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>
22 BH_Uptodate, /* Contains valid data */
23 BH_Dirty, /* Is dirty */
24 BH_Lock, /* Is locked */
25 BH_Req, /* Has been submitted for I/O */
27 BH_Mapped, /* Has a disk mapping */
28 BH_New, /* Disk mapping was newly created by get_block */
29 BH_Async_Read, /* Is under end_buffer_async_read I/O */
30 BH_Async_Write, /* Is under end_buffer_async_write I/O */
31 BH_Delay, /* Buffer is not yet allocated on disk */
32 BH_Boundary, /* Block is followed by a discontiguity */
33 BH_Write_EIO, /* I/O error on write */
34 BH_Unwritten, /* Buffer is allocated on disk but not written */
35 BH_Quiet, /* Buffer Error Prinks to be quiet */
36 BH_Meta, /* Buffer contains metadata */
37 BH_Prio, /* Buffer should be submitted with REQ_PRIO */
38 BH_Defer_Completion, /* Defer AIO completion to workqueue */
40 BH_PrivateStart,/* not a state bit, but the first bit available
41 * for private allocation by other entities
45 #define MAX_BUF_PER_PAGE (PAGE_SIZE / 512)
50 typedef void (bh_end_io_t)(struct buffer_head *bh, int uptodate);
53 * Historically, a buffer_head was used to map a single block
54 * within a page, and of course as the unit of I/O through the
55 * filesystem and block layers. Nowadays the basic I/O unit
56 * is the bio, and buffer_heads are used for extracting block
57 * mappings (via a get_block_t call), for tracking state within
58 * a page (via a page_mapping) and for wrapping bio submission
59 * for backward compatibility reasons (e.g. submit_bh).
62 unsigned long b_state; /* buffer state bitmap (see above) */
63 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 struct folio *b_folio; /* the folio this bh is mapped to */
69 sector_t b_blocknr; /* start block number */
70 size_t b_size; /* size of mapping */
71 char *b_data; /* pointer to data within the page */
73 struct block_device *b_bdev;
74 bh_end_io_t *b_end_io; /* I/O completion */
75 void *b_private; /* reserved for b_end_io */
76 struct list_head b_assoc_buffers; /* associated with another mapping */
77 struct address_space *b_assoc_map; /* mapping this buffer is
79 atomic_t b_count; /* users using this buffer_head */
80 spinlock_t b_uptodate_lock; /* Used by the first bh in a page, to
81 * serialise IO completion of other
82 * buffers in the page */
86 * macro tricks to expand the set_buffer_foo(), clear_buffer_foo()
87 * and buffer_foo() functions.
88 * To avoid reset buffer flags that are already set, because that causes
89 * a costly cache line transition, check the flag first.
91 #define BUFFER_FNS(bit, name) \
92 static __always_inline void set_buffer_##name(struct buffer_head *bh) \
94 if (!test_bit(BH_##bit, &(bh)->b_state)) \
95 set_bit(BH_##bit, &(bh)->b_state); \
97 static __always_inline void clear_buffer_##name(struct buffer_head *bh) \
99 clear_bit(BH_##bit, &(bh)->b_state); \
101 static __always_inline int buffer_##name(const struct buffer_head *bh) \
103 return test_bit(BH_##bit, &(bh)->b_state); \
107 * test_set_buffer_foo() and test_clear_buffer_foo()
109 #define TAS_BUFFER_FNS(bit, name) \
110 static __always_inline int test_set_buffer_##name(struct buffer_head *bh) \
112 return test_and_set_bit(BH_##bit, &(bh)->b_state); \
114 static __always_inline int test_clear_buffer_##name(struct buffer_head *bh) \
116 return test_and_clear_bit(BH_##bit, &(bh)->b_state); \
120 * Emit the buffer bitops functions. Note that there are also functions
121 * of the form "mark_buffer_foo()". These are higher-level functions which
122 * do something in addition to setting a b_state bit.
124 BUFFER_FNS(Dirty, dirty)
125 TAS_BUFFER_FNS(Dirty, dirty)
126 BUFFER_FNS(Lock, locked)
128 TAS_BUFFER_FNS(Req, req)
129 BUFFER_FNS(Mapped, mapped)
131 BUFFER_FNS(Async_Read, async_read)
132 BUFFER_FNS(Async_Write, async_write)
133 BUFFER_FNS(Delay, delay)
134 BUFFER_FNS(Boundary, boundary)
135 BUFFER_FNS(Write_EIO, write_io_error)
136 BUFFER_FNS(Unwritten, unwritten)
137 BUFFER_FNS(Meta, meta)
138 BUFFER_FNS(Prio, prio)
139 BUFFER_FNS(Defer_Completion, defer_completion)
141 static __always_inline void set_buffer_uptodate(struct buffer_head *bh)
144 * If somebody else already set this uptodate, they will
145 * have done the memory barrier, and a reader will thus
146 * see *some* valid buffer state.
148 * Any other serialization (with IO errors or whatever that
149 * might clear the bit) has to come from other state (eg BH_Lock).
151 if (test_bit(BH_Uptodate, &bh->b_state))
155 * make it consistent with folio_mark_uptodate
156 * pairs with smp_load_acquire in buffer_uptodate
158 smp_mb__before_atomic();
159 set_bit(BH_Uptodate, &bh->b_state);
162 static __always_inline void clear_buffer_uptodate(struct buffer_head *bh)
164 clear_bit(BH_Uptodate, &bh->b_state);
167 static __always_inline int buffer_uptodate(const struct buffer_head *bh)
170 * make it consistent with folio_test_uptodate
171 * pairs with smp_mb__before_atomic in set_buffer_uptodate
173 return test_bit_acquire(BH_Uptodate, &bh->b_state);
176 #define bh_offset(bh) ((unsigned long)(bh)->b_data & ~PAGE_MASK)
178 /* If we *know* page->private refers to buffer_heads */
179 #define page_buffers(page) \
181 BUG_ON(!PagePrivate(page)); \
182 ((struct buffer_head *)page_private(page)); \
184 #define page_has_buffers(page) PagePrivate(page)
185 #define folio_buffers(folio) folio_get_private(folio)
187 void buffer_check_dirty_writeback(struct folio *folio,
188 bool *dirty, bool *writeback);
194 void mark_buffer_dirty(struct buffer_head *bh);
195 void mark_buffer_write_io_error(struct buffer_head *bh);
196 void touch_buffer(struct buffer_head *bh);
197 void set_bh_page(struct buffer_head *bh,
198 struct page *page, unsigned long offset);
199 void folio_set_bh(struct buffer_head *bh, struct folio *folio,
200 unsigned long offset);
201 bool try_to_free_buffers(struct folio *);
202 struct buffer_head *folio_alloc_buffers(struct folio *folio, unsigned long size,
204 struct buffer_head *alloc_page_buffers(struct page *page, unsigned long size,
206 void create_empty_buffers(struct page *, unsigned long,
207 unsigned long b_state);
208 void folio_create_empty_buffers(struct folio *folio, unsigned long blocksize,
209 unsigned long b_state);
210 void end_buffer_read_sync(struct buffer_head *bh, int uptodate);
211 void end_buffer_write_sync(struct buffer_head *bh, int uptodate);
212 void end_buffer_async_write(struct buffer_head *bh, int uptodate);
214 /* Things to do with buffers at mapping->private_list */
215 void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode);
216 int inode_has_buffers(struct inode *);
217 void invalidate_inode_buffers(struct inode *);
218 int remove_inode_buffers(struct inode *inode);
219 int sync_mapping_buffers(struct address_space *mapping);
220 void clean_bdev_aliases(struct block_device *bdev, sector_t block,
222 static inline void clean_bdev_bh_alias(struct buffer_head *bh)
224 clean_bdev_aliases(bh->b_bdev, bh->b_blocknr, 1);
227 void mark_buffer_async_write(struct buffer_head *bh);
228 void __wait_on_buffer(struct buffer_head *);
229 wait_queue_head_t *bh_waitq_head(struct buffer_head *bh);
230 struct buffer_head *__find_get_block(struct block_device *bdev, sector_t block,
232 struct buffer_head *__getblk_gfp(struct block_device *bdev, sector_t block,
233 unsigned size, gfp_t gfp);
234 void __brelse(struct buffer_head *);
235 void __bforget(struct buffer_head *);
236 void __breadahead(struct block_device *, sector_t block, unsigned int size);
237 struct buffer_head *__bread_gfp(struct block_device *,
238 sector_t block, unsigned size, gfp_t gfp);
239 void invalidate_bh_lrus(void);
240 void invalidate_bh_lrus_cpu(void);
241 bool has_bh_in_lru(int cpu, void *dummy);
242 struct buffer_head *alloc_buffer_head(gfp_t gfp_flags);
243 void free_buffer_head(struct buffer_head * bh);
244 void unlock_buffer(struct buffer_head *bh);
245 void __lock_buffer(struct buffer_head *bh);
246 int sync_dirty_buffer(struct buffer_head *bh);
247 int __sync_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags);
248 void write_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags);
249 void submit_bh(blk_opf_t, struct buffer_head *);
250 void write_boundary_block(struct block_device *bdev,
251 sector_t bblock, unsigned blocksize);
252 int bh_uptodate_or_lock(struct buffer_head *bh);
253 int __bh_read(struct buffer_head *bh, blk_opf_t op_flags, bool wait);
254 void __bh_read_batch(int nr, struct buffer_head *bhs[],
255 blk_opf_t op_flags, bool force_lock);
257 extern int buffer_heads_over_limit;
260 * Generic address_space_operations implementations for buffer_head-backed
263 void block_invalidate_folio(struct folio *folio, size_t offset, size_t length);
264 int block_write_full_page(struct page *page, get_block_t *get_block,
265 struct writeback_control *wbc);
266 int __block_write_full_page(struct inode *inode, struct page *page,
267 get_block_t *get_block, struct writeback_control *wbc,
268 bh_end_io_t *handler);
269 int block_read_full_folio(struct folio *, get_block_t *);
270 bool block_is_partially_uptodate(struct folio *, size_t from, size_t count);
271 int block_write_begin(struct address_space *mapping, loff_t pos, unsigned len,
272 struct page **pagep, get_block_t *get_block);
273 int __block_write_begin(struct page *page, loff_t pos, unsigned len,
274 get_block_t *get_block);
275 int block_write_end(struct file *, struct address_space *,
276 loff_t, unsigned, unsigned,
277 struct page *, void *);
278 int generic_write_end(struct file *, struct address_space *,
279 loff_t, unsigned, unsigned,
280 struct page *, void *);
281 void page_zero_new_buffers(struct page *page, unsigned from, unsigned to);
282 void clean_page_buffers(struct page *page);
283 int cont_write_begin(struct file *, struct address_space *, loff_t,
284 unsigned, struct page **, void **,
285 get_block_t *, loff_t *);
286 int generic_cont_expand_simple(struct inode *inode, loff_t size);
287 int block_commit_write(struct page *page, unsigned from, unsigned to);
288 int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
289 get_block_t get_block);
290 /* Convert errno to return value from ->page_mkwrite() call */
291 static inline vm_fault_t block_page_mkwrite_return(int err)
294 return VM_FAULT_LOCKED;
295 if (err == -EFAULT || err == -EAGAIN)
296 return VM_FAULT_NOPAGE;
299 /* -ENOSPC, -EDQUOT, -EIO ... */
300 return VM_FAULT_SIGBUS;
302 sector_t generic_block_bmap(struct address_space *, sector_t, get_block_t *);
303 int block_truncate_page(struct address_space *, loff_t, get_block_t *);
305 #ifdef CONFIG_MIGRATION
306 extern int buffer_migrate_folio(struct address_space *,
307 struct folio *dst, struct folio *src, enum migrate_mode);
308 extern int buffer_migrate_folio_norefs(struct address_space *,
309 struct folio *dst, struct folio *src, enum migrate_mode);
311 #define buffer_migrate_folio NULL
312 #define buffer_migrate_folio_norefs NULL
315 void buffer_init(void);
321 static inline void get_bh(struct buffer_head *bh)
323 atomic_inc(&bh->b_count);
326 static inline void put_bh(struct buffer_head *bh)
328 smp_mb__before_atomic();
329 atomic_dec(&bh->b_count);
332 static inline void brelse(struct buffer_head *bh)
338 static inline void bforget(struct buffer_head *bh)
344 static inline struct buffer_head *
345 sb_bread(struct super_block *sb, sector_t block)
347 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE);
350 static inline struct buffer_head *
351 sb_bread_unmovable(struct super_block *sb, sector_t block)
353 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, 0);
357 sb_breadahead(struct super_block *sb, sector_t block)
359 __breadahead(sb->s_bdev, block, sb->s_blocksize);
362 static inline struct buffer_head *
363 sb_getblk(struct super_block *sb, sector_t block)
365 return __getblk_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE);
369 static inline struct buffer_head *
370 sb_getblk_gfp(struct super_block *sb, sector_t block, gfp_t gfp)
372 return __getblk_gfp(sb->s_bdev, block, sb->s_blocksize, gfp);
375 static inline struct buffer_head *
376 sb_find_get_block(struct super_block *sb, sector_t block)
378 return __find_get_block(sb->s_bdev, block, sb->s_blocksize);
382 map_bh(struct buffer_head *bh, struct super_block *sb, sector_t block)
384 set_buffer_mapped(bh);
385 bh->b_bdev = sb->s_bdev;
386 bh->b_blocknr = block;
387 bh->b_size = sb->s_blocksize;
390 static inline void wait_on_buffer(struct buffer_head *bh)
393 if (buffer_locked(bh))
394 __wait_on_buffer(bh);
397 static inline int trylock_buffer(struct buffer_head *bh)
399 return likely(!test_and_set_bit_lock(BH_Lock, &bh->b_state));
402 static inline void lock_buffer(struct buffer_head *bh)
405 if (!trylock_buffer(bh))
409 static inline struct buffer_head *getblk_unmovable(struct block_device *bdev,
413 return __getblk_gfp(bdev, block, size, 0);
416 static inline struct buffer_head *__getblk(struct block_device *bdev,
420 return __getblk_gfp(bdev, block, size, __GFP_MOVABLE);
423 static inline void bh_readahead(struct buffer_head *bh, blk_opf_t op_flags)
425 if (!buffer_uptodate(bh) && trylock_buffer(bh)) {
426 if (!buffer_uptodate(bh))
427 __bh_read(bh, op_flags, false);
433 static inline void bh_read_nowait(struct buffer_head *bh, blk_opf_t op_flags)
435 if (!bh_uptodate_or_lock(bh))
436 __bh_read(bh, op_flags, false);
439 /* Returns 1 if buffer uptodated, 0 on success, and -EIO on error. */
440 static inline int bh_read(struct buffer_head *bh, blk_opf_t op_flags)
442 if (bh_uptodate_or_lock(bh))
444 return __bh_read(bh, op_flags, true);
447 static inline void bh_read_batch(int nr, struct buffer_head *bhs[])
449 __bh_read_batch(nr, bhs, 0, true);
452 static inline void bh_readahead_batch(int nr, struct buffer_head *bhs[],
455 __bh_read_batch(nr, bhs, op_flags, false);
459 * __bread() - reads a specified block and returns the bh
460 * @bdev: the block_device to read from
461 * @block: number of block
462 * @size: size (in bytes) to read
464 * Reads a specified block, and returns buffer head that contains it.
465 * The page cache is allocated from movable area so that it can be migrated.
466 * It returns NULL if the block was unreadable.
468 static inline struct buffer_head *
469 __bread(struct block_device *bdev, sector_t block, unsigned size)
471 return __bread_gfp(bdev, block, size, __GFP_MOVABLE);
474 bool block_dirty_folio(struct address_space *mapping, struct folio *folio);
476 #else /* CONFIG_BLOCK */
478 static inline void buffer_init(void) {}
479 static inline bool try_to_free_buffers(struct folio *folio) { return true; }
480 static inline int inode_has_buffers(struct inode *inode) { return 0; }
481 static inline void invalidate_inode_buffers(struct inode *inode) {}
482 static inline int remove_inode_buffers(struct inode *inode) { return 1; }
483 static inline int sync_mapping_buffers(struct address_space *mapping) { return 0; }
484 static inline void invalidate_bh_lrus_cpu(void) {}
485 static inline bool has_bh_in_lru(int cpu, void *dummy) { return false; }
486 #define buffer_heads_over_limit 0
488 #endif /* CONFIG_BLOCK */
489 #endif /* _LINUX_BUFFER_HEAD_H */