The Single UNIX ® Specification, Version 2
Copyright © 1997 The Open Group

NAME

t_bind - bind an address to a transport endpoint

SYNOPSIS



#include <xti.h>

int t_bind(
    int fd,
    const struct t_bind *req,
    struct t_bind *ret)

DESCRIPTION

Parameters Before call After call
fd x /
req-> addr.maxlen =
req-> addr.len x0
req-> addr.buf x (x)
req-> qlen x0
ret-> addr.maxlen x
ret-> addr.len /
ret-> addr.buf ?
ret-> qlen /

ret->addr.buf (the pointer itself, not the buffer it points to) is also unchanged.

This function associates a protocol address with the transport endpoint specified by fd and activates that transport endpoint. In connection mode, the transport provider may begin enqueuing incoming connect indications, or servicing a connection request on the transport endpoint. In connectionless-mode, the transport user may send or receive data units through the transport endpoint.

The req and ret arguments point to a t_bind structure containing the following members:


struct netbuf addr;
unsigned qlen;

The addr field of the t_bind structure specifies a protocol address, and the qlen field is used to indicate the maximum number of outstanding connection indications.

The parameter req is used to request that an address, represented by the netbuf structure, be bound to the given transport endpoint. The parameter len specifies the number of bytes in the address, and buf points to the address buffer. The parameter maxlen has no meaning for the req argument. On return, ret contains an encoding for the address that the transport provider actually bound to the transport endpoint; if an address was specified in req, this will be an encoding of the same address. In ret, the user specifies maxlen, which is the maximum size of the address buffer, and buf which points to the buffer where the address is to be placed. On return, len specifies the number of bytes in the bound address, and buf points to the bound address. If maxlen equals zero, no address is returned. If maxlen is greater than zero and less than the length of the address, t_bind() fails with t_errno set to [TBUFOVFLW].

If the requested address is not available, t_bind() will return -1 with t_errno set as appropriate. If no address is specified in req (the len field of addr in req is zero or req is NULL), the transport provider will assign an appropriate address to be bound, and will return that address in the addr field of ret. If the transport provider could not allocate an address, t_bind() will fail with t_errno set to [TNOADDR].

The parameter req may be a null pointer if the user does not wish to specify an address to be bound. Here, the value of qlen is assumed to be zero, and the transport provider will assign an address to the transport endpoint. Similarly, ret may be a null pointer if the user does not care what address was bound by the provider and is not interested in the negotiated value of qlen. It is valid to set req and ret to the null pointer for the same call, in which case the provider chooses the address to bind to the transport endpoint and does not return that information to the user.

The qlen field has meaning only when initialising a connection-mode service. It specifies the number of outstanding connection indications that the transport provider should support for the given transport endpoint. An outstanding connection indication is one that has been passed to the transport user by the transport provider but which has not been accepted or rejected. A value of qlen greater than zero is only meaningful when issued by a passive transport user that expects other users to call it. The value of qlen will be negotiated by the transport provider and may be changed if the transport provider cannot support the specified number of outstanding connection indications. However, this value of qlen will never be negotiated from a requested value greater than zero to zero. This is a requirement on transport providers; see CAVEATS below. On return, the qlen field in ret will contain the negotiated value.

If fd refers to a connection-mode service, this function allows more than one transport endpoint to be bound to the same protocol address (however, the transport provider must also support this capability), but it is not possible to bind more than one protocol address to the same transport endpoint. If a user binds more than one transport endpoint to the same protocol address, only one endpoint can be used to listen for connection indications associated with that protocol address. In other words, only one t_bind() for a given protocol address may specify a value of qlen greater than zero. In this way, the transport provider can identify which transport endpoint should be notified of an incoming connection indication. If a user attempts to bind a protocol address to a second transport endpoint with a value of qlen greater than zero, t_bind() will return -1 and set t_errno to [TADDRBUSY]. When a user accepts a connection on the transport endpoint that is being used as the listening endpoint, the bound protocol address will be found to be busy for the duration of the connection, until a t_unbind() or t_close() call has been issued. No other transport endpoints may be bound for listening on that same protocol address while that initial listening endpoint is active (in the data transfer phase or in the T_IDLE state). This will prevent more than one transport endpoint bound to the same protocol address from accepting connection indications.

If fd refers to connectionless mode service, this function allows for more than one transport endpoint to be associated with a protocol address, where the underlying transport provider supports this capability (often in conjunction with value of a protocol-specific option). If a user attempts to bind a second transport endpoint to an already bound protocol address when such capability is not supported for a transport provider, t_bind() will return -1 and set t_errno to [TADDRBUSY].

VALID STATES

T_UNBND

ERRORS

On failure, t_errno is set to one of the following:
[TACCES]
The user does not have permission to use the specified address.
[TADDRBUSY]
The requested address is in use.
[TBADADDR]
The specified protocol address was in an incorrect format or contained illegal information.
[TBADF]
The specified file descriptor does not refer to a transport endpoint.
[TBUFOVFLW]
The number of bytes allowed for an incoming argument (maxlen) is greater than 0 but not sufficient to store the value of that argument. The provider's state will change to T_IDLE and the information to be returned in ret will be discarded.
[TOUTSTATE]
The communications endpoint referenced by fd is not in one of the states in which a call to this function is valid.
[TNOADDR]
The transport provider could not allocate an address.
[TPROTO]
This error indicates that a communication problem has been detected between XTI and the transport provider for which there is no other suitable XTI error (t_errno).
[TSYSERR]
A system error has occurred during execution of this function.

RETURN VALUE

Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and t_errno is set to indicate an error.

SEE ALSO

t_accept(), t_alloc(), t_close(), t_connect(), t_unbind().

CAVEATS

The requirement that the value of qlen never be negotiated from a requested value greater than zero to zero implies that transport providers, rather than the XTI implementation itself, accept this restriction.

An implementation need not allow an application explicitly to bind more than one communications endpoint to a single protocol address, while permitting more than one connection to be accepted to the same protocol address. That means that although an attempt to bind a communications endpoint to some address with qlen=0 might be rejected with [TADDRBUSY], the user may nevertheless use this (unbound) endpoint as a responding endpoint in a call to t_accept(). To become independent of such implementation differences, the user should supply unbound responding endpoints to t_accept().

The local address bound to an endpoint may change as result of a t_accept() or t_connect() call. Such changes are not necessarily reversed when the connection is released.


UNIX ® is a registered Trademark of The Open Group.
Copyright © 1997 The Open Group
[ Main Index | XSH | XCU | XBD | XCURSES | XNS ]