[openib-general] SMI comments

[Hal Rosenstock]

Hi again,

I have another comment on the retrun value from smi_handle_dr_smp_send.
In most cases, it means the output port.
In other cases, 0 means discard. This is an issue for switches
where 0 is a valid port number.

As far as the initialization of outgoing and returning DR SMPs, the
assumption is that this is done outside of this code. The same is true
for most of the LID handling, although there needs to be more cases and
support for non permissive LIDs.

-- Hal


-----Forwarded Message-----

From: Hal Rosenstock <halr at voltaire.com>
To: Sean Hefty <mshefty at ichips.intel.com>
Cc: openib-general at openib.org
Subject: [openib-general] SMI comments
Date: 08 Oct 2004 14:27:28 -0400

Hi,

I started looking at the SMI code in detail and have a few
questions/comments below. 

I am assuming the implementation is based on the IBA 1.1 spec
description. 

It looks to me like there is no SMP initialization code (per 14.2.2.1
and 14.2.2.3). 14.2.2.1 doesn't matter right now as we don't initiate
any outgoing SMPs, but 14.2.2.3 does. I think this is why I needed the
hop pointer workaround.

I also didn't see the LID handling as described in this section of the
spec. 

In terms of switches, the port is the input or output (physical) port
and not the port of the MAD registration (base or enhanced port 0).

More detailed comments and one or more patches to follow based on these
comments.

Thanks.

-- Hal

_______________________________________________
openib-general mailing list
openib-general at openib.org
http://openib.org/mailman/listinfo/openib-general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general