17.7 getpeername System Call

Team-Fly     TCP/IP Illustrated, Volume 2: The Implementation By Gary R. Wright, W. Richard Stevens Table of Contents Chapter 17.  Socket Options 17.7 getpeername System Call The prototype for this system call is: int getpeername(int fd, caddr_t asa, int *alen); The getpeername system call returns the address of the remote end of the connection associated with the specified socket. This function is often called when a server is invoked through a fork and exec by the process that calls accept (i.e., any server started by inetd). The server doesn't have access to the peer address returned by accept and must use getpeername. The returned address is often checked against an access list for the application, and the connection is closed if the address is not on the list. Some protocols, such as TP4, utilize this function to determine if an incoming connection should be rejected or confirmed. In TP4, the connection associated with a socket returned by accept is not yet complete and must be confirmed before the connection completes. Based on the address returned by getpeername, the server can close the connection or implicitly confirm the connection by sending or receiving data. This feature is irrelevant for TCP, since TCP doesn't make a connection available to accept until the three-way handshake is complete. Figure 17.23 shows the getpeername function. Figure 17.23. getpeername system call. 719-753 The code here is almost identical to the getsockname code. getsock locates the socket and ENOTCONN is returned if the socket is not yet connected to a peer or if the connection is not in a confirmation state (e.g., TP4). If it is connected, the size of the buffer is copied in from the process and an mbuf is allocated to hold the address. The PRU_PEERADDR request is issued to get the remote address from the protocol layer. The address and the length of the address are copied from the kernel mbuf to the buffer in the process. The mbuf is released and the function returns. Team-Fly     Top