[openib-general] Re: [PATCH][iWARP] Another CM Verbs Approach

[Roland Dreier]

    Tom> I think there is general concensus that having CM verbs in
    Tom> the ib_device structure is disgusting. This patch moves these
    Tom> verbs to the ib_cm.h file and adds a device capabilities flag
    Tom> that indicates whether or not the device has CM verbs.

    Tom> I know this is not exactly what everyone is looking for, but
    Tom> I hope it is workable given the practical realities we face
    Tom> relative to host stack integration and the need to support
    Tom> devices that establish connections internally.

How about creating an iwarp_cm.h file instead of using ib_cm.h?  That
way it's clear that these new methods are for iWARP devices.

I'd like to understand that last bit above about devices that
establish connections internally.  Does that mean that there are RNICs
that cannot support the iWARP verbs' notion of migrating a connection?

 - R.