dhcpd and infiniband (address type 32)

[Bryan Green]

"David W. Hankins" writes:
> On Thu, Jun 12, 2008 at 11:09:15PM -0700, Bryan Green wrote:
> > +                     case ARPHRD_INFINIBAND:
> > +                       tmp -> hw_address.hlen = 1;
> > +                       tmp -> hw_address.hbuf [0] = ARPHRD_INFINIBAND;
> > +                       memcpy (&tmp -> hw_address.hbuf [1], sa.sa_data,
> > 20);
> > +                       break;
> 
> we'll need to evaluate how the client transmits packets in this case,
> since the code is common, and how or if the relay works.
> 
> i'm also not clear on how this works with the default BPF/LPF/raw
> sockets code, since it doesn't add infiniband framing.  does this
> only work with BSD sockets?  if so it should be #if defined in.

Oh, I left out something that is in that patch:

============================================================================
--- dhcp-3.0.4.orig/includes/site.h     2002-03-12 20:33:39.000000000 +0200
+++ dhcp-3.0.4/includes/site.h  2006-05-23 11:34:38.000000000 +0300
@@ -135,7 +135,7 @@
    the aforementioned problems do not matter to you, or if no other
        API is supported for your system, you may want to go with it. */
         
         -/* #define USE_SOCKETS */
         +#define USE_SOCKETS
============================================================================

It sounds like that is an important component, from what you say.
Sorry for not including that in the original message.

> and we'll have to verify the above against ietf rfc on the subject
> of dhcp over infiniband, if there is any (i know there are rfc's
> on the subject for usb and firewire, although i'm not sure i agree
> with their authors).

The only one that I know of (and I'm no expert) is RFC 4390.

> but do send the patch to dhcp-suggest at isc.org, our ticketing system.

Okay, I'll do that.

Thanks,
-Bryan