1ACCEPT(2) Linux Programmer's Manual ACCEPT(2)
2
6 accept - accept a connection on a socket
7
9 #include <sys/types.h> /* See NOTES */
10 #include <sys/socket.h>
11 int accept(int sockfd, struct sockaddr *addr, socklen_t *addrlen);
13 #define _GNU_SOURCE
15 #include <sys/socket.h>
16 int accept4(int sockfd, struct sockaddr *addr,
18 socklen_t *addrlen, int flags);
19
21 The accept() system call is used with connection-based socket types
22 (SOCK_STREAM, SOCK_SEQPACKET). It extracts the first connection
23 request on the queue of pending connections for the listening socket,
24 sockfd, creates a new connected socket, and returns a new file descrip‐
25 tor referring to that socket. The newly created socket is not in the
26 listening state. The original socket sockfd is unaffected by this
27 call.
28 The argument sockfd is a socket that has been created with socket(2),
30 bound to a local address with bind(2), and is listening for connections
31 after a listen(2).
32 The argument addr is a pointer to a sockaddr structure. This structure
34 is filled in with the address of the peer socket, as known to the com‐
35 munications layer. The exact format of the address returned addr is
36 determined by the socket's address family (see socket(2) and the
37 respective protocol man pages). When addr is NULL, nothing is filled
38 in; in this case, addrlen is not used, and should also be NULL.
39 The addrlen argument is a value-result argument: the caller must ini‐
41 tialize it to contain the size (in bytes) of the structure pointed to
42 by addr; on return it will contain the actual size of the peer address.
43 The returned address is truncated if the buffer provided is too small;
45 in this case, addrlen will return a value greater than was supplied to
46 the call.
47 If no pending connections are present on the queue, and the socket is
49 not marked as nonblocking, accept() blocks the caller until a connec‐
50 tion is present. If the socket is marked nonblocking and no pending
51 connections are present on the queue, accept() fails with the error
52 EAGAIN or EWOULDBLOCK.
53 In order to be notified of incoming connections on a socket, you can
55 use select(2) or poll(2). A readable event will be delivered when a
56 new connection is attempted and you may then call accept() to get a
57 socket for that connection. Alternatively, you can set the socket to
58 deliver SIGIO when activity occurs on a socket; see socket(7) for
59 details.
60 For certain protocols which require an explicit confirmation, such as
62 DECNet, accept() can be thought of as merely dequeuing the next connec‐
63 tion request and not implying confirmation. Confirmation can be
64 implied by a normal read or write on the new file descriptor, and
65 rejection can be implied by closing the new socket. Currently only
66 DECNet has these semantics on Linux.
67 If flags is 0, then accept4() is the same as accept(). The following
69 values can be bitwise ORed in flags to obtain different behavior:
70 SOCK_NONBLOCK Set the O_NONBLOCK file status flag on the new open
72 file description. Using this flag saves extra calls to
73 fcntl(2) to achieve the same result.
74 SOCK_CLOEXEC Set the close-on-exec (FD_CLOEXEC) flag on the new file
76 descriptor. See the description of the O_CLOEXEC flag
77 in open(2) for reasons why this may be useful.
78
80 On success, these system calls return a nonnegative integer that is a
81 descriptor for the accepted socket. On error, -1 is returned, and
82 errno is set appropriately.
83 Error Handling
85 Linux accept() (and accept4()) passes already-pending network errors on
86 the new socket as an error code from accept(). This behavior differs
87 from other BSD socket implementations. For reliable operation the
88 application should detect the network errors defined for the protocol
89 after accept() and treat them like EAGAIN by retrying. In case of
90 TCP/IP these are ENETDOWN, EPROTO, ENOPROTOOPT, EHOSTDOWN, ENONET,
91 EHOSTUNREACH, EOPNOTSUPP, and ENETUNREACH.
92
94 EAGAIN or EWOULDBLOCK
95 The socket is marked nonblocking and no connections are present
96 to be accepted. POSIX.1-2001 allows either error to be returned
97 for this case, and does not require these constants to have the
98 same value, so a portable application should check for both pos‐
99 sibilities.
100 EBADF The descriptor is invalid.
102 ECONNABORTED
104 A connection has been aborted.
105 EFAULT The addr argument is not in a writable part of the user address
107 space.
108 EINTR The system call was interrupted by a signal that was caught
110 before a valid connection arrived; see signal(7).
111 EINVAL Socket is not listening for connections, or addrlen is invalid
113 (e.g., is negative).
114 EINVAL (accept4()) invalid value in flags.
116 EMFILE The per-process limit of open file descriptors has been reached.
118 ENFILE The system limit on the total number of open files has been
120 reached.
121 ENOBUFS, ENOMEM
123 Not enough free memory. This often means that the memory allo‐
124 cation is limited by the socket buffer limits, not by the system
125 memory.
126 ENOTSOCK
128 The descriptor references a file, not a socket.
129 EOPNOTSUPP
131 The referenced socket is not of type SOCK_STREAM.
132 EPROTO Protocol error.
134 In addition, Linux accept() may fail if:
136 EPERM Firewall rules forbid connection.
138 In addition, network errors for the new socket and as defined for the
140 protocol may be returned. Various Linux kernels can return other
141 errors such as ENOSR, ESOCKTNOSUPPORT, EPROTONOSUPPORT, ETIMEDOUT. The
142 value ERESTARTSYS may be seen during a trace.
143
145 The accept4() system call is available starting with Linux 2.6.28; sup‐
146 port in glibc is available starting with version 2.10.
147
149 accept(): POSIX.1-2001, SVr4, 4.4BSD, (accept() first appeared in
150 4.2BSD).
151 accept4() is a nonstandard Linux extension.
153 On Linux, the new socket returned by accept() does not inherit file
155 status flags such as O_NONBLOCK and O_ASYNC from the listening socket.
156 This behavior differs from the canonical BSD sockets implementation.
157 Portable programs should not rely on inheritance or noninheritance of
158 file status flags and always explicitly set all required flags on the
159 socket returned from accept().
160
162 POSIX.1-2001 does not require the inclusion of <sys/types.h>, and this
163 header file is not required on Linux. However, some historical (BSD)
164 implementations required this header file, and portable applications
165 are probably wise to include it.
166 There may not always be a connection waiting after a SIGIO is delivered
168 or select(2) or poll(2) return a readability event because the connec‐
169 tion might have been removed by an asynchronous network error or
170 another thread before accept() is called. If this happens then the
171 call will block waiting for the next connection to arrive. To ensure
172 that accept() never blocks, the passed socket sockfd needs to have the
173 O_NONBLOCK flag set (see socket(7)).
174 The socklen_t type
176 The third argument of accept() was originally declared as an int * (and
177 is that under libc4 and libc5 and on many other systems like 4.x BSD,
178 SunOS 4, SGI); a POSIX.1g draft standard wanted to change it into a
179 size_t *, and that is what it is for SunOS 5. Later POSIX drafts have
180 socklen_t *, and so do the Single Unix Specification and glibc2. Quot‐
181 ing Linus Torvalds:
182 "_Any_ sane library _must_ have "socklen_t" be the same size as int.
184 Anything else breaks any BSD socket layer stuff. POSIX initially did
185 make it a size_t, and I (and hopefully others, but obviously not too
186 many) complained to them very loudly indeed. Making it a size_t is
187 completely broken, exactly because size_t very seldom is the same size
188 as "int" on 64-bit architectures, for example. And it has to be the
189 same size as "int" because that's what the BSD socket interface is.
190 Anyway, the POSIX people eventually got a clue, and created
191 "socklen_t". They shouldn't have touched it in the first place, but
192 once they did they felt it had to have a named type for some unfath‐
193 omable reason (probably somebody didn't like losing face over having
194 done the original stupid thing, so they silently just renamed their
195 blunder)."
196
198 See bind(2).
199
201 bind(2), connect(2), listen(2), select(2), socket(2), socket(7)
202
204 This page is part of release 3.25 of the Linux man-pages project. A
205 description of the project, and information about reporting bugs, can
206 be found at http://www.kernel.org/doc/man-pages/.
207Linux 2009-02-23 ACCEPT(2)