A DCC installation had an odd problem last week. It seemed that the one of a pair of private DCC servers was receiving malformed DCC requests from all over the Internet. The owner noticed that the IP addresses were of DNS servers and made the connection with a BIND patch intended to vary DNS client port numbers.
Sure enough, http://www.isc.org/index.pl?/sw/bind/view/?release=9.3.5-P2 says in part: 2396. [bug] Don't set SO_REUSEADDR for randomized ports. It seems that 9.3.5-P1 is crazy about varying its source UDP port numbers. It not only doesn't care about playing nice, but actively tries to step on ports in use by other applications with setsockopt(SO_REUSEADDR). When a DCC server (dccd) or client (dccproc, dccifd, or dccm) gets a screwball DNS packet that is really a DNS response, the DNS recursing server does not get the DNS packet it is waiting for. Conversely, when the DNS recursing server gets a screwball DNS response that is really a DCC packet, the DCC client or server does not get its DCC packet. I don't know if more applications than recursing DNS servers and DCC clients and servers are affected by that bug, but I wouldn't be surprised. Installing BIND 9.3.5-P2 (perhaps via a Sun patch) fixed the problem. Vernon Schryver [EMAIL PROTECTED] _______________________________________________ DCC mailing list [email protected] http://www.rhyolite.com/mailman/listinfo/dcc
