[ofw] Porting OFA stack to WDK

[Sean Hefty]

>I'd vote we do the 'right' thing, for both short and long term....  :-)

If we break the existing userspace APIs, WinOF should move to a 2.x release.  If
we move to a 2.x release, is this the only thing about the interface that we
would want to change?

>That said, accomplish both (short & long term) goals.
>
>The earlier in the release chain of WinOF we push the application recompile,
>the easier (less impact) it will be.

I don't think it's *just* a recompile.  I'm guessing some minor code changes
will be needed as well - if only to remove some type casts.

>The assumption being, once the __ptr64 is removed (applications recompiled),
>then the longer term goal of splitting user-kernel header files will not
>force an application recompile?

The suggesting of splitting userspace from kernel is to prevent userspace from
needing to recompile.  Only the kernel would need that, but OFA owns those
pieces.  (I.e. the existing userspace APIs should be cast in stone for 1.x, but
not for the kernel.)

>Does splitting user-kernel header files makes sense in light of WinVerbs?

I do not plan on having winverbs share the interface/implementation between
userspace and the kernel.  For the short term, I don't really plan on having a
formal kernel winverb interface, since there wouldn't be any kernel consumers.
For longer term, I hadn't given it much thought.

- Sean