Hi, 2004-07-07, sze keltezéssel 07:50-kor Andrew Ivins ezt írta:
Bit of a low level question, again probably one for Krisztian.
One of our developers is trying to ascertain what the original peer addresses are on a tproxy socket. In the code snippet below, he prints out the original source and destination addresses. However they are always zero.
Does anybody know the correct way of doing this? We are still using cttproxy-2.4.25-1.9.3.
Here's the code sample
struct in_origaddrs orig; ... ... if ( getsockopt(fd, SOL_IP, IP_ORIGADDRS, &orig, &orig_len) ) { debug(5, 1) ("commConnectStart: Couldn't get original address for fd\n"); } else { debug(5, 1) ("commConnectStart: FD = %d\n", fd); debug(5, 1) ("commConnectStart: Original source address is %s\n", inet_ntoa(orig.ioa_srcaddr)); debug(5, 1) ("commConnectStart: Original destination address is %s\n", inet_ntoa(orig.ioa_dstaddr)); }
If you are using TCP sockets, use the facilities provided by Netfilter, that is, the SO_ORIGINAL_DST sockopt. For example: struct sockaddr_in sin; socklen_t sinlen = sizeof(sin); if (getsockopt(fd, SOL_IP, SO_ORIGINAL_DST, &sin, &sinlen) != 0) { /* handle error */ } else { /* success, address is in sin */ } TProxy's original address sockopt is useful for UDP sockets, where every packet may have different original destination addresses. In this case, you have to be able to receive all information atomically, with one system call. So, you have to enable receiving of original address information with a setsockopt(), and then use recvmsg() to receive the message. Then, the necessary information should be in the auxiliary information block of the msghdr structure. But you need this only for UDP, the TCP case is much more simple. -- Regards, Krisztian KOVACS