1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef LINUX_IOMAP_H 3 #define LINUX_IOMAP_H 1 4 5 #include <linux/atomic.h> 6 #include <linux/bitmap.h> 7 #include <linux/blk_types.h> 8 #include <linux/mm.h> 9 #include <linux/types.h> 10 #include <linux/mm_types.h> 11 #include <linux/blkdev.h> 12 13 struct address_space; 14 struct fiemap_extent_info; 15 struct inode; 16 struct iomap_iter; 17 struct iomap_dio; 18 struct iomap_writepage_ctx; 19 struct iov_iter; 20 struct kiocb; 21 struct page; 22 struct vm_area_struct; 23 struct vm_fault; 24 25 /* 26 * Types of block ranges for iomap mappings: 27 */ 28 #define IOMAP_HOLE 0 /* no blocks allocated, need allocation */ 29 #define IOMAP_DELALLOC 1 /* delayed allocation blocks */ 30 #define IOMAP_MAPPED 2 /* blocks allocated at @addr */ 31 #define IOMAP_UNWRITTEN 3 /* blocks allocated at @addr in unwritten state */ 32 #define IOMAP_INLINE 4 /* data inline in the inode */ 33 34 /* 35 * Flags reported by the file system from iomap_begin: 36 * 37 * IOMAP_F_NEW indicates that the blocks have been newly allocated and need 38 * zeroing for areas that no data is copied to. 39 * 40 * IOMAP_F_DIRTY indicates the inode has uncommitted metadata needed to access 41 * written data and requires fdatasync to commit them to persistent storage. 42 * This needs to take into account metadata changes that *may* be made at IO 43 * completion, such as file size updates from direct IO. 44 * 45 * IOMAP_F_SHARED indicates that the blocks are shared, and will need to be 46 * unshared as part a write. 47 * 48 * IOMAP_F_MERGED indicates that the iomap contains the merge of multiple block 49 * mappings. 50 * 51 * IOMAP_F_BUFFER_HEAD indicates that the file system requires the use of 52 * buffer heads for this mapping. 53 * 54 * IOMAP_F_XATTR indicates that the iomap is for an extended attribute extent 55 * rather than a file data extent. 56 */ 57 #define IOMAP_F_NEW (1U << 0) 58 #define IOMAP_F_DIRTY (1U << 1) 59 #define IOMAP_F_SHARED (1U << 2) 60 #define IOMAP_F_MERGED (1U << 3) 61 #define IOMAP_F_BUFFER_HEAD (1U << 4) 62 #define IOMAP_F_XATTR (1U << 5) 63 64 /* 65 * Flags set by the core iomap code during operations: 66 * 67 * IOMAP_F_SIZE_CHANGED indicates to the iomap_end method that the file size 68 * has changed as the result of this write operation. 69 * 70 * IOMAP_F_STALE indicates that the iomap is not valid any longer and the file 71 * range it covers needs to be remapped by the high level before the operation 72 * can proceed. 73 */ 74 #define IOMAP_F_SIZE_CHANGED (1U << 8) 75 #define IOMAP_F_STALE (1U << 9) 76 77 /* 78 * Flags from 0x1000 up are for file system specific usage: 79 */ 80 #define IOMAP_F_PRIVATE (1U << 12) 81 82 83 /* 84 * Magic value for addr: 85 */ 86 #define IOMAP_NULL_ADDR -1ULL /* addr is not valid */ 87 88 struct iomap_folio_ops; 89 90 struct iomap { 91 u64 addr; /* disk offset of mapping, bytes */ 92 loff_t offset; /* file offset of mapping, bytes */ 93 u64 length; /* length of mapping, bytes */ 94 u16 type; /* type of mapping */ 95 u16 flags; /* flags for mapping */ 96 struct block_device *bdev; /* block device for I/O */ 97 struct dax_device *dax_dev; /* dax_dev for dax operations */ 98 void *inline_data; 99 void *private; /* filesystem private */ 100 const struct iomap_folio_ops *folio_ops; 101 u64 validity_cookie; /* used with .iomap_valid() */ 102 }; 103 104 static inline sector_t iomap_sector(const struct iomap *iomap, loff_t pos) 105 { 106 return (iomap->addr + pos - iomap->offset) >> SECTOR_SHIFT; 107 } 108 109 /* 110 * Returns the inline data pointer for logical offset @pos. 111 */ 112 static inline void *iomap_inline_data(const struct iomap *iomap, loff_t pos) 113 { 114 return iomap->inline_data + pos - iomap->offset; 115 } 116 117 /* 118 * Check if the mapping's length is within the valid range for inline data. 119 * This is used to guard against accessing data beyond the page inline_data 120 * points at. 121 */ 122 static inline bool iomap_inline_data_valid(const struct iomap *iomap) 123 { 124 return iomap->length <= PAGE_SIZE - offset_in_page(iomap->inline_data); 125 } 126 127 /* 128 * When a filesystem sets folio_ops in an iomap mapping it returns, get_folio 129 * and put_folio will be called for each folio written to. This only applies 130 * to buffered writes as unbuffered writes will not typically have folios 131 * associated with them. 132 * 133 * When get_folio succeeds, put_folio will always be called to do any 134 * cleanup work necessary. put_folio is responsible for unlocking and putting 135 * @folio. 136 */ 137 struct iomap_folio_ops { 138 struct folio *(*get_folio)(struct iomap_iter *iter, loff_t pos, 139 unsigned len); 140 void (*put_folio)(struct inode *inode, loff_t pos, unsigned copied, 141 struct folio *folio); 142 143 /* 144 * Check that the cached iomap still maps correctly to the filesystem's 145 * internal extent map. FS internal extent maps can change while iomap 146 * is iterating a cached iomap, so this hook allows iomap to detect that 147 * the iomap needs to be refreshed during a long running write 148 * operation. 149 * 150 * The filesystem can store internal state (e.g. a sequence number) in 151 * iomap->validity_cookie when the iomap is first mapped to be able to 152 * detect changes between mapping time and whenever .iomap_valid() is 153 * called. 154 * 155 * This is called with the folio over the specified file position held 156 * locked by the iomap code. 157 */ 158 bool (*iomap_valid)(struct inode *inode, const struct iomap *iomap); 159 }; 160 161 /* 162 * Flags for iomap_begin / iomap_end. No flag implies a read. 163 */ 164 #define IOMAP_WRITE (1 << 0) /* writing, must allocate blocks */ 165 #define IOMAP_ZERO (1 << 1) /* zeroing operation, may skip holes */ 166 #define IOMAP_REPORT (1 << 2) /* report extent status, e.g. FIEMAP */ 167 #define IOMAP_FAULT (1 << 3) /* mapping for page fault */ 168 #define IOMAP_DIRECT (1 << 4) /* direct I/O */ 169 #define IOMAP_NOWAIT (1 << 5) /* do not block */ 170 #define IOMAP_OVERWRITE_ONLY (1 << 6) /* only pure overwrites allowed */ 171 #define IOMAP_UNSHARE (1 << 7) /* unshare_file_range */ 172 #ifdef CONFIG_FS_DAX 173 #define IOMAP_DAX (1 << 8) /* DAX mapping */ 174 #else 175 #define IOMAP_DAX 0 176 #endif /* CONFIG_FS_DAX */ 177 178 struct iomap_ops { 179 /* 180 * Return the existing mapping at pos, or reserve space starting at 181 * pos for up to length, as long as we can do it as a single mapping. 182 * The actual length is returned in iomap->length. 183 */ 184 int (*iomap_begin)(struct inode *inode, loff_t pos, loff_t length, 185 unsigned flags, struct iomap *iomap, 186 struct iomap *srcmap); 187 188 /* 189 * Commit and/or unreserve space previous allocated using iomap_begin. 190 * Written indicates the length of the successful write operation which 191 * needs to be commited, while the rest needs to be unreserved. 192 * Written might be zero if no data was written. 193 */ 194 int (*iomap_end)(struct inode *inode, loff_t pos, loff_t length, 195 ssize_t written, unsigned flags, struct iomap *iomap); 196 }; 197 198 /** 199 * struct iomap_iter - Iterate through a range of a file 200 * @inode: Set at the start of the iteration and should not change. 201 * @pos: The current file position we are operating on. It is updated by 202 * calls to iomap_iter(). Treat as read-only in the body. 203 * @len: The remaining length of the file segment we're operating on. 204 * It is updated at the same time as @pos. 205 * @processed: The number of bytes processed by the body in the most recent 206 * iteration, or a negative errno. 0 causes the iteration to stop. 207 * @flags: Zero or more of the iomap_begin flags above. 208 * @iomap: Map describing the I/O iteration 209 * @srcmap: Source map for COW operations 210 */ 211 struct iomap_iter { 212 struct inode *inode; 213 loff_t pos; 214 u64 len; 215 s64 processed; 216 unsigned flags; 217 struct iomap iomap; 218 struct iomap srcmap; 219 void *private; 220 }; 221 222 int iomap_iter(struct iomap_iter *iter, const struct iomap_ops *ops); 223 224 /** 225 * iomap_length - length of the current iomap iteration 226 * @iter: iteration structure 227 * 228 * Returns the length that the operation applies to for the current iteration. 229 */ 230 static inline u64 iomap_length(const struct iomap_iter *iter) 231 { 232 u64 end = iter->iomap.offset + iter->iomap.length; 233 234 if (iter->srcmap.type != IOMAP_HOLE) 235 end = min(end, iter->srcmap.offset + iter->srcmap.length); 236 return min(iter->len, end - iter->pos); 237 } 238 239 /** 240 * iomap_iter_srcmap - return the source map for the current iomap iteration 241 * @i: iteration structure 242 * 243 * Write operations on file systems with reflink support might require a 244 * source and a destination map. This function retourns the source map 245 * for a given operation, which may or may no be identical to the destination 246 * map in &i->iomap. 247 */ 248 static inline const struct iomap *iomap_iter_srcmap(const struct iomap_iter *i) 249 { 250 if (i->srcmap.type != IOMAP_HOLE) 251 return &i->srcmap; 252 return &i->iomap; 253 } 254 255 ssize_t iomap_file_buffered_write(struct kiocb *iocb, struct iov_iter *from, 256 const struct iomap_ops *ops); 257 int iomap_file_buffered_write_punch_delalloc(struct inode *inode, 258 struct iomap *iomap, loff_t pos, loff_t length, ssize_t written, 259 int (*punch)(struct inode *inode, loff_t pos, loff_t length)); 260 261 int iomap_read_folio(struct folio *folio, const struct iomap_ops *ops); 262 void iomap_readahead(struct readahead_control *, const struct iomap_ops *ops); 263 bool iomap_is_partially_uptodate(struct folio *, size_t from, size_t count); 264 struct folio *iomap_get_folio(struct iomap_iter *iter, loff_t pos); 265 bool iomap_release_folio(struct folio *folio, gfp_t gfp_flags); 266 void iomap_invalidate_folio(struct folio *folio, size_t offset, size_t len); 267 int iomap_file_unshare(struct inode *inode, loff_t pos, loff_t len, 268 const struct iomap_ops *ops); 269 int iomap_zero_range(struct inode *inode, loff_t pos, loff_t len, 270 bool *did_zero, const struct iomap_ops *ops); 271 int iomap_truncate_page(struct inode *inode, loff_t pos, bool *did_zero, 272 const struct iomap_ops *ops); 273 vm_fault_t iomap_page_mkwrite(struct vm_fault *vmf, 274 const struct iomap_ops *ops); 275 int iomap_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, 276 u64 start, u64 len, const struct iomap_ops *ops); 277 loff_t iomap_seek_hole(struct inode *inode, loff_t offset, 278 const struct iomap_ops *ops); 279 loff_t iomap_seek_data(struct inode *inode, loff_t offset, 280 const struct iomap_ops *ops); 281 sector_t iomap_bmap(struct address_space *mapping, sector_t bno, 282 const struct iomap_ops *ops); 283 284 /* 285 * Structure for writeback I/O completions. 286 */ 287 struct iomap_ioend { 288 struct list_head io_list; /* next ioend in chain */ 289 u16 io_type; 290 u16 io_flags; /* IOMAP_F_* */ 291 u32 io_folios; /* folios added to ioend */ 292 struct inode *io_inode; /* file being written to */ 293 size_t io_size; /* size of the extent */ 294 loff_t io_offset; /* offset in the file */ 295 sector_t io_sector; /* start sector of ioend */ 296 struct bio *io_bio; /* bio being built */ 297 struct bio io_inline_bio; /* MUST BE LAST! */ 298 }; 299 300 struct iomap_writeback_ops { 301 /* 302 * Required, maps the blocks so that writeback can be performed on 303 * the range starting at offset. 304 */ 305 int (*map_blocks)(struct iomap_writepage_ctx *wpc, struct inode *inode, 306 loff_t offset); 307 308 /* 309 * Optional, allows the file systems to perform actions just before 310 * submitting the bio and/or override the bio end_io handler for complex 311 * operations like copy on write extent manipulation or unwritten extent 312 * conversions. 313 */ 314 int (*prepare_ioend)(struct iomap_ioend *ioend, int status); 315 316 /* 317 * Optional, allows the file system to discard state on a page where 318 * we failed to submit any I/O. 319 */ 320 void (*discard_folio)(struct folio *folio, loff_t pos); 321 }; 322 323 struct iomap_writepage_ctx { 324 struct iomap iomap; 325 struct iomap_ioend *ioend; 326 const struct iomap_writeback_ops *ops; 327 }; 328 329 void iomap_finish_ioends(struct iomap_ioend *ioend, int error); 330 void iomap_ioend_try_merge(struct iomap_ioend *ioend, 331 struct list_head *more_ioends); 332 void iomap_sort_ioends(struct list_head *ioend_list); 333 int iomap_writepages(struct address_space *mapping, 334 struct writeback_control *wbc, struct iomap_writepage_ctx *wpc, 335 const struct iomap_writeback_ops *ops); 336 337 /* 338 * Flags for direct I/O ->end_io: 339 */ 340 #define IOMAP_DIO_UNWRITTEN (1 << 0) /* covers unwritten extent(s) */ 341 #define IOMAP_DIO_COW (1 << 1) /* covers COW extent(s) */ 342 343 struct iomap_dio_ops { 344 int (*end_io)(struct kiocb *iocb, ssize_t size, int error, 345 unsigned flags); 346 void (*submit_io)(const struct iomap_iter *iter, struct bio *bio, 347 loff_t file_offset); 348 349 /* 350 * Filesystems wishing to attach private information to a direct io bio 351 * must provide a ->submit_io method that attaches the additional 352 * information to the bio and changes the ->bi_end_io callback to a 353 * custom function. This function should, at a minimum, perform any 354 * relevant post-processing of the bio and end with a call to 355 * iomap_dio_bio_end_io. 356 */ 357 struct bio_set *bio_set; 358 }; 359 360 /* 361 * Wait for the I/O to complete in iomap_dio_rw even if the kiocb is not 362 * synchronous. 363 */ 364 #define IOMAP_DIO_FORCE_WAIT (1 << 0) 365 366 /* 367 * Do not allocate blocks or zero partial blocks, but instead fall back to 368 * the caller by returning -EAGAIN. Used to optimize direct I/O writes that 369 * are not aligned to the file system block size. 370 */ 371 #define IOMAP_DIO_OVERWRITE_ONLY (1 << 1) 372 373 /* 374 * When a page fault occurs, return a partial synchronous result and allow 375 * the caller to retry the rest of the operation after dealing with the page 376 * fault. 377 */ 378 #define IOMAP_DIO_PARTIAL (1 << 2) 379 380 ssize_t iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter, 381 const struct iomap_ops *ops, const struct iomap_dio_ops *dops, 382 unsigned int dio_flags, void *private, size_t done_before); 383 struct iomap_dio *__iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter, 384 const struct iomap_ops *ops, const struct iomap_dio_ops *dops, 385 unsigned int dio_flags, void *private, size_t done_before); 386 ssize_t iomap_dio_complete(struct iomap_dio *dio); 387 void iomap_dio_bio_end_io(struct bio *bio); 388 389 #ifdef CONFIG_SWAP 390 struct file; 391 struct swap_info_struct; 392 393 int iomap_swapfile_activate(struct swap_info_struct *sis, 394 struct file *swap_file, sector_t *pagespan, 395 const struct iomap_ops *ops); 396 #else 397 # define iomap_swapfile_activate(sis, swapfile, pagespan, ops) (-EIO) 398 #endif /* CONFIG_SWAP */ 399 400 #endif /* LINUX_IOMAP_H */ 401