1 /* SPDX-License-Identifier: GPL-2.0 */
3 #define LINUX_IOMAP_H 1
5 #include <linux/atomic.h>
6 #include <linux/bitmap.h>
7 #include <linux/blk_types.h>
9 #include <linux/types.h>
10 #include <linux/mm_types.h>
11 #include <linux/blkdev.h>
14 struct fiemap_extent_info;
17 struct iomap_writepage_ctx;
21 struct vm_area_struct;
25 * Types of block ranges for iomap mappings:
27 #define IOMAP_HOLE 0 /* no blocks allocated, need allocation */
28 #define IOMAP_DELALLOC 1 /* delayed allocation blocks */
29 #define IOMAP_MAPPED 2 /* blocks allocated at @addr */
30 #define IOMAP_UNWRITTEN 3 /* blocks allocated at @addr in unwritten state */
31 #define IOMAP_INLINE 4 /* data inline in the inode */
34 * Flags reported by the file system from iomap_begin:
36 * IOMAP_F_NEW indicates that the blocks have been newly allocated and need
37 * zeroing for areas that no data is copied to.
39 * IOMAP_F_DIRTY indicates the inode has uncommitted metadata needed to access
40 * written data and requires fdatasync to commit them to persistent storage.
41 * This needs to take into account metadata changes that *may* be made at IO
42 * completion, such as file size updates from direct IO.
44 * IOMAP_F_SHARED indicates that the blocks are shared, and will need to be
45 * unshared as part a write.
47 * IOMAP_F_MERGED indicates that the iomap contains the merge of multiple block
50 * IOMAP_F_BUFFER_HEAD indicates that the file system requires the use of
51 * buffer heads for this mapping.
53 * IOMAP_F_XATTR indicates that the iomap is for an extended attribute extent
54 * rather than a file data extent.
56 #define IOMAP_F_NEW (1U << 0)
57 #define IOMAP_F_DIRTY (1U << 1)
58 #define IOMAP_F_SHARED (1U << 2)
59 #define IOMAP_F_MERGED (1U << 3)
60 #define IOMAP_F_BUFFER_HEAD (1U << 4)
61 #define IOMAP_F_ZONE_APPEND (1U << 5)
62 #define IOMAP_F_XATTR (1U << 6)
65 * Flags set by the core iomap code during operations:
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.
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
74 #define IOMAP_F_SIZE_CHANGED (1U << 8)
75 #define IOMAP_F_STALE (1U << 9)
78 * Flags from 0x1000 up are for file system specific usage:
80 #define IOMAP_F_PRIVATE (1U << 12)
84 * Magic value for addr:
86 #define IOMAP_NULL_ADDR -1ULL /* addr is not valid */
88 struct iomap_page_ops;
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 */
99 void *private; /* filesystem private */
100 const struct iomap_page_ops *page_ops;
101 u64 validity_cookie; /* used with .iomap_valid() */
104 static inline sector_t iomap_sector(const struct iomap *iomap, loff_t pos)
106 return (iomap->addr + pos - iomap->offset) >> SECTOR_SHIFT;
110 * Returns the inline data pointer for logical offset @pos.
112 static inline void *iomap_inline_data(const struct iomap *iomap, loff_t pos)
114 return iomap->inline_data + pos - iomap->offset;
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
122 static inline bool iomap_inline_data_valid(const struct iomap *iomap)
124 return iomap->length <= PAGE_SIZE - offset_in_page(iomap->inline_data);
128 * When a filesystem sets page_ops in an iomap mapping it returns, page_prepare
129 * and page_done will be called for each page written to. This only applies to
130 * buffered writes as unbuffered writes will not typically have pages
131 * associated with them.
133 * When page_prepare succeeds, page_done will always be called to do any
134 * cleanup work necessary. In that page_done call, @page will be NULL if the
135 * associated page could not be obtained.
137 struct iomap_page_ops {
138 int (*page_prepare)(struct inode *inode, loff_t pos, unsigned len);
139 void (*page_done)(struct inode *inode, loff_t pos, unsigned copied,
143 * Check that the cached iomap still maps correctly to the filesystem's
144 * internal extent map. FS internal extent maps can change while iomap
145 * is iterating a cached iomap, so this hook allows iomap to detect that
146 * the iomap needs to be refreshed during a long running write
149 * The filesystem can store internal state (e.g. a sequence number) in
150 * iomap->validity_cookie when the iomap is first mapped to be able to
151 * detect changes between mapping time and whenever .iomap_valid() is
154 * This is called with the folio over the specified file position held
155 * locked by the iomap code.
157 bool (*iomap_valid)(struct inode *inode, const struct iomap *iomap);
161 * Flags for iomap_begin / iomap_end. No flag implies a read.
163 #define IOMAP_WRITE (1 << 0) /* writing, must allocate blocks */
164 #define IOMAP_ZERO (1 << 1) /* zeroing operation, may skip holes */
165 #define IOMAP_REPORT (1 << 2) /* report extent status, e.g. FIEMAP */
166 #define IOMAP_FAULT (1 << 3) /* mapping for page fault */
167 #define IOMAP_DIRECT (1 << 4) /* direct I/O */
168 #define IOMAP_NOWAIT (1 << 5) /* do not block */
169 #define IOMAP_OVERWRITE_ONLY (1 << 6) /* only pure overwrites allowed */
170 #define IOMAP_UNSHARE (1 << 7) /* unshare_file_range */
172 #define IOMAP_DAX (1 << 8) /* DAX mapping */
175 #endif /* CONFIG_FS_DAX */
179 * Return the existing mapping at pos, or reserve space starting at
180 * pos for up to length, as long as we can do it as a single mapping.
181 * The actual length is returned in iomap->length.
183 int (*iomap_begin)(struct inode *inode, loff_t pos, loff_t length,
184 unsigned flags, struct iomap *iomap,
185 struct iomap *srcmap);
188 * Commit and/or unreserve space previous allocated using iomap_begin.
189 * Written indicates the length of the successful write operation which
190 * needs to be commited, while the rest needs to be unreserved.
191 * Written might be zero if no data was written.
193 int (*iomap_end)(struct inode *inode, loff_t pos, loff_t length,
194 ssize_t written, unsigned flags, struct iomap *iomap);
198 * struct iomap_iter - Iterate through a range of a file
199 * @inode: Set at the start of the iteration and should not change.
200 * @pos: The current file position we are operating on. It is updated by
201 * calls to iomap_iter(). Treat as read-only in the body.
202 * @len: The remaining length of the file segment we're operating on.
203 * It is updated at the same time as @pos.
204 * @processed: The number of bytes processed by the body in the most recent
205 * iteration, or a negative errno. 0 causes the iteration to stop.
206 * @flags: Zero or more of the iomap_begin flags above.
207 * @iomap: Map describing the I/O iteration
208 * @srcmap: Source map for COW operations
221 int iomap_iter(struct iomap_iter *iter, const struct iomap_ops *ops);
224 * iomap_length - length of the current iomap iteration
225 * @iter: iteration structure
227 * Returns the length that the operation applies to for the current iteration.
229 static inline u64 iomap_length(const struct iomap_iter *iter)
231 u64 end = iter->iomap.offset + iter->iomap.length;
233 if (iter->srcmap.type != IOMAP_HOLE)
234 end = min(end, iter->srcmap.offset + iter->srcmap.length);
235 return min(iter->len, end - iter->pos);
239 * iomap_iter_srcmap - return the source map for the current iomap iteration
240 * @i: iteration structure
242 * Write operations on file systems with reflink support might require a
243 * source and a destination map. This function retourns the source map
244 * for a given operation, which may or may no be identical to the destination
247 static inline const struct iomap *iomap_iter_srcmap(const struct iomap_iter *i)
249 if (i->srcmap.type != IOMAP_HOLE)
254 ssize_t iomap_file_buffered_write(struct kiocb *iocb, struct iov_iter *from,
255 const struct iomap_ops *ops);
256 int iomap_file_buffered_write_punch_delalloc(struct inode *inode,
257 struct iomap *iomap, loff_t pos, loff_t length, ssize_t written,
258 int (*punch)(struct inode *inode, loff_t pos, loff_t length));
260 int iomap_read_folio(struct folio *folio, const struct iomap_ops *ops);
261 void iomap_readahead(struct readahead_control *, const struct iomap_ops *ops);
262 bool iomap_is_partially_uptodate(struct folio *, size_t from, size_t count);
263 bool iomap_release_folio(struct folio *folio, gfp_t gfp_flags);
264 void iomap_invalidate_folio(struct folio *folio, size_t offset, size_t len);
265 int iomap_file_unshare(struct inode *inode, loff_t pos, loff_t len,
266 const struct iomap_ops *ops);
267 int iomap_zero_range(struct inode *inode, loff_t pos, loff_t len,
268 bool *did_zero, const struct iomap_ops *ops);
269 int iomap_truncate_page(struct inode *inode, loff_t pos, bool *did_zero,
270 const struct iomap_ops *ops);
271 vm_fault_t iomap_page_mkwrite(struct vm_fault *vmf,
272 const struct iomap_ops *ops);
273 int iomap_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
274 u64 start, u64 len, const struct iomap_ops *ops);
275 loff_t iomap_seek_hole(struct inode *inode, loff_t offset,
276 const struct iomap_ops *ops);
277 loff_t iomap_seek_data(struct inode *inode, loff_t offset,
278 const struct iomap_ops *ops);
279 sector_t iomap_bmap(struct address_space *mapping, sector_t bno,
280 const struct iomap_ops *ops);
283 * Structure for writeback I/O completions.
286 struct list_head io_list; /* next ioend in chain */
288 u16 io_flags; /* IOMAP_F_* */
289 u32 io_folios; /* folios added to ioend */
290 struct inode *io_inode; /* file being written to */
291 size_t io_size; /* size of the extent */
292 loff_t io_offset; /* offset in the file */
293 sector_t io_sector; /* start sector of ioend */
294 struct bio *io_bio; /* bio being built */
295 struct bio io_inline_bio; /* MUST BE LAST! */
298 struct iomap_writeback_ops {
300 * Required, maps the blocks so that writeback can be performed on
301 * the range starting at offset.
303 int (*map_blocks)(struct iomap_writepage_ctx *wpc, struct inode *inode,
307 * Optional, allows the file systems to perform actions just before
308 * submitting the bio and/or override the bio end_io handler for complex
309 * operations like copy on write extent manipulation or unwritten extent
312 int (*prepare_ioend)(struct iomap_ioend *ioend, int status);
315 * Optional, allows the file system to discard state on a page where
316 * we failed to submit any I/O.
318 void (*discard_folio)(struct folio *folio, loff_t pos);
321 struct iomap_writepage_ctx {
323 struct iomap_ioend *ioend;
324 const struct iomap_writeback_ops *ops;
327 void iomap_finish_ioends(struct iomap_ioend *ioend, int error);
328 void iomap_ioend_try_merge(struct iomap_ioend *ioend,
329 struct list_head *more_ioends);
330 void iomap_sort_ioends(struct list_head *ioend_list);
331 int iomap_writepages(struct address_space *mapping,
332 struct writeback_control *wbc, struct iomap_writepage_ctx *wpc,
333 const struct iomap_writeback_ops *ops);
336 * Flags for direct I/O ->end_io:
338 #define IOMAP_DIO_UNWRITTEN (1 << 0) /* covers unwritten extent(s) */
339 #define IOMAP_DIO_COW (1 << 1) /* covers COW extent(s) */
341 struct iomap_dio_ops {
342 int (*end_io)(struct kiocb *iocb, ssize_t size, int error,
344 void (*submit_io)(const struct iomap_iter *iter, struct bio *bio,
348 * Filesystems wishing to attach private information to a direct io bio
349 * must provide a ->submit_io method that attaches the additional
350 * information to the bio and changes the ->bi_end_io callback to a
351 * custom function. This function should, at a minimum, perform any
352 * relevant post-processing of the bio and end with a call to
353 * iomap_dio_bio_end_io.
355 struct bio_set *bio_set;
359 * Wait for the I/O to complete in iomap_dio_rw even if the kiocb is not
362 #define IOMAP_DIO_FORCE_WAIT (1 << 0)
365 * Do not allocate blocks or zero partial blocks, but instead fall back to
366 * the caller by returning -EAGAIN. Used to optimize direct I/O writes that
367 * are not aligned to the file system block size.
369 #define IOMAP_DIO_OVERWRITE_ONLY (1 << 1)
372 * When a page fault occurs, return a partial synchronous result and allow
373 * the caller to retry the rest of the operation after dealing with the page
376 #define IOMAP_DIO_PARTIAL (1 << 2)
379 * The caller will sync the write if needed; do not sync it within
380 * iomap_dio_rw. Overrides IOMAP_DIO_FORCE_WAIT.
382 #define IOMAP_DIO_NOSYNC (1 << 3)
384 ssize_t iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
385 const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
386 unsigned int dio_flags, void *private, size_t done_before);
387 struct iomap_dio *__iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
388 const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
389 unsigned int dio_flags, void *private, size_t done_before);
390 ssize_t iomap_dio_complete(struct iomap_dio *dio);
391 void iomap_dio_bio_end_io(struct bio *bio);
395 struct swap_info_struct;
397 int iomap_swapfile_activate(struct swap_info_struct *sis,
398 struct file *swap_file, sector_t *pagespan,
399 const struct iomap_ops *ops);
401 # define iomap_swapfile_activate(sis, swapfile, pagespan, ops) (-EIO)
402 #endif /* CONFIG_SWAP */
404 #endif /* LINUX_IOMAP_H */