[Openib-windows] [PATCH] MTHCA: always treat RKEY in network order

[Leonid Keller]

About the patch: if changing, I think, we need to change both rkey and
lkey to be consistent.
About the below stuff: did you try using the new driver on both sides ?

> -----Original Message-----
> From: ftillier.sst at gmail.com [mailto:ftillier.sst at gmail.com] 
> On Behalf Of Fabian Tillier
> Sent: Saturday, April 01, 2006 2:30 AM
> To: Leonid Keller
> Cc: openib-windows at openib.org
> Subject: Re: [Openib-windows] [PATCH] MTHCA: always treat 
> RKEY in network order
> 
> On 3/30/06, Fab Tillier <ftillier at silverstorm.com> wrote:
> > Hi Leo,
> >
> > In testing WSD between the old driver and the new, I 
> discovered that 
> > the RKEY is not handled quite right.  The stack expects the RKEY to 
> > always be returned as a network-byte-order token.  Here's a 
> patch that 
> > swaps the RKEY returned to users, as well as eliminates the swaps 
> > required when posting RDMA read or write (or
> > atomic) requests.
> >
> > With this patch, WSD will work with RDMA reads disabled 
> (IBWSD_NO_READ 
> > environment variable).
> 
> To follow up, with this patch, WSD works with MTHCA as the 
> initiator of all RDMA operations.  If MTHCA is the target of 
> an RDMA, the request fails with an invalid request error 
> (IB_WCS_REM_INVALID_REQ_ERR).  So if sending data to a system 
> with MTHCA installed, using RDMA read will work, if sending 
> data from a system with MTHCA installed, using RDMA write 
> will work.  In either of the preceding cases, the peer system 
> must be running the mt23108 driver.
> 
> - Fab
>