1RECVMMSG(2) Linux Programmer's Manual RECVMMSG(2)
2
3
4
6 recvmmsg - receive multiple messages on a socket
7
9 #define _GNU_SOURCE /* See feature_test_macros(7) */
10 #include <sys/socket.h>
11
12 int recvmmsg(int sockfd, struct mmsghdr *msgvec, unsigned int vlen,
13 int flags, struct timespec *timeout);
14
16 The recvmmsg() system call is an extension of recvmsg(2) that allows
17 the caller to receive multiple messages from a socket using a single
18 system call. (This has performance benefits for some applications.) A
19 further extension over recvmsg(2) is support for a timeout on the
20 receive operation.
21
22 The sockfd argument is the file descriptor of the socket to receive
23 data from.
24
25 The msgvec argument is a pointer to an array of mmsghdr structures.
26 The size of this array is specified in vlen.
27
28 The mmsghdr structure is defined in <sys/socket.h> as:
29
30 struct mmsghdr {
31 struct msghdr msg_hdr; /* Message header */
32 unsigned int msg_len; /* Number of received bytes for header */
33 };
34
35 The msg_hdr field is a msghdr structure, as described in recvmsg(2).
36 The msg_len field is the number of bytes returned for the message in
37 the entry. This field has the same value as the return value of a sin‐
38 gle recvmsg(2) on the header.
39
40 The flags argument contains flags ORed together. The flags are the
41 same as documented for recvmsg(2), with the following addition:
42
43 MSG_WAITFORONE (since Linux 2.6.34)
44 Turns on MSG_DONTWAIT after the first message has been received.
45
46 The timeout argument points to a struct timespec (see clock_gettime(2))
47 defining a timeout (seconds plus nanoseconds) for the receive operation
48 (but see BUGS!). (This interval will be rounded up to the system clock
49 granularity, and kernel scheduling delays mean that the blocking inter‐
50 val may overrun by a small amount.) If timeout is NULL, then the oper‐
51 ation blocks indefinitely.
52
53 A blocking recvmmsg() call blocks until vlen messages have been
54 received or until the timeout expires. A nonblocking call reads as
55 many messages as are available (up to the limit specified by vlen) and
56 returns immediately.
57
58 On return from recvmmsg(), successive elements of msgvec are updated to
59 contain information about each received message: msg_len contains the
60 size of the received message; the subfields of msg_hdr are updated as
61 described in recvmsg(2). The return value of the call indicates the
62 number of elements of msgvec that have been updated.
63
65 On success, recvmmsg() returns the number of messages received in
66 msgvec; on error, -1 is returned, and errno is set to indicate the
67 error.
68
70 Errors are as for recvmsg(2). In addition, the following error can
71 occur:
72
73 EINVAL timeout is invalid.
74
75 See also BUGS.
76
78 The recvmmsg() system call was added in Linux 2.6.33. Support in glibc
79 was added in version 2.12.
80
82 recvmmsg() is Linux-specific.
83
85 The timeout argument does not work as intended. The timeout is checked
86 only after the receipt of each datagram, so that if up to vlen-1 data‐
87 grams are received before the timeout expires, but then no further
88 datagrams are received, the call will block forever.
89
90 If an error occurs after at least one message has been received, the
91 call succeeds, and returns the number of messages received. The error
92 code is expected to be returned on a subsequent call to recvmmsg(). In
93 the current implementation, however, the error code can be overwritten
94 in the meantime by an unrelated network event on a socket, for example
95 an incoming ICMP packet.
96
98 The following program uses recvmmsg() to receive multiple messages on a
99 socket and stores them in multiple buffers. The call returns if all
100 buffers are filled or if the timeout specified has expired.
101
102 The following snippet periodically generates UDP datagrams containing a
103 random number:
104
105 $ while true; do echo $RANDOM > /dev/udp/127.0.0.1/1234;
106 sleep 0.25; done
107
108 These datagrams are read by the example application, which can give the
109 following output:
110
111 $ ./a.out
112 5 messages received
113 1 11782
114 2 11345
115 3 304
116 4 13514
117 5 28421
118
119 Program source
120
121 #define _GNU_SOURCE
122 #include <netinet/ip.h>
123 #include <stdio.h>
124 #include <stdlib.h>
125 #include <string.h>
126 #include <sys/socket.h>
127
128 int
129 main(void)
130 {
131 #define VLEN 10
132 #define BUFSIZE 200
133 #define TIMEOUT 1
134 int sockfd, retval, i;
135 struct sockaddr_in addr;
136 struct mmsghdr msgs[VLEN];
137 struct iovec iovecs[VLEN];
138 char bufs[VLEN][BUFSIZE+1];
139 struct timespec timeout;
140
141 sockfd = socket(AF_INET, SOCK_DGRAM, 0);
142 if (sockfd == -1) {
143 perror("socket()");
144 exit(EXIT_FAILURE);
145 }
146
147 addr.sin_family = AF_INET;
148 addr.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
149 addr.sin_port = htons(1234);
150 if (bind(sockfd, (struct sockaddr *) &addr, sizeof(addr)) == -1) {
151 perror("bind()");
152 exit(EXIT_FAILURE);
153 }
154
155 memset(msgs, 0, sizeof(msgs));
156 for (i = 0; i < VLEN; i++) {
157 iovecs[i].iov_base = bufs[i];
158 iovecs[i].iov_len = BUFSIZE;
159 msgs[i].msg_hdr.msg_iov = &iovecs[i];
160 msgs[i].msg_hdr.msg_iovlen = 1;
161 }
162
163 timeout.tv_sec = TIMEOUT;
164 timeout.tv_nsec = 0;
165
166 retval = recvmmsg(sockfd, msgs, VLEN, 0, &timeout);
167 if (retval == -1) {
168 perror("recvmmsg()");
169 exit(EXIT_FAILURE);
170 }
171
172 printf("%d messages received\n", retval);
173 for (i = 0; i < retval; i++) {
174 bufs[i][msgs[i].msg_len] = 0;
175 printf("%d %s", i+1, bufs[i]);
176 }
177 exit(EXIT_SUCCESS);
178 }
179
181 clock_gettime(2), recvmsg(2), sendmmsg(2), sendmsg(2), socket(2),
182 socket(7)
183
185 This page is part of release 5.02 of the Linux man-pages project. A
186 description of the project, information about reporting bugs, and the
187 latest version of this page, can be found at
188 https://www.kernel.org/doc/man-pages/.
189
190
191
192Linux 2019-03-06 RECVMMSG(2)