Diff client IDs -> multiple leases

[didi at cs.tau.ac.il]

Hi all,

On Sun, Mar 12, 2006 at 10:47:59AM -0600, Jesse Saletan wrote:
> I run a lab of about 250 systems and have been using ISC's DHCP ever since 
> I replaced the MS DHCP server set up by my predecessor.  It's been working 
> beautifully, but now I find myself coming up against a problem that has 
> been discussed on this list before.
> 
> There are a number of systems on my network that multiboot between Windows 
> and Linux.  When a Windows system requests a lease it sends its MAC 
> address as the client ID.  When that same system is rebooted into linux 
> the DHCP client does not send a client ID, and it is therefore issued a 
> new lease with a different IP address.  I would like to limit clients 
> to a single lease and IP address.
> 
> Back in 2004/2005 a guy named Didi posted some patches to change the way 

I simply took Ted Lemon's advice at
<http://marc.theaimsgroup.com/?l=dhcp-server&m=109423002406763&w=2>
and moved to a job that does not need that :-)

> the dhcp server handles client IDs in order to address this problem.  Some 
> discussion followed about whether it was proper to violate the the RFC to 
> produce the desired result, and it was proposed that an option be added to 
> dhcpd.conf to control this behavior, with appropriate warnings that it was 
> not RFC compliant.
> 
> Can anyone bring me up to date on the status of this issue?  Does the 
> current dhcp server include any option for how to handle client IDs?  Or 
> is there some other way to produce the desired result, other than 
> modifying all the dhcp clients' settings?

All I know is still here:
<http://www.cs.tau.ac.il/~didi/dhcp>

Summary of answers to your questions (but do search the archives for
more details):

My original patch, that ignores UID completely, worked well (and still
does as far as I know) for me. It was considered too intrusive so was
never discussed for inclusion.

There were two main paths:
1. Create a larger (potentially much larger) patch that makes the
behaviour runtime configurable (from the conf file). There is more than
one relevant wanted behaviour. After spending a few hours reading the
sources I decided this was too big for me and dropped it, and apparently
noone else (at least as far as I know) took this.
2. Write a smaller patch, that
2a. is not configurable (and therefore requires much less work) and
2b. does not ignore UID completely, but only behaves as if (no UID) is
the same as (UID is the MAC address). It was implied by others (Ted and
maybe others, I don't remember) that such behaviour will be considered
"necessary evil" which isn't too much against the RFC.
I wrote such a patch, but tested it very briefly, and got no response of
anyone else that tested it, so I do not know if it's good. You are
welcome to try it and give your recommendation on this list, and who
knows - maybe it (or a modified version) will be finally included and
lower the traffic on this list by a few percent per year (as this seems
one of the most recurring subjects).

As I said, I no longer need it for myself, and do not have a diverse
enough environment that will make my testing helpful, but am willing to
help otherwise if you need it (although it seems there are others on
this list more qualified).
-- 
Didi