[openib-general] Re: [Fwd: Re: [Fwd: Re: OpenSM Bugs]]

[Hal Rosenstock]

On Thu, 2005-01-20 at 16:07, sandip barua wrote:
> I did find the bug that causes the incorrect base version for outgoing
> MADs while in the receiver termination loop. I fixed the software on
> Tom's solaris system. 

I'm glad to hear you found (and fixed) it.

> I still do see the incoming ACK, 

Can you provide the MAD header for this ? In particular, what is the
method ? Maybe this gets back to OpenSM and some response is attempted.

> but not many
> other unexpected packets from a brief look at the traces.

Are there any unexpected packets ? 

> >> so, ibmf drops them and
> >>
> >>>returns ACKs if in the receiver termination loop.
> > 
> > 
> > Looks to me like these should also be ABORTed with BadT in the receiver
> > termination loop.
> 
> My interpretation is that if a DATA packet arrives in the receiver
> termination loop, an ACK should be sent. If an ACK arrives, and the
> transaction is not double-sided, an ABORT should be sent. If any other
> packet arrives, an ABORT should be sent.

You are right. I misread the termination flow diagram. Sorry.

-- Hal