[openfabrics-ewg] QoS in opensm will not be part of OFED 1.2

Michael S. Tsirkin mst at mellanox.co.il
Tue Feb 6 06:41:03 PST 2007


> > > Quoting Michael S. Tsirkin <mst at mellanox.co.il>:
> > > Subject: Re: QoS in opensm will not be part of OFED 1.2
> > > 
> > > > > > > I had an AI to check the QoS status with OSM.
> > > > > > > Conclusions are that QoS support in OpenSM will not be part of OFED 1.2 
> > > > > > > (I updated the plan on the Wiki)
> > > > > > > 
> > > > > > > The reasons for this are:
> > > > > > > 1. Code not ready at code freeze.
> > > > > > > 2. There are technical discussion in the list regarding some 
> > > > > > >    implementation details (e.g. XML or text syntax).
> > > > > > > 3. SPEC is not published by IBTA yet.
> > > > > > 
> > > > > > I think this last reason also applies to the end client QoS changes as
> > > > > > well.
> > > > > 
> > > > > Yes. But the other 2 don't.
> > > > 
> > > > Right but I think that precludes it from being included in OFED right
> > > > now.
> > > 
> > > Since the code is already included in OFED, moving it out would violate the feature
> > > freeze rules, unless there's an actual bug this would fix.
> > 
> > OTOH, you are right in that without SM support we can't claim to have this
> > feature at all. So, to avoid controversy, I have just removed the QoS patches
> > from IB core and pushed the code out.
> 
> I think that the mthca patch to encode SL in sched_queue field to
> improve hardware QoS guarantees for connected QPs is useful as this can
> be exercised by IPoIB-CM. If so, should/can this be included ?

OK. Note this is still untested, and off by default.

-- 
MST




More information about the ewg mailing list