]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - include/linux/iomap.h
nfs: don't atempt blocking locks on nfs reexports
[mirror_ubuntu-jammy-kernel.git] / include / linux / iomap.h
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_dio;
17 struct iomap_writepage_ctx;
18 struct iov_iter;
19 struct kiocb;
20 struct page;
21 struct vm_area_struct;
22 struct vm_fault;
23
24 /*
25 * Types of block ranges for iomap mappings:
26 */
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 */
32
33 /*
34 * Flags reported by the file system from iomap_begin:
35 *
36 * IOMAP_F_NEW indicates that the blocks have been newly allocated and need
37 * zeroing for areas that no data is copied to.
38 *
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.
43 *
44 * IOMAP_F_SHARED indicates that the blocks are shared, and will need to be
45 * unshared as part a write.
46 *
47 * IOMAP_F_MERGED indicates that the iomap contains the merge of multiple block
48 * mappings.
49 *
50 * IOMAP_F_BUFFER_HEAD indicates that the file system requires the use of
51 * buffer heads for this mapping.
52 */
53 #define IOMAP_F_NEW 0x01
54 #define IOMAP_F_DIRTY 0x02
55 #define IOMAP_F_SHARED 0x04
56 #define IOMAP_F_MERGED 0x08
57 #define IOMAP_F_BUFFER_HEAD 0x10
58 #define IOMAP_F_ZONE_APPEND 0x20
59
60 /*
61 * Flags set by the core iomap code during operations:
62 *
63 * IOMAP_F_SIZE_CHANGED indicates to the iomap_end method that the file size
64 * has changed as the result of this write operation.
65 */
66 #define IOMAP_F_SIZE_CHANGED 0x100
67
68 /*
69 * Flags from 0x1000 up are for file system specific usage:
70 */
71 #define IOMAP_F_PRIVATE 0x1000
72
73
74 /*
75 * Magic value for addr:
76 */
77 #define IOMAP_NULL_ADDR -1ULL /* addr is not valid */
78
79 struct iomap_page_ops;
80
81 struct iomap {
82 u64 addr; /* disk offset of mapping, bytes */
83 loff_t offset; /* file offset of mapping, bytes */
84 u64 length; /* length of mapping, bytes */
85 u16 type; /* type of mapping */
86 u16 flags; /* flags for mapping */
87 struct block_device *bdev; /* block device for I/O */
88 struct dax_device *dax_dev; /* dax_dev for dax operations */
89 void *inline_data;
90 void *private; /* filesystem private */
91 const struct iomap_page_ops *page_ops;
92 };
93
94 static inline sector_t
95 iomap_sector(struct iomap *iomap, loff_t pos)
96 {
97 return (iomap->addr + pos - iomap->offset) >> SECTOR_SHIFT;
98 }
99
100 /*
101 * When a filesystem sets page_ops in an iomap mapping it returns, page_prepare
102 * and page_done will be called for each page written to. This only applies to
103 * buffered writes as unbuffered writes will not typically have pages
104 * associated with them.
105 *
106 * When page_prepare succeeds, page_done will always be called to do any
107 * cleanup work necessary. In that page_done call, @page will be NULL if the
108 * associated page could not be obtained.
109 */
110 struct iomap_page_ops {
111 int (*page_prepare)(struct inode *inode, loff_t pos, unsigned len,
112 struct iomap *iomap);
113 void (*page_done)(struct inode *inode, loff_t pos, unsigned copied,
114 struct page *page, struct iomap *iomap);
115 };
116
117 /*
118 * Flags for iomap_begin / iomap_end. No flag implies a read.
119 */
120 #define IOMAP_WRITE (1 << 0) /* writing, must allocate blocks */
121 #define IOMAP_ZERO (1 << 1) /* zeroing operation, may skip holes */
122 #define IOMAP_REPORT (1 << 2) /* report extent status, e.g. FIEMAP */
123 #define IOMAP_FAULT (1 << 3) /* mapping for page fault */
124 #define IOMAP_DIRECT (1 << 4) /* direct I/O */
125 #define IOMAP_NOWAIT (1 << 5) /* do not block */
126 #define IOMAP_OVERWRITE_ONLY (1 << 6) /* only pure overwrites allowed */
127
128 struct iomap_ops {
129 /*
130 * Return the existing mapping at pos, or reserve space starting at
131 * pos for up to length, as long as we can do it as a single mapping.
132 * The actual length is returned in iomap->length.
133 */
134 int (*iomap_begin)(struct inode *inode, loff_t pos, loff_t length,
135 unsigned flags, struct iomap *iomap,
136 struct iomap *srcmap);
137
138 /*
139 * Commit and/or unreserve space previous allocated using iomap_begin.
140 * Written indicates the length of the successful write operation which
141 * needs to be commited, while the rest needs to be unreserved.
142 * Written might be zero if no data was written.
143 */
144 int (*iomap_end)(struct inode *inode, loff_t pos, loff_t length,
145 ssize_t written, unsigned flags, struct iomap *iomap);
146 };
147
148 /*
149 * Main iomap iterator function.
150 */
151 typedef loff_t (*iomap_actor_t)(struct inode *inode, loff_t pos, loff_t len,
152 void *data, struct iomap *iomap, struct iomap *srcmap);
153
154 loff_t iomap_apply(struct inode *inode, loff_t pos, loff_t length,
155 unsigned flags, const struct iomap_ops *ops, void *data,
156 iomap_actor_t actor);
157
158 ssize_t iomap_file_buffered_write(struct kiocb *iocb, struct iov_iter *from,
159 const struct iomap_ops *ops);
160 int iomap_readpage(struct page *page, const struct iomap_ops *ops);
161 void iomap_readahead(struct readahead_control *, const struct iomap_ops *ops);
162 int iomap_is_partially_uptodate(struct page *page, unsigned long from,
163 unsigned long count);
164 int iomap_releasepage(struct page *page, gfp_t gfp_mask);
165 void iomap_invalidatepage(struct page *page, unsigned int offset,
166 unsigned int len);
167 #ifdef CONFIG_MIGRATION
168 int iomap_migrate_page(struct address_space *mapping, struct page *newpage,
169 struct page *page, enum migrate_mode mode);
170 #else
171 #define iomap_migrate_page NULL
172 #endif
173 int iomap_file_unshare(struct inode *inode, loff_t pos, loff_t len,
174 const struct iomap_ops *ops);
175 int iomap_zero_range(struct inode *inode, loff_t pos, loff_t len,
176 bool *did_zero, const struct iomap_ops *ops);
177 int iomap_truncate_page(struct inode *inode, loff_t pos, bool *did_zero,
178 const struct iomap_ops *ops);
179 vm_fault_t iomap_page_mkwrite(struct vm_fault *vmf,
180 const struct iomap_ops *ops);
181 int iomap_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
182 u64 start, u64 len, const struct iomap_ops *ops);
183 loff_t iomap_seek_hole(struct inode *inode, loff_t offset,
184 const struct iomap_ops *ops);
185 loff_t iomap_seek_data(struct inode *inode, loff_t offset,
186 const struct iomap_ops *ops);
187 sector_t iomap_bmap(struct address_space *mapping, sector_t bno,
188 const struct iomap_ops *ops);
189
190 /*
191 * Structure for writeback I/O completions.
192 */
193 struct iomap_ioend {
194 struct list_head io_list; /* next ioend in chain */
195 u16 io_type;
196 u16 io_flags; /* IOMAP_F_* */
197 struct inode *io_inode; /* file being written to */
198 size_t io_size; /* size of the extent */
199 loff_t io_offset; /* offset in the file */
200 struct bio *io_bio; /* bio being built */
201 struct bio io_inline_bio; /* MUST BE LAST! */
202 };
203
204 struct iomap_writeback_ops {
205 /*
206 * Required, maps the blocks so that writeback can be performed on
207 * the range starting at offset.
208 */
209 int (*map_blocks)(struct iomap_writepage_ctx *wpc, struct inode *inode,
210 loff_t offset);
211
212 /*
213 * Optional, allows the file systems to perform actions just before
214 * submitting the bio and/or override the bio end_io handler for complex
215 * operations like copy on write extent manipulation or unwritten extent
216 * conversions.
217 */
218 int (*prepare_ioend)(struct iomap_ioend *ioend, int status);
219
220 /*
221 * Optional, allows the file system to discard state on a page where
222 * we failed to submit any I/O.
223 */
224 void (*discard_page)(struct page *page, loff_t fileoff);
225 };
226
227 struct iomap_writepage_ctx {
228 struct iomap iomap;
229 struct iomap_ioend *ioend;
230 const struct iomap_writeback_ops *ops;
231 };
232
233 void iomap_finish_ioends(struct iomap_ioend *ioend, int error);
234 void iomap_ioend_try_merge(struct iomap_ioend *ioend,
235 struct list_head *more_ioends);
236 void iomap_sort_ioends(struct list_head *ioend_list);
237 int iomap_writepage(struct page *page, struct writeback_control *wbc,
238 struct iomap_writepage_ctx *wpc,
239 const struct iomap_writeback_ops *ops);
240 int iomap_writepages(struct address_space *mapping,
241 struct writeback_control *wbc, struct iomap_writepage_ctx *wpc,
242 const struct iomap_writeback_ops *ops);
243
244 /*
245 * Flags for direct I/O ->end_io:
246 */
247 #define IOMAP_DIO_UNWRITTEN (1 << 0) /* covers unwritten extent(s) */
248 #define IOMAP_DIO_COW (1 << 1) /* covers COW extent(s) */
249
250 struct iomap_dio_ops {
251 int (*end_io)(struct kiocb *iocb, ssize_t size, int error,
252 unsigned flags);
253 blk_qc_t (*submit_io)(struct inode *inode, struct iomap *iomap,
254 struct bio *bio, loff_t file_offset);
255 };
256
257 /*
258 * Wait for the I/O to complete in iomap_dio_rw even if the kiocb is not
259 * synchronous.
260 */
261 #define IOMAP_DIO_FORCE_WAIT (1 << 0)
262
263 /*
264 * Do not allocate blocks or zero partial blocks, but instead fall back to
265 * the caller by returning -EAGAIN. Used to optimize direct I/O writes that
266 * are not aligned to the file system block size.
267 */
268 #define IOMAP_DIO_OVERWRITE_ONLY (1 << 1)
269
270 ssize_t iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
271 const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
272 unsigned int dio_flags);
273 struct iomap_dio *__iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
274 const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
275 unsigned int dio_flags);
276 ssize_t iomap_dio_complete(struct iomap_dio *dio);
277 int iomap_dio_iopoll(struct kiocb *kiocb, bool spin);
278
279 #ifdef CONFIG_SWAP
280 struct file;
281 struct swap_info_struct;
282
283 int iomap_swapfile_activate(struct swap_info_struct *sis,
284 struct file *swap_file, sector_t *pagespan,
285 const struct iomap_ops *ops);
286 #else
287 # define iomap_swapfile_activate(sis, swapfile, pagespan, ops) (-EIO)
288 #endif /* CONFIG_SWAP */
289
290 #endif /* LINUX_IOMAP_H */