1RECV(2) Linux Programmer's Manual RECV(2)
2
6 recv, recvfrom, recvmsg - receive a message from a socket
7
9 #include <sys/socket.h>
10 ssize_t recv(int sockfd, void *buf, size_t len, int flags);
12 ssize_t recvfrom(int sockfd, void *restrict buf, size_t len, int flags,
13 struct sockaddr *restrict src_addr,
14 socklen_t *restrict addrlen);
15 ssize_t recvmsg(int sockfd, struct msghdr *msg, int flags);
16
18 The recv(), recvfrom(), and recvmsg() calls are used to receive mes‐
19 sages from a socket. They may be used to receive data on both connec‐
20 tionless and connection-oriented sockets. This page first describes
21 common features of all three system calls, and then describes the dif‐
22 ferences between the calls.
23 The only difference between recv() and read(2) is the presence of
25 flags. With a zero flags argument, recv() is generally equivalent to
26 read(2) (but see NOTES). Also, the following call
27 recv(sockfd, buf, len, flags);
29 is equivalent to
31 recvfrom(sockfd, buf, len, flags, NULL, NULL);
33 All three calls return the length of the message on successful comple‐
35 tion. If a message is too long to fit in the supplied buffer, excess
36 bytes may be discarded depending on the type of socket the message is
37 received from.
38 If no messages are available at the socket, the receive calls wait for
40 a message to arrive, unless the socket is nonblocking (see fcntl(2)),
41 in which case the value -1 is returned and errno is set to EAGAIN or
42 EWOULDBLOCK. The receive calls normally return any data available, up
43 to the requested amount, rather than waiting for receipt of the full
44 amount requested.
45 An application can use select(2), poll(2), or epoll(7) to determine
47 when more data arrives on a socket.
48 The flags argument
50 The flags argument is formed by ORing one or more of the following val‐
51 ues:
52 MSG_CMSG_CLOEXEC (recvmsg() only; since Linux 2.6.23)
54 Set the close-on-exec flag for the file descriptor received via
55 a UNIX domain file descriptor using the SCM_RIGHTS operation
56 (described in unix(7)). This flag is useful for the same rea‐
57 sons as the O_CLOEXEC flag of open(2).
58 MSG_DONTWAIT (since Linux 2.2)
60 Enables nonblocking operation; if the operation would block, the
61 call fails with the error EAGAIN or EWOULDBLOCK. This provides
62 similar behavior to setting the O_NONBLOCK flag (via the fc‐
63 ntl(2) F_SETFL operation), but differs in that MSG_DONTWAIT is a
64 per-call option, whereas O_NONBLOCK is a setting on the open
65 file description (see open(2)), which will affect all threads in
66 the calling process and as well as other processes that hold
67 file descriptors referring to the same open file description.
68 MSG_ERRQUEUE (since Linux 2.2)
70 This flag specifies that queued errors should be received from
71 the socket error queue. The error is passed in an ancillary
72 message with a type dependent on the protocol (for IPv4
73 IP_RECVERR). The user should supply a buffer of sufficient
74 size. See cmsg(3) and ip(7) for more information. The payload
75 of the original packet that caused the error is passed as normal
76 data via msg_iovec. The original destination address of the
77 datagram that caused the error is supplied via msg_name.
78 The error is supplied in a sock_extended_err structure:
80 #define SO_EE_ORIGIN_NONE 0
82 #define SO_EE_ORIGIN_LOCAL 1
83 #define SO_EE_ORIGIN_ICMP 2
84 #define SO_EE_ORIGIN_ICMP6 3
85 struct sock_extended_err
87 {
88 uint32_t ee_errno; /* Error number */
89 uint8_t ee_origin; /* Where the error originated */
90 uint8_t ee_type; /* Type */
91 uint8_t ee_code; /* Code */
92 uint8_t ee_pad; /* Padding */
93 uint32_t ee_info; /* Additional information */
94 uint32_t ee_data; /* Other data */
95 /* More data may follow */
96 };
97 struct sockaddr *SO_EE_OFFENDER(struct sock_extended_err *);
99 ee_errno contains the errno number of the queued error. ee_ori‐
101 gin is the origin code of where the error originated. The other
102 fields are protocol-specific. The macro SO_EE_OFFENDER returns
103 a pointer to the address of the network object where the error
104 originated from given a pointer to the ancillary message. If
105 this address is not known, the sa_family member of the sockaddr
106 contains AF_UNSPEC and the other fields of the sockaddr are un‐
107 defined. The payload of the packet that caused the error is
108 passed as normal data.
109 For local errors, no address is passed (this can be checked with
111 the cmsg_len member of the cmsghdr). For error receives, the
112 MSG_ERRQUEUE flag is set in the msghdr. After an error has been
113 passed, the pending socket error is regenerated based on the
114 next queued error and will be passed on the next socket opera‐
115 tion.
116 MSG_OOB
118 This flag requests receipt of out-of-band data that would not be
119 received in the normal data stream. Some protocols place expe‐
120 dited data at the head of the normal data queue, and thus this
121 flag cannot be used with such protocols.
122 MSG_PEEK
124 This flag causes the receive operation to return data from the
125 beginning of the receive queue without removing that data from
126 the queue. Thus, a subsequent receive call will return the same
127 data.
128 MSG_TRUNC (since Linux 2.2)
130 For raw (AF_PACKET), Internet datagram (since Linux
131 2.4.27/2.6.8), netlink (since Linux 2.6.22), and UNIX datagram
132 (since Linux 3.4) sockets: return the real length of the packet
133 or datagram, even when it was longer than the passed buffer.
134 For use with Internet stream sockets, see tcp(7).
136 MSG_WAITALL (since Linux 2.2)
138 This flag requests that the operation block until the full re‐
139 quest is satisfied. However, the call may still return less
140 data than requested if a signal is caught, an error or discon‐
141 nect occurs, or the next data to be received is of a different
142 type than that returned. This flag has no effect for datagram
143 sockets.
144 recvfrom()
146 recvfrom() places the received message into the buffer buf. The caller
147 must specify the size of the buffer in len.
148 If src_addr is not NULL, and the underlying protocol provides the
150 source address of the message, that source address is placed in the
151 buffer pointed to by src_addr. In this case, addrlen is a value-result
152 argument. Before the call, it should be initialized to the size of the
153 buffer associated with src_addr. Upon return, addrlen is updated to
154 contain the actual size of the source address. The returned address is
155 truncated if the buffer provided is too small; in this case, addrlen
156 will return a value greater than was supplied to the call.
157 If the caller is not interested in the source address, src_addr and ad‐
159 drlen should be specified as NULL.
160 recv()
162 The recv() call is normally used only on a connected socket (see con‐
163 nect(2)). It is equivalent to the call:
164 recvfrom(fd, buf, len, flags, NULL, 0);
166 recvmsg()
168 The recvmsg() call uses a msghdr structure to minimize the number of
169 directly supplied arguments. This structure is defined as follows in
170 <sys/socket.h>:
171 struct iovec { /* Scatter/gather array items */
173 void *iov_base; /* Starting address */
174 size_t iov_len; /* Number of bytes to transfer */
175 };
176 struct msghdr {
178 void *msg_name; /* Optional address */
179 socklen_t msg_namelen; /* Size of address */
180 struct iovec *msg_iov; /* Scatter/gather array */
181 size_t msg_iovlen; /* # elements in msg_iov */
182 void *msg_control; /* Ancillary data, see below */
183 size_t msg_controllen; /* Ancillary data buffer len */
184 int msg_flags; /* Flags on received message */
185 };
186 The msg_name field points to a caller-allocated buffer that is used to
188 return the source address if the socket is unconnected. The caller
189 should set msg_namelen to the size of this buffer before this call;
190 upon return from a successful call, msg_namelen will contain the length
191 of the returned address. If the application does not need to know the
192 source address, msg_name can be specified as NULL.
193 The fields msg_iov and msg_iovlen describe scatter-gather locations, as
195 discussed in readv(2).
196 The field msg_control, which has length msg_controllen, points to a
198 buffer for other protocol control-related messages or miscellaneous an‐
199 cillary data. When recvmsg() is called, msg_controllen should contain
200 the length of the available buffer in msg_control; upon return from a
201 successful call it will contain the length of the control message se‐
202 quence.
203 The messages are of the form:
205 struct cmsghdr {
207 size_t cmsg_len; /* Data byte count, including header
208 (type is socklen_t in POSIX) */
209 int cmsg_level; /* Originating protocol */
210 int cmsg_type; /* Protocol-specific type */
211 /* followed by
212 unsigned char cmsg_data[]; */
213 };
214 Ancillary data should be accessed only by the macros defined in
216 cmsg(3).
217 As an example, Linux uses this ancillary data mechanism to pass ex‐
219 tended errors, IP options, or file descriptors over UNIX domain sock‐
220 ets. For further information on the use of ancillary data in various
221 socket domains, see unix(7) and ip(7).
222 The msg_flags field in the msghdr is set on return of recvmsg(). It
224 can contain several flags:
225 MSG_EOR
227 indicates end-of-record; the data returned completed a record
228 (generally used with sockets of type SOCK_SEQPACKET).
229 MSG_TRUNC
231 indicates that the trailing portion of a datagram was discarded
232 because the datagram was larger than the buffer supplied.
233 MSG_CTRUNC
235 indicates that some control data was discarded due to lack of
236 space in the buffer for ancillary data.
237 MSG_OOB
239 is returned to indicate that expedited or out-of-band data was
240 received.
241 MSG_ERRQUEUE
243 indicates that no data was received but an extended error from
244 the socket error queue.
245
247 These calls return the number of bytes received, or -1 if an error oc‐
248 curred. In the event of an error, errno is set to indicate the error.
249 When a stream socket peer has performed an orderly shutdown, the return
251 value will be 0 (the traditional "end-of-file" return).
252 Datagram sockets in various domains (e.g., the UNIX and Internet do‐
254 mains) permit zero-length datagrams. When such a datagram is received,
255 the return value is 0.
256 The value 0 may also be returned if the requested number of bytes to
258 receive from a stream socket was 0.
259
261 These are some standard errors generated by the socket layer. Addi‐
262 tional errors may be generated and returned from the underlying proto‐
263 col modules; see their manual pages.
264 EAGAIN or EWOULDBLOCK
266 The socket is marked nonblocking and the receive operation would
267 block, or a receive timeout had been set and the timeout expired
268 before data was received. POSIX.1 allows either error to be re‐
269 turned for this case, and does not require these constants to
270 have the same value, so a portable application should check for
271 both possibilities.
272 EBADF The argument sockfd is an invalid file descriptor.
274 ECONNREFUSED
276 A remote host refused to allow the network connection (typically
277 because it is not running the requested service).
278 EFAULT The receive buffer pointer(s) point outside the process's ad‐
280 dress space.
281 EINTR The receive was interrupted by delivery of a signal before any
283 data was available; see signal(7).
284 EINVAL Invalid argument passed.
286 ENOMEM Could not allocate memory for recvmsg().
288 ENOTCONN
290 The socket is associated with a connection-oriented protocol and
291 has not been connected (see connect(2) and accept(2)).
292 ENOTSOCK
294 The file descriptor sockfd does not refer to a socket.
295
297 POSIX.1-2001, POSIX.1-2008, 4.4BSD (these interfaces first appeared in
298 4.2BSD).
299 POSIX.1 describes only the MSG_OOB, MSG_PEEK, and MSG_WAITALL flags.
301
303 If a zero-length datagram is pending, read(2) and recv() with a flags
304 argument of zero provide different behavior. In this circumstance,
305 read(2) has no effect (the datagram remains pending), while recv() con‐
306 sumes the pending datagram.
307 The socklen_t type was invented by POSIX. See also accept(2).
309 According to POSIX.1, the msg_controllen field of the msghdr structure
311 should be typed as socklen_t, and the msg_iovlen field should be typed
312 as int, but glibc currently types both as size_t.
313 See recvmmsg(2) for information about a Linux-specific system call that
315 can be used to receive multiple datagrams in a single call.
316
318 An example of the use of recvfrom() is shown in getaddrinfo(3).
319
321 fcntl(2), getsockopt(2), read(2), recvmmsg(2), select(2), shutdown(2),
322 socket(2), cmsg(3), sockatmark(3), ip(7), ipv6(7), socket(7), tcp(7),
323 udp(7), unix(7)
324
326 This page is part of release 5.13 of the Linux man-pages project. A
327 description of the project, information about reporting bugs, and the
328 latest version of this page, can be found at
329 https://www.kernel.org/doc/man-pages/.
330Linux 2021-03-22 RECV(2)