Main index | Section 3 | Options |
#include <netgraph/netgraph.h>
The NgMkSockNode() function should be called first, to create a new socket type netgraph node with associated control and data sockets. If name is non- NULL, the node will have that global name assigned to it. The csp and dsp arguments will be set to the newly opened control and data sockets associated with the node; either csp or dsp may be NULL if only one socket is desired. The NgMkSockNode() function loads the socket node type KLD if it is not already loaded.
The NgNameNode() function assigns a global name to the node addressed by path.
The NgSendMsg() function sends a binary control message from the socket node associated with control socket cs to the node addressed by path. The cookie indicates how to interpret cmd, which indicates a specific command. Extra argument data (if any) is specified by arg and arglen. The cookie, ,, cmd, and argument data are defined by the header file corresponding to the type of the node being addressed. The unique, non-negative token value chosen for use in the message header is returned. This value is typically used to associate replies.
Use NgSendReplyMsg() to send reply to a previously received control message. The original message header should be pointed to by msg.
The NgSendAsciiMsg() function performs the same function as NgSendMsg(), but adds support for ASCII encoding of control messages. The NgSendAsciiMsg() function formats its input a la printf(3) and then sends the resulting ASCII string to the node in a NGM_ASCII2BINARY control message. The node returns a binary version of the message, which is then sent back to the node just as with NgSendMsg(). As with NgSendMsg(), the message token value is returned. Note that ASCII conversion may not be supported by all node types.
The NgRecvMsg() function reads the next control message received by the node associated with control socket cs. The message and any extra argument data must fit in replen bytes. If path is non- NULL, it must point to a buffer of at least NG_PATHSIZ bytes, which will be filled in (and NUL terminated) with the path to the node from which the message was received.
The length of the control message is returned. A return value of zero indicates that the socket was closed.
The NgAllocRecvMsg() function works exactly like NgRecvMsg(), except that the buffer for a message is dynamically allocated to guarantee that a message is not truncated. The size of the buffer is equal to the socket's receive buffer size. The caller is responsible for freeing the buffer when it is no longer required.
The NgRecvAsciiMsg() function works exactly like NgRecvMsg(), except that after the message is received, any binary arguments are converted to ASCII by sending a NGM_BINARY2ASCII request back to the originating node. The result is the same as NgRecvMsg(), with the exception that the reply arguments field will contain a NUL-terminated ASCII version of the arguments (and the reply header argument length field will be adjusted).
The NgAllocRecvAsciiMsg() function works exactly like NgRecvAsciiMsg(), except that the buffer for a message is dynamically allocated to guarantee that a message is not truncated. The size of the buffer is equal to the socket's receive buffer size. The caller is responsible for freeing the buffer when it is no longer required.
The NgSendData() function writes a data packet out on the specified hook of the node corresponding to data socket ds. The node must already be connected to some other node via that hook.
The NgRecvData() function reads the next data packet (of up to len bytes) received by the node corresponding to data socket ds and stores it in buf, which must be large enough to hold the entire packet. If hook is non- NULL, it must point to a buffer of at least NG_HOOKSIZ bytes, which will be filled in (and NUL terminated) with the name of the hook on which the data was received.
The length of the packet is returned. A return value of zero indicates that the socket was closed.
The NgAllocRecvData() function works exactly like NgRecvData(), except that the buffer for a data packet is dynamically allocated to guarantee that a data packet is not truncated. The size of the buffer is equal to the socket's receive buffer size. The caller is responsible for freeing the buffer when it is no longer required.
The NgSetDebug() and NgSetErrLog() functions are used for debugging. The NgSetDebug() function sets the debug level (if non-negative), and returns the old setting. Higher debug levels result in more verbosity. The default is zero. All debug and error messages are logged via the functions specified in the most recent call to NgSetErrLog(). The default logging functions are vwarn(3) and vwarnx(3).
At debug level 3, the library attempts to display control message arguments in ASCII format; however, this results in additional messages being sent which may interfere with debugging. At even higher levels, even these additional messages will be displayed, etc.
Note that select(2) can be used on the data and the control sockets to detect the presence of incoming data and control messages, respectively. Data and control packets are always written and read atomically, i.e., in one whole piece.
User mode programs must be linked with the
The NgSetErrLog() function has no return value.
All other functions return -1 if there was an error and set errno accordingly.
A return value of zero from NgRecvMsg() or NgRecvData() indicates that the netgraph socket has been closed.
For NgSendAsciiMsg() and NgRecvAsciiMsg(), the following additional errors are possible:
[ENOSYS] | |
The node type does not know how to encode or decode the control message. | |
[ERANGE] | |
The encoded or decoded arguments were too long for the supplied buffer. | |
[ENOENT] | |
An unknown structure field was seen in an ASCII control message. | |
[EALREADY] | |
The same structure field was specified twice in an ASCII control message. | |
[EINVAL] | |
ASCII control message parse error or illegal value. | |
[E2BIG] | |
ASCII control message array or fixed width string buffer overflow. | |
NETGRAPH (3) | November 25, 2013 |
Main index | Section 3 | Options |
Please direct any comments about this manual page service to Ben Bullock. Privacy policy.
“ | Computer science would have progressed much further and faster if all of the time and effort that has been spent maintaining and nurturing Unix had been spent on a sounder operating system. | ” |
— The Unix Haters' handbook |