dhcpd related issue - not giving up

Michael Powell nightrecon at hotmail.com
Sun Nov 1 17:28:32 UTC 2009

Dánielisz László wrote:

> I don't give it up, doing some tcpdump on my BSD I can see the dhcp
> request reaches the machine, the dhcpd is running, but why doesn't gives
> any IP?
> # tcpdump -i rl1 -n port 67 or port 68
> tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
> listening on rl1, link-type EN10MB (Ethernet), capture size 96 bytes
> 11:51:43.086597 IP > BOOTP/DHCP, Request
> from 00:24:03:f1:bd:36, length 300 11:51:45.102260 IP >
> BOOTP/DHCP, Request from 00:24:03:f1:bd:36, length 300

I only have a couple if ideas. First, is it possible to substitute some 
other non rl or re NIC for rl1? I seem to recall something about these cards 
having some sort of problem like this. This test would eliminate that idea.

Also, right after a client machine requests a lease examine your arp tables 
on both machines. Maybe the dhcpd server is confused and sending the reply 
out the wrong interface? sockstat -4l can confirm which/what interface dhcpd 
is listening on, compare with arp results. Theoretically if dhcpd is bound 
to and listening on rl1 there shouldn't be any replies going out rl0. Check 
to eliminate.

Wrt to a managed switch blocking ports, I think you probably ruled this out 
by connecting the machines to each other. Note that for GigE, or NICs that 
do MDI-X properly any cable will work. However, on many older 100baseTX 
cards this would need to be using a crossover cable to function correctly.

You can also broaden your tcpdump to include arp traffic. When the output 
files become cumbersome to examine it's easier to look at them in Wireshark. 
I have a hunch if rl1 could be replaced with some old fxp or sk card lying 
around it might work. YMMV


More information about the freebsd-questions mailing list