[PATCH V2 5/6] x86: add CONFIG_ARM_AMBA, selected by STA2X11

[H. Peter Anvin]

There is no problem with adding ARM or !X86 dependencies to drivers now and fixing them later or as required.

Alessandro Rubini <rubini at gnudd.com> wrote:

>>> The sta2x11 I/O Hub is a bridge from PCIe to AMBA. It reuses a
>number
>>> of amba drivers and needs to activate core bus support.
>
>> Nacked-by: H. Peter Anvin <hpa at zytor.com>
>> 
>> This brings in tons of code into the x86 all{yes,mod}config builds,
>and 
>> as perhaps one could expect some of it doesn't compile.  For example:
>
>Ok. I apologize for not checking this. Actually, I only fixed the ones
>that I needed to compile.
>
>How should I address the problem? (original code, published on
>sourceforge was simply replicating a number of amba drivers into pci
>drivers, but I don't think massive code duplication is ever sensible,
>thus I preferred to reuse the existing drivers).
>
>I doubt all amba drivers can easily be made x86-compatible.
>I see these possible approaces, all of them with their cons
>
> 1- add STA2X11 dependencies to ARM_AMBA _and_ the drivers that do work
>
>  2- leave CONFIG_ARM_AMBA undefined and play select file in Makefiles
>   using obj-$(CONFIG_STA2X11)
>
>  3- add ARM dependencies to the drivers that do not compile on x86
>
>  any other option?
>
>(1) and (2) bring chipset-specific stuff into the amba world, which is
>bad, especially since I suspect other similar bridges will appear soon.
>
>(3) looks like a temporary hack and stuff can be fixed over time, but
>it has the disadvantage of adding a big number of needless drivers in
>the dsitributions' packages -- like my nacked patch did.
>
>thanks
>/alessandro

-- 
Sent from my mobile phone. Please excuse brevity and lack of formatting.