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/android_kabi.h>
13
14 struct address_space;
15 struct fiemap_extent_info;
16 struct inode;
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 #define IOMAP_F_NEW 0x01
55 #define IOMAP_F_DIRTY 0x02
56 #define IOMAP_F_SHARED 0x04
57 #define IOMAP_F_MERGED 0x08
58 #define IOMAP_F_BUFFER_HEAD 0x10
59 #define IOMAP_F_ZONE_APPEND 0x20
60
61 /*
62 * Flags set by the core iomap code during operations:
63 *
64 * IOMAP_F_SIZE_CHANGED indicates to the iomap_end method that the file size
65 * has changed as the result of this write operation.
66 */
67 #define IOMAP_F_SIZE_CHANGED 0x100
68
69 /*
70 * Flags from 0x1000 up are for file system specific usage:
71 */
72 #define IOMAP_F_PRIVATE 0x1000
73
74
75 /*
76 * Magic value for addr:
77 */
78 #define IOMAP_NULL_ADDR -1ULL /* addr is not valid */
79
80 struct iomap_page_ops;
81
82 struct iomap {
83 u64 addr; /* disk offset of mapping, bytes */
84 loff_t offset; /* file offset of mapping, bytes */
85 u64 length; /* length of mapping, bytes */
86 u16 type; /* type of mapping */
87 u16 flags; /* flags for mapping */
88 struct block_device *bdev; /* block device for I/O */
89 struct dax_device *dax_dev; /* dax_dev for dax operations */
90 void *inline_data;
91 void *private; /* filesystem private */
92 const struct iomap_page_ops *page_ops;
93
94 ANDROID_KABI_RESERVE(1);
95 };
96
iomap_sector(const struct iomap * iomap,loff_t pos)97 static inline sector_t iomap_sector(const struct iomap *iomap, loff_t pos)
98 {
99 return (iomap->addr + pos - iomap->offset) >> SECTOR_SHIFT;
100 }
101
102 /*
103 * Returns the inline data pointer for logical offset @pos.
104 */
iomap_inline_data(const struct iomap * iomap,loff_t pos)105 static inline void *iomap_inline_data(const struct iomap *iomap, loff_t pos)
106 {
107 return iomap->inline_data + pos - iomap->offset;
108 }
109
110 /*
111 * Check if the mapping's length is within the valid range for inline data.
112 * This is used to guard against accessing data beyond the page inline_data
113 * points at.
114 */
iomap_inline_data_valid(const struct iomap * iomap)115 static inline bool iomap_inline_data_valid(const struct iomap *iomap)
116 {
117 return iomap->length <= PAGE_SIZE - offset_in_page(iomap->inline_data);
118 }
119
120 /*
121 * When a filesystem sets page_ops in an iomap mapping it returns, page_prepare
122 * and page_done will be called for each page written to. This only applies to
123 * buffered writes as unbuffered writes will not typically have pages
124 * associated with them.
125 *
126 * When page_prepare succeeds, page_done will always be called to do any
127 * cleanup work necessary. In that page_done call, @page will be NULL if the
128 * associated page could not be obtained.
129 */
130 struct iomap_page_ops {
131 int (*page_prepare)(struct inode *inode, loff_t pos, unsigned len);
132 void (*page_done)(struct inode *inode, loff_t pos, unsigned copied,
133 struct page *page);
134 };
135
136 /*
137 * Flags for iomap_begin / iomap_end. No flag implies a read.
138 */
139 #define IOMAP_WRITE (1 << 0) /* writing, must allocate blocks */
140 #define IOMAP_ZERO (1 << 1) /* zeroing operation, may skip holes */
141 #define IOMAP_REPORT (1 << 2) /* report extent status, e.g. FIEMAP */
142 #define IOMAP_FAULT (1 << 3) /* mapping for page fault */
143 #define IOMAP_DIRECT (1 << 4) /* direct I/O */
144 #define IOMAP_NOWAIT (1 << 5) /* do not block */
145 #define IOMAP_OVERWRITE_ONLY (1 << 6) /* only pure overwrites allowed */
146 #define IOMAP_UNSHARE (1 << 7) /* unshare_file_range */
147
148 struct iomap_ops {
149 /*
150 * Return the existing mapping at pos, or reserve space starting at
151 * pos for up to length, as long as we can do it as a single mapping.
152 * The actual length is returned in iomap->length.
153 */
154 int (*iomap_begin)(struct inode *inode, loff_t pos, loff_t length,
155 unsigned flags, struct iomap *iomap,
156 struct iomap *srcmap);
157
158 /*
159 * Commit and/or unreserve space previous allocated using iomap_begin.
160 * Written indicates the length of the successful write operation which
161 * needs to be commited, while the rest needs to be unreserved.
162 * Written might be zero if no data was written.
163 */
164 int (*iomap_end)(struct inode *inode, loff_t pos, loff_t length,
165 ssize_t written, unsigned flags, struct iomap *iomap);
166
167 ANDROID_KABI_RESERVE(1);
168 ANDROID_KABI_RESERVE(2);
169 };
170
171 /**
172 * struct iomap_iter - Iterate through a range of a file
173 * @inode: Set at the start of the iteration and should not change.
174 * @pos: The current file position we are operating on. It is updated by
175 * calls to iomap_iter(). Treat as read-only in the body.
176 * @len: The remaining length of the file segment we're operating on.
177 * It is updated at the same time as @pos.
178 * @processed: The number of bytes processed by the body in the most recent
179 * iteration, or a negative errno. 0 causes the iteration to stop.
180 * @flags: Zero or more of the iomap_begin flags above.
181 * @iomap: Map describing the I/O iteration
182 * @srcmap: Source map for COW operations
183 */
184 struct iomap_iter {
185 struct inode *inode;
186 loff_t pos;
187 u64 len;
188 s64 processed;
189 unsigned flags;
190 struct iomap iomap;
191 struct iomap srcmap;
192 };
193
194 int iomap_iter(struct iomap_iter *iter, const struct iomap_ops *ops);
195
196 /**
197 * iomap_length - length of the current iomap iteration
198 * @iter: iteration structure
199 *
200 * Returns the length that the operation applies to for the current iteration.
201 */
iomap_length(const struct iomap_iter * iter)202 static inline u64 iomap_length(const struct iomap_iter *iter)
203 {
204 u64 end = iter->iomap.offset + iter->iomap.length;
205
206 if (iter->srcmap.type != IOMAP_HOLE)
207 end = min(end, iter->srcmap.offset + iter->srcmap.length);
208 return min(iter->len, end - iter->pos);
209 }
210
211 /**
212 * iomap_iter_srcmap - return the source map for the current iomap iteration
213 * @i: iteration structure
214 *
215 * Write operations on file systems with reflink support might require a
216 * source and a destination map. This function retourns the source map
217 * for a given operation, which may or may no be identical to the destination
218 * map in &i->iomap.
219 */
iomap_iter_srcmap(const struct iomap_iter * i)220 static inline const struct iomap *iomap_iter_srcmap(const struct iomap_iter *i)
221 {
222 if (i->srcmap.type != IOMAP_HOLE)
223 return &i->srcmap;
224 return &i->iomap;
225 }
226
227 ssize_t iomap_file_buffered_write(struct kiocb *iocb, struct iov_iter *from,
228 const struct iomap_ops *ops);
229 int iomap_readpage(struct page *page, const struct iomap_ops *ops);
230 void iomap_readahead(struct readahead_control *, const struct iomap_ops *ops);
231 int iomap_is_partially_uptodate(struct page *page, unsigned long from,
232 unsigned long count);
233 int iomap_releasepage(struct page *page, gfp_t gfp_mask);
234 void iomap_invalidatepage(struct page *page, unsigned int offset,
235 unsigned int len);
236 #ifdef CONFIG_MIGRATION
237 int iomap_migrate_page(struct address_space *mapping, struct page *newpage,
238 struct page *page, enum migrate_mode mode);
239 #else
240 #define iomap_migrate_page NULL
241 #endif
242 int iomap_file_unshare(struct inode *inode, loff_t pos, loff_t len,
243 const struct iomap_ops *ops);
244 int iomap_zero_range(struct inode *inode, loff_t pos, loff_t len,
245 bool *did_zero, const struct iomap_ops *ops);
246 int iomap_truncate_page(struct inode *inode, loff_t pos, bool *did_zero,
247 const struct iomap_ops *ops);
248 vm_fault_t iomap_page_mkwrite(struct vm_fault *vmf,
249 const struct iomap_ops *ops);
250 int iomap_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
251 u64 start, u64 len, const struct iomap_ops *ops);
252 loff_t iomap_seek_hole(struct inode *inode, loff_t offset,
253 const struct iomap_ops *ops);
254 loff_t iomap_seek_data(struct inode *inode, loff_t offset,
255 const struct iomap_ops *ops);
256 sector_t iomap_bmap(struct address_space *mapping, sector_t bno,
257 const struct iomap_ops *ops);
258
259 /*
260 * Structure for writeback I/O completions.
261 */
262 struct iomap_ioend {
263 struct list_head io_list; /* next ioend in chain */
264 u16 io_type;
265 u16 io_flags; /* IOMAP_F_* */
266 struct inode *io_inode; /* file being written to */
267 size_t io_size; /* size of the extent */
268 loff_t io_offset; /* offset in the file */
269 struct bio *io_bio; /* bio being built */
270 struct bio io_inline_bio; /* MUST BE LAST! */
271 };
272
273 struct iomap_writeback_ops {
274 /*
275 * Required, maps the blocks so that writeback can be performed on
276 * the range starting at offset.
277 */
278 int (*map_blocks)(struct iomap_writepage_ctx *wpc, struct inode *inode,
279 loff_t offset);
280
281 /*
282 * Optional, allows the file systems to perform actions just before
283 * submitting the bio and/or override the bio end_io handler for complex
284 * operations like copy on write extent manipulation or unwritten extent
285 * conversions.
286 */
287 int (*prepare_ioend)(struct iomap_ioend *ioend, int status);
288
289 /*
290 * Optional, allows the file system to discard state on a page where
291 * we failed to submit any I/O.
292 */
293 void (*discard_page)(struct page *page, loff_t fileoff);
294 };
295
296 struct iomap_writepage_ctx {
297 struct iomap iomap;
298 struct iomap_ioend *ioend;
299 const struct iomap_writeback_ops *ops;
300 };
301
302 void iomap_finish_ioends(struct iomap_ioend *ioend, int error);
303 void iomap_ioend_try_merge(struct iomap_ioend *ioend,
304 struct list_head *more_ioends);
305 void iomap_sort_ioends(struct list_head *ioend_list);
306 int iomap_writepage(struct page *page, struct writeback_control *wbc,
307 struct iomap_writepage_ctx *wpc,
308 const struct iomap_writeback_ops *ops);
309 int iomap_writepages(struct address_space *mapping,
310 struct writeback_control *wbc, struct iomap_writepage_ctx *wpc,
311 const struct iomap_writeback_ops *ops);
312
313 /*
314 * Flags for direct I/O ->end_io:
315 */
316 #define IOMAP_DIO_UNWRITTEN (1 << 0) /* covers unwritten extent(s) */
317 #define IOMAP_DIO_COW (1 << 1) /* covers COW extent(s) */
318
319 struct iomap_dio_ops {
320 int (*end_io)(struct kiocb *iocb, ssize_t size, int error,
321 unsigned flags);
322 blk_qc_t (*submit_io)(const struct iomap_iter *iter, struct bio *bio,
323 loff_t file_offset);
324 };
325
326 /*
327 * Wait for the I/O to complete in iomap_dio_rw even if the kiocb is not
328 * synchronous.
329 */
330 #define IOMAP_DIO_FORCE_WAIT (1 << 0)
331
332 /*
333 * Do not allocate blocks or zero partial blocks, but instead fall back to
334 * the caller by returning -EAGAIN. Used to optimize direct I/O writes that
335 * are not aligned to the file system block size.
336 */
337 #define IOMAP_DIO_OVERWRITE_ONLY (1 << 1)
338
339 /*
340 * When a page fault occurs, return a partial synchronous result and allow
341 * the caller to retry the rest of the operation after dealing with the page
342 * fault.
343 */
344 #define IOMAP_DIO_PARTIAL (1 << 2)
345
346 ssize_t iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
347 const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
348 unsigned int dio_flags, size_t done_before);
349 struct iomap_dio *__iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
350 const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
351 unsigned int dio_flags, size_t done_before);
352 ssize_t iomap_dio_complete(struct iomap_dio *dio);
353 int iomap_dio_iopoll(struct kiocb *kiocb, bool spin);
354
355 #ifdef CONFIG_SWAP
356 struct file;
357 struct swap_info_struct;
358
359 int iomap_swapfile_activate(struct swap_info_struct *sis,
360 struct file *swap_file, sector_t *pagespan,
361 const struct iomap_ops *ops);
362 #else
363 # define iomap_swapfile_activate(sis, swapfile, pagespan, ops) (-EIO)
364 #endif /* CONFIG_SWAP */
365
366 #endif /* LINUX_IOMAP_H */
367