[ofa-general] [PATCH 1/4] ib/ipoib: specify TClassand FlowLabelwith PR queries for QoS support

[Eitan Zahavi]

I support your proposal Sean. Only flowlabel and Tclass.
MTU can be provided by the SA too.

Did we discuss PKey? I am not sure if we need to provide it or not, but
it is the identifier of the 
IPoIB subnet, so the SA will get the IPoIB context of the request.

Eitan Zahavi
Senior Engineering Director, Software Architect
Mellanox Technologies LTD
Tel:+972-4-9097208
Fax:+972-4-9593245
P.O. Box 586 Yokneam 20692 ISRAEL

 

> -----Original Message-----
> From: Sean Hefty [mailto:sean.hefty at intel.com] 
> Sent: Tuesday, August 07, 2007 10:01 PM
> To: 'Jason Gunthorpe'
> Cc: Eitan Zahavi; OpenIB
> Subject: RE: [ofa-general] [PATCH 1/4] ib/ipoib: specify 
> TClassand FlowLabelwith PR queries for QoS support
> 
> >MTU - at a minimum the PR must request a MTU >= than the 
> interface MTU, 
> >having a MTU that is exactly equal to the broadcast group MTU seems 
> >fine to me. The SM should be able to do >= internally.
> 
> My thoughts on MTU (and rate) are that since it isn't carried 
> in any of the headers, we should be able to derive the 
> maximum MTU from the other fields.
> Plus, we've had situations in the past where a smaller MTU 
> than the maximum performed better than the max, so I'm 
> hesitant to add MTU to the query.
> 
> >Flowlabel - IBA hasn't really specified what to do with this 
> field. In
> >IPv6 FlowLabel is assigned randomly by the host on a per-flow (ie
> >socket) basis to allow routers to group packets into flows without 
> >doing deeper inspection. This becomes important when 
> encryption is used 
> >and the routers can no longer access the TCP header to find flows.
> 
> Dropping the flow label from the query is simple enough for 
> IPoIB.  We'll have to consider what to do with the flowinfo 
> field of an IPv6 address for the rdma_cm, however.
> 
> I'd like the minimal number of fields necessary on the host 
> side to give greater flexibility to the SA.  Right now, we're 
> only looking at adding traffic class and, possibly, MTU to 
> the IPoIB PR query.
> 
> - Sean
>