Re: [PATCH 0/7] Kconfig cleanup part II

[David Gibson <david@gibson.dropbear.id.au>]

On Fri, Apr 13, 2007 at 01:37:07PM +0200, Arnd Bergmann wrote:
> On Friday 13 April 2007, Kumar Gala wrote:
> > Here the second part of the Kconfig cleanup to move all the
> > platform/*/Kconfig to be sourced by platform/Kconfig.  Plus some other
> > minor cleanups.
> 
> Nice patches, I had thought about doing these as well after
> I have cleaned up the CPU selection.
> 
> Two questions though:
> - Is there a technical reason why we want to disallow building
>   a kernel that contains a combination of 6xx, 83xx, 86xx and
>   embedded6xx platforms, or are there only bugs that make it
>   impossible?
> 
> - I'm still not sure whether platform selection or cpu selection
>   should come first, three options I can think of are:
>   1. CPU first, then platform:
>      you might choose to build a kernel for 970 exclusively, which
>      would allow you to select powermac, pseries and maple but no
>      other platforms.
>   2. Platform first, the CPU:
>      You might select pseries as the only platform, and then get
>      presented with the choice between power3 to power6, rs64
>      and 970, but not cell or pa6t.
>   3. platform and CPU depend on instruction set, but not on each other,
>      You only choose between powerpc64, 4xx, 6xx, e200, e500 or 8xx.
>      This would allow nonworking setups though, like a 83xx platform
>      with a 601 cpu or a ps3 with power5 cpu.

I would suggest CPU family first (where "family" mostly corresponds to
MMU design: 4xx, 44x, classic32, ppc64 etc.).  Then platforms, with
any combination allowable from the selected CPU family.  Specific CPU
models would usually be implied by platform selection.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson