[Dnsmasq-discuss] Conditional sending of different DHCP-Options

[joerg jungermann]

Hi !

Simon Kelley wrote:
>>>> Then (stage2) uboot.0 is started an request IP + bootoptions via DHCP.
>>>> The vendor class is "DBOX2, EMPTY" (obtained via tcpdump/wireshark)
>>>> , so
>>>> the line
>>>> dhcp-vendorclass=dbox2-kernel,DBOX2
>>>> should match.
>>>
>>> dnsmasq logs the vendor-class when log-dhcp is on, and it's not doing so
>>> here, so the vendor class is not being picked up. That's enough to
>>> explain the problems you are having. Could you send me a dump of the
>>> packet which has the vendor-class ID in it, so I can check what's
>>> happening?
>>
>> I see. Have a look at http://[...] please.
> Got it, thanks.
> 
> The obvious thing there is that the DHCP packets from the client are
> subtly wrong, they are messing the "end" option. I think that this is
> interacting with a bug in dnsmasq which means that the last option in
> the packet does not get recognised. Normally this is the "end" option,
> so it doesn't matter, but in this case it's the vendor-class option, so
> that is being missed.
> 
> Can you easily upgrade dnsmasq to version 2.45, the latest? I believe
> that this bug was fixed in 2.43 and just doing the upgrade should fix
> the problem.
Upgrading to 2.45 fixes this problem.

Thanks.

cya