Re: [PATCH v2 1/3] ARM: tegra: Add AHB driver

[Hiroshi Doyu <hdoyu-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>]

On Wed, 25 Apr 2012 17:59:40 +0200
Stephen Warren <swarren-3lzwWm7+Weoh9ZMKESR00Q@public.gmane.org> wrote:

> On 04/25/2012 06:15 AM, Arnd Bergmann wrote:
> > On Wednesday 25 April 2012, Felipe Balbi wrote:
> ...
> >>> Can this driver be located under "drivers/misc"? Or is there any better place?
> >>
> >> maybe drivers/platform/arm/ ??
> > 
> > I really wouldn't want to add that directory: It has the risk of letting
> > people add random crap there that may or may not be related to ARM (the
> > company) and/or ARM (the architecture).
> > 
> > I think it could go into drivers/amba/ along with the primecell bus driver.
> > The two drivers are for two different aspects of the AMBA spec and don't
> > actually depend on one another as far as I can tell, but it's at least
> > a fitting name, and it doesn't depend on the ARM architecture, which is
> > important because the driver could be used on other architectures that
> > are connected to an AHB bus.
> 
> Hiroshi, is this driver for something (registers/features) that AMBA
> actually specifies? AHB might be part of AMBA (I'm not familiar enough
> to know), but I don't think this aspect of Tegra's AHB bus is part of
> any core AHB/AMBA specification, rather than being some Tegra-specific
> control over the bus (given that the registers and bits correspond to
> Tegra-specific devices on the bus).

Agree.

Taking a look at AHB spec, this tegra-ahb part is a bit beyond that
spec from S/W POV. If there's other similiar AHB drivers from othere
SoC, we may be able to generalize them, but at the moment, it's hard
to imagine how much/little part is common for similar AHB drivers,
like registers/bit definitions. Most of the part may depends on
SoC/Tegra.

> I'm not really convinced that low-level platform-specific drivers like
> this shouldn't be in the arch/arm/mach-* directories.

So I think that keeping this kind of drivers under
"arch/arm/mach-*" or "drivers/amba" as 'tegra'-ahb.c name. If similier
drivers comes, then, what about thinking the generalization again?
Either way would be ok for me.