I have had the same problem, where I could not pxeboot from an alias of an interface. eg pxeboot worked from hosts connected to eth0, but not to hosts connected to eth0:1.
tcpdump shows what's happening: > 18:41:00.618219 ARP, Request who-has 10.70.200.1 tell 10.70.200.167, length 46 > 18:41:00.618231 ARP, Reply 10.70.200.1 is-at 00:20:4c:e2:10:8c, length 28 > 18:41:00.618484 IP 10.70.200.167.2070 > 50.10.1.69.69: 27 RRQ "pxelinux.0" > octet tsize 0 > 18:41:00.618794 IP 10.70.200.1.53553 > 10.70.200.167.2070: UDP, length 14 > 18:41:00.619097 IP 10.70.200.167.2070 > 50.10.1.69.53553: UDP, length 17 > 18:41:00.619128 IP 50.10.1.69 > 10.70.200.167: ICMP 50.10.1.69 udp port 53553 > unreachable, length 53 50.10.1.69 is host A, eth0 10.70.200.1 is host A, eth0:1 10.70.200.167 is host B So the packets go: 1. 10.70.200.167.2070 > 50.10.1.69.69 (client contacts server on eth0) 2. 10.70.200.1.53553 > 10.70.200.167.2070 (server binds with different interface, eth0:1, and replies) 3. 10.70.200.167.2070 > 50.10.1.69.53553 (client tries to respond to server on original IP address, eth0, and fails because atftpd is bound to wrong IP) 4. IP 50.10.1.69 > 10.70.200.167: ICMP 50.10.1.69 udp port 53553 unreachable I do not program much in C, so I cannot evaluate Paul Szabo's patch except to say that I tried it and it solves my problem. Therefore I vote for including the patch asap, pending DD review. Thanks, Chris -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
