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 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 */ 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 45struct page; 46struct buffer_head; 47struct address_space; 48typedef 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 page (via a page_mapping) and for wrapping bio submission 57 * for backward compatibility reasons (e.g. submit_bh). 58 */ 59struct 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) \ 90static __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} \ 95static __always_inline void clear_buffer_##name(struct buffer_head *bh) \ 96{ \ 97 clear_bit(BH_##bit, &(bh)->b_state); \ 98} \ 99static __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) \ 108static __always_inline int test_set_buffer_##name(struct buffer_head *bh) \ 109{ \ 110 return test_and_set_bit(BH_##bit, &(bh)->b_state); \ 111} \ 112static __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 */ 122BUFFER_FNS(Dirty, dirty) 123TAS_BUFFER_FNS(Dirty, dirty) 124BUFFER_FNS(Lock, locked) 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) 134BUFFER_FNS(Unwritten, unwritten) 135BUFFER_FNS(Meta, meta) 136BUFFER_FNS(Prio, prio) 137BUFFER_FNS(Defer_Completion, defer_completion) 138 139static __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 160static __always_inline void clear_buffer_uptodate(struct buffer_head *bh) 161{ 162 clear_bit(BH_Uptodate, &bh->b_state); 163} 164 165static __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 174static 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 188void buffer_check_dirty_writeback(struct folio *folio, 189 bool *dirty, bool *writeback); 190 191/* 192 * Declarations 193 */ 194 195void mark_buffer_dirty(struct buffer_head *bh); 196void mark_buffer_write_io_error(struct buffer_head *bh); 197void touch_buffer(struct buffer_head *bh); 198void folio_set_bh(struct buffer_head *bh, struct folio *folio, 199 unsigned long offset); 200struct buffer_head *folio_alloc_buffers(struct folio *folio, unsigned long size, 201 gfp_t gfp); 202struct buffer_head *alloc_page_buffers(struct page *page, unsigned long size, 203 bool retry); 204struct buffer_head *create_empty_buffers(struct folio *folio, 205 unsigned long blocksize, unsigned long b_state); 206void end_buffer_read_sync(struct buffer_head *bh, int uptodate); 207void end_buffer_write_sync(struct buffer_head *bh, int uptodate); 208 209/* Things to do with buffers at mapping->private_list */ 210void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode); 211int generic_buffers_fsync_noflush(struct file *file, loff_t start, loff_t end, 212 bool datasync); 213int generic_buffers_fsync(struct file *file, loff_t start, loff_t end, 214 bool datasync); 215void clean_bdev_aliases(struct block_device *bdev, sector_t block, 216 sector_t len); 217static inline void clean_bdev_bh_alias(struct buffer_head *bh) 218{ 219 clean_bdev_aliases(bh->b_bdev, bh->b_blocknr, 1); 220} 221 222void mark_buffer_async_write(struct buffer_head *bh); 223void __wait_on_buffer(struct buffer_head *); 224wait_queue_head_t *bh_waitq_head(struct buffer_head *bh); 225struct buffer_head *__find_get_block(struct block_device *bdev, sector_t block, 226 unsigned size); 227struct buffer_head *bdev_getblk(struct block_device *bdev, sector_t block, 228 unsigned size, gfp_t gfp); 229void __brelse(struct buffer_head *); 230void __bforget(struct buffer_head *); 231void __breadahead(struct block_device *, sector_t block, unsigned int size); 232struct buffer_head *__bread_gfp(struct block_device *, 233 sector_t block, unsigned size, gfp_t gfp); 234struct buffer_head *alloc_buffer_head(gfp_t gfp_flags); 235void free_buffer_head(struct buffer_head * bh); 236void unlock_buffer(struct buffer_head *bh); 237void __lock_buffer(struct buffer_head *bh); 238int sync_dirty_buffer(struct buffer_head *bh); 239int __sync_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags); 240void write_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags); 241void submit_bh(blk_opf_t, struct buffer_head *); 242void write_boundary_block(struct block_device *bdev, 243 sector_t bblock, unsigned blocksize); 244int bh_uptodate_or_lock(struct buffer_head *bh); 245int __bh_read(struct buffer_head *bh, blk_opf_t op_flags, bool wait); 246void __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 */ 253void block_invalidate_folio(struct folio *folio, size_t offset, size_t length); 254int block_write_full_folio(struct folio *folio, struct writeback_control *wbc, 255 void *get_block); 256int __block_write_full_folio(struct inode *inode, struct folio *folio, 257 get_block_t *get_block, struct writeback_control *wbc); 258int block_read_full_folio(struct folio *, get_block_t *); 259bool block_is_partially_uptodate(struct folio *, size_t from, size_t count); 260int block_write_begin(struct address_space *mapping, loff_t pos, unsigned len, 261 struct page **pagep, get_block_t *get_block); 262int __block_write_begin(struct page *page, loff_t pos, unsigned len, 263 get_block_t *get_block); 264int block_write_end(struct file *, struct address_space *, 265 loff_t, unsigned, unsigned, 266 struct page *, void *); 267int generic_write_end(struct file *, struct address_space *, 268 loff_t, unsigned, unsigned, 269 struct page *, void *); 270void folio_zero_new_buffers(struct folio *folio, size_t from, size_t to); 271int cont_write_begin(struct file *, struct address_space *, loff_t, 272 unsigned, struct page **, void **, 273 get_block_t *, loff_t *); 274int generic_cont_expand_simple(struct inode *inode, loff_t size); 275void block_commit_write(struct page *page, unsigned int from, unsigned int to); 276int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf, 277 get_block_t get_block); 278sector_t generic_block_bmap(struct address_space *, sector_t, get_block_t *); 279int block_truncate_page(struct address_space *, loff_t, get_block_t *); 280 281#ifdef CONFIG_MIGRATION 282extern int buffer_migrate_folio(struct address_space *, 283 struct folio *dst, struct folio *src, enum migrate_mode); 284extern 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 295static inline void get_bh(struct buffer_head *bh) 296{ 297 atomic_inc(&bh->b_count); 298} 299 300static inline void put_bh(struct buffer_head *bh) 301{ 302 smp_mb__before_atomic(); 303 atomic_dec(&bh->b_count); 304} 305 306static inline void brelse(struct buffer_head *bh) 307{ 308 if (bh) 309 __brelse(bh); 310} 311 312static inline void bforget(struct buffer_head *bh) 313{ 314 if (bh) 315 __bforget(bh); 316} 317 318static inline struct buffer_head * 319sb_bread(struct super_block *sb, sector_t block) 320{ 321 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE); 322} 323 324static inline struct buffer_head * 325sb_bread_unmovable(struct super_block *sb, sector_t block) 326{ 327 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, 0); 328} 329 330static inline void 331sb_breadahead(struct super_block *sb, sector_t block) 332{ 333 __breadahead(sb->s_bdev, block, sb->s_blocksize); 334} 335 336static inline struct buffer_head *getblk_unmovable(struct block_device *bdev, 337 sector_t block, unsigned size) 338{ 339 gfp_t gfp; 340 341 gfp = mapping_gfp_constraint(bdev->bd_inode->i_mapping, ~__GFP_FS); 342 gfp |= __GFP_NOFAIL; 343 344 return bdev_getblk(bdev, block, size, gfp); 345} 346 347static inline struct buffer_head *__getblk(struct block_device *bdev, 348 sector_t block, unsigned size) 349{ 350 gfp_t gfp; 351 352 gfp = mapping_gfp_constraint(bdev->bd_inode->i_mapping, ~__GFP_FS); 353 gfp |= __GFP_MOVABLE | __GFP_NOFAIL; 354 355 return bdev_getblk(bdev, block, size, gfp); 356} 357 358static inline struct buffer_head *sb_getblk(struct super_block *sb, 359 sector_t block) 360{ 361 return __getblk(sb->s_bdev, block, sb->s_blocksize); 362} 363 364static inline struct buffer_head *sb_getblk_gfp(struct super_block *sb, 365 sector_t block, gfp_t gfp) 366{ 367 return bdev_getblk(sb->s_bdev, block, sb->s_blocksize, gfp); 368} 369 370static inline struct buffer_head * 371sb_find_get_block(struct super_block *sb, sector_t block) 372{ 373 return __find_get_block(sb->s_bdev, block, sb->s_blocksize); 374} 375 376static inline void 377map_bh(struct buffer_head *bh, struct super_block *sb, sector_t block) 378{ 379 set_buffer_mapped(bh); 380 bh->b_bdev = sb->s_bdev; 381 bh->b_blocknr = block; 382 bh->b_size = sb->s_blocksize; 383} 384 385static inline void wait_on_buffer(struct buffer_head *bh) 386{ 387 might_sleep(); 388 if (buffer_locked(bh)) 389 __wait_on_buffer(bh); 390} 391 392static inline int trylock_buffer(struct buffer_head *bh) 393{ 394 return likely(!test_and_set_bit_lock(BH_Lock, &bh->b_state)); 395} 396 397static inline void lock_buffer(struct buffer_head *bh) 398{ 399 might_sleep(); 400 if (!trylock_buffer(bh)) 401 __lock_buffer(bh); 402} 403 404static inline void bh_readahead(struct buffer_head *bh, blk_opf_t op_flags) 405{ 406 if (!buffer_uptodate(bh) && trylock_buffer(bh)) { 407 if (!buffer_uptodate(bh)) 408 __bh_read(bh, op_flags, false); 409 else 410 unlock_buffer(bh); 411 } 412} 413 414static inline void bh_read_nowait(struct buffer_head *bh, blk_opf_t op_flags) 415{ 416 if (!bh_uptodate_or_lock(bh)) 417 __bh_read(bh, op_flags, false); 418} 419 420/* Returns 1 if buffer uptodated, 0 on success, and -EIO on error. */ 421static inline int bh_read(struct buffer_head *bh, blk_opf_t op_flags) 422{ 423 if (bh_uptodate_or_lock(bh)) 424 return 1; 425 return __bh_read(bh, op_flags, true); 426} 427 428static inline void bh_read_batch(int nr, struct buffer_head *bhs[]) 429{ 430 __bh_read_batch(nr, bhs, 0, true); 431} 432 433static inline void bh_readahead_batch(int nr, struct buffer_head *bhs[], 434 blk_opf_t op_flags) 435{ 436 __bh_read_batch(nr, bhs, op_flags, false); 437} 438 439/** 440 * __bread() - reads a specified block and returns the bh 441 * @bdev: the block_device to read from 442 * @block: number of block 443 * @size: size (in bytes) to read 444 * 445 * Reads a specified block, and returns buffer head that contains it. 446 * The page cache is allocated from movable area so that it can be migrated. 447 * It returns NULL if the block was unreadable. 448 */ 449static inline struct buffer_head * 450__bread(struct block_device *bdev, sector_t block, unsigned size) 451{ 452 return __bread_gfp(bdev, block, size, __GFP_MOVABLE); 453} 454 455/** 456 * get_nth_bh - Get a reference on the n'th buffer after this one. 457 * @bh: The buffer to start counting from. 458 * @count: How many buffers to skip. 459 * 460 * This is primarily useful for finding the nth buffer in a folio; in 461 * that case you pass the head buffer and the byte offset in the folio 462 * divided by the block size. It can be used for other purposes, but 463 * it will wrap at the end of the folio rather than returning NULL or 464 * proceeding to the next folio for you. 465 * 466 * Return: The requested buffer with an elevated refcount. 467 */ 468static inline __must_check 469struct buffer_head *get_nth_bh(struct buffer_head *bh, unsigned int count) 470{ 471 while (count--) 472 bh = bh->b_this_page; 473 get_bh(bh); 474 return bh; 475} 476 477bool block_dirty_folio(struct address_space *mapping, struct folio *folio); 478 479#ifdef CONFIG_BUFFER_HEAD 480 481void buffer_init(void); 482bool try_to_free_buffers(struct folio *folio); 483int inode_has_buffers(struct inode *inode); 484void invalidate_inode_buffers(struct inode *inode); 485int remove_inode_buffers(struct inode *inode); 486int sync_mapping_buffers(struct address_space *mapping); 487void invalidate_bh_lrus(void); 488void invalidate_bh_lrus_cpu(void); 489bool has_bh_in_lru(int cpu, void *dummy); 490extern int buffer_heads_over_limit; 491 492#else /* CONFIG_BUFFER_HEAD */ 493 494static inline void buffer_init(void) {} 495static inline bool try_to_free_buffers(struct folio *folio) { return true; } 496static inline int inode_has_buffers(struct inode *inode) { return 0; } 497static inline void invalidate_inode_buffers(struct inode *inode) {} 498static inline int remove_inode_buffers(struct inode *inode) { return 1; } 499static inline int sync_mapping_buffers(struct address_space *mapping) { return 0; } 500static inline void invalidate_bh_lrus(void) {} 501static inline void invalidate_bh_lrus_cpu(void) {} 502static inline bool has_bh_in_lru(int cpu, void *dummy) { return false; } 503#define buffer_heads_over_limit 0 504 505#endif /* CONFIG_BUFFER_HEAD */ 506#endif /* _LINUX_BUFFER_HEAD_H */ 507