Re: [RFC PATCH 2/3] pinctrl: imx: add pinmux-imx53 support

[Linus Walleij <linus.walleij@linaro.org>]

On Tue, Dec 6, 2011 at 8:33 AM, Lothar Waßmann <LW@karo-electronics.de> wrote:
> Shawn Guo writes:
>> On Mon, Dec 05, 2011 at 10:18:38PM +0100, Sascha Hauer wrote:

>> > This brings me to the point that currently we have the pins described as
>> >
>> > #define MX53_PAD_<name>__<function>
>> >
>> But that's also the reason why we have so many lengthy iomux-mx*.h on
>> imx.  Taking iomux-mx53.h for example, it's a 109K header with 1219
>> LOC, but probably only 10% of the definitions will actually be used.
>>
> Which has the benefit of having correct pin definitions for everyone
> to use. If developers who need to use currently unused pindefs have to
> create them on their own, there will always be a good chance in
> getting them wrong.
>
>> > which means that you don't have to look into the datasheet to get the
>> > different options for a pin
>>
>> Looking at the datasheet when we write code is a pretty natural thing
>> to me.
>>
> The pindefs are like interrupt numbers or IO addresses for which there
> also is a complete list of definitions in the kernel no matter whether
> they are actually all in use.

In both cases I'd say it's not the business of the pin control
implementation to worry about size of .h files in arch/arm/*

Getting rid of such defines and board data is the business of the
device tree and nothing else, if I understand the way people are
thinking about this.

So I would prefer to keep these two concepts separate:
1) get something in place that integrates nicely with pinctrl
2) get device tree in place
3) get rid of old board files and huge .h define lists including
   I/O, irqs, pinmux lists...
4) ...
5) profit

So don't try to solve all things at once or you'll end up trying
to drink the ocean.

Yours,
Linus Walleij