xref: /linux/include/linux/iomap.h (revision 0f00132132937ca01a99feaf8985109a9087c9ff)
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 #include <linux/pagevec.h>
13 
14 struct address_space;
15 struct fiemap_extent_info;
16 struct inode;
17 struct iomap_iter;
18 struct iomap_dio;
19 struct iomap_writepage_ctx;
20 struct iomap_read_folio_ctx;
21 struct iov_iter;
22 struct kiocb;
23 struct page;
24 struct vm_area_struct;
25 struct vm_fault;
26 
27 /*
28  * Types of block ranges for iomap mappings:
29  */
30 #define IOMAP_HOLE	0	/* no blocks allocated, need allocation */
31 #define IOMAP_DELALLOC	1	/* delayed allocation blocks */
32 #define IOMAP_MAPPED	2	/* blocks allocated at @addr */
33 #define IOMAP_UNWRITTEN	3	/* blocks allocated at @addr in unwritten state */
34 #define IOMAP_INLINE	4	/* data inline in the inode */
35 
36 /*
37  * Flags reported by the file system from iomap_begin:
38  *
39  * IOMAP_F_NEW indicates that the blocks have been newly allocated and need
40  * zeroing for areas that no data is copied to.
41  *
42  * IOMAP_F_DIRTY indicates the inode has uncommitted metadata needed to access
43  * written data and requires fdatasync to commit them to persistent storage.
44  * This needs to take into account metadata changes that *may* be made at IO
45  * completion, such as file size updates from direct IO.
46  *
47  * IOMAP_F_SHARED indicates that the blocks are shared, and will need to be
48  * unshared as part a write.
49  *
50  * IOMAP_F_MERGED indicates that the iomap contains the merge of multiple block
51  * mappings.
52  *
53  * IOMAP_F_BUFFER_HEAD indicates that the file system requires the use of
54  * buffer heads for this mapping.
55  *
56  * IOMAP_F_XATTR indicates that the iomap is for an extended attribute extent
57  * rather than a file data extent.
58  *
59  * IOMAP_F_BOUNDARY indicates that I/O and I/O completions for this iomap must
60  * never be merged with the mapping before it.
61  *
62  * IOMAP_F_ANON_WRITE indicates that (write) I/O does not have a target block
63  * assigned to it yet and the file system will do that in the bio submission
64  * handler, splitting the I/O as needed.
65  *
66  * IOMAP_F_ATOMIC_BIO indicates that (write) I/O will be issued as an atomic
67  * bio, i.e. set REQ_ATOMIC.
68  *
69  * IOMAP_F_INTEGRITY indicates that the filesystems handles integrity metadata.
70  */
71 #define IOMAP_F_NEW		(1U << 0)
72 #define IOMAP_F_DIRTY		(1U << 1)
73 #define IOMAP_F_SHARED		(1U << 2)
74 #define IOMAP_F_MERGED		(1U << 3)
75 #ifdef CONFIG_BUFFER_HEAD
76 #define IOMAP_F_BUFFER_HEAD	(1U << 4)
77 #else
78 #define IOMAP_F_BUFFER_HEAD	0
79 #endif /* CONFIG_BUFFER_HEAD */
80 #define IOMAP_F_XATTR		(1U << 5)
81 #define IOMAP_F_BOUNDARY	(1U << 6)
82 #define IOMAP_F_ANON_WRITE	(1U << 7)
83 #define IOMAP_F_ATOMIC_BIO	(1U << 8)
84 #ifdef CONFIG_BLK_DEV_INTEGRITY
85 #define IOMAP_F_INTEGRITY	(1U << 9)
86 #else
87 #define IOMAP_F_INTEGRITY	0
88 #endif /* CONFIG_BLK_DEV_INTEGRITY */
89 
90 /*
91  * Flag reserved for file system specific usage
92  */
93 #define IOMAP_F_PRIVATE		(1U << 12)
94 
95 /*
96  * Flags set by the core iomap code during operations:
97  *
98  * IOMAP_F_FOLIO_BATCH indicates that the folio batch mechanism is active
99  * for this operation, set by iomap_fill_dirty_folios().
100  *
101  * IOMAP_F_SIZE_CHANGED indicates to the iomap_end method that the file size
102  * has changed as the result of this write operation.
103  *
104  * IOMAP_F_STALE indicates that the iomap is not valid any longer and the file
105  * range it covers needs to be remapped by the high level before the operation
106  * can proceed.
107  */
108 #define IOMAP_F_FOLIO_BATCH	(1U << 13)
109 #define IOMAP_F_SIZE_CHANGED	(1U << 14)
110 #define IOMAP_F_STALE		(1U << 15)
111 
112 /*
113  * Magic value for addr:
114  */
115 #define IOMAP_NULL_ADDR -1ULL	/* addr is not valid */
116 
117 struct iomap {
118 	u64			addr; /* disk offset of mapping, bytes */
119 	loff_t			offset;	/* file offset of mapping, bytes */
120 	u64			length;	/* length of mapping, bytes */
121 	u16			type;	/* type of mapping */
122 	u16			flags;	/* flags for mapping */
123 	struct block_device	*bdev;	/* block device for I/O */
124 	struct dax_device	*dax_dev; /* dax_dev for dax operations */
125 	void			*inline_data;
126 	void			*private; /* filesystem private */
127 	u64			validity_cookie; /* used with .iomap_valid() */
128 };
129 
iomap_sector(const struct iomap * iomap,loff_t pos)130 static inline sector_t iomap_sector(const struct iomap *iomap, loff_t pos)
131 {
132 	if (iomap->flags & IOMAP_F_ANON_WRITE)
133 		return U64_MAX; /* invalid */
134 	return (iomap->addr + pos - iomap->offset) >> SECTOR_SHIFT;
135 }
136 
137 /*
138  * Returns the inline data pointer for logical offset @pos.
139  */
iomap_inline_data(const struct iomap * iomap,loff_t pos)140 static inline void *iomap_inline_data(const struct iomap *iomap, loff_t pos)
141 {
142 	return iomap->inline_data + pos - iomap->offset;
143 }
144 
145 /*
146  * Check if the mapping's length is within the valid range for inline data.
147  * This is used to guard against accessing data beyond the page inline_data
148  * points at.
149  */
iomap_inline_data_valid(const struct iomap * iomap)150 static inline bool iomap_inline_data_valid(const struct iomap *iomap)
151 {
152 	return iomap->length <= PAGE_SIZE - offset_in_page(iomap->inline_data);
153 }
154 
155 /*
156  * When get_folio succeeds, put_folio will always be called to do any
157  * cleanup work necessary.  put_folio is responsible for unlocking and putting
158  * @folio.
159  */
160 struct iomap_write_ops {
161 	struct folio *(*get_folio)(struct iomap_iter *iter, loff_t pos,
162 			unsigned len);
163 	void (*put_folio)(struct inode *inode, loff_t pos, unsigned copied,
164 			struct folio *folio);
165 
166 	/*
167 	 * Check that the cached iomap still maps correctly to the filesystem's
168 	 * internal extent map. FS internal extent maps can change while iomap
169 	 * is iterating a cached iomap, so this hook allows iomap to detect that
170 	 * the iomap needs to be refreshed during a long running write
171 	 * operation.
172 	 *
173 	 * The filesystem can store internal state (e.g. a sequence number) in
174 	 * iomap->validity_cookie when the iomap is first mapped to be able to
175 	 * detect changes between mapping time and whenever .iomap_valid() is
176 	 * called.
177 	 *
178 	 * This is called with the folio over the specified file position held
179 	 * locked by the iomap code.
180 	 */
181 	bool (*iomap_valid)(struct inode *inode, const struct iomap *iomap);
182 
183 	/*
184 	 * Optional if the filesystem wishes to provide a custom handler for
185 	 * reading in the contents of a folio, otherwise iomap will default to
186 	 * submitting a bio read request.
187 	 *
188 	 * The read must be done synchronously.
189 	 */
190 	int (*read_folio_range)(const struct iomap_iter *iter,
191 			struct folio *folio, loff_t pos, size_t len);
192 };
193 
194 /*
195  * Flags for iomap_begin / iomap_end.  No flag implies a read.
196  */
197 #define IOMAP_WRITE		(1 << 0) /* writing, must allocate blocks */
198 #define IOMAP_ZERO		(1 << 1) /* zeroing operation, may skip holes */
199 #define IOMAP_REPORT		(1 << 2) /* report extent status, e.g. FIEMAP */
200 #define IOMAP_FAULT		(1 << 3) /* mapping for page fault */
201 #define IOMAP_DIRECT		(1 << 4) /* direct I/O */
202 #define IOMAP_NOWAIT		(1 << 5) /* do not block */
203 #define IOMAP_OVERWRITE_ONLY	(1 << 6) /* only pure overwrites allowed */
204 #define IOMAP_UNSHARE		(1 << 7) /* unshare_file_range */
205 #ifdef CONFIG_FS_DAX
206 #define IOMAP_DAX		(1 << 8) /* DAX mapping */
207 #else
208 #define IOMAP_DAX		0
209 #endif /* CONFIG_FS_DAX */
210 #define IOMAP_ATOMIC		(1 << 9) /* torn-write protection */
211 #define IOMAP_DONTCACHE		(1 << 10)
212 
213 struct iomap_ops {
214 	/*
215 	 * Return the existing mapping at pos, or reserve space starting at
216 	 * pos for up to length, as long as we can do it as a single mapping.
217 	 * The actual length is returned in iomap->length.
218 	 */
219 	int (*iomap_begin)(struct inode *inode, loff_t pos, loff_t length,
220 			unsigned flags, struct iomap *iomap,
221 			struct iomap *srcmap);
222 
223 	/*
224 	 * Commit and/or unreserve space previous allocated using iomap_begin.
225 	 * Written indicates the length of the successful write operation which
226 	 * needs to be commited, while the rest needs to be unreserved.
227 	 * Written might be zero if no data was written.
228 	 */
229 	int (*iomap_end)(struct inode *inode, loff_t pos, loff_t length,
230 			ssize_t written, unsigned flags, struct iomap *iomap);
231 };
232 
233 /**
234  * struct iomap_iter - Iterate through a range of a file
235  * @inode: Set at the start of the iteration and should not change.
236  * @pos: The current file position we are operating on.  It is updated by
237  *	calls to iomap_iter().  Treat as read-only in the body.
238  * @len: The remaining length of the file segment we're operating on.
239  *	It is updated at the same time as @pos.
240  * @iter_start_pos: The original start pos for the current iomap. Used for
241  *	incremental iter advance.
242  * @status: Status of the most recent iteration. Zero on success or a negative
243  *	errno on error.
244  * @flags: Zero or more of the iomap_begin flags above.
245  * @iomap: Map describing the I/O iteration
246  * @srcmap: Source map for COW operations
247  */
248 struct iomap_iter {
249 	struct inode *inode;
250 	loff_t pos;
251 	u64 len;
252 	loff_t iter_start_pos;
253 	int status;
254 	unsigned flags;
255 	struct iomap iomap;
256 	struct iomap srcmap;
257 	struct folio_batch *fbatch;
258 	void *private;
259 };
260 
261 int iomap_iter(struct iomap_iter *iter, const struct iomap_ops *ops);
262 int iomap_iter_advance(struct iomap_iter *iter, u64 count);
263 
264 /**
265  * iomap_length_trim - trimmed length of the current iomap iteration
266  * @iter: iteration structure
267  * @pos: File position to trim from.
268  * @len: Length of the mapping to trim to.
269  *
270  * Returns a trimmed length that the operation applies to for the current
271  * iteration.
272  */
iomap_length_trim(const struct iomap_iter * iter,loff_t pos,u64 len)273 static inline u64 iomap_length_trim(const struct iomap_iter *iter, loff_t pos,
274 		u64 len)
275 {
276 	u64 end = iter->iomap.offset + iter->iomap.length;
277 
278 	if (iter->srcmap.type != IOMAP_HOLE)
279 		end = min(end, iter->srcmap.offset + iter->srcmap.length);
280 	return min(len, end - pos);
281 }
282 
283 /**
284  * iomap_length - length of the current iomap iteration
285  * @iter: iteration structure
286  *
287  * Returns the length that the operation applies to for the current iteration.
288  */
iomap_length(const struct iomap_iter * iter)289 static inline u64 iomap_length(const struct iomap_iter *iter)
290 {
291 	return iomap_length_trim(iter, iter->pos, iter->len);
292 }
293 
294 /**
295  * iomap_iter_advance_full - advance by the full length of current map
296  */
iomap_iter_advance_full(struct iomap_iter * iter)297 static inline int iomap_iter_advance_full(struct iomap_iter *iter)
298 {
299 	return iomap_iter_advance(iter, iomap_length(iter));
300 }
301 
302 /**
303  * iomap_iter_srcmap - return the source map for the current iomap iteration
304  * @i: iteration structure
305  *
306  * Write operations on file systems with reflink support might require a
307  * source and a destination map.  This function retourns the source map
308  * for a given operation, which may or may no be identical to the destination
309  * map in &i->iomap.
310  */
iomap_iter_srcmap(const struct iomap_iter * i)311 static inline const struct iomap *iomap_iter_srcmap(const struct iomap_iter *i)
312 {
313 	if (i->srcmap.type != IOMAP_HOLE)
314 		return &i->srcmap;
315 	return &i->iomap;
316 }
317 
318 /*
319  * Return the file offset for the first unchanged block after a short write.
320  *
321  * If nothing was written, round @pos down to point at the first block in
322  * the range, else round up to include the partially written block.
323  */
iomap_last_written_block(struct inode * inode,loff_t pos,ssize_t written)324 static inline loff_t iomap_last_written_block(struct inode *inode, loff_t pos,
325 		ssize_t written)
326 {
327 	if (unlikely(!written))
328 		return round_down(pos, i_blocksize(inode));
329 	return round_up(pos + written, i_blocksize(inode));
330 }
331 
332 /*
333  * Check if the range needs to be unshared for a FALLOC_FL_UNSHARE_RANGE
334  * operation.
335  *
336  * Don't bother with blocks that are not shared to start with; or mappings that
337  * cannot be shared, such as inline data, delalloc reservations, holes or
338  * unwritten extents.
339  *
340  * Note that we use srcmap directly instead of iomap_iter_srcmap as unsharing
341  * requires providing a separate source map, and the presence of one is a good
342  * indicator that unsharing is needed, unlike IOMAP_F_SHARED which can be set
343  * for any data that goes into the COW fork for XFS.
344  */
iomap_want_unshare_iter(const struct iomap_iter * iter)345 static inline bool iomap_want_unshare_iter(const struct iomap_iter *iter)
346 {
347 	return (iter->iomap.flags & IOMAP_F_SHARED) &&
348 		iter->srcmap.type == IOMAP_MAPPED;
349 }
350 
351 ssize_t iomap_file_buffered_write(struct kiocb *iocb, struct iov_iter *from,
352 		const struct iomap_ops *ops,
353 		const struct iomap_write_ops *write_ops, void *private);
354 void iomap_read_folio(const struct iomap_ops *ops,
355 		struct iomap_read_folio_ctx *ctx, void *private);
356 void iomap_readahead(const struct iomap_ops *ops,
357 		struct iomap_read_folio_ctx *ctx, void *private);
358 bool iomap_is_partially_uptodate(struct folio *, size_t from, size_t count);
359 struct folio *iomap_get_folio(struct iomap_iter *iter, loff_t pos, size_t len);
360 bool iomap_release_folio(struct folio *folio, gfp_t gfp_flags);
361 void iomap_invalidate_folio(struct folio *folio, size_t offset, size_t len);
362 bool iomap_dirty_folio(struct address_space *mapping, struct folio *folio);
363 int iomap_file_unshare(struct inode *inode, loff_t pos, loff_t len,
364 		const struct iomap_ops *ops,
365 		const struct iomap_write_ops *write_ops);
366 unsigned int iomap_fill_dirty_folios(struct iomap_iter *iter, loff_t *start,
367 		loff_t end, unsigned int *iomap_flags);
368 int iomap_zero_range(struct inode *inode, loff_t pos, loff_t len,
369 		bool *did_zero, const struct iomap_ops *ops,
370 		const struct iomap_write_ops *write_ops, void *private);
371 int iomap_truncate_page(struct inode *inode, loff_t pos, bool *did_zero,
372 		const struct iomap_ops *ops,
373 		const struct iomap_write_ops *write_ops, void *private);
374 vm_fault_t iomap_page_mkwrite(struct vm_fault *vmf, const struct iomap_ops *ops,
375 		void *private);
376 typedef void (*iomap_punch_t)(struct inode *inode, loff_t offset, loff_t length,
377 		struct iomap *iomap);
378 void iomap_write_delalloc_release(struct inode *inode, loff_t start_byte,
379 		loff_t end_byte, unsigned flags, struct iomap *iomap,
380 		iomap_punch_t punch);
381 
382 int iomap_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
383 		u64 start, u64 len, const struct iomap_ops *ops);
384 loff_t iomap_seek_hole(struct inode *inode, loff_t offset,
385 		const struct iomap_ops *ops);
386 loff_t iomap_seek_data(struct inode *inode, loff_t offset,
387 		const struct iomap_ops *ops);
388 sector_t iomap_bmap(struct address_space *mapping, sector_t bno,
389 		const struct iomap_ops *ops);
390 
391 /*
392  * Flags for iomap_ioend->io_flags.
393  */
394 /* shared COW extent */
395 #define IOMAP_IOEND_SHARED		(1U << 0)
396 /* unwritten extent */
397 #define IOMAP_IOEND_UNWRITTEN		(1U << 1)
398 /* don't merge into previous ioend */
399 #define IOMAP_IOEND_BOUNDARY		(1U << 2)
400 /* is direct I/O */
401 #define IOMAP_IOEND_DIRECT		(1U << 3)
402 /* is DONTCACHE I/O */
403 #define IOMAP_IOEND_DONTCACHE		(1U << 4)
404 
405 /*
406  * Flags that if set on either ioend prevent the merge of two ioends.
407  * (IOMAP_IOEND_BOUNDARY also prevents merges, but only one-way)
408  */
409 #define IOMAP_IOEND_NOMERGE_FLAGS \
410 	(IOMAP_IOEND_SHARED | IOMAP_IOEND_UNWRITTEN | IOMAP_IOEND_DIRECT | \
411 	 IOMAP_IOEND_DONTCACHE)
412 
413 /*
414  * Structure for writeback I/O completions.
415  *
416  * File systems can split a bio generated by iomap.  In that case the parent
417  * ioend it was split from is recorded in ioend->io_parent.
418  */
419 struct iomap_ioend {
420 	struct list_head	io_list;	/* next ioend in chain */
421 	u16			io_flags;	/* IOMAP_IOEND_* */
422 	struct inode		*io_inode;	/* file being written to */
423 	size_t			io_size;	/* size of the extent */
424 	atomic_t		io_remaining;	/* completetion defer count */
425 	int			io_error;	/* stashed away status */
426 	struct iomap_ioend	*io_parent;	/* parent for completions */
427 	loff_t			io_offset;	/* offset in the file */
428 	sector_t		io_sector;	/* start sector of ioend */
429 	void			*io_private;	/* file system private data */
430 	struct bio		io_bio;		/* MUST BE LAST! */
431 };
432 
iomap_ioend_from_bio(struct bio * bio)433 static inline struct iomap_ioend *iomap_ioend_from_bio(struct bio *bio)
434 {
435 	return container_of(bio, struct iomap_ioend, io_bio);
436 }
437 
438 struct iomap_writeback_ops {
439 	/*
440 	 * Performs writeback on the passed in range
441 	 *
442 	 * Can map arbitrarily large regions, but we need to call into it at
443 	 * least once per folio to allow the file systems to synchronize with
444 	 * the write path that could be invalidating mappings.
445 	 *
446 	 * An existing mapping from a previous call to this method can be reused
447 	 * by the file system if it is still valid.
448 	 *
449 	 * If this succeeds, iomap_finish_folio_write() must be called once
450 	 * writeback completes for the range, regardless of whether the
451 	 * writeback succeeded or failed.
452 	 *
453 	 * Returns the number of bytes processed or a negative errno.
454 	 */
455 	ssize_t (*writeback_range)(struct iomap_writepage_ctx *wpc,
456 			struct folio *folio, u64 pos, unsigned int len,
457 			u64 end_pos);
458 
459 	/*
460 	 * Submit a writeback context previously build up by ->writeback_range.
461 	 *
462 	 * Returns 0 if the context was successfully submitted, or a negative
463 	 * error code if not.  If @error is non-zero a failure occurred, and
464 	 * the writeback context should be completed with an error.
465 	 */
466 	int (*writeback_submit)(struct iomap_writepage_ctx *wpc, int error);
467 };
468 
469 struct iomap_writepage_ctx {
470 	struct iomap		iomap;
471 	struct inode		*inode;
472 	struct writeback_control *wbc;
473 	const struct iomap_writeback_ops *ops;
474 	u32			nr_folios;	/* folios added to the ioend */
475 	void			*wb_ctx;	/* pending writeback context */
476 };
477 
478 struct iomap_ioend *iomap_init_ioend(struct inode *inode, struct bio *bio,
479 		loff_t file_offset, u16 ioend_flags);
480 struct iomap_ioend *iomap_split_ioend(struct iomap_ioend *ioend,
481 		unsigned int max_len, bool is_append);
482 void iomap_finish_ioends(struct iomap_ioend *ioend, int error);
483 void iomap_ioend_try_merge(struct iomap_ioend *ioend,
484 		struct list_head *more_ioends);
485 void iomap_sort_ioends(struct list_head *ioend_list);
486 ssize_t iomap_add_to_ioend(struct iomap_writepage_ctx *wpc, struct folio *folio,
487 		loff_t pos, loff_t end_pos, unsigned int dirty_len);
488 int iomap_ioend_writeback_submit(struct iomap_writepage_ctx *wpc, int error);
489 
490 void iomap_finish_folio_read(struct folio *folio, size_t off, size_t len,
491 		int error);
492 void iomap_finish_folio_write(struct inode *inode, struct folio *folio,
493 		size_t len);
494 
495 int iomap_writeback_folio(struct iomap_writepage_ctx *wpc, struct folio *folio);
496 int iomap_writepages(struct iomap_writepage_ctx *wpc);
497 
498 struct iomap_read_folio_ctx {
499 	const struct iomap_read_ops *ops;
500 	struct folio		*cur_folio;
501 	struct readahead_control *rac;
502 	void			*read_ctx;
503 	loff_t			read_ctx_file_offset;
504 };
505 
506 struct iomap_read_ops {
507 	/*
508 	 * Read in a folio range.
509 	 *
510 	 * If this succeeds, iomap_finish_folio_read() must be called after the
511 	 * range is read in, regardless of whether the read succeeded or failed.
512 	 *
513 	 * Returns 0 on success or a negative error on failure.
514 	 */
515 	int (*read_folio_range)(const struct iomap_iter *iter,
516 			struct iomap_read_folio_ctx *ctx, size_t len);
517 
518 	/*
519 	 * Submit any pending read requests.
520 	 *
521 	 * This is optional.
522 	 */
523 	void (*submit_read)(const struct iomap_iter *iter,
524 			struct iomap_read_folio_ctx *ctx);
525 
526 	/*
527 	 * Optional, allows filesystem to specify own bio_set, so new bio's
528 	 * can be allocated from the provided bio_set.
529 	 */
530 	struct bio_set *bio_set;
531 };
532 
533 /*
534  * Flags for direct I/O ->end_io:
535  */
536 #define IOMAP_DIO_UNWRITTEN	(1 << 0)	/* covers unwritten extent(s) */
537 #define IOMAP_DIO_COW		(1 << 1)	/* covers COW extent(s) */
538 
539 struct iomap_dio_ops {
540 	int (*end_io)(struct kiocb *iocb, ssize_t size, int error,
541 		      unsigned flags);
542 	void (*submit_io)(const struct iomap_iter *iter, struct bio *bio,
543 		          loff_t file_offset);
544 
545 	/*
546 	 * Filesystems wishing to attach private information to a direct io bio
547 	 * must provide a ->submit_io method that attaches the additional
548 	 * information to the bio and changes the ->bi_end_io callback to a
549 	 * custom function.  This function should, at a minimum, perform any
550 	 * relevant post-processing of the bio and end with a call to
551 	 * iomap_dio_bio_end_io.
552 	 */
553 	struct bio_set *bio_set;
554 };
555 
556 /*
557  * Wait for the I/O to complete in iomap_dio_rw even if the kiocb is not
558  * synchronous.
559  */
560 #define IOMAP_DIO_FORCE_WAIT	(1 << 0)
561 
562 /*
563  * Do not allocate blocks or zero partial blocks, but instead fall back to
564  * the caller by returning -EAGAIN.  Used to optimize direct I/O writes that
565  * are not aligned to the file system block size.
566   */
567 #define IOMAP_DIO_OVERWRITE_ONLY	(1 << 1)
568 
569 /*
570  * When a page fault occurs, return a partial synchronous result and allow
571  * the caller to retry the rest of the operation after dealing with the page
572  * fault.
573  */
574 #define IOMAP_DIO_PARTIAL		(1 << 2)
575 
576 /*
577  * Ensure each bio is aligned to fs block size.
578  *
579  * For filesystems which need to calculate/verify the checksum of each fs
580  * block. Otherwise they may not be able to handle unaligned bios.
581  */
582 #define IOMAP_DIO_FSBLOCK_ALIGNED	(1 << 3)
583 
584 /*
585  * Bounce buffer instead of using zero copy access.
586  *
587  * This is needed if the device needs stable data to checksum or generate
588  * parity.  The file system must hook into the I/O submission and offload
589  * completions to user context for reads when this is set.
590  */
591 #define IOMAP_DIO_BOUNCE		(1 << 4)
592 
593 ssize_t iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
594 		const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
595 		unsigned int dio_flags, void *private, size_t done_before);
596 struct iomap_dio *__iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
597 		const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
598 		unsigned int dio_flags, void *private, size_t done_before);
599 ssize_t iomap_dio_complete(struct iomap_dio *dio);
600 void iomap_dio_bio_end_io(struct bio *bio);
601 
602 #ifdef CONFIG_SWAP
603 struct file;
604 struct swap_info_struct;
605 
606 int iomap_swapfile_activate(struct swap_info_struct *sis,
607 		struct file *swap_file, sector_t *pagespan,
608 		const struct iomap_ops *ops);
609 #else
610 # define iomap_swapfile_activate(sis, swapfile, pagespan, ops)	(-EIO)
611 #endif /* CONFIG_SWAP */
612 
613 extern struct bio_set iomap_ioend_bioset;
614 
615 #ifdef CONFIG_BLOCK
616 int iomap_bio_read_folio_range(const struct iomap_iter *iter,
617 		struct iomap_read_folio_ctx *ctx, size_t plen);
618 
619 extern const struct iomap_read_ops iomap_bio_read_ops;
620 
iomap_bio_read_folio(struct folio * folio,const struct iomap_ops * ops)621 static inline void iomap_bio_read_folio(struct folio *folio,
622 		const struct iomap_ops *ops)
623 {
624 	struct iomap_read_folio_ctx ctx = {
625 		.ops		= &iomap_bio_read_ops,
626 		.cur_folio	= folio,
627 	};
628 
629 	iomap_read_folio(ops, &ctx, NULL);
630 }
631 
iomap_bio_readahead(struct readahead_control * rac,const struct iomap_ops * ops)632 static inline void iomap_bio_readahead(struct readahead_control *rac,
633 		const struct iomap_ops *ops)
634 {
635 	struct iomap_read_folio_ctx ctx = {
636 		.ops		= &iomap_bio_read_ops,
637 		.rac		= rac,
638 	};
639 
640 	iomap_readahead(ops, &ctx, NULL);
641 }
642 #endif /* CONFIG_BLOCK */
643 
644 #endif /* LINUX_IOMAP_H */
645