[ipxe-devel] custom dhcp client_id option
Wissam Shoukair
wissams at mellanox.com
Wed Apr 13 07:12:28 UTC 2016
+YossefE
-----Original Message-----
From: Wissam Shoukair
Sent: Wednesday, April 13, 2016 10:12
To: 'Michael Brown' <mcb30 at ipxe.org>; Maule Mark <mark_maule at yahoo.com>; ipxe-devel at lists.ipxe.org
Subject: RE: [ipxe-devel] custom dhcp client_id option
Hi,
I will need to get back to you on this.
But anyway, a change like this is risky I think, although option 61 doesn't include the correct (or spec compliant) value, changing it might break existing deployments.
I remember that we did add a feature that takes the client identifier according to a NV configuration (or default behavior if non)
-----Original Message-----
From: Michael Brown [mailto:mcb30 at ipxe.org]
Sent: Wednesday, April 13, 2016 00:19
To: Maule Mark <mark_maule at yahoo.com>; ipxe-devel at lists.ipxe.org; Wissam Shoukair <wissams at mellanox.com>
Subject: Re: [ipxe-devel] custom dhcp client_id option
On 12/04/16 17:08, Maule Mark wrote:
> Yea, they use client-identifier a lot, but I don't see that they ever
> say what the id should be, only that it is unique on the subnet.
> Since in the general case a system might want to have independent
> ipaddr's from the same subnet assigned to dhcp clients running
> per-port, wouldn't that mean we need in general for the client id to
> be unique per port (with the ability to override in special case instances)?
Some DHCP identifiers are intended to be unique to the machine, rather than unique to the network interface. It's unclear (to me) which is intended by the RFC.
> On a separate note, is it even required that the client identifier for
> the purposes of address assignment come from DHCP option 61? The pxe
> spec seems to use the ambiguous "DHCP option tags" terminology,
> implying that maybe it is the dhcp server that should be configurable.
Section 2.3 (page 21) of the PXE spec states:
The Client UUID field specifies a universally unique ID (UUID), retrieved
from the client system. The client must have a UUID and must report it in
tag #97 and #61.
so the PXE spec does require that we put the (SMBIOS) UUID into DHCP option 61. There's nothing in the PXE spec that requires the DHCP server to use option 61 for the purposes of address assignment, however.
Michael
More information about the ipxe-devel
mailing list