Hello,

> Question to the audience: could it be that the IBM firmware tries to set up 
> synchronous transfers on its own, when asked to identify itself? This would
> explain it since the driver (almost) certainly didn't request it. It doesn't seem
> too likely though - I had an IBM drive in my S900 for a while ( a 4.5GB DCAS )
> - no problems.

I got some confirmation about this, apparently some IBM drives indeed initiate sync negotiation on their own. The mesh driver's config flags only keep the driver from requesting sync negotiation, not from honouring answers - I doubt this is the intended behaviour. 
This leaves the question how to deal with it - either add more config flags to reject sync negotiation on a per-target base or change the behaviour for the existing flags to effectively prevent synchronous mode - both should be trivial to code, I don't really see a reason to add new flags though.
The third possibility would be to fix the driver to support synchronous transfers, but I didn't find anything remotely resembling documentation for the mesh controller - does anyone know if the Linux folks got synchronous transfers to work on it?

have fun
Michael