sendfile64(2)

1sendfile(2)                   System Calls Manual                  sendfile(2)
2
3
4

NAME

6       sendfile - transfer data between file descriptors
7

LIBRARY

9       Standard C library (libc, -lc)
10

SYNOPSIS

12       #include <sys/sendfile.h>
13
14       ssize_t sendfile(int out_fd, int in_fd, off_t *_Nullable offset,
15                        size_t count);
16

DESCRIPTION

18       sendfile()  copies  data  between one file descriptor and another.  Be‐
19       cause this copying is done within the kernel, sendfile() is more  effi‐
20       cient than the combination of read(2) and write(2), which would require
21       transferring data to and from user space.
22
23       in_fd should be a file descriptor opened for reading and out_fd  should
24       be a descriptor opened for writing.
25
26       If  offset  is  not NULL, then it points to a variable holding the file
27       offset from which sendfile() will start reading data from in_fd.   When
28       sendfile() returns, this variable will be set to the offset of the byte
29       following the last byte that was read.  If offset  is  not  NULL,  then
30       sendfile() does not modify the file offset of in_fd; otherwise the file
31       offset is adjusted to reflect the number of bytes read from in_fd.
32
33       If offset is NULL, then data will be read from in_fd  starting  at  the
34       file offset, and the file offset will be updated by the call.
35
36       count is the number of bytes to copy between the file descriptors.
37
38       The   in_fd   argument   must  correspond  to  a  file  which  supports
39       mmap(2)-like operations (i.e., it cannot be a socket).
40
41       Before Linux 2.6.33, out_fd must refer to a socket.  Since Linux 2.6.33
42       it  can  be any file.  If it is a regular file, then sendfile() changes
43       the file offset appropriately.
44

RETURN VALUE

46       If the transfer was successful, the number of bytes written  to  out_fd
47       is returned.  Note that a successful call to sendfile() may write fewer
48       bytes than requested; the caller should be prepared to retry  the  call
49       if there were unsent bytes.  See also NOTES.
50
51       On error, -1 is returned, and errno is set to indicate the error.
52

ERRORS

54       EAGAIN Nonblocking I/O has been selected using O_NONBLOCK and the write
55              would block.
56
57       EBADF  The input file was not opened for reading or the output file was
58              not opened for writing.
59
60       EFAULT Bad address.
61
62       EINVAL Descriptor  is not valid or locked, or an mmap(2)-like operation
63              is not available for in_fd, or count is negative.
64
65       EINVAL out_fd has the O_APPEND flag set.  This is  not  currently  sup‐
66              ported by sendfile().
67
68       EIO    Unspecified error while reading from in_fd.
69
70       ENOMEM Insufficient memory to read from in_fd.
71
72       EOVERFLOW
73              count  is too large, the operation would result in exceeding the
74              maximum size of either the input file or the output file.
75
76       ESPIPE offset is not NULL but the input file is not seekable.
77

VERSIONS

79       Other UNIX systems implement sendfile() with  different  semantics  and
80       prototypes.  It should not be used in portable programs.
81

STANDARDS

83       None.
84

HISTORY

86       Linux 2.2, glibc 2.1.
87
88       In  Linux  2.4  and earlier, out_fd could also refer to a regular file;
89       this possibility went away in the Linux 2.6.x kernel  series,  but  was
90       restored in Linux 2.6.33.
91
92       The  original  Linux  sendfile() system call was not designed to handle
93       large file offsets.  Consequently, Linux 2.4 added sendfile64(), with a
94       wider type for the offset argument.  The glibc sendfile() wrapper func‐
95       tion transparently deals with the kernel differences.
96

NOTES

98       sendfile() will transfer at most 0x7ffff000 (2,147,479,552) bytes,  re‐
99       turning  the  number  of  bytes actually transferred.  (This is true on
100       both 32-bit and 64-bit systems.)
101
102       If you plan to use sendfile() for sending files to a  TCP  socket,  but
103       need  to  send some header data in front of the file contents, you will
104       find it useful to employ the TCP_CORK option, described in  tcp(7),  to
105       minimize the number of packets and to tune performance.
106
107       Applications  may wish to fall back to read(2) and write(2) in the case
108       where sendfile() fails with EINVAL or ENOSYS.
109
110       If out_fd refers to a socket or pipe with  zero-copy  support,  callers
111       must  ensure  the transferred portions of the file referred to by in_fd
112       remain unmodified until the reader on the other end of out_fd has  con‐
113       sumed the transferred data.
114
115       The  Linux-specific  splice(2)  call supports transferring data between
116       arbitrary file descriptors provided one (or both) of them is a pipe.
117