Merge tag 'pci-v6.16-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/pci/pci
[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
LT
18
19enum 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 */
33a266dd 32 BH_Unwritten, /* Buffer is allocated on disk but not written */
08bafc03 33 BH_Quiet, /* Buffer Error Prinks to be quiet */
877f962c
TT
34 BH_Meta, /* Buffer contains metadata */
35 BH_Prio, /* Buffer should be submitted with REQ_PRIO */
7b7a8665 36 BH_Defer_Completion, /* Defer AIO completion to workqueue */
2d900eff 37 BH_Migrate, /* Buffer is being migrated (norefs) */
1da177e4
LT
38
39 BH_PrivateStart,/* not a state bit, but the first bit available
40 * for private allocation by other entities
41 */
42};
43
09cbfeaf 44#define MAX_BUF_PER_PAGE (PAGE_SIZE / 512)
1da177e4
LT
45
46struct page;
47struct buffer_head;
48struct address_space;
49typedef void (bh_end_io_t)(struct buffer_head *bh, int uptodate);
50
51/*
205f87f6
BP
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
06668257 57 * a folio (via a folio_mapping) and for wrapping bio submission
205f87f6 58 * for backward compatibility reasons (e.g. submit_bh).
1da177e4
LT
59 */
60struct buffer_head {
1da177e4
LT
61 unsigned long b_state; /* buffer state bitmap (see above) */
62 struct buffer_head *b_this_page;/* circular list of page's buffers */
d685c668
MWO
63 union {
64 struct page *b_page; /* the page this bh is mapped to */
65 struct folio *b_folio; /* the folio this bh is mapped to */
66 };
1da177e4 67
205f87f6
BP
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 */
1da177e4
LT
71
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 */
58ff407b
JK
76 struct address_space *b_assoc_map; /* mapping this buffer is
77 associated with */
205f87f6 78 atomic_t b_count; /* users using this buffer_head */
f1e67e35
TG
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 */
1da177e4
LT
82};
83
84/*
85 * macro tricks to expand the set_buffer_foo(), clear_buffer_foo()
86 * and buffer_foo() functions.
60f91826
KW
87 * To avoid reset buffer flags that are already set, because that causes
88 * a costly cache line transition, check the flag first.
1da177e4
LT
89 */
90#define BUFFER_FNS(bit, name) \
ee91ef61 91static __always_inline void set_buffer_##name(struct buffer_head *bh) \
1da177e4 92{ \
60f91826
KW
93 if (!test_bit(BH_##bit, &(bh)->b_state)) \
94 set_bit(BH_##bit, &(bh)->b_state); \
1da177e4 95} \
ee91ef61 96static __always_inline void clear_buffer_##name(struct buffer_head *bh) \
1da177e4
LT
97{ \
98 clear_bit(BH_##bit, &(bh)->b_state); \
99} \
ee91ef61 100static __always_inline int buffer_##name(const struct buffer_head *bh) \
1da177e4
LT
101{ \
102 return test_bit(BH_##bit, &(bh)->b_state); \
103}
104
105/*
106 * test_set_buffer_foo() and test_clear_buffer_foo()
107 */
108#define TAS_BUFFER_FNS(bit, name) \
ee91ef61 109static __always_inline int test_set_buffer_##name(struct buffer_head *bh) \
1da177e4
LT
110{ \
111 return test_and_set_bit(BH_##bit, &(bh)->b_state); \
112} \
ee91ef61 113static __always_inline int test_clear_buffer_##name(struct buffer_head *bh) \
1da177e4
LT
114{ \
115 return test_and_clear_bit(BH_##bit, &(bh)->b_state); \
116} \
117
118/*
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.
122 */
1da177e4
LT
123BUFFER_FNS(Dirty, dirty)
124TAS_BUFFER_FNS(Dirty, dirty)
125BUFFER_FNS(Lock, locked)
1da177e4
LT
126BUFFER_FNS(Req, req)
127TAS_BUFFER_FNS(Req, req)
128BUFFER_FNS(Mapped, mapped)
129BUFFER_FNS(New, new)
130BUFFER_FNS(Async_Read, async_read)
131BUFFER_FNS(Async_Write, async_write)
132BUFFER_FNS(Delay, delay)
133BUFFER_FNS(Boundary, boundary)
134BUFFER_FNS(Write_EIO, write_io_error)
33a266dd 135BUFFER_FNS(Unwritten, unwritten)
877f962c
TT
136BUFFER_FNS(Meta, meta)
137BUFFER_FNS(Prio, prio)
7b7a8665 138BUFFER_FNS(Defer_Completion, defer_completion)
1da177e4 139
d4252071
MP
140static __always_inline void set_buffer_uptodate(struct buffer_head *bh)
141{
2f79cdfe
LT
142 /*
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.
146 *
147 * Any other serialization (with IO errors or whatever that
148 * might clear the bit) has to come from other state (eg BH_Lock).
149 */
150 if (test_bit(BH_Uptodate, &bh->b_state))
151 return;
152
d4252071
MP
153 /*
154 * make it consistent with folio_mark_uptodate
155 * pairs with smp_load_acquire in buffer_uptodate
156 */
157 smp_mb__before_atomic();
158 set_bit(BH_Uptodate, &bh->b_state);
159}
160
161static __always_inline void clear_buffer_uptodate(struct buffer_head *bh)
162{
163 clear_bit(BH_Uptodate, &bh->b_state);
164}
165
166static __always_inline int buffer_uptodate(const struct buffer_head *bh)
167{
168 /*
169 * make it consistent with folio_test_uptodate
170 * pairs with smp_mb__before_atomic in set_buffer_uptodate
171 */
8238b457 172 return test_bit_acquire(BH_Uptodate, &bh->b_state);
d4252071
MP
173}
174
f94cf220
MWO
175static inline unsigned long bh_offset(const struct buffer_head *bh)
176{
177 return (unsigned long)(bh)->b_data & (page_size(bh->b_page) - 1);
178}
1da177e4
LT
179
180/* If we *know* page->private refers to buffer_heads */
181#define page_buffers(page) \
182 ({ \
4c21e2f2
HD
183 BUG_ON(!PagePrivate(page)); \
184 ((struct buffer_head *)page_private(page)); \
1da177e4 185 })
cd1067be 186#define folio_buffers(folio) folio_get_private(folio)
1da177e4 187
520f301c 188void buffer_check_dirty_writeback(struct folio *folio,
b4597226
MG
189 bool *dirty, bool *writeback);
190
1da177e4
LT
191/*
192 * Declarations
193 */
194
b3c97528 195void mark_buffer_dirty(struct buffer_head *bh);
87354e5d 196void mark_buffer_write_io_error(struct buffer_head *bh);
f0059afd 197void touch_buffer(struct buffer_head *bh);
465e5e6a
PR
198void folio_set_bh(struct buffer_head *bh, struct folio *folio,
199 unsigned long offset);
c71124a8 200struct buffer_head *folio_alloc_buffers(struct folio *folio, unsigned long size,
2a418157 201 gfp_t gfp);
5c40e050 202struct buffer_head *alloc_page_buffers(struct page *page, unsigned long size);
0a88810d 203struct buffer_head *create_empty_buffers(struct folio *folio,
3decb856 204 unsigned long blocksize, unsigned long b_state);
1da177e4
LT
205void end_buffer_read_sync(struct buffer_head *bh, int uptodate);
206void end_buffer_write_sync(struct buffer_head *bh, int uptodate);
1da177e4
LT
207
208/* Things to do with buffers at mapping->private_list */
209void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode);
31b2ebc0
RHI
210int generic_buffers_fsync_noflush(struct file *file, loff_t start, loff_t end,
211 bool datasync);
212int generic_buffers_fsync(struct file *file, loff_t start, loff_t end,
213 bool datasync);
29f3ad7d
JK
214void clean_bdev_aliases(struct block_device *bdev, sector_t block,
215 sector_t len);
e64855c6
JK
216static inline void clean_bdev_bh_alias(struct buffer_head *bh)
217{
218 clean_bdev_aliases(bh->b_bdev, bh->b_blocknr, 1);
219}
1da177e4
LT
220
221void mark_buffer_async_write(struct buffer_head *bh);
1da177e4
LT
222void __wait_on_buffer(struct buffer_head *);
223wait_queue_head_t *bh_waitq_head(struct buffer_head *bh);
3991d3bd
TK
224struct buffer_head *__find_get_block(struct block_device *bdev, sector_t block,
225 unsigned size);
2814a7d3
DB
226struct buffer_head *__find_get_block_nonatomic(struct block_device *bdev,
227 sector_t block, unsigned size);
3ed65f04
MWO
228struct buffer_head *bdev_getblk(struct block_device *bdev, sector_t block,
229 unsigned size, gfp_t gfp);
1da177e4
LT
230void __brelse(struct buffer_head *);
231void __bforget(struct buffer_head *);
3991d3bd 232void __breadahead(struct block_device *, sector_t block, unsigned int size);
3b5e6454
GK
233struct buffer_head *__bread_gfp(struct block_device *,
234 sector_t block, unsigned size, gfp_t gfp);
dd0fc66f 235struct buffer_head *alloc_buffer_head(gfp_t gfp_flags);
1da177e4 236void free_buffer_head(struct buffer_head * bh);
b3c97528
HH
237void unlock_buffer(struct buffer_head *bh);
238void __lock_buffer(struct buffer_head *bh);
1da177e4 239int sync_dirty_buffer(struct buffer_head *bh);
3ae72869
BVA
240int __sync_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags);
241void write_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags);
5bdf402a 242void submit_bh(blk_opf_t, struct buffer_head *);
1da177e4
LT
243void write_boundary_block(struct block_device *bdev,
244 sector_t bblock, unsigned blocksize);
389d1b08 245int bh_uptodate_or_lock(struct buffer_head *bh);
fdee117e
ZY
246int __bh_read(struct buffer_head *bh, blk_opf_t op_flags, bool wait);
247void __bh_read_batch(int nr, struct buffer_head *bhs[],
248 blk_opf_t op_flags, bool force_lock);
1da177e4 249
1da177e4
LT
250/*
251 * Generic address_space_operations implementations for buffer_head-backed
252 * address_spaces.
253 */
7ba13abb 254void block_invalidate_folio(struct folio *folio, size_t offset, size_t length);
17bf23a9
MWO
255int block_write_full_folio(struct folio *folio, struct writeback_control *wbc,
256 void *get_block);
53418a18 257int __block_write_full_folio(struct inode *inode, struct folio *folio,
14059f66 258 get_block_t *get_block, struct writeback_control *wbc);
2c69e205 259int block_read_full_folio(struct folio *, get_block_t *);
2e7e80f7 260bool block_is_partially_uptodate(struct folio *, size_t from, size_t count);
155130a4 261int block_write_begin(struct address_space *mapping, loff_t pos, unsigned len,
1da86618 262 struct folio **foliop, get_block_t *get_block);
9f04609f 263int __block_write_begin(struct folio *folio, loff_t pos, unsigned len,
6e1db88d 264 get_block_t *get_block);
afddba49 265int block_write_end(struct file *, struct address_space *,
97edbc02
MWO
266 loff_t, unsigned len, unsigned copied,
267 struct folio *, void *);
afddba49 268int generic_write_end(struct file *, struct address_space *,
a225800f
MWO
269 loff_t, unsigned len, unsigned copied,
270 struct folio *, void *);
4a9622f2 271void folio_zero_new_buffers(struct folio *folio, size_t from, size_t to);
89e10787 272int cont_write_begin(struct file *, struct address_space *, loff_t,
1da86618 273 unsigned, struct folio **, void **,
89e10787 274 get_block_t *, loff_t *);
05eb0b51 275int generic_cont_expand_simple(struct inode *inode, loff_t size);
63a23847 276void block_commit_write(struct folio *folio, size_t from, size_t to);
c2ec175c 277int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
54171690 278 get_block_t get_block);
1da177e4 279sector_t generic_block_bmap(struct address_space *, sector_t, get_block_t *);
1da177e4 280int block_truncate_page(struct address_space *, loff_t, get_block_t *);
1da177e4 281
67235182
MWO
282#ifdef CONFIG_MIGRATION
283extern int buffer_migrate_folio(struct address_space *,
284 struct folio *dst, struct folio *src, enum migrate_mode);
285extern int buffer_migrate_folio_norefs(struct address_space *,
286 struct folio *dst, struct folio *src, enum migrate_mode);
287#else
288#define buffer_migrate_folio NULL
289#define buffer_migrate_folio_norefs NULL
290#endif
1da177e4 291
1da177e4
LT
292/*
293 * inline definitions
294 */
295
1da177e4
LT
296static inline void get_bh(struct buffer_head *bh)
297{
298 atomic_inc(&bh->b_count);
299}
300
301static inline void put_bh(struct buffer_head *bh)
302{
4e857c58 303 smp_mb__before_atomic();
1da177e4
LT
304 atomic_dec(&bh->b_count);
305}
306
66924fda
MWO
307/**
308 * brelse - Release a buffer.
309 * @bh: The buffer to release.
310 *
311 * Decrement a buffer_head's reference count. If @bh is NULL, this
312 * function is a no-op.
313 *
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
319 * be reattached).
320 *
321 * Context: Any context.
322 */
1da177e4
LT
323static inline void brelse(struct buffer_head *bh)
324{
325 if (bh)
326 __brelse(bh);
327}
328
b73a936f
MWO
329/**
330 * bforget - Discard any dirty data in a buffer.
331 * @bh: The buffer to forget.
332 *
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.
336 *
337 * Context: Any context.
338 */
1da177e4
LT
339static inline void bforget(struct buffer_head *bh)
340{
341 if (bh)
342 __bforget(bh);
343}
344
345static inline struct buffer_head *
346sb_bread(struct super_block *sb, sector_t block)
347{
3b5e6454
GK
348 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE);
349}
350
351static inline struct buffer_head *
352sb_bread_unmovable(struct super_block *sb, sector_t block)
353{
354 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, 0);
1da177e4
LT
355}
356
357static inline void
358sb_breadahead(struct super_block *sb, sector_t block)
359{
360 __breadahead(sb->s_bdev, block, sb->s_blocksize);
361}
362
c645e65c
MWO
363static inline struct buffer_head *getblk_unmovable(struct block_device *bdev,
364 sector_t block, unsigned size)
365{
366 gfp_t gfp;
367
224941e8 368 gfp = mapping_gfp_constraint(bdev->bd_mapping, ~__GFP_FS);
c645e65c
MWO
369 gfp |= __GFP_NOFAIL;
370
371 return bdev_getblk(bdev, block, size, gfp);
372}
373
374static inline struct buffer_head *__getblk(struct block_device *bdev,
375 sector_t block, unsigned size)
376{
377 gfp_t gfp;
378
224941e8 379 gfp = mapping_gfp_constraint(bdev->bd_mapping, ~__GFP_FS);
c645e65c
MWO
380 gfp |= __GFP_MOVABLE | __GFP_NOFAIL;
381
382 return bdev_getblk(bdev, block, size, gfp);
383}
384
4b9c8b19
MWO
385static inline struct buffer_head *sb_getblk(struct super_block *sb,
386 sector_t block)
1da177e4 387{
4b9c8b19 388 return __getblk(sb->s_bdev, block, sb->s_blocksize);
1da177e4
LT
389}
390
8a83ac54
MWO
391static inline struct buffer_head *sb_getblk_gfp(struct super_block *sb,
392 sector_t block, gfp_t gfp)
bd7ade3c 393{
8a83ac54 394 return bdev_getblk(sb->s_bdev, block, sb->s_blocksize, gfp);
bd7ade3c
NB
395}
396
1da177e4
LT
397static inline struct buffer_head *
398sb_find_get_block(struct super_block *sb, sector_t block)
399{
400 return __find_get_block(sb->s_bdev, block, sb->s_blocksize);
401}
402
2814a7d3
DB
403static inline struct buffer_head *
404sb_find_get_block_nonatomic(struct super_block *sb, sector_t block)
405{
406 return __find_get_block_nonatomic(sb->s_bdev, block, sb->s_blocksize);
407}
408
1da177e4
LT
409static inline void
410map_bh(struct buffer_head *bh, struct super_block *sb, sector_t block)
411{
412 set_buffer_mapped(bh);
413 bh->b_bdev = sb->s_bdev;
414 bh->b_blocknr = block;
b0cf2321 415 bh->b_size = sb->s_blocksize;
1da177e4
LT
416}
417
1da177e4
LT
418static inline void wait_on_buffer(struct buffer_head *bh)
419{
420 might_sleep();
a9877cc2 421 if (buffer_locked(bh))
1da177e4
LT
422 __wait_on_buffer(bh);
423}
424
ca5de404
NP
425static inline int trylock_buffer(struct buffer_head *bh)
426{
51b07fc3 427 return likely(!test_and_set_bit_lock(BH_Lock, &bh->b_state));
ca5de404
NP
428}
429
1da177e4
LT
430static inline void lock_buffer(struct buffer_head *bh)
431{
432 might_sleep();
ca5de404 433 if (!trylock_buffer(bh))
1da177e4
LT
434 __lock_buffer(bh);
435}
436
fdee117e
ZY
437static inline void bh_readahead(struct buffer_head *bh, blk_opf_t op_flags)
438{
439 if (!buffer_uptodate(bh) && trylock_buffer(bh)) {
440 if (!buffer_uptodate(bh))
441 __bh_read(bh, op_flags, false);
442 else
443 unlock_buffer(bh);
444 }
445}
446
447static inline void bh_read_nowait(struct buffer_head *bh, blk_opf_t op_flags)
448{
449 if (!bh_uptodate_or_lock(bh))
450 __bh_read(bh, op_flags, false);
451}
452
453/* Returns 1 if buffer uptodated, 0 on success, and -EIO on error. */
454static inline int bh_read(struct buffer_head *bh, blk_opf_t op_flags)
455{
456 if (bh_uptodate_or_lock(bh))
457 return 1;
458 return __bh_read(bh, op_flags, true);
459}
460
461static inline void bh_read_batch(int nr, struct buffer_head *bhs[])
462{
463 __bh_read_batch(nr, bhs, 0, true);
464}
465
466static inline void bh_readahead_batch(int nr, struct buffer_head *bhs[],
467 blk_opf_t op_flags)
468{
469 __bh_read_batch(nr, bhs, op_flags, false);
470}
471
3b5e6454 472/**
324ecaee
MWO
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.
3b5e6454 477 *
324ecaee
MWO
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.
482 *
483 * Context: May sleep waiting for I/O.
484 * Return: NULL if the block was unreadable.
3b5e6454 485 */
324ecaee
MWO
486static inline struct buffer_head *__bread(struct block_device *bdev,
487 sector_t block, unsigned size)
3b5e6454
GK
488{
489 return __bread_gfp(bdev, block, size, __GFP_MOVABLE);
490}
491
0217fbb0
MWO
492/**
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.
496 *
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.
502 *
503 * Return: The requested buffer with an elevated refcount.
504 */
505static inline __must_check
506struct buffer_head *get_nth_bh(struct buffer_head *bh, unsigned int count)
507{
508 while (count--)
509 bh = bh->b_this_page;
510 get_bh(bh);
511 return bh;
512}
513
e621900a 514bool block_dirty_folio(struct address_space *mapping, struct folio *folio);
9361401e 515
925c86a1
CH
516#ifdef CONFIG_BUFFER_HEAD
517
518void buffer_init(void);
519bool try_to_free_buffers(struct folio *folio);
520int inode_has_buffers(struct inode *inode);
521void invalidate_inode_buffers(struct inode *inode);
522int remove_inode_buffers(struct inode *inode);
523int sync_mapping_buffers(struct address_space *mapping);
524void invalidate_bh_lrus(void);
525void invalidate_bh_lrus_cpu(void);
526bool has_bh_in_lru(int cpu, void *dummy);
527extern int buffer_heads_over_limit;
528
529#else /* CONFIG_BUFFER_HEAD */
9361401e
DH
530
531static inline void buffer_init(void) {}
68189fef 532static inline bool try_to_free_buffers(struct folio *folio) { return true; }
9361401e
DH
533static inline int inode_has_buffers(struct inode *inode) { return 0; }
534static inline void invalidate_inode_buffers(struct inode *inode) {}
535static inline int remove_inode_buffers(struct inode *inode) { return 1; }
536static inline int sync_mapping_buffers(struct address_space *mapping) { return 0; }
925c86a1 537static inline void invalidate_bh_lrus(void) {}
243418e3 538static inline void invalidate_bh_lrus_cpu(void) {}
6de522d1 539static inline bool has_bh_in_lru(int cpu, void *dummy) { return false; }
d2de7ea4 540#define buffer_heads_over_limit 0
9361401e 541
925c86a1 542#endif /* CONFIG_BUFFER_HEAD */
1da177e4 543#endif /* _LINUX_BUFFER_HEAD_H */