Bad udp checksum, says tcpdump
Hi,
i have a vanilla Sarge mailserver running kernel 2.4.27-2-686 and
postfix, no iptables and no special routing rules.
While i was investigating a DNS resolution problem, i discovered that
the outgoing udp packets have a bad udp checksum. Here's a snapshot:
12:19:21.946897 IP (tos 0x0, ttl 64, id 37082, offset 0, flags [DF],
length: 73) 192.168.2.80.33170 > 192.168.2.254.domain: [bad udp cksum
266b!] 22058+ PTR? 10.130.148.21 3.in-addr.arpa. (45)
12:19:21.951073 IP (tos 0x0, ttl 64, id 32476, offset 0, flags [none],
length: 99) 192.168.2.254.domain > 192.168.2.80.33170: 22058 q: PTR?
10.130.148.213.in-addr.arpa. 1/0/0 10.130.148.213.in-addr.arpa. (71)
It's quite strange that i get the answer although my udp packet has a
bad checksum. Could it be a tcpdump/libpcap problem?
Furthermore, the checksum error comes and goes. I tried with a kernel
2.6.8 and a different nic (from a different brand) but nothing changed.
Tcpdump version is 3.8.3-5, the one that comes with Sarge.
The problem happens when postfix does a DNS query but also when i use
host/dig.
Can somebody point me in the right direction?
MatB
Reply to: