dhcpinform request from Vista gets wrong Name servers
Glenn.Satchell at uniq.com.au
Thu Aug 16 13:01:28 UTC 2007
It sort of does do that, but in the DHCPINFORM there is often not
enough information to work out which pool, for example, a request
belongs to. So you wind up with the most specific level being the
subnet information, which may not be the same as what is in the pool.
Or something like that.
>Date: Wed, 15 Aug 2007 20:54:30 +0200
>From: Sten Carlsen <sten at s-carlsen.dk>
>How about going from general to specific?
>Consider options mentioned in global scope first, then look more
>specific places like subnet declarations, range declarations and host
>declarations. Each time an option is encountered replace whatever was
>picked up by the new value, that should make things work as expected.
>E.g. parsing dhcpinform would pick up the general DNS option and when
>the specific subnet matching this client is parsed, the value specific
>to that subnet is now placed in the packet. This means that you will get
>the generic value if no more specific is defined.
>David W. Hankins wrote:
>> On Wed, Aug 15, 2007 at 12:05:16PM -0400, Randall C Grimshaw wrote:
>>> The Vista machines initially get the correct DNS settings and query for
isatap (which resolves to the registration system) then queries for
teredo.ipv6.microsoft.com (which resolves to the registration system).... then
the rub... the Vista machine will perform a dhcpinfo query for name servers
among other things and get the default name servers in the dhcpack - the default
(production) name servers cannot be reached from the captive subnets and
wouldn't direct users to the registration system if they were.
>>> I am not sure if this is a bug in ISC dhcp by RFC or not.
>> It's (very) cloudy. The final answer is that ISC DHCP should provide
>> the nameservers you /want/ it to provide in DHCPINFORM. We don't.
>> I would say we need to update ISC DHCP to address this, but I'm not
>> precisely sure how...an exclusion list of 'things not to send in
>> DHCPINFORM', sourcing the client's active binding in DHCPINFORM to
>> bring in all that you know about it...
>>> Does snyone know if the dhcpinfo is a response to unexpected answer to the
other two queries?
>> It's hard to say...Windows boxes (XP and Vista alike) will DHCPINFORM
>> for many reasons, none of them can be reliably predicted (windows
>> updater, Flash Proxy Autodetection, etc).
>No improvements come from shouting:
> "MALE BOVINE MANURE!!!"
More information about the dhcp-users