[openib-general] Re: [PATCH] Allow setting of NodeDescription

[Hal Rosenstock]

On Thu, 2005-09-15 at 06:58, Michael S. Tsirkin wrote:
> Quoting Hal Rosenstock <halr at voltaire.com>:
> > Subject: Re: [openib-general] Re: [PATCH] Allow setting of NodeDescription
> > 
> > On Thu, 2005-09-15 at 03:36, Michael S. Tsirkin wrote:
> > > Would make more sense to initialize node_desc to an empty string
> > > instead of whatever is programmed in the device flash?
> > > This creates a way for the remote user to find out that the "echo"
> > > script above did not run yet, and try again later.
> > 
> > I think this is the wrong direction to go as the SM may now see many
> > nodes without any NodeDescription. I don't think the tradeoff to know
> > that it wasn't set this way is worth it on what it potentially causes to
> > SM and any management/diagnosic tools.
> 
> Sorry, I dont really understand. Does the spec imply NodeDescription
> must have non-zero bytes? If not, maybe the management/diagnosic tools
> should be fixed.

NodeDescription itself is always 64 bytes. Whether it is null terminated
at 0 bytes and the implications of this is another story.

This approach could result in nodes not having a 0 byte null terminated
NodeDescription. Since this is being used as a convenient administrator
(human) node identification string to say it is that node over there, it
loses that meaning when it is not filled in. That was the problem I was
referring to. It's the human side of using the tools.

-- Hal