head *bh);
237 void __lock_buffer(struct buffer_head *bh);
238 int sync_dirty_buffer(struct buffer_head *bh);
239 int __sync_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags);
240 void write_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags);
241 void submit_bh(blk_opf_t, struct buffer_head *);
242 void write_boundary_block(struct block_device *bdev,
243 sector_t bblock, unsigned blocksize);
244 int bh_uptodate_or_lock(struct buffer_head *bh);
245 int __bh_read(struct buffer_head *bh, blk_opf_t op_flags, bool wait);
246 void __bh_read_batch(int nr, struct buffer_head *bhs[],
247 blk_opf_t op_flags, bool force_lock);
248
249 /*
250 * Generic address_space_operations implementations for buffer_head-backed
251 * address_spaces.
252 */
253 void block_invalidate_folio(struct folio *folio, size_t offset, size_t length);
254 int block_write_full_folio(struct folio *folio, struct writeback_control *wbc,
255 void *get_block);
256 int __block_write_full_folio(struct inode *inode, struct folio *folio,
257 get_block_t *get_block, struct writeback_control *wbc);
258 int block_read_full_folio(struct folio *, get_block_t *);
259 bool block_is_partially_uptodate(struct folio *, size_t from, size_t count);
260 int block_write_begin(struct address_space *mapping, loff_t pos, unsigned len,
261 struct page **pagep, get_block_t *get_block);
262 int __block_write_begin(struct page *page, loff_t pos, unsigned len,
263 get_block_t *get_block);
264 int block_write_end(struct file *, struct address_space *,
265 loff_t, unsigned, unsigned,
266 struct page *, void *);
267 int generic_write_end(struct file *, struct address_space *,
268 loff_t, unsigned, unsigned,
269 struct page *, void *);
270 void folio_zero_new_buffers(struct folio *folio, size_t from, size_t to);
271 int cont_write_begin(struct file *, struct address_space *, loff_t,
272 unsigned, struct page **, void **,
273 get_block_t *, loff_t *);
274 int generic_cont_expand_simple(struct inode *inode, loff_t size);
275 void block_commit_write(struct page *page, unsigned int from, unsigned int to);
276 int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
277 get_block_t get_block);
278 sector_t generic_block_bmap(struct address_space *, sector_t, get_block_t *);
279 int block_truncate_page(struct address_space *, loff_t, get_block_t *);
280
281 #ifdef CONFIG_MIGRATION
282 extern int buffer_migrate_folio(struct address_space *,
283 struct folio *dst, struct folio *src, enum migrate_mode);
284 extern int buffer_migrate_folio_norefs(struct address_space *,
285 struct folio *dst, struct folio *src, enum migrate_mode);
286 #else
287 #define buffer_migrate_folio NULL
288 #define buffer_migrate_folio_norefs NULL
289 #endif
290
291 /*
292 * inline definitions
293 */
294
295 static inline void get_bh(struct buffer_head *bh)
296 {
297 atomic_inc(&bh->b_count);
298 }
299
300 static inline void put_bh(struct buffer_head *bh)
301 {
302 smp_mb__before_atomic();
303 atomic_dec(&bh->b_count);
304 }
305
306 /**
307 * brelse - Release a buffer.
308 * @bh: The buffer to release.
309 *
310 * Decrement a buffer_head's reference count. If @bh is NULL, this
311 * function is a no-op.
312 *
313 * If all buffers on a folio have zero reference count, are clean
314 * and unlocked, and if the folio is unlocked and not under writeback
315 * then try_to_free_buffers() may strip the buffers from the folio in
316 * preparation for freeing it (sometimes, rarely, buffers are removed
317 * from a folio but it ends up not being freed, and buffers may later
318 * be reattached).
319 *
320 * Context: Any context.
321 */
322 static inline void brelse(struct buffer_head *bh)
323 {
324 if (bh)
325 __brelse(bh);
326 }
327
328 /**
329 * bforget - Discard any dirty data in a buffer.
330 * @bh: The buffer to forget.
331 *
332 * Call this function instead of brelse() if the data written to a buffer
333 * no longer needs to be written back. It will clear the buffer's dirty
334 * flag so writeback of this buffer will be skipped.
335 *
336 * Context: Any context.
337 */
338 static inline void bforget(struct buffer_head *bh)
339 {
340 if (bh)
341 __bforget(bh);
342 }
343
344 static inline struct buffer_head *
345 sb_bread(struct super_block *sb, sector_t block)
346 {
347 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE);
348 }
349
350 static inline struct buffer_head *
351 sb_bread_unmovable(struct super_block *sb, sector_t block)
352 {
353 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, 0);
354 }
355
356 static inline void
357 sb_breadahead(struct super_block *sb, sector_t block)
358 {
359 __breadahead(sb->s_bdev, block, sb->s_blocksize);
360 }
361
362 static inline struct buffer_head *getblk_unmovable(struct block_device *bdev,
363 sector_t block, unsigned size)
364 {
365 gfp_t gfp;
366
367 gfp = mapping_gfp_constraint(bdev->bd_mapping, ~__GFP_FS);
368 gfp |= __GFP_NOFAIL;
369
370 return bdev_getblk(bdev, block, size, gfp);
371 }
372
373 static inline struct buffer_head *__getblk(struct block_device *bdev,
374 sector_t block, unsigned size)
375 {
376 gfp_t gfp;
377
378 gfp = mapping_gfp_constraint(bdev->bd_mapping, ~__GFP_FS);
379 gfp |= __GFP_MOVABLE | __GFP_NOFAIL;
380
381 return bdev_getblk(bdev, block, size, gfp);
382 }
383
384 static inline struct buffer_head *sb_getblk(struct super_block *sb,
385 sector_t block)
386 {
387 return __getblk(sb->s_bdev, block, sb->s_blocksize);
388 }
389
390 static inline struct buffer_head *sb_getblk_gfp(struct super_block *sb,
391 sector_t block, gfp_t gfp)
392 {
393 return bdev_getblk(sb->s_bdev, block, sb->s_blocksize, gfp);
394 }
395
396 static inline struct buffer_head *
397 sb_find_get_block(struct super_block *sb, sector_t block)
398 {
399 return __find_get_block(sb->s_bdev, block, sb->s_blocksize);
400 }
401
402 static inline void
403 map_bh(struct buffer_head *bh, struct super_block *sb, sector_t block)
404 {
405 set_buffer_mapped(bh);
406 bh->b_bdev = sb->s_bdev;
407 bh->b_blocknr = block;
408 bh->b_size = sb->s_blocksize;
409 }
410
411 static inline void wait_on_buffer(struct buffer_head *bh)
412 {
413 might_sleep();
414 if (buffer_locked(bh))
415 __wait_on_buffer(bh);
416 }
417
418 static inline int trylock_buffer(struct buffer_head *bh)
419 {
420 return likely(!test_and_set_bit_lock(BH_Lock, &bh->b_state));
421 }
422
423 static inline void lock_buffer(struct buffer_head *bh)
424 {
425 might_sleep();
426 if (!trylock_buffer(bh))
427 __lock_buffer(bh);
428 }
429
430 static inline void bh_readahead(struct buffer_head *bh, blk_opf_t op_flags)
431 {
432 if (!buffer_uptodate(bh) && trylock_buffer(bh)) {
433 if (!buffer_uptodate(bh))
434 __bh_read(bh, op_flags, false);
435 else
436 unlock_buffer(bh);
437 }
438 }
439
440 static inline void bh_read_nowait(struct buffer_head *bh, blk_opf_t op_flags)
441 {
442 if (!bh_uptodate_or_lock(bh))
443 __bh_read(bh, op_flags, false);
444 }
445
446 /* Returns 1 if buffer uptodated, 0 on success, and -EIO on error. */
447 static inline int bh_read(struct buffer_head *bh, blk_opf_t op_flags)
448 {
449 if (bh_uptodate_or_lock(bh))
450 return 1;
451 return __bh_read(bh, op_flags, true);
452 }
453
454 static inline void bh_read_batch(int nr, struct buffer_head *bhs[])
455 {
456 __bh_read_batch(nr, bhs, 0, true);
457 }
458
459 static inline void bh_readahead_batch(int nr, struct buffer_head *bhs[],
460 blk_opf_t op_flags)
461 {
462 __bh_read_batch(nr, bhs, op_flags, false);
463 }
464
465 /**
466 * __bread() - Read a block.
467 * @bdev: The block device to read from.
468 * @block: Block number in units of block size.
469 * @size: The block size of this device in bytes.
470 *
471 * Read a specified block, and return the buffer head that refers
472 * to it. The memory is allocated from the movable area so that it can
473 * be migrated. The returned buffer head has its refcount increased.
474 * The caller should call brelse() when it has finished with the buffer.
475 *
476 * Context: May sleep waiting for I/O.
477 * Return: NULL if the block was unreadable.
478 */
479 static inline struct buffer_head *__bread(struct block_device *bdev,
480 sector_t block, unsigned size)
481 {
482 return __bread_gfp(bdev, block, size, __GFP_MOVABLE);
483 }
484
485 /**
486 * get_nth_bh - Get a reference on the n'th buffer after this one.
487 * @bh: The buffer to start counting from.
488 * @count: How many buffers to skip.
489 *
490 * This is primarily useful for finding the nth buffer in a folio; in
491 * that case you pass the head buffer and the byte offset in the folio
492 * divided by the block size. It can be used for other purposes, but
493 * it will wrap at the end of the folio rather than returning NULL or
494 * proceeding to the next folio for you.
495 *
496 * Return: The requested buffer with an elevated refcount.
497 */
498 static inline __must_check
499 struct buffer_head *get_nth_bh(struct buffer_head *bh, unsigned int count)
500 {
501 while (count--)
502 bh = bh->b_this_page;
503 get_bh(bh);
504 return bh;
505 }
506
507 bool block_dirty_folio(struct address_space *mapping, struct folio *folio);
508
509 #ifdef CONFIG_BUFFER_HEAD
510
511 void buffer_init(void);
512 bool try_to_free_buffers(struct folio *folio);
513 int inode_has_buffers(struct inode *inode);
514 void invalidate_inode_buffers(struct inode *inode);
515 int remove_inode_buffers(struct inode *inode);
516 int sync_mapping_buffers(struct address_space *mapping);
517 void invalidate_bh_lrus(void);
518 void invalidate_bh_lrus_cpu(void);
519 bool has_bh_in_lru(int cpu, void *dummy);
520 extern int buffer_heads_over_limit;
521
522 #else /* CONFIG_BUFFER_HEAD */
523
524 static inline void buffer_init(void) {}
525 static inline bool try_to_free_buffers(struct folio *folio) { return true; }
526 static inline int inode_has_buffers(struct inode *inode) { return 0; }
527 static inline void invalidate_inode_buffers(struct inode *inode) {}
528 static inline int remove_inode_buffers(struct inode *inode) { return 1; }
529 static inline int sync_mapping_buffers(struct address_space *mapping) { return 0; }
530 static inline void invalidate_bh_lrus(void) {}
531 static inline void invalidate_bh_lrus_cpu(void) {}
532 static inline bool has_bh_in_lru(int cpu, void *dummy) { return false; }
533 #define buffer_heads_over_limit 0
534
535 #endif /* CONFIG_BUFFER_HEAD */
536 #endif /* _LINUX_BUFFER_HEAD_H */
537
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.