[openib-general] [PATCH][iWARP] Added provider CM verbsandquery provider methods

[Bob Sharp]

>From NetEffect's perspective, the per device approach is simple to
implement and I do not see it as an Ammasso specific approach.  As Caitlin
described, existing code needs to be reorganized but this aspect of our port
is not a major effort.

Bob

-----Original Message-----
From:	openib-general-bounces at openib.org on behalf of Caitlin Bestler
Sent:	Thu 8/25/2005 6:35 PM
To:	openib-general at openib.org
Cc:	
Subject:	RE: [openib-general] [PATCH][iWARP] Added provider CM verbsandquery provider methods
 

> -----Original Message-----
> From: openib-general-bounces at openib.org 
> [mailto:openib-general-bounces at openib.org] On Behalf Of Roland Dreier
> Sent: Thursday, August 25, 2005 4:19 PM
> To: Steve Wise
> Cc: openib-general at openib.org
> Subject: Re: [openib-general] [PATCH][iWARP] Added provider 
> CM verbs andquery provider methods
> 
>     Steve> Would you rather we put this in a CM device struct of some
>     Steve> sort?  And have rnic devices export a CM device that has
>     Steve> these sorts of methods?  To me that's basically the same as
>     Steve> adding it to the struct ib_device.
> 
> I think if we really need these new methods, then they might 
> as well be in the rdma_device structure.
> 
>     Steve> Well, i haven't seen any other iwarp device at this detail
>     Steve> (i only know of three iwarp devices in existance), so i
>     Steve> cannot tell you.  Perhaps the other vendors can respond.
> 
> I think it's important for us to understand whether this is 
> an Amasso-specific interface or something that will work in 
> the general case.  Perhaps someone from Broadcom and/or 
> NetEffect (or some other
> vendor) can comment?
> 

The per-device connection methods proposed by Ammasso are definitely
implementable for every iWARP RNIC that I am aware of. It does move
some code around from where we have it currently, so we're not ready
to release our versions yet. But that's because things take time, not
because we think there's anything wrong with the interface.

This interface needs to be complemented by the existing IB-specific
interface and by an eventual TCP-specific interface that is compatible
with the DAT Socket Service Point and/or IT-API's socket convert options.
That interface would also deal with interoperability with pre-IETF MPA.
But only a handful of applications need those optinons, the DAT-style
interface accomodates the vast majority of applications that we are
aware of.



Caitlin Bestler
Principal Software Scientist
Broadcom
caitlinb at broadcom.com

 

_______________________________________________
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