[openib-general] Re: CMA: compliancy issue?

[Michael S. Tsirkin]

Quoting r. Or Gerlitz <ogerlitz at voltaire.com>:
> >>You are suggesting a design change in the CMA which would effect also 
> >>the current CMA ULP consumers: iSER, RDS, NFSoRDMA and Lustre.
> 
> >I don't see how it will affect these ULPs: don't see how ULPs that don't
> >check private data care whether RTU is sent before, or after, the handler.
> 
> I ment to say that today they just get an ESTABLISHED event, and with 
> the change you suggest they will get CONNECT_RESPONSE event and will 
> need to call either rdma_accept or rdma_reject, that's a little change 
> in the ULP state machine.

That would work too, and it does not sound like a huge change.

I was originally thinking along the lines of still using ESTABLISHED, and simply
delaying RTU till after the handler is called.  We would then need to teach CMA
to perform reject instead of RTU if handler returns an error code.

We even can have a flag to select the required behaviour, or even
behave specially for SDP, although I don't think this makes a lot of sense.

Sean, what looks best to you?

-- 
MST