[openib-general] version #defines for the kernel

[Roland Dreier]

 > > No other kernel subsystem has one, so I don't think it's realistic to
 > > expect one for IB.

 > Don't you think it would be useful?  Even if only to make API changes
 > explicit?

Sure, I admit it would be useful for out-of-tree code.  But it would
also be an unmaintainable mess to actually try and have a set of
feature flags, so I don't think we can do it.

 - R.