1accept(2) System Calls Manual accept(2)
2
6 accept, accept4 - accept a connection on a socket
7
9 Standard C library (libc, -lc)
10
12 #include <sys/socket.h>
13 int accept(int sockfd, struct sockaddr *_Nullable restrict addr,
15 socklen_t *_Nullable restrict addrlen);
16 #define _GNU_SOURCE /* See feature_test_macros(7) */
18 #include <sys/socket.h>
19 int accept4(int sockfd, struct sockaddr *_Nullable restrict addr,
21 socklen_t *_Nullable restrict addrlen, int flags);
22
24 The accept() system call is used with connection-based socket types
25 (SOCK_STREAM, SOCK_SEQPACKET). It extracts the first connection re‐
26 quest on the queue of pending connections for the listening socket,
27 sockfd, creates a new connected socket, and returns a new file descrip‐
28 tor referring to that socket. The newly created socket is not in the
29 listening state. The original socket sockfd is unaffected by this
30 call.
31 The argument sockfd is a socket that has been created with socket(2),
33 bound to a local address with bind(2), and is listening for connections
34 after a listen(2).
35 The argument addr is a pointer to a sockaddr structure. This structure
37 is filled in with the address of the peer socket, as known to the com‐
38 munications layer. The exact format of the address returned addr is
39 determined by the socket's address family (see socket(2) and the re‐
40 spective protocol man pages). When addr is NULL, nothing is filled in;
41 in this case, addrlen is not used, and should also be NULL.
42 The addrlen argument is a value-result argument: the caller must ini‐
44 tialize it to contain the size (in bytes) of the structure pointed to
45 by addr; on return it will contain the actual size of the peer address.
46 The returned address is truncated if the buffer provided is too small;
48 in this case, addrlen will return a value greater than was supplied to
49 the call.
50 If no pending connections are present on the queue, and the socket is
52 not marked as nonblocking, accept() blocks the caller until a connec‐
53 tion is present. If the socket is marked nonblocking and no pending
54 connections are present on the queue, accept() fails with the error EA‐
55 GAIN or EWOULDBLOCK.
56 In order to be notified of incoming connections on a socket, you can
58 use select(2), poll(2), or epoll(7). A readable event will be deliv‐
59 ered when a new connection is attempted and you may then call accept()
60 to get a socket for that connection. Alternatively, you can set the
61 socket to deliver SIGIO when activity occurs on a socket; see socket(7)
62 for details.
63 If flags is 0, then accept4() is the same as accept(). The following
65 values can be bitwise ORed in flags to obtain different behavior:
66 SOCK_NONBLOCK Set the O_NONBLOCK file status flag on the open file
68 description (see open(2)) referred to by the new file
69 descriptor. Using this flag saves extra calls to fc‐
70 ntl(2) to achieve the same result.
71 SOCK_CLOEXEC Set the close-on-exec (FD_CLOEXEC) flag on the new file
73 descriptor. See the description of the O_CLOEXEC flag
74 in open(2) for reasons why this may be useful.
75
77 On success, these system calls return a file descriptor for the ac‐
78 cepted socket (a nonnegative integer). On error, -1 is returned, errno
79 is set to indicate the error, and addrlen is left unchanged.
80 Error handling
82 Linux accept() (and accept4()) passes already-pending network errors on
83 the new socket as an error code from accept(). This behavior differs
84 from other BSD socket implementations. For reliable operation the ap‐
85 plication should detect the network errors defined for the protocol af‐
86 ter accept() and treat them like EAGAIN by retrying. In the case of
87 TCP/IP, these are ENETDOWN, EPROTO, ENOPROTOOPT, EHOSTDOWN, ENONET,
88 EHOSTUNREACH, EOPNOTSUPP, and ENETUNREACH.
89
91 EAGAIN or EWOULDBLOCK
92 The socket is marked nonblocking and no connections are present
93 to be accepted. POSIX.1-2001 and POSIX.1-2008 allow either er‐
94 ror to be returned for this case, and do not require these con‐
95 stants to have the same value, so a portable application should
96 check for both possibilities.
97 EBADF sockfd is not an open file descriptor.
99 ECONNABORTED
101 A connection has been aborted.
102 EFAULT The addr argument is not in a writable part of the user address
104 space.
105 EINTR The system call was interrupted by a signal that was caught be‐
107 fore a valid connection arrived; see signal(7).
108 EINVAL Socket is not listening for connections, or addrlen is invalid
110 (e.g., is negative).
111 EINVAL (accept4()) invalid value in flags.
113 EMFILE The per-process limit on the number of open file descriptors has
115 been reached.
116 ENFILE The system-wide limit on the total number of open files has been
118 reached.
119 ENOBUFS, ENOMEM
121 Not enough free memory. This often means that the memory allo‐
122 cation is limited by the socket buffer limits, not by the system
123 memory.
124 ENOTSOCK
126 The file descriptor sockfd does not refer to a socket.
127 EOPNOTSUPP
129 The referenced socket is not of type SOCK_STREAM.
130 EPERM Firewall rules forbid connection.
132 EPROTO Protocol error.
134 In addition, network errors for the new socket and as defined for the
136 protocol may be returned. Various Linux kernels can return other er‐
137 rors such as ENOSR, ESOCKTNOSUPPORT, EPROTONOSUPPORT, ETIMEDOUT. The
138 value ERESTARTSYS may be seen during a trace.
139
141 On Linux, the new socket returned by accept() does not inherit file
142 status flags such as O_NONBLOCK and O_ASYNC from the listening socket.
143 This behavior differs from the canonical BSD sockets implementation.
144 Portable programs should not rely on inheritance or noninheritance of
145 file status flags and always explicitly set all required flags on the
146 socket returned from accept().
147
149 accept()
150 POSIX.1-2008.
151 accept4()
153 Linux.
154
156 accept()
157 POSIX.1-2001, SVr4, 4.4BSD (accept() first appeared in 4.2BSD).
158 accept4()
160 Linux 2.6.28, glibc 2.10.
161
163 There may not always be a connection waiting after a SIGIO is delivered
164 or select(2), poll(2), or epoll(7) return a readability event because
165 the connection might have been removed by an asynchronous network error
166 or another thread before accept() is called. If this happens, then the
167 call will block waiting for the next connection to arrive. To ensure
168 that accept() never blocks, the passed socket sockfd needs to have the
169 O_NONBLOCK flag set (see socket(7)).
170 For certain protocols which require an explicit confirmation, such as
172 DECnet, accept() can be thought of as merely dequeuing the next connec‐
173 tion request and not implying confirmation. Confirmation can be im‐
174 plied by a normal read or write on the new file descriptor, and rejec‐
175 tion can be implied by closing the new socket. Currently, only DECnet
176 has these semantics on Linux.
177 The socklen_t type
179 In the original BSD sockets implementation (and on other older systems)
180 the third argument of accept() was declared as an int *. A POSIX.1g
181 draft standard wanted to change it into a size_t *C; later POSIX stan‐
182 dards and glibc 2.x have socklen_t * .
183
185 See bind(2).
186
188 bind(2), connect(2), listen(2), select(2), socket(2), socket(7)
189Linux man-pages 6.04 2023-03-30 accept(2)