Package: dhcp3-relay
Severity: important
Running version 3.0.1-2 of dhcp3-relay, I needed to run a development
DHCP server alongside it. I decided to bind the dhcp server to the
local loopback interface (lo) and configure dhcp3-relay with:
SERVERS="127.0.0.1"
INTERFACES="eth0.95 eth0.96 eth0.97"
OPTIONS="-m discard"
The packets arrive okay, but when dhcp3-relay passes them on the udp
checksum that //leaves// on the udp packet is incorrect.
If you look at the attached packet capture (dhcp-fail.dump) you can see
the packets coming in okay, but then when dhcp3-relay passes them on the
udp checksum is corrupt. I first tested with a 127.0.0.1<->127.0.0.1
setup and when that failed I thought about trying to 'real' public IP's
incase it was some strange loopback issue. I got the exact same
problem. :(
Its only when the source address matches the destination address, it
seems, that the checksum is incorrect calculated.
I have attached a capture (dhcp-good.dump) of it behaving correctly when
its relaying to a non-local dhcp server on another computer.
Okay, its extremely unlikely that someone would want to run a relay on
the same machine as the dhcp server it-self but in a development
environment is saves 'borrowing' another computer; it also points to a
slight issue in the underlying checksum generator[1] which should be
corrected.
Meanwhile the dhcp server keeps filling up the syslog with:
Feb 16 14:07:07 troll dhcpd: 5 bad udp checksums in 5 packets
Feb 16 14:07:24 troll dhcpd: 5 bad udp checksums in 5 packets
Feb 16 14:07:33 troll dhcpd: 5 bad udp checksums in 5 packets
Feb 16 14:07:53 troll dhcpd: 5 bad udp checksums in 5 packets
Feb 16 14:08:44 troll dhcpd: 5 bad udp checksums in 5 packets
Feb 16 14:08:57 troll dhcpd: 5 bad udp checksums in 5 packets
Feb 16 14:09:23 troll dhcpd: 5 bad udp checksums in 5 packets
Which is correct, as the checksum is indeed incorrect.
Cheers
Alex
[1] I'm assuming that the dhcp relay server opens a raw socket and so
generates its own udp checksum?
-- System Information (altered to reflect settings on actual server):
Debian Release: testing/unstable
APT prefers stable
Architecture: i386 (i686)
Shell: /bin/sh linked to /bin/bash
Kernel: Linux 2.6.15-ck1
Locale: LANG=C, LC_CTYPE=C (charmap=ANSI_X3.4-1968)
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
