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 19 enum 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 45 struct page; 46 struct buffer_head; 47 struct address_space; 48 typedef 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 */ 59 struct 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) \ 90 static __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 } \ 95 static __always_inline void clear_buffer_##name(struct buffer_head *bh) \ 96 { \ 97 clear_bit(BH_##bit, &(bh)->b_state); \ 98 } \ 99 static __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) \ 108 static __always_inline int test_set_buffer_##name(struct buffer_head *bh) \ 109 { \ 110 return test_and_set_bit(BH_##bit, &(bh)->b_state); \ 111 } \ 112 static __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 */ BUFFER_FNS(Dirty,dirty)122 BUFFER_FNS(Dirty, dirty) 123 TAS_BUFFER_FNS(Dirty, dirty) 124 BUFFER_FNS(Lock, locked) 125 BUFFER_FNS(Req, req) 126 TAS_BUFFER_FNS(Req, req) 127 BUFFER_FNS(Mapped, mapped) 128 BUFFER_FNS(New, new) 129 BUFFER_FNS(Async_Read, async_read) 130 BUFFER_FNS(Async_Write, async_write) 131 BUFFER_FNS(Delay, delay) 132 BUFFER_FNS(Boundary, boundary) 133 BUFFER_FNS(Write_EIO, write_io_error) 134 BUFFER_FNS(Unwritten, unwritten) 135 BUFFER_FNS(Meta, meta) 136 BUFFER_FNS(Prio, prio) 137 BUFFER_FNS(Defer_Completion, defer_completion) 138 139 static __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 clear_buffer_uptodate(struct buffer_head * bh)160 static __always_inline void clear_buffer_uptodate(struct buffer_head *bh) 161 { 162 clear_bit(BH_Uptodate, &bh->b_state); 163 } 164 buffer_uptodate(const struct buffer_head * bh)165 static __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 bh_offset(const struct buffer_head * bh)174 static 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) 187 188 void buffer_check_dirty_writeback(struct folio *folio, 189 bool *dirty, bool *writeback); 190 191 /* 192 * Declarations 193 */ 194 195 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, 201 bool retry); 202 struct buffer_head *alloc_page_buffers(struct page *page, unsigned long size, 203 bool retry); 204 void create_empty_buffers(struct page *, unsigned long, 205 unsigned long b_state); 206 void folio_create_empty_buffers(struct folio *folio, unsigned long blocksize, 207 unsigned long b_state); 208 void end_buffer_read_sync(struct buffer_head *bh, int uptodate); 209 void end_buffer_write_sync(struct buffer_head *bh, int uptodate); 210 void end_buffer_async_write(struct buffer_head *bh, int uptodate); 211 212 /* Things to do with buffers at mapping->private_list */ 213 void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode); 214 int generic_buffers_fsync_noflush(struct file *file, loff_t start, loff_t end, 215 bool datasync); 216 int generic_buffers_fsync(struct file *file, loff_t start, loff_t end, 217 bool datasync); 218 void clean_bdev_aliases(struct block_device *bdev, sector_t block, 219 sector_t len); clean_bdev_bh_alias(struct buffer_head * bh)220 static inline void clean_bdev_bh_alias(struct buffer_head *bh) 221 { 222 clean_bdev_aliases(bh->b_bdev, bh->b_blocknr, 1); 223 } 224 225 void mark_buffer_async_write(struct buffer_head *bh); 226 void __wait_on_buffer(struct buffer_head *); 227 wait_queue_head_t *bh_waitq_head(struct buffer_head *bh); 228 struct buffer_head *__find_get_block(struct block_device *bdev, sector_t block, 229 unsigned size); 230 struct buffer_head *__getblk_gfp(struct block_device *bdev, sector_t block, 231 unsigned size, gfp_t gfp); 232 void __brelse(struct buffer_head *); 233 void __bforget(struct buffer_head *); 234 void __breadahead(struct block_device *, sector_t block, unsigned int size); 235 struct buffer_head *__bread_gfp(struct block_device *, 236 sector_t block, unsigned size, gfp_t gfp); 237 struct buffer_head *alloc_buffer_head(gfp_t gfp_flags); 238 void free_buffer_head(struct buffer_head * bh); 239 void unlock_buffer(struct buffer_head *bh); 240 void __lock_buffer(struct buffer_head *bh); 241 int sync_dirty_buffer(struct buffer_head *bh); 242 int __sync_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags); 243 void write_dirty_buffer(struct buffer_head *bh, blk_opf_t op_flags); 244 void submit_bh(blk_opf_t, struct buffer_head *); 245 void write_boundary_block(struct block_device *bdev, 246 sector_t bblock, unsigned blocksize); 247 int bh_uptodate_or_lock(struct buffer_head *bh); 248 int __bh_read(struct buffer_head *bh, blk_opf_t op_flags, bool wait); 249 void __bh_read_batch(int nr, struct buffer_head *bhs[], 250 blk_opf_t op_flags, bool force_lock); 251 252 /* 253 * Generic address_space_operations implementations for buffer_head-backed 254 * address_spaces. 255 */ 256 void block_invalidate_folio(struct folio *folio, size_t offset, size_t length); 257 int block_write_full_page(struct page *page, get_block_t *get_block, 258 struct writeback_control *wbc); 259 int __block_write_full_folio(struct inode *inode, struct folio *folio, 260 get_block_t *get_block, struct writeback_control *wbc, 261 bh_end_io_t *handler); 262 int block_read_full_folio(struct folio *, get_block_t *); 263 bool block_is_partially_uptodate(struct folio *, size_t from, size_t count); 264 int block_write_begin(struct address_space *mapping, loff_t pos, unsigned len, 265 struct page **pagep, get_block_t *get_block); 266 int __block_write_begin(struct page *page, loff_t pos, unsigned len, 267 get_block_t *get_block); 268 int block_write_end(struct file *, struct address_space *, 269 loff_t, unsigned, unsigned, 270 struct page *, void *); 271 int generic_write_end(struct file *, struct address_space *, 272 loff_t, unsigned, unsigned, 273 struct page *, void *); 274 void folio_zero_new_buffers(struct folio *folio, size_t from, size_t to); 275 void clean_page_buffers(struct page *page); 276 int cont_write_begin(struct file *, struct address_space *, loff_t, 277 unsigned, struct page **, void **, 278 get_block_t *, loff_t *); 279 int generic_cont_expand_simple(struct inode *inode, loff_t size); 280 void block_commit_write(struct page *page, unsigned int from, unsigned int to); 281 int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf, 282 get_block_t get_block); 283 sector_t generic_block_bmap(struct address_space *, sector_t, get_block_t *); 284 int block_truncate_page(struct address_space *, loff_t, get_block_t *); 285 286 #ifdef CONFIG_MIGRATION 287 extern int buffer_migrate_folio(struct address_space *, 288 struct folio *dst, struct folio *src, enum migrate_mode); 289 extern int buffer_migrate_folio_norefs(struct address_space *, 290 struct folio *dst, struct folio *src, enum migrate_mode); 291 #else 292 #define buffer_migrate_folio NULL 293 #define buffer_migrate_folio_norefs NULL 294 #endif 295 296 /* 297 * inline definitions 298 */ 299 get_bh(struct buffer_head * bh)300 static inline void get_bh(struct buffer_head *bh) 301 { 302 atomic_inc(&bh->b_count); 303 } 304 put_bh(struct buffer_head * bh)305 static inline void put_bh(struct buffer_head *bh) 306 { 307 smp_mb__before_atomic(); 308 atomic_dec(&bh->b_count); 309 } 310 brelse(struct buffer_head * bh)311 static inline void brelse(struct buffer_head *bh) 312 { 313 if (bh) 314 __brelse(bh); 315 } 316 bforget(struct buffer_head * bh)317 static inline void bforget(struct buffer_head *bh) 318 { 319 if (bh) 320 __bforget(bh); 321 } 322 323 static inline struct buffer_head * sb_bread(struct super_block * sb,sector_t block)324 sb_bread(struct super_block *sb, sector_t block) 325 { 326 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE); 327 } 328 329 static inline struct buffer_head * sb_bread_unmovable(struct super_block * sb,sector_t block)330 sb_bread_unmovable(struct super_block *sb, sector_t block) 331 { 332 return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, 0); 333 } 334 335 static inline void sb_breadahead(struct super_block * sb,sector_t block)336 sb_breadahead(struct super_block *sb, sector_t block) 337 { 338 __breadahead(sb->s_bdev, block, sb->s_blocksize); 339 } 340 341 static inline struct buffer_head * sb_getblk(struct super_block * sb,sector_t block)342 sb_getblk(struct super_block *sb, sector_t block) 343 { 344 return __getblk_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE); 345 } 346 347 348 static inline struct buffer_head * sb_getblk_gfp(struct super_block * sb,sector_t block,gfp_t gfp)349 sb_getblk_gfp(struct super_block *sb, sector_t block, gfp_t gfp) 350 { 351 return __getblk_gfp(sb->s_bdev, block, sb->s_blocksize, gfp); 352 } 353 354 static inline struct buffer_head * sb_find_get_block(struct super_block * sb,sector_t block)355 sb_find_get_block(struct super_block *sb, sector_t block) 356 { 357 return __find_get_block(sb->s_bdev, block, sb->s_blocksize); 358 } 359 360 static inline void map_bh(struct buffer_head * bh,struct super_block * sb,sector_t block)361 map_bh(struct buffer_head *bh, struct super_block *sb, sector_t block) 362 { 363 set_buffer_mapped(bh); 364 bh->b_bdev = sb->s_bdev; 365 bh->b_blocknr = block; 366 bh->b_size = sb->s_blocksize; 367 } 368 wait_on_buffer(struct buffer_head * bh)369 static inline void wait_on_buffer(struct buffer_head *bh) 370 { 371 might_sleep(); 372 if (buffer_locked(bh)) 373 __wait_on_buffer(bh); 374 } 375 trylock_buffer(struct buffer_head * bh)376 static inline int trylock_buffer(struct buffer_head *bh) 377 { 378 return likely(!test_and_set_bit_lock(BH_Lock, &bh->b_state)); 379 } 380 lock_buffer(struct buffer_head * bh)381 static inline void lock_buffer(struct buffer_head *bh) 382 { 383 might_sleep(); 384 if (!trylock_buffer(bh)) 385 __lock_buffer(bh); 386 } 387 getblk_unmovable(struct block_device * bdev,sector_t block,unsigned size)388 static inline struct buffer_head *getblk_unmovable(struct block_device *bdev, 389 sector_t block, 390 unsigned size) 391 { 392 return __getblk_gfp(bdev, block, size, 0); 393 } 394 __getblk(struct block_device * bdev,sector_t block,unsigned size)395 static inline struct buffer_head *__getblk(struct block_device *bdev, 396 sector_t block, 397 unsigned size) 398 { 399 return __getblk_gfp(bdev, block, size, __GFP_MOVABLE); 400 } 401 bh_readahead(struct buffer_head * bh,blk_opf_t op_flags)402 static inline void bh_readahead(struct buffer_head *bh, blk_opf_t op_flags) 403 { 404 if (!buffer_uptodate(bh) && trylock_buffer(bh)) { 405 if (!buffer_uptodate(bh)) 406 __bh_read(bh, op_flags, false); 407 else 408 unlock_buffer(bh); 409 } 410 } 411 bh_read_nowait(struct buffer_head * bh,blk_opf_t op_flags)412 static inline void bh_read_nowait(struct buffer_head *bh, blk_opf_t op_flags) 413 { 414 if (!bh_uptodate_or_lock(bh)) 415 __bh_read(bh, op_flags, false); 416 } 417 418 /* Returns 1 if buffer uptodated, 0 on success, and -EIO on error. */ bh_read(struct buffer_head * bh,blk_opf_t op_flags)419 static inline int bh_read(struct buffer_head *bh, blk_opf_t op_flags) 420 { 421 if (bh_uptodate_or_lock(bh)) 422 return 1; 423 return __bh_read(bh, op_flags, true); 424 } 425 bh_read_batch(int nr,struct buffer_head * bhs[])426 static inline void bh_read_batch(int nr, struct buffer_head *bhs[]) 427 { 428 __bh_read_batch(nr, bhs, 0, true); 429 } 430 bh_readahead_batch(int nr,struct buffer_head * bhs[],blk_opf_t op_flags)431 static inline void bh_readahead_batch(int nr, struct buffer_head *bhs[], 432 blk_opf_t op_flags) 433 { 434 __bh_read_batch(nr, bhs, op_flags, false); 435 } 436 437 /** 438 * __bread() - reads a specified block and returns the bh 439 * @bdev: the block_device to read from 440 * @block: number of block 441 * @size: size (in bytes) to read 442 * 443 * Reads a specified block, and returns buffer head that contains it. 444 * The page cache is allocated from movable area so that it can be migrated. 445 * It returns NULL if the block was unreadable. 446 */ 447 static inline struct buffer_head * __bread(struct block_device * bdev,sector_t block,unsigned size)448 __bread(struct block_device *bdev, sector_t block, unsigned size) 449 { 450 return __bread_gfp(bdev, block, size, __GFP_MOVABLE); 451 } 452 453 bool block_dirty_folio(struct address_space *mapping, struct folio *folio); 454 455 #ifdef CONFIG_BUFFER_HEAD 456 457 void buffer_init(void); 458 bool try_to_free_buffers(struct folio *folio); 459 int inode_has_buffers(struct inode *inode); 460 void invalidate_inode_buffers(struct inode *inode); 461 int remove_inode_buffers(struct inode *inode); 462 int sync_mapping_buffers(struct address_space *mapping); 463 void invalidate_bh_lrus(void); 464 void invalidate_bh_lrus_cpu(void); 465 bool has_bh_in_lru(int cpu, void *dummy); 466 extern int buffer_heads_over_limit; 467 468 #else /* CONFIG_BUFFER_HEAD */ 469 buffer_init(void)470 static inline void buffer_init(void) {} try_to_free_buffers(struct folio * folio)471 static inline bool try_to_free_buffers(struct folio *folio) { return true; } inode_has_buffers(struct inode * inode)472 static inline int inode_has_buffers(struct inode *inode) { return 0; } invalidate_inode_buffers(struct inode * inode)473 static inline void invalidate_inode_buffers(struct inode *inode) {} remove_inode_buffers(struct inode * inode)474 static inline int remove_inode_buffers(struct inode *inode) { return 1; } sync_mapping_buffers(struct address_space * mapping)475 static inline int sync_mapping_buffers(struct address_space *mapping) { return 0; } invalidate_bh_lrus(void)476 static inline void invalidate_bh_lrus(void) {} invalidate_bh_lrus_cpu(void)477 static inline void invalidate_bh_lrus_cpu(void) {} has_bh_in_lru(int cpu,void * dummy)478 static inline bool has_bh_in_lru(int cpu, void *dummy) { return false; } 479 #define buffer_heads_over_limit 0 480 481 #endif /* CONFIG_BUFFER_HEAD */ 482 #endif /* _LINUX_BUFFER_HEAD_H */ 483
