1 /* SPDX-License-Identifier: GPL-2.0 */
3 * include/linux/buffer_head.h
5 * Everything to do with buffer_heads.
8 #ifndef _LINUX_BUFFER_HEAD_H
9 #define _LINUX_BUFFER_HEAD_H
11 #include <linux/types.h>
12 #include <linux/blk_types.h>
14 #include <linux/linkage.h>
15 #include <linux/pagemap.h>
16 #include <linux/wait.h>
17 #include <linux/atomic.h>
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 */
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 BH_Migrate, /* Buffer is being migrated (norefs) */
39 BH_PrivateStart,/* not a state bit, but the first bit available
40 * for private allocation by other entities
44 #define MAX_BUF_PER_PAGE (PAGE_SIZE / 512)
49 typedef void (bh_end_io_t)(struct buffer_head *bh, int uptodate);
52 * Historically, a buffer_head was used to map a single block
53 * within a page, and of course as the unit of I/O through the
54 * filesystem and block layers. Nowadays the basic I/O unit
55 * is the bio, and buffer_heads are used for extracting block
56 * mappings (via a get_block_t call), for tracking state within
57 * a folio (via a folio_mapping) and for wrapping bio submission
58 * for backward compatibility reasons (e.g. submit_bh).
61 unsigned long b_state; /* buffer state bitmap (see above) */
62 struct buffer_head *b_this_page;/* circular list of page's buffers */
64 struct page *b_page; /* the page this bh is mapped to */
65 struct folio *b_folio; /* the folio this bh is mapped to */
68 sector_t b_blocknr; /* start block number */
69 size_t b_size; /* size of mapping */
70 char *b_data; /* pointer to data within the page */
72 struct block_device *b_bdev;
73 bh_end_io_t *b_end_io; /* I/O completion */
74 void *b_private; /* reserved for b_end_io */
75 struct list_head b_assoc_buffers; /* associated with another mapping */
76 struct address_space *b_assoc_map; /* mapping this buffer is
78 atomic_t b_count; /* users using this buffer_head */
79 spinlock_t b_uptodate_lock; /* Used by the first bh in a page, to
80 * serialise IO completion of other
81 * buffers in the page */
85 * macro tricks to expand the set_buffer_foo(), clear_buffer_foo()
86 * and buffer_foo() functions.
87 * To avoid reset buffer flags that are already set, because that causes
88 * a costly cache line transition, check the flag first.
90 #define BUFFER_FNS(bit, name) \
91 static __always_inline void set_buffer_##name(struct buffer_head *bh) \
93 if (!test_bit(BH_##bit, &(bh)->b_state)) \
94 set_bit(BH_##bit, &(bh)->b_state); \
96 static __always_inline void clear_buffer_##name(struct buffer_head *bh) \
98 clear_bit(BH_##bit, &(bh)->b_state); \
100 static __always_inline int buffer_##name(const struct buffer_head *bh) \
102 return test_bit(BH_##bit, &(bh)->b_state); \
106 * test_set_buffer_foo() and test_clear_buffer_foo()
108 #define TAS_BUFFER_FNS(bit, name) \
109 static __always_inline int test_set_buffer_##name(struct buffer_head *bh) \
111 return test_and_set_bit(BH_##bit, &(bh)->b_state); \
113 static __always_inline int test_clear_buffer_##name(struct buffer_head *bh) \
115 return test_and_clear_bit(BH_##bit, &(bh)->b_state); \
119 * Emit the buffer bitops functions. Note that there are also functions
120 * of the form "mark_buffer_foo()". These are higher-level functions which
121 * do something in addition to setting a b_state bit.
123 BUFFER_FNS(Dirty, dirty)
124 TAS_BUFFER_FNS(Dirty, dirty)
125 BUFFER_FNS(Lock, locked)
127 TAS_BUFFER_FNS(Req, req)
128 BUFFER_FNS(Mapped, mapped)
130 BUFFER_FNS(Async_Read, async_read)
131 BUFFER_FNS(Async_Write, async_write)
132 BUFFER_FNS(Delay, delay)
133 BUFFER_FNS(Boundary, boundary)
134 BUFFER_FNS(Write_EIO, write_io_error)
135 BUFFER_FNS(Unwritten, unwritten)
136 BUFFER_FNS(Meta, meta)
137 BUFFER_FNS(Prio, prio)
138 BUFFER_FNS(Defer_Completion, defer_completion)
140 static __always_inline void set_buffer_uptodate(struct buffer_head *bh)
143 * If somebody else already set this uptodate, they will
144 * have done the memory barrier, and a reader will thus
145 * see *some* valid buffer state.
147 * Any other serialization (with IO errors or whatever that
148 * might clear the bit) has to come from other state (eg BH_Lock).
150 if (test_bit(BH_Uptodate, &bh->b_state))
154 * make it consistent with folio_mark_uptodate
155 * pairs with smp_load_acquire in buffer_uptodate
157 smp_mb__before_atomic();
158 set_bit(BH_Uptodate, &bh->b_state);
161 static __always_inline void clear_buffer_uptodate(struct buffer_head *bh)
163 clear_bit(BH_Uptodate, &bh->b_state);
166 static __always_inline int buffer_uptodate(const struct buffer_head *bh)
169 * make it consistent with folio_test_uptodate
170 * pairs with smp_mb__before_atomic in set_buffer_uptodate
172 return test_bit_acquire(BH_Uptodate, &bh->b_state);
175 static inline unsigned long bh_offset(const struct buffer_head *bh)
177 return (unsigned long)(bh)->b_data & (page_size(bh->b_page) - 1);
180 /* If we *know* page->private refers to buffer_heads */
181 #define page_buffers(page) \
183 BUG_ON(!PagePrivate(page)); \
184 ((struct buffer_head *)page_private(page)); \
186 #define folio_buffers(folio) folio_get_private(folio)
188 void buffer_check_dirty_writeback(struct folio *folio,
189 bool *dirty, bool *writeback);
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,
202 struct buffer_head *alloc_page_buffers(struct page *page, unsigned long size);
203 struct buffer_head *create_empty_buffers(struct folio *folio,
204 unsigned long blocksize, unsigned long b_state);
205 void end_buffer_read_sync(struct buffer_head *bh, int uptodate);
206 void end_buffer_write_sync(struct buffer_head *bh, int uptodate);
208 /* Things to do with buffers at mapping->private_list */
209 void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode);
210 int generic_buffers_fsync_noflush(struct file *file, loff_t start, loff_t end,
212 int generic_buffers_fsync(struct file *file, loff_t start, loff_t end,
214 void clean_bdev_aliases(struct block_device *bdev, sector_t block,
216 static inline void clean_bdev_bh_alias(struct buffer_head *bh)
218 clean_bdev_aliases(bh->b_bdev, bh->b_blocknr, 1);
221 void mark_buffer_async_write(struct buffer_head *bh);
222 void __wait_on_buffer(struct buffer_head *);
223 wait_queue_head_t *bh_waitq_head(struct buffer_head *bh);
224 struct buffer_head *__find_get_block(struct block_device *bdev, sector_t block,
226 struct buffer_head *__find_get_block_nonatomic(struct block_device *bdev,
227 sector_t block, unsigned size);
228 struct buffer_head *bdev_getblk(struct block_device *bdev, sector_t block,
229 unsigned size, gfp_t gfp);
230 void __brelse(struct buffer_head *);
231 void __bforget(struct buffer_head *);
232 void __breadahead(struct block_device *, sector_t block, unsigned int size);
233 struct buffer_head *__bread_gfp(struct block_device *,
234 sector_t block, unsigned size, gfp_t gfp);
235 struct buffer_head *alloc_buffer_head(gfp_t gfp_flags);
236 void free_buffer_head(struct buffer_head * bh);
237 void unlock_buffer(struct buffer_head *bh);
238 void __lock_buffer(struct buffer_head *bh);
239 int sync_dirty_buffer(struct buffer_head *bh);
240 int __sync_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags);
241 void write_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags);
242 void submit_bh(blk_opf_t, struct buffer_head *);
243 void write_boundary_block(struct block_device *bdev,
244 sector_t bblock, unsigned blocksize);
245 int bh_uptodate_or_lock(struct buffer_head *bh);
246 int __bh_read(struct buffer_head *bh, blk_opf_t op_flags, bool wait);
247 void __bh_read_batch(int nr, struct buffer_head *bhs[],
248 blk_opf_t op_flags, bool force_lock);
251 * Generic address_space_operations implementations for buffer_head-backed
254 void block_invalidate_folio(struct folio *folio, size_t offset, size_t length);
255 int block_write_full_folio(struct folio *folio, struct writeback_control *wbc,
257 int __block_write_full_folio(struct inode *inode, struct folio *folio,
258 get_block_t *get_block, struct writeback_control *wbc);
259 int block_read_full_folio(struct folio *, get_block_t *);
260 bool block_is_partially_uptodate(struct folio *, size_t from, size_t count);
261 int block_write_begin(struct address_space *mapping, loff_t pos, unsigned len,
262 struct folio **foliop, get_block_t *get_block);
263 int __block_write_begin(struct folio *folio, loff_t pos, unsigned len,
264 get_block_t *get_block);
265 int block_write_end(struct file *, struct address_space *,
266 loff_t, unsigned len, unsigned copied,
267 struct folio *, void *);
268 int generic_write_end(struct file *, struct address_space *,
269 loff_t, unsigned len, unsigned copied,
270 struct folio *, void *);
271 void folio_zero_new_buffers(struct folio *folio, size_t from, size_t to);
272 int cont_write_begin(struct file *, struct address_space *, loff_t,
273 unsigned, struct folio **, void **,
274 get_block_t *, loff_t *);
275 int generic_cont_expand_simple(struct inode *inode, loff_t size);
276 void block_commit_write(struct folio *folio, size_t from, size_t to);
277 int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
278 get_block_t get_block);
279 sector_t generic_block_bmap(struct address_space *, sector_t, get_block_t *);
280 int block_truncate_page(struct address_space *, loff_t, get_block_t *);
282 #ifdef CONFIG_MIGRATION
283 extern int buffer_migrate_folio(struct address_space *,
284 struct folio *dst, struct folio *src, enum migrate_mode);
285 extern int buffer_migrate_folio_norefs(struct address_space *,
286 struct folio *dst, struct folio *src, enum migrate_mode);
288 #define buffer_migrate_folio NULL
289 #define buffer_migrate_folio_norefs NULL
296 static inline void get_bh(struct buffer_head *bh)
298 atomic_inc(&bh->b_count);
301 static inline void put_bh(struct buffer_head *bh)
303 smp_mb__before_atomic();
304 atomic_dec(&bh->b_count);
308 * brelse - Release a buffer.
309 * @bh: The buffer to release.
311 * Decrement a buffer_head's reference count. If @bh is NULL, this
312 * function is a no-op.
314 * If all buffers on a folio have zero reference count, are clean
315 * and unlocked, and if the folio is unlocked and not under writeback
316 * then try_to_free_buffers() may strip the buffers from the folio in
317 * preparation for freeing it (sometimes, rarely, buffers are removed
318 * from a folio but it ends up not being freed, and buffers may later
321 * Context: Any context.
323 static inline void brelse(struct buffer_head *bh)
330 * bforget - Discard any dirty data in a buffer.
331 * @bh: The buffer to forget.
333 * Call this function instead of brelse() if the data written to a buffer
334 * no longer needs to be written back. It will clear the buffer's dirty
335 * flag so writeback of this buffer will be skipped.
337 * Context: Any context.
339 static inline void bforget(struct buffer_head *bh)
345 static inline struct buffer_head *
346 sb_bread(struct super_block *sb, sector_t block)
348 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE);
351 static inline struct buffer_head *
352 sb_bread_unmovable(struct super_block *sb, sector_t block)
354 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, 0);
358 sb_breadahead(struct super_block *sb, sector_t block)
360 __breadahead(sb->s_bdev, block, sb->s_blocksize);
363 static inline struct buffer_head *getblk_unmovable(struct block_device *bdev,
364 sector_t block, unsigned size)
368 gfp = mapping_gfp_constraint(bdev->bd_mapping, ~__GFP_FS);
371 return bdev_getblk(bdev, block, size, gfp);
374 static inline struct buffer_head *__getblk(struct block_device *bdev,
375 sector_t block, unsigned size)
379 gfp = mapping_gfp_constraint(bdev->bd_mapping, ~__GFP_FS);
380 gfp |= __GFP_MOVABLE | __GFP_NOFAIL;
382 return bdev_getblk(bdev, block, size, gfp);
385 static inline struct buffer_head *sb_getblk(struct super_block *sb,
388 return __getblk(sb->s_bdev, block, sb->s_blocksize);
391 static inline struct buffer_head *sb_getblk_gfp(struct super_block *sb,
392 sector_t block, gfp_t gfp)
394 return bdev_getblk(sb->s_bdev, block, sb->s_blocksize, gfp);
397 static inline struct buffer_head *
398 sb_find_get_block(struct super_block *sb, sector_t block)
400 return __find_get_block(sb->s_bdev, block, sb->s_blocksize);
403 static inline struct buffer_head *
404 sb_find_get_block_nonatomic(struct super_block *sb, sector_t block)
406 return __find_get_block_nonatomic(sb->s_bdev, block, sb->s_blocksize);
410 map_bh(struct buffer_head *bh, struct super_block *sb, sector_t block)
412 set_buffer_mapped(bh);
413 bh->b_bdev = sb->s_bdev;
414 bh->b_blocknr = block;
415 bh->b_size = sb->s_blocksize;
418 static inline void wait_on_buffer(struct buffer_head *bh)
421 if (buffer_locked(bh))
422 __wait_on_buffer(bh);
425 static inline int trylock_buffer(struct buffer_head *bh)
427 return likely(!test_and_set_bit_lock(BH_Lock, &bh->b_state));
430 static inline void lock_buffer(struct buffer_head *bh)
433 if (!trylock_buffer(bh))
437 static inline void bh_readahead(struct buffer_head *bh, blk_opf_t op_flags)
439 if (!buffer_uptodate(bh) && trylock_buffer(bh)) {
440 if (!buffer_uptodate(bh))
441 __bh_read(bh, op_flags, false);
447 static inline void bh_read_nowait(struct buffer_head *bh, blk_opf_t op_flags)
449 if (!bh_uptodate_or_lock(bh))
450 __bh_read(bh, op_flags, false);
453 /* Returns 1 if buffer uptodated, 0 on success, and -EIO on error. */
454 static inline int bh_read(struct buffer_head *bh, blk_opf_t op_flags)
456 if (bh_uptodate_or_lock(bh))
458 return __bh_read(bh, op_flags, true);
461 static inline void bh_read_batch(int nr, struct buffer_head *bhs[])
463 __bh_read_batch(nr, bhs, 0, true);
466 static inline void bh_readahead_batch(int nr, struct buffer_head *bhs[],
469 __bh_read_batch(nr, bhs, op_flags, false);
473 * __bread() - Read a block.
474 * @bdev: The block device to read from.
475 * @block: Block number in units of block size.
476 * @size: The block size of this device in bytes.
478 * Read a specified block, and return the buffer head that refers
479 * to it. The memory is allocated from the movable area so that it can
480 * be migrated. The returned buffer head has its refcount increased.
481 * The caller should call brelse() when it has finished with the buffer.
483 * Context: May sleep waiting for I/O.
484 * Return: NULL if the block was unreadable.
486 static inline struct buffer_head *__bread(struct block_device *bdev,
487 sector_t block, unsigned size)
489 return __bread_gfp(bdev, block, size, __GFP_MOVABLE);
493 * get_nth_bh - Get a reference on the n'th buffer after this one.
494 * @bh: The buffer to start counting from.
495 * @count: How many buffers to skip.
497 * This is primarily useful for finding the nth buffer in a folio; in
498 * that case you pass the head buffer and the byte offset in the folio
499 * divided by the block size. It can be used for other purposes, but
500 * it will wrap at the end of the folio rather than returning NULL or
501 * proceeding to the next folio for you.
503 * Return: The requested buffer with an elevated refcount.
505 static inline __must_check
506 struct buffer_head *get_nth_bh(struct buffer_head *bh, unsigned int count)
509 bh = bh->b_this_page;
514 bool block_dirty_folio(struct address_space *mapping, struct folio *folio);
516 #ifdef CONFIG_BUFFER_HEAD
518 void buffer_init(void);
519 bool try_to_free_buffers(struct folio *folio);
520 int inode_has_buffers(struct inode *inode);
521 void invalidate_inode_buffers(struct inode *inode);
522 int remove_inode_buffers(struct inode *inode);
523 int sync_mapping_buffers(struct address_space *mapping);
524 void invalidate_bh_lrus(void);
525 void invalidate_bh_lrus_cpu(void);
526 bool has_bh_in_lru(int cpu, void *dummy);
527 extern int buffer_heads_over_limit;
529 #else /* CONFIG_BUFFER_HEAD */
531 static inline void buffer_init(void) {}
532 static inline bool try_to_free_buffers(struct folio *folio) { return true; }
533 static inline int inode_has_buffers(struct inode *inode) { return 0; }
534 static inline void invalidate_inode_buffers(struct inode *inode) {}
535 static inline int remove_inode_buffers(struct inode *inode) { return 1; }
536 static inline int sync_mapping_buffers(struct address_space *mapping) { return 0; }
537 static inline void invalidate_bh_lrus(void) {}
538 static inline void invalidate_bh_lrus_cpu(void) {}
539 static inline bool has_bh_in_lru(int cpu, void *dummy) { return false; }
540 #define buffer_heads_over_limit 0
542 #endif /* CONFIG_BUFFER_HEAD */
543 #endif /* _LINUX_BUFFER_HEAD_H */