[openib-general] OFED 1.1 and OpenSM on SLES 10 for PPC64

[Hal Rosenstock]

On Wed, 2006-09-20 at 16:42, Michael S. Tsirkin wrote:
> Quoting r. Hal Rosenstock <halr at voltaire.com>:
> > Subject: RE: OFED 1.1 and OpenSM on SLES 10 for PPC64
> > 
> > On Wed, 2006-09-20 at 11:11, Eitan Zahavi wrote:
> > > I will try to get to that tomorrow
> > 
> > It's not an OpenSM issue. See the latest info in the bug report:
> > http://openib.org/bugzilla/show_bug.cgi?id=241
> 
> I dug in a bit and I'm not sure what's the root cause,
> but what is triggering the problem is that the saquery diag
> utility depends on opensm,

No, it depends on the opensm library not opensm. This was all fine until
the libraries were broken into a separate RPM for OFED to attempt to
decouple them from OpenSM.

> which makes a mess of dependencies,

It requires opensm library for the SA client interface and complib for
portability. I believe this is no different than some other IB utilities
in OFED too.

>  and at some point libtool goes berserk.

Huh ?

> Short term, can we just skip saquery utility in OFED 1.1?
> Hal, can you approve this please?

I would prefer that is not the case and this is part of OFED 1.1.

> Longer term, I think saquery should be fixed not to depend on opensm - opensm is
> a large tool, complicated by portability requirements etc, and it is a waste to
> need parts of it on endnodes just to be able to run some diagnostics.
> With RMPP support in kernel, we really sholdn't need an extra depenency
> just to push a query and get a response.
> Comments?

It could depend on Sean's new user space SA client API (which perhaps
needs some more infrastructure) but we are not there yet.

-- Hal