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

[Hal Rosenstock]

On Wed, 2005-01-19 at 16:30, Tom Duffy wrote:
> On Tue, 2005-01-18 at 14:06 -0500, Hal Rosenstock wrote:
> > The OpenIB MAD layer is complaining about received MADs with base
> > versions it doesn't understand (other than base version 1). Perhaps are
> > some RMPP MADs sent to it with this not set to 1 ?
> 
> Who is doing this?  You think Solaris, or is it openSM?

Solaris as OpenIB was the receiving side.

> > > On the Solaris side:
> > > 
> > > Jan 18 10:00:20 dongon.SFBay.Sun.COM ibmf: ibmf_i_rmpp_recvr_term_flow(): Received ACK while in recv_term state for single sided trans, msgp = 0x30003770580
> > That's weird. OpenIB wouldn't send any ACKs so I'm not sure what this
> > is. What version were you running with ? I found another issue with 
> > certain SA records where the PayloadLength was miscalculated and would
> > confuse the receiver. This was fixed in rev 1560.
> 
> This is happening on the latest version of Solaris 10 with a fixed ibmf
> (as of today fixed a hang when first packet segment value of 0 instead
> of 1 is sent to it), the latest OpenIB running kernel 2.6.10 and the
> latest openSM (1597).

Latest OpenIB uses Segment Number 1 now so that shouldn't matter. The
older one didn't set this (so it was 0) so basically this was fixed on
both side.

The OpenIB side wouldn't send an ACK so this message must indicate
something else. Any chance you can get an IB trace of what is going on ?
Alternatively can you dump out the MAD where the unsupported base
version is printed out ? It will take me a little bit before I am setup
to try to recreate this.

-- Hal

> -tduffy