User`s guide

KwikNet TCP/IP Sockets

DAK

183

Non-Blocking Sockets

When operations are performed using a socket, the caller requesting the action is usually

forced to wait until the operation completes. The socket is said to be in blocking mode.

KwikNet offers an alternative mode of operation in which the socket action is allowed to

proceed without blocking the caller. In this case, the socket is said to be in non-blocking

mode.

When a socket is created it is placed in blocking mode. Thereafter, your application can

call the sockets procedure kn_setsockopt() to alter the operating mode of the socket.

KwikNet socket option SO_NONBLOCK can be used to place the socket in non-blocking

mode. The same option can be used to restore the socket to blocking mode. Since

SO_NONBLOCK is a unique KwikNet option, it is flagged as non-standard in the summary of

sockets options presented in Chapter 5.3.

KwikNet Error Codes

KwikNet socket procedures return a positive value or 0 if the call is successful or -1 if the

call fails. Procedures which return a socket descriptor return a non-negative value if the

call is successful.

Standard sockets procedures also return additional error information in the UNIX global

variable errno. This is the same variable which many standard C libraries use to save

error results. However, KwikNet must coexist with these libraries. Furthermore, since

KwikNet must operate in multitasking environments, it cannot use errno since the results

of operations by different tasks would be indeterminate.

For these reasons, global variable errno is not altered by KwikNet. Instead, the

completion status of each socket operation is recorded within the socket descriptor upon

completion of the operation. This error status is then made available to the caller through

the kn_errno() procedure. Be sure to use kn_errno() (or its Treck equivalent) instead

of the BSD errno() function.

Unfortunately, procedures kn_socket() and kn_select() have no socket descriptor in

which to save the error code. Hence, you cannot use kn_errno() to retrieve error

information after calls to these procedures.

Error codes returned by kn_errno() are integer values which are a subset of the standard

Berkeley error codes. These error codes are summarized in Appendix B.