[ipxe-devel] custom dhcp client_id option

Maule Mark mark_maule at yahoo.com
Tue Apr 12 14:29:44 UTC 2016


But wouldn't that cause issues in the general case?  If all of the IB ports will then carry the same client_id, then you can only use dhcp to assign an ipaddr to one of them (in the general case).
In my application, I can guarantee that only one port on a host will be issuing dhcp discover requests, although the port which does the dhcp discovery for the host may change from boot to boot.
Mark
 
      From: Michael Brown <mcb30 at ipxe.org>
 To: Maule Mark <mark_maule at yahoo.com>; "ipxe-devel at lists.ipxe.org" <ipxe-devel at lists.ipxe.org>; Wissam Shoukair <wissams at mellanox.com> 
 Sent: Tuesday, April 12, 2016 9:21 AM
 Subject: Re: [ipxe-devel] custom dhcp client_id option
   
On 12/04/16 15:04, Maule Mark wrote:
> I noticed that.  However, dnsmasq only considers client_id (if present)
> followed by hwaddr when looking up configs in its hostsfile.  I've tried
> hacking around that in dnsmasq (by trying to match client_uuid if
> present and no client_id match), but the code is pretty tangled and it's
> not clear to me that this would be spec compliant anyway.

It looks as though our use of the link-layer address as the DHCP client 
identifier (option 61) may be obsolete anyway, since IPoIB devices now 
use 6-byte eIPoIB LEMACs in ll_addr (rather than the 20-byte IPoIB MAC).

Wissam: any objections to switching all link layers (including eIPoIB) 
to place the client UUID in option 61?

Michael


   
 

Sent from Yahoo Mail. Get it now 
      From: Michael Brown <mcb30 at ipxe.org>
 To: Maule Mark <mark_maule at yahoo.com>; "ipxe-devel at lists.ipxe.org" <ipxe-devel at lists.ipxe.org>; Wissam Shoukair <wissams at mellanox.com> 
 Sent: Tuesday, April 12, 2016 9:21 AM
 Subject: Re: [ipxe-devel] custom dhcp client_id option
   
On 12/04/16 15:04, Maule Mark wrote:
> I noticed that.  However, dnsmasq only considers client_id (if present)
> followed by hwaddr when looking up configs in its hostsfile.  I've tried
> hacking around that in dnsmasq (by trying to match client_uuid if
> present and no client_id match), but the code is pretty tangled and it's
> not clear to me that this would be spec compliant anyway.

It looks as though our use of the link-layer address as the DHCP client 
identifier (option 61) may be obsolete anyway, since IPoIB devices now 
use 6-byte eIPoIB LEMACs in ll_addr (rather than the 20-byte IPoIB MAC).

Wissam: any objections to switching all link layers (including eIPoIB) 
to place the client UUID in option 61?

Michael


   
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ipxe.org/pipermail/ipxe-devel/attachments/20160412/0001f329/attachment.htm>


More information about the ipxe-devel mailing list