[Dnsmasq-discuss] question on tftp service

[Simon Kelley]

Carlos Carvalho wrote:
> I tried the tftp service and got this in the log:
> 
> dnsmasq[26150]: DHCPDISCOVER(eth0.58) 00:11:2f:84:22:cc 
> dnsmasq[26150]: DHCPOFFER(eth0.58) 192.168.58.146 00:11:2f:84:22:cc 
> dnsmasq[26150]: DHCPREQUEST(eth0.58) 192.168.58.146 00:11:2f:84:22:cc 
> dnsmasq[26150]: Ignoring DHCP host name ankarana.fisica.ufpr.br because it has an illegal domain part
> ???
> dnsmasq[26150]: DHCPACK(eth0.58) 192.168.58.146 00:11:2f:84:22:cc 
> dnsmasq[26150]: TFTP sent /var/remoteboot/pxelinux.0 to 192.168.58.146
> dnsmasq[26150]: TFTP error 0 TFTP Aborted received from 192.168.58.146
> dnsmasq[26150]: TFTP failed sending /var/remoteboot/pxelinux.0 to 192.168.58.146
> ???
> dnsmasq[26150]: TFTP sent /var/remoteboot/pxelinux.0 to 192.168.58.146
> dnsmasq[26150]: TFTP sent /var/remoteboot/pxelinux.cfg/...
> 
> Two questions: the first is the illegal domain nonsense. It has
> nothing to do with tftp, I get it since 2.35. Why is it saying this?
> All machines are defined in the addn-hosts file; the domain is not
> sent by dhcp, it's explicitly removed.

As far as I can see, you will only get that message if the client sends
the FQDN in the DHCPREQUEST. My guess is that
DHCPOFFER is sending ankarana.fisica.ufpr.br in the hostname or FQDN
option and the client is echoing this back. There's not check that a
hostname matches the hosts-file, just that if it has a domain part, it
must match the domain given in the dnsmasq "domain=" config. You can
safely ignore this, but I might try and fix the problem for the next
release. BTW do you have a domain set in the dnsmasq config, and if so
what is it?

> 
> Second, the tftp error is repeatable, which makes it unlikely to be a
> hardware problem. What could cause it?

The client (The PXE boot-ROM) is aborting the first connection. I saw
this in testing. The ROM  makes one connection to get test for file
existence and to get the length, then abandons that connection and makes
another one to get the file. Maybe it's a workaround from some buggy
server which doesn't handle the file-length option correctly, or it
might just be a PXE-ROM bug - there is some amazingly buggy PXE firmware
out there.

Again, as far as know, it's not a problem as long as the netboot succeeds.

Cheers,

Simon.

> 
> _______________________________________________
> Dnsmasq-discuss mailing list
> Dnsmasq-discuss at lists.thekelleys.org.uk
> http://lists.thekelleys.org.uk/mailman/listinfo/dnsmasq-discuss
>