Manual Pages  — UNIMSG

uni_msg_len, uni_msg_space, uni_msg_leading, uni_msg_size, uni_msg_ensure, uni_msg_append, uni_msg_extend, uni_msg_alloc, uni_msg_build, uni_msg_destroy, uni_msg_strip32, uni_msg_get32, uni_msg_append32, uni_msg_append8, uni_msg_trail32, uni_msg_dup – ATM signalling library - message buffers

LIBRARY SYNOPSIS DESCRIPTION SEE ALSO AUTHORS

Begemot ATM signalling library (libunimsg, -lunimsg)

#include <uni4/unimsg.h>

size_t
uni_msg_len(const struct uni_msg *msg);

size_t
uni_msg_space(const struct uni_msg *msg);

size_t
uni_msg_leading(const struct uni_msg *msg);

size_t
uni_msg_size(const struct uni_msg *msg);

int
uni_msg_ensure(struct uni_msg *msg, size_t bytes);

int
uni_msg_append(struct uni_msg *msg, void *buf, size_t buflen);

int
uni_msg_extend(struct uni_msg *msg, size_t bytes);

struct uni_msg *
uni_msg_alloc(size_t space);

struct uni_msg *
uni_msg_build(void *buf, ...);

void
uni_msg_destroy(struct uni_msg *msg);

u_int
uni_msg_strip32(struct uni_msg *msg);

u_int
uni_msg_get32(struct uni_msg *msg);

int
uni_msg_append32(struct uni_msg *msg, u_int value);

int
uni_msg_append8(struct uni_msg *msg, u_int byte);

u_int
uni_msg_trail32(const struct uni_msg *msg, int n);

struct uni_msg *
uni_msg_dup(const struct uni_msg *msg);

These functions are used to manipulate variable sized message. They are inspired by BSD mbufs and SysV stream buffers, but somewhat simplified because signalling generally is a low bandwidth task. All the functions operation on a uni_msg data structure:

struct uni_msg {
        u_char  *b_wptr;        /* tail pointer */
        u_char  *b_rptr;        /* head pointer */
        u_char  *b_buf;         /* data buffer */
        u_char  *b_lim;         /* end of data buffer */
};

The field b_buf points to the begin of a memory block that is used to store the actual message and the field b_lim points just to the first byte behind that buffer. This buffer is allocated separate from the structure itself and the user calling any of the above functions with a non const struct uni_msg argument should expect the buffer to be reallocated and hence not hold pointers into the buffer accross call to these functions. The pointer b_rptr points to the first used byte in the message and b_wptr to the first unused byte behind all used bytes. If the message is empty, both pointers point to the same place somewhere in the allocated buffer.

There are several functions and macros that return various sizes and lengths. The macro uni_msg_len() returns the actual size of the message (the number of used bytes). The macro uni_msg_space() returns the number of bytes that are left unused behind the used space. The macro uni_msg_leading() returns the number of bytes that are unused before the used space and the macro uni_msg_size() returns the maximum size of the message (that is the size of the allocated buffer).

Two functions may be used to create new messages: The function uni_msg_alloc() allocates the message structure and a buffer to hold at least space bytes (In fact it allocates a couple of bytes more). If the allocation fails NULL is returned. The pointers are setup so that there is no leading space in the buffer. The function uni_msg_build() constructs a new message from a variable number of buffers. The arguments are pairs of void * pointers to buffers and size_t buffer sizes, terminated by a NULL pointer. The function computes the total resulting message size, allocates a message and copies all the buffers into the message. The message is built to have no leading space. If the allocation fails, NULL is returned.

The function uni_msg_destroy() deallocates the buffer pointed to by the message and the message itself. It is save to pass a message with a NULL buffer, but not a NULL message.

The function uni_msg_dup() returns a copy of a message with exact the same leading space.

A number of functions are used to add bytes to an existing message. The function uni_msg_extend() extends the message buffer to have space for at least bytes additional byte at the end. The leading space does not change. This function may reallocate the message buffer. The function returns 0 on success and ENOMEM if the reallocation fails. In this case the message buffer is not changed. The macro uni_msg_ensure() checks whether the message has space for additional bytes bytes. If not it calls uni_msg_extend() to make the message buffer larger. The macro returns 0 on success or ENOMEM if there is not enough space and the reallocation fails. In this case the message buffer is not changed. The function uni_msg_append() appends buflen bytes from the buffer pointed to by buf to the message. The function uni_msg_append8() appends one byte to the message and the function uni_msg_append32() appends a 32-bit value in network byte order to the message (, b_wptr needs not to be aligned). All three functions call uni_msg_ensure() to make sure, that the buffer contents fit into the message. They return 0 on success and ENOMEM if the buffer is too small and the reallocation fails. In this case the message buffer is not changed.

A number of functions can be used to retrieve parts of the message. The function uni_msg_strip32() returns the last four bytes of the message as a 32-bit integer assumed to be in network byte order. It adjusts b_wptr to remove these four bytes from the message. b_wptr does not need to be aligned. The function uni_msg_get32() returns the first four bytes of the message as a 32-bit integer assumed to be in network byte order. It adjusts b_rptr to remove these four bytes from the message. b_rptr does not need to be aligned. The function uni_msg_trail32() returns the n, 'th 32-bit integer from the buffer counted from the end of the buffer. The integer is assumed to be in network byte order. A value of -1 for n returns the last four bytes of the buffer, a value of -2 the four bytes just before the last four bytes and so on. All three functions do not check that the message is large enough.

Hartmut Brandt <harti@FreeBSD.org>