Merge tag 'pull-work.iov_iter-rebased' of git://git.kernel.org/pub/scm/linux/kernel...
[linux-block.git] / include / linux / buffer_head.h
CommitLineData
b2441318 1/* SPDX-License-Identifier: GPL-2.0 */
1da177e4
LT
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>
3ae72869 12#include <linux/blk_types.h>
1da177e4
LT
13#include <linux/fs.h>
14#include <linux/linkage.h>
15#include <linux/pagemap.h>
16#include <linux/wait.h>
60063497 17#include <linux/atomic.h>
1da177e4 18
9361401e
DH
19#ifdef CONFIG_BLOCK
20
1da177e4
LT
21enum bh_state_bits {
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 */
26
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 */
33a266dd 34 BH_Unwritten, /* Buffer is allocated on disk but not written */
08bafc03 35 BH_Quiet, /* Buffer Error Prinks to be quiet */
877f962c
TT
36 BH_Meta, /* Buffer contains metadata */
37 BH_Prio, /* Buffer should be submitted with REQ_PRIO */
7b7a8665 38 BH_Defer_Completion, /* Defer AIO completion to workqueue */
1da177e4
LT
39
40 BH_PrivateStart,/* not a state bit, but the first bit available
41 * for private allocation by other entities
42 */
43};
44
09cbfeaf 45#define MAX_BUF_PER_PAGE (PAGE_SIZE / 512)
1da177e4
LT
46
47struct page;
48struct buffer_head;
49struct address_space;
50typedef void (bh_end_io_t)(struct buffer_head *bh, int uptodate);
51
52/*
205f87f6
BP
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).
1da177e4
LT
60 */
61struct buffer_head {
1da177e4
LT
62 unsigned long b_state; /* buffer state bitmap (see above) */
63 struct buffer_head *b_this_page;/* circular list of page's buffers */
64 struct page *b_page; /* the page this bh is mapped to */
1da177e4 65
205f87f6
BP
66 sector_t b_blocknr; /* start block number */
67 size_t b_size; /* size of mapping */
68 char *b_data; /* pointer to data within the page */
1da177e4
LT
69
70 struct block_device *b_bdev;
71 bh_end_io_t *b_end_io; /* I/O completion */
72 void *b_private; /* reserved for b_end_io */
73 struct list_head b_assoc_buffers; /* associated with another mapping */
58ff407b
JK
74 struct address_space *b_assoc_map; /* mapping this buffer is
75 associated with */
205f87f6 76 atomic_t b_count; /* users using this buffer_head */
f1e67e35
TG
77 spinlock_t b_uptodate_lock; /* Used by the first bh in a page, to
78 * serialise IO completion of other
79 * buffers in the page */
1da177e4
LT
80};
81
82/*
83 * macro tricks to expand the set_buffer_foo(), clear_buffer_foo()
84 * and buffer_foo() functions.
60f91826
KW
85 * To avoid reset buffer flags that are already set, because that causes
86 * a costly cache line transition, check the flag first.
1da177e4
LT
87 */
88#define BUFFER_FNS(bit, name) \
ee91ef61 89static __always_inline void set_buffer_##name(struct buffer_head *bh) \
1da177e4 90{ \
60f91826
KW
91 if (!test_bit(BH_##bit, &(bh)->b_state)) \
92 set_bit(BH_##bit, &(bh)->b_state); \
1da177e4 93} \
ee91ef61 94static __always_inline void clear_buffer_##name(struct buffer_head *bh) \
1da177e4
LT
95{ \
96 clear_bit(BH_##bit, &(bh)->b_state); \
97} \
ee91ef61 98static __always_inline int buffer_##name(const struct buffer_head *bh) \
1da177e4
LT
99{ \
100 return test_bit(BH_##bit, &(bh)->b_state); \
101}
102
103/*
104 * test_set_buffer_foo() and test_clear_buffer_foo()
105 */
106#define TAS_BUFFER_FNS(bit, name) \
ee91ef61 107static __always_inline int test_set_buffer_##name(struct buffer_head *bh) \
1da177e4
LT
108{ \
109 return test_and_set_bit(BH_##bit, &(bh)->b_state); \
110} \
ee91ef61 111static __always_inline int test_clear_buffer_##name(struct buffer_head *bh) \
1da177e4
LT
112{ \
113 return test_and_clear_bit(BH_##bit, &(bh)->b_state); \
114} \
115
116/*
117 * Emit the buffer bitops functions. Note that there are also functions
118 * of the form "mark_buffer_foo()". These are higher-level functions which
119 * do something in addition to setting a b_state bit.
120 */
121BUFFER_FNS(Uptodate, uptodate)
122BUFFER_FNS(Dirty, dirty)
123TAS_BUFFER_FNS(Dirty, dirty)
124BUFFER_FNS(Lock, locked)
1da177e4
LT
125BUFFER_FNS(Req, req)
126TAS_BUFFER_FNS(Req, req)
127BUFFER_FNS(Mapped, mapped)
128BUFFER_FNS(New, new)
129BUFFER_FNS(Async_Read, async_read)
130BUFFER_FNS(Async_Write, async_write)
131BUFFER_FNS(Delay, delay)
132BUFFER_FNS(Boundary, boundary)
133BUFFER_FNS(Write_EIO, write_io_error)
33a266dd 134BUFFER_FNS(Unwritten, unwritten)
877f962c
TT
135BUFFER_FNS(Meta, meta)
136BUFFER_FNS(Prio, prio)
7b7a8665 137BUFFER_FNS(Defer_Completion, defer_completion)
1da177e4
LT
138
139#define bh_offset(bh) ((unsigned long)(bh)->b_data & ~PAGE_MASK)
1da177e4
LT
140
141/* If we *know* page->private refers to buffer_heads */
142#define page_buffers(page) \
143 ({ \
4c21e2f2
HD
144 BUG_ON(!PagePrivate(page)); \
145 ((struct buffer_head *)page_private(page)); \
1da177e4
LT
146 })
147#define page_has_buffers(page) PagePrivate(page)
cd1067be 148#define folio_buffers(folio) folio_get_private(folio)
1da177e4 149
520f301c 150void buffer_check_dirty_writeback(struct folio *folio,
b4597226
MG
151 bool *dirty, bool *writeback);
152
1da177e4
LT
153/*
154 * Declarations
155 */
156
b3c97528 157void mark_buffer_dirty(struct buffer_head *bh);
87354e5d 158void mark_buffer_write_io_error(struct buffer_head *bh);
f0059afd 159void touch_buffer(struct buffer_head *bh);
1da177e4
LT
160void set_bh_page(struct buffer_head *bh,
161 struct page *page, unsigned long offset);
68189fef 162bool try_to_free_buffers(struct folio *);
1da177e4 163struct buffer_head *alloc_page_buffers(struct page *page, unsigned long size,
640ab98f 164 bool retry);
1da177e4
LT
165void create_empty_buffers(struct page *, unsigned long,
166 unsigned long b_state);
167void end_buffer_read_sync(struct buffer_head *bh, int uptodate);
168void end_buffer_write_sync(struct buffer_head *bh, int uptodate);
35c80d5f 169void end_buffer_async_write(struct buffer_head *bh, int uptodate);
1da177e4
LT
170
171/* Things to do with buffers at mapping->private_list */
172void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode);
173int inode_has_buffers(struct inode *);
174void invalidate_inode_buffers(struct inode *);
175int remove_inode_buffers(struct inode *inode);
176int sync_mapping_buffers(struct address_space *mapping);
29f3ad7d
JK
177void clean_bdev_aliases(struct block_device *bdev, sector_t block,
178 sector_t len);
e64855c6
JK
179static inline void clean_bdev_bh_alias(struct buffer_head *bh)
180{
181 clean_bdev_aliases(bh->b_bdev, bh->b_blocknr, 1);
182}
1da177e4
LT
183
184void mark_buffer_async_write(struct buffer_head *bh);
1da177e4
LT
185void __wait_on_buffer(struct buffer_head *);
186wait_queue_head_t *bh_waitq_head(struct buffer_head *bh);
3991d3bd
TK
187struct buffer_head *__find_get_block(struct block_device *bdev, sector_t block,
188 unsigned size);
3b5e6454
GK
189struct buffer_head *__getblk_gfp(struct block_device *bdev, sector_t block,
190 unsigned size, gfp_t gfp);
1da177e4
LT
191void __brelse(struct buffer_head *);
192void __bforget(struct buffer_head *);
3991d3bd 193void __breadahead(struct block_device *, sector_t block, unsigned int size);
d87f6392
RG
194void __breadahead_gfp(struct block_device *, sector_t block, unsigned int size,
195 gfp_t gfp);
3b5e6454
GK
196struct buffer_head *__bread_gfp(struct block_device *,
197 sector_t block, unsigned size, gfp_t gfp);
f9a14399 198void invalidate_bh_lrus(void);
243418e3 199void invalidate_bh_lrus_cpu(void);
8cc621d2 200bool has_bh_in_lru(int cpu, void *dummy);
dd0fc66f 201struct buffer_head *alloc_buffer_head(gfp_t gfp_flags);
1da177e4 202void free_buffer_head(struct buffer_head * bh);
b3c97528
HH
203void unlock_buffer(struct buffer_head *bh);
204void __lock_buffer(struct buffer_head *bh);
1420c4a5 205void ll_rw_block(blk_opf_t, int, struct buffer_head * bh[]);
1da177e4 206int sync_dirty_buffer(struct buffer_head *bh);
3ae72869
BVA
207int __sync_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags);
208void write_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags);
1420c4a5 209int submit_bh(blk_opf_t, struct buffer_head *);
1da177e4
LT
210void write_boundary_block(struct block_device *bdev,
211 sector_t bblock, unsigned blocksize);
389d1b08
AK
212int bh_uptodate_or_lock(struct buffer_head *bh);
213int bh_submit_read(struct buffer_head *bh);
1da177e4
LT
214
215extern int buffer_heads_over_limit;
216
217/*
218 * Generic address_space_operations implementations for buffer_head-backed
219 * address_spaces.
220 */
7ba13abb 221void block_invalidate_folio(struct folio *folio, size_t offset, size_t length);
1da177e4
LT
222int block_write_full_page(struct page *page, get_block_t *get_block,
223 struct writeback_control *wbc);
b4bba389
BM
224int __block_write_full_page(struct inode *inode, struct page *page,
225 get_block_t *get_block, struct writeback_control *wbc,
226 bh_end_io_t *handler);
2c69e205 227int block_read_full_folio(struct folio *, get_block_t *);
2e7e80f7 228bool block_is_partially_uptodate(struct folio *, size_t from, size_t count);
155130a4 229int block_write_begin(struct address_space *mapping, loff_t pos, unsigned len,
b3992d1e 230 struct page **pagep, get_block_t *get_block);
6e1db88d
CH
231int __block_write_begin(struct page *page, loff_t pos, unsigned len,
232 get_block_t *get_block);
afddba49
NP
233int block_write_end(struct file *, struct address_space *,
234 loff_t, unsigned, unsigned,
235 struct page *, void *);
236int generic_write_end(struct file *, struct address_space *,
237 loff_t, unsigned, unsigned,
238 struct page *, void *);
239void page_zero_new_buffers(struct page *page, unsigned from, unsigned to);
f892760a 240void clean_page_buffers(struct page *page);
89e10787 241int cont_write_begin(struct file *, struct address_space *, loff_t,
be3bbbc5 242 unsigned, struct page **, void **,
89e10787 243 get_block_t *, loff_t *);
05eb0b51 244int generic_cont_expand_simple(struct inode *inode, loff_t size);
1da177e4 245int block_commit_write(struct page *page, unsigned from, unsigned to);
c2ec175c 246int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
54171690 247 get_block_t get_block);
24da4fab 248/* Convert errno to return value from ->page_mkwrite() call */
401b25aa 249static inline vm_fault_t block_page_mkwrite_return(int err)
24da4fab
JK
250{
251 if (err == 0)
252 return VM_FAULT_LOCKED;
0911d004 253 if (err == -EFAULT || err == -EAGAIN)
24da4fab
JK
254 return VM_FAULT_NOPAGE;
255 if (err == -ENOMEM)
256 return VM_FAULT_OOM;
257 /* -ENOSPC, -EDQUOT, -EIO ... */
258 return VM_FAULT_SIGBUS;
259}
1da177e4 260sector_t generic_block_bmap(struct address_space *, sector_t, get_block_t *);
1da177e4 261int block_truncate_page(struct address_space *, loff_t, get_block_t *);
1da177e4 262
67235182
MWO
263#ifdef CONFIG_MIGRATION
264extern int buffer_migrate_folio(struct address_space *,
265 struct folio *dst, struct folio *src, enum migrate_mode);
266extern int buffer_migrate_folio_norefs(struct address_space *,
267 struct folio *dst, struct folio *src, enum migrate_mode);
268#else
269#define buffer_migrate_folio NULL
270#define buffer_migrate_folio_norefs NULL
271#endif
1da177e4 272
b6cd0b77 273void buffer_init(void);
1da177e4
LT
274
275/*
276 * inline definitions
277 */
278
1da177e4
LT
279static inline void get_bh(struct buffer_head *bh)
280{
281 atomic_inc(&bh->b_count);
282}
283
284static inline void put_bh(struct buffer_head *bh)
285{
4e857c58 286 smp_mb__before_atomic();
1da177e4
LT
287 atomic_dec(&bh->b_count);
288}
289
290static inline void brelse(struct buffer_head *bh)
291{
292 if (bh)
293 __brelse(bh);
294}
295
296static inline void bforget(struct buffer_head *bh)
297{
298 if (bh)
299 __bforget(bh);
300}
301
302static inline struct buffer_head *
303sb_bread(struct super_block *sb, sector_t block)
304{
3b5e6454
GK
305 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE);
306}
307
308static inline struct buffer_head *
309sb_bread_unmovable(struct super_block *sb, sector_t block)
310{
311 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, 0);
1da177e4
LT
312}
313
314static inline void
315sb_breadahead(struct super_block *sb, sector_t block)
316{
317 __breadahead(sb->s_bdev, block, sb->s_blocksize);
318}
319
d87f6392
RG
320static inline void
321sb_breadahead_unmovable(struct super_block *sb, sector_t block)
322{
323 __breadahead_gfp(sb->s_bdev, block, sb->s_blocksize, 0);
324}
325
1da177e4
LT
326static inline struct buffer_head *
327sb_getblk(struct super_block *sb, sector_t block)
328{
3b5e6454 329 return __getblk_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE);
1da177e4
LT
330}
331
bd7ade3c
NB
332
333static inline struct buffer_head *
334sb_getblk_gfp(struct super_block *sb, sector_t block, gfp_t gfp)
335{
336 return __getblk_gfp(sb->s_bdev, block, sb->s_blocksize, gfp);
337}
338
1da177e4
LT
339static inline struct buffer_head *
340sb_find_get_block(struct super_block *sb, sector_t block)
341{
342 return __find_get_block(sb->s_bdev, block, sb->s_blocksize);
343}
344
345static inline void
346map_bh(struct buffer_head *bh, struct super_block *sb, sector_t block)
347{
348 set_buffer_mapped(bh);
349 bh->b_bdev = sb->s_bdev;
350 bh->b_blocknr = block;
b0cf2321 351 bh->b_size = sb->s_blocksize;
1da177e4
LT
352}
353
1da177e4
LT
354static inline void wait_on_buffer(struct buffer_head *bh)
355{
356 might_sleep();
a9877cc2 357 if (buffer_locked(bh))
1da177e4
LT
358 __wait_on_buffer(bh);
359}
360
ca5de404
NP
361static inline int trylock_buffer(struct buffer_head *bh)
362{
51b07fc3 363 return likely(!test_and_set_bit_lock(BH_Lock, &bh->b_state));
ca5de404
NP
364}
365
1da177e4
LT
366static inline void lock_buffer(struct buffer_head *bh)
367{
368 might_sleep();
ca5de404 369 if (!trylock_buffer(bh))
1da177e4
LT
370 __lock_buffer(bh);
371}
372
3b5e6454
GK
373static inline struct buffer_head *getblk_unmovable(struct block_device *bdev,
374 sector_t block,
375 unsigned size)
376{
377 return __getblk_gfp(bdev, block, size, 0);
378}
379
380static inline struct buffer_head *__getblk(struct block_device *bdev,
381 sector_t block,
382 unsigned size)
383{
384 return __getblk_gfp(bdev, block, size, __GFP_MOVABLE);
385}
386
387/**
388 * __bread() - reads a specified block and returns the bh
389 * @bdev: the block_device to read from
390 * @block: number of block
391 * @size: size (in bytes) to read
392 *
393 * Reads a specified block, and returns buffer head that contains it.
394 * The page cache is allocated from movable area so that it can be migrated.
395 * It returns NULL if the block was unreadable.
396 */
397static inline struct buffer_head *
398__bread(struct block_device *bdev, sector_t block, unsigned size)
399{
400 return __bread_gfp(bdev, block, size, __GFP_MOVABLE);
401}
402
e621900a 403bool block_dirty_folio(struct address_space *mapping, struct folio *folio);
9361401e
DH
404
405#else /* CONFIG_BLOCK */
406
407static inline void buffer_init(void) {}
68189fef 408static inline bool try_to_free_buffers(struct folio *folio) { return true; }
9361401e
DH
409static inline int inode_has_buffers(struct inode *inode) { return 0; }
410static inline void invalidate_inode_buffers(struct inode *inode) {}
411static inline int remove_inode_buffers(struct inode *inode) { return 1; }
412static inline int sync_mapping_buffers(struct address_space *mapping) { return 0; }
243418e3 413static inline void invalidate_bh_lrus_cpu(void) {}
6de522d1 414static inline bool has_bh_in_lru(int cpu, void *dummy) { return false; }
d2de7ea4 415#define buffer_heads_over_limit 0
9361401e
DH
416
417#endif /* CONFIG_BLOCK */
1da177e4 418#endif /* _LINUX_BUFFER_HEAD_H */