[openib-general] RFC: detecting duplicate MAD requests

Michael S. Tsirkin mst at mellanox.co.il
Wed Jun 14 14:37:50 PDT 2006


Quoting r. Michael S. Tsirkin <mst at mellanox.co.il>:
> Subject: Re: RFC: detecting duplicate MAD requests
> 
> Quoting r. Hal Rosenstock <halr at voltaire.com>:
> > Subject: Re: RFC: detecting duplicate MAD requests
> > 
> > On Wed, 2006-06-14 at 17:22, Michael S. Tsirkin wrote:
> > > Quoting r. Sean Hefty <mshefty at ichips.intel.com>:
> > > > Subject: Re: RFC: detecting duplicate MAD requests
> > > > 
> > > > Michael S. Tsirkin wrote:
> > > > >>We're kind of left with the same issue of trying to determine if a received
> > > > >>MAD will generate a response.
> > > > > 
> > > > > 
> > > > > How do you mean? We have IsDS=1 flag for dual-sided, don't we? Dual-sided
> > > > > transfer always has a response, doesn't it?
> > > 
> > > I mean, the flag in the application that says that the transfer is dual-sided.
> > > The spec seems to imply that user can figure *from the method* that IsDS=1, so I
> > > assume users will have this logic:
> > > 
> > > "2)
> > > Begin the initial transfer by starting the send operation at the point labelled
> > > Send. The method or other indication should be interpreted on
> > > the other side as initiating a double-sided transfer, causing the receive
> > > context to set IsDS=1."
> > > 
> > > 
> > > So why does the MAD layer care whether a received MAD will generate a
> > > resonse?  A request arrives - we pass it up. Now the ACK for the direction
> > > switch arrives - we pass it up too, application should be waiting for it, it
> > > should take the window and pass the response back to us.
> > 
> > The ACKs are transparent to the application/user.
> 
> Well the ACK for the direction switch is special, isn't it?
> All I'm saying, let's pass it up to the application.

I suggest a rule along the lines of "if an ACK arrives with segment number of 0
this means sender is requesting dual sided RMPP, pass it up to the application".

What's the problem with this approach? I think this does not break existing apps
since these don't do DS RMPP and so never get such an ACK. Right?

-- 
MST




More information about the general mailing list