[openib-general] Re: [Rdma-developers] Meeting (07/22)summary:OpenRDMA community development discussion

[Steve Wise]

> -----Original Message-----
> From: openib-general-bounces at openib.org 
> [mailto:openib-general-bounces at openib.org] On Behalf Of 
> Christoph Hellwig
> Sent: Monday, August 01, 2005 11:14 AM
> To: Talpey, Thomas
> Cc: rdma-developers at lists.sourceforge.net; openib-general at openib.org
> Subject: Re: [openib-general] Re: [Rdma-developers] Meeting 
> (07/22)summary:OpenRDMA community development discussion
> 
> On Mon, Aug 01, 2005 at 07:00:48AM -0400, Talpey, Thomas wrote:
> > At 06:49 AM 8/1/2005, Caitlin Bestler wrote:
> > >After that we need a standardized way to implement "modify qp
> > >to RTS" in a iWARP-centric fashion.
> > 
> > I will add that RPC/RDMA (NFS-RDMA) does not currently require this
> > functionality, it begins all connections in RDMA mode and 
> the current
> > DAPL semantics are fine for now. This functionality would 
> be of great
> > importance to iSER however, of course.
> > 
> > In the future, NFSv4/sessions has an exchange which can to make
> > use of the iWARP step-up mode, but it is not required and we have
> > deferred implementing it in the session establishment so far.
> 
> I think taking over an existing TCP connection is a horrible idea
> and we should avoid it if possible.  The state for a TCP connection
> is very complex and it's doubtfull we can make it work 100%.
> 

Christoph, 

Can you provide more details on exactly why you think this is a horrible
idea?  I agree it will be complex, but it _could_ be scoped such that the
complexity is reduced.  For instance, the "offload" function could fail
(with EBUSY or something) if there is _any_ data pending on the socket.
Thus removing any requirement to pass down pending unacked outgoing data, or
pending data that has been received but not yet "read" by the application.
The idea here is that the applications at the top "know" they are going into
RDMA mode and have effectively quiesced the connection before attempting to
move the connection into RDMA mode.  We could, in fact, _require_ the
connect be quiesced to keep things simpler.  I'm quickly sinking into gory
details, but I want to know if you have other reasons (other than the
complextity) for why this is a bad idea.


Stevo