[Lse-tech] Re: [RFC] modularization of i386 setup_arch and mem_init in 2.4.18

[Greg KH <gr...@kr...>]

On Fri, Mar 08, 2002 at 10:33:30PM +0100, Dave Jones wrote:
>  As a sidenote (sort of related topic) :
>  An idea being kicked around a little right now is x86 subarch
>  support for 2.5. With so many of the niche x86 spin-offs appearing
>  lately, all fighting for their own piece of various files in
>  arch/i386/kernel/, it may be time to do the same as the ARM folks did,
>  and have..
> 
>   arch/i386/generic/
>   arch/i386/numaq/
>   arch/i386/visws
>   arch/i386/voyager/
>   etc..

YES!!!
I've been working on the Foster patches and keep thinking that this
would be the best solution to our current #ifdef hell.

>  I've been meaning to find some time to move the necessary bits around,
>  and jiggle configs to see how it would work out, but with a pending
>  house move, I haven't got around to it yet.. Maybe next week.
> 
>  The downsides to this:
>  - Code duplication.
>    Some routines will likely be very similar if not identical.
>  - Bug propagation.
>    If something is fixed in one subarch, theres a high possibility
>    it needs fixing in other subarchs

Make sure that every subarch has a maintainer/someone to blame who needs
to make sure their subarch also keeps up to date with the "generic" one
would help out a lot with this problem.

>   The plus sides of this:
>   - Removal of #ifdef noise
>     With more and more of these subarchs appearing, this is getting
> 	more of an issue.
>   - subarchs are free to do things 'their way' without affecting the
>     common case.

I think Martin's recent CONFIG_MULTIQUAD patches prove that the plus
side would outweigh any possible downside :)

thanks,

greg k-h