Debian Codegraph @

upstream/debian-unstable/man2/sync_file_range.2

raw
1 .\" Copyright 2006, Michael Kerrisk <[email protected]>
2 .\" Copyright, the authors of the Linux man-pages project
3 .\"
4 .\" SPDX-License-Identifier: Linux-man-pages-copyleft
5 .\"
6 .TH sync_file_range 2 2026-02-08 "Linux man-pages 6.18"
7 .SH NAME
8 sync_file_range \- sync a file segment with disk
9 .SH LIBRARY
10 Standard C library
11 .RI ( libc ,\~ \-lc )
12 .SH SYNOPSIS
13 .nf
14 .BR "#define _GNU_SOURCE" " /* See feature_test_macros(7) */"
15 .B #define _FILE_OFFSET_BITS 64
16 .B #include <fcntl.h>
17 .P
18 .BI "int sync_file_range(int " fd ", off_t " offset ", off_t " nbytes ,
19 .BI " unsigned int " flags );
20 .fi
21 .SH DESCRIPTION
22 .BR sync_file_range ()
23 permits fine control when synchronizing the open file referred to by the
24 file descriptor
25 .I fd
26 with disk.
27 .P
28 .I offset
29 is the starting byte of the file range to be synchronized.
30 .I nbytes
31 specifies the length of the range to be synchronized, in bytes;
32 if
33 .I nbytes
34 is zero, then all bytes from
35 .I offset
36 through to the end of file are synchronized.
37 Synchronization is in units of the system page size:
38 .I offset
39 is rounded down to a page boundary;
40 .I (offset+nbytes\-1)
41 is rounded up to a page boundary.
42 .P
43 The
44 .I flags
45 bit-mask argument can include any of the following values:
46 .TP
47 .B SYNC_FILE_RANGE_WAIT_BEFORE
48 Wait upon write-out of all pages in the specified range
49 that have already been submitted to the device driver for write-out
50 before performing any write.
51 .TP
52 .B SYNC_FILE_RANGE_WRITE
53 Initiate write-out of all dirty pages in the specified
54 range which are not presently submitted write-out.
55 Note that even this may block if you attempt to
56 write more than request queue size.
57 .TP
58 .B SYNC_FILE_RANGE_WAIT_AFTER
59 Wait upon write-out of all pages in the range
60 after performing any write.
61 .P
62 Specifying
63 .I flags
64 as 0 is permitted, as a no-op.
65 .SS Warning
66 This system call is extremely dangerous and should not be used in portable
67 programs.
68 None of these operations writes out the file's metadata.
69 Therefore, unless the application is strictly performing overwrites of
70 already-instantiated disk blocks, there are no guarantees that the data will
71 be available after a crash.
72 There is no user interface to know if a write is purely an overwrite.
73 On filesystems using copy-on-write semantics (e.g.,
74 .IR btrfs )
75 an overwrite of existing allocated blocks is impossible.
76 When writing into preallocated space,
77 many filesystems also require calls into the block
78 allocator, which this system call does not sync out to disk.
79 This system call does not flush disk write caches and thus does not provide
80 any data integrity on systems with volatile disk write caches.
81 .SS Some details
82 .B SYNC_FILE_RANGE_WAIT_BEFORE
83 and
84 .B SYNC_FILE_RANGE_WAIT_AFTER
85 will detect any
86 I/O errors or
87 .B ENOSPC
88 conditions and will return these to the caller.
89 .P
90 Useful combinations of the
91 .I flags
92 bits are:
93 .TP
94 .B SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE
95 Ensures that all pages
96 in the specified range which were dirty when
97 .BR sync_file_range ()
98 was called are placed
99 under write-out.
100 This is a start-write-for-data-integrity operation.
101 .TP
102 .B SYNC_FILE_RANGE_WRITE
103 Start write-out of all dirty pages in the specified range which
104 are not presently under write-out.
105 This is an asynchronous flush-to-disk
106 operation.
107 This is not suitable for data integrity operations.
108 .TP
109 .BR SYNC_FILE_RANGE_WAIT_BEFORE " (or " SYNC_FILE_RANGE_WAIT_AFTER )
110 Wait for
111 completion of write-out of all pages in the specified range.
112 This can be used after an earlier
113 .B SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE
114 operation to wait for completion of that operation, and obtain its result.
115 .TP
116 .B SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE | \
117 SYNC_FILE_RANGE_WAIT_AFTER
118 This is a write-for-data-integrity operation
119 that will ensure that all pages in the specified range which were dirty when
120 .BR sync_file_range ()
121 was called are committed to disk.
122 .SH RETURN VALUE
123 On success,
124 .BR sync_file_range ()
125 returns 0;
126 on failure \-1 is returned and
127 .I errno
128 is set to indicate the error.
129 .SH ERRORS
130 .TP
131 .B EBADF
132 .I fd
133 is not a valid file descriptor.
134 .TP
135 .B EINVAL
136 .I flags
137 specifies an invalid bit;
138 or
139 .I offset
140 or
141 .I nbytes
142 is invalid.
143 .TP
144 .B EIO
145 I/O error.
146 .TP
147 .B ENOMEM
148 Out of memory.
149 .TP
150 .B ENOSPC
151 Out of disk space.
152 .TP
153 .B ESPIPE
154 .I fd
155 refers to something other than a regular file, a block device, or
156 a directory.
157 .SH VERSIONS
158 .SS sync_file_range2()
159 Some architectures (e.g., PowerPC, ARM)
160 need 64-bit arguments to be aligned in a suitable pair of registers.
161 .\" See kernel commit edd5cd4a9424f22b0fa08bef5e299d41befd5622
162 On such architectures, the call signature of
163 .BR sync_file_range ()
164 shown in the SYNOPSIS would force
165 a register to be wasted as padding between the
166 .I fd
167 and
168 .I offset
169 arguments.
170 (See
171 .BR syscall (2)
172 for details.)
173 Therefore, these architectures define a different
174 system call that orders the arguments suitably:
175 .P
176 .in +4n
177 .EX
178 .BI "int sync_file_range2(int " fd ", unsigned int " flags ,
179 .BI " off_t " offset ", off_t " nbytes );
180 .EE
181 .in
182 .P
183 The behavior of this system call is otherwise exactly the same as
184 .BR sync_file_range ().
185 .SH STANDARDS
186 Linux.
187 .SH HISTORY
188 Linux 2.6.17.
189 .SS sync_file_range2()
190 A system call with this signature first appeared on the ARM architecture
191 in Linux 2.6.20, with the name
192 .BR arm_sync_file_range ().
193 It was renamed in Linux 2.6.22,
194 when the analogous system call was added for PowerPC.
195 On architectures where glibc support is provided,
196 glibc transparently wraps
197 .BR sync_file_range2 ()
198 under the name
199 .BR sync_file_range ().
200 .SH NOTES
201 .B _FILE_OFFSET_BITS
202 should be defined to be 64 in code that takes the address of
203 .BR sync_file_range ,
204 if the code is intended to be portable
205 to traditional 32-bit x86 and ARM platforms where
206 .BR off_t 's
207 width defaults to 32 bits.
208 .SH SEE ALSO
209 .BR fdatasync (2),
210 .BR fsync (2),
211 .BR msync (2),
212 .BR sync (2)