Message-ID: <20081030232452.GA15089@yookeroo.seuss>
Date:	Fri, 31 Oct 2008 10:24:52 +1100
From:	David Gibson <david@...son.dropbear.id.au>
To:	Anton Vorontsov <avorontsov@...mvista.com>
Cc:	Grant Likely <grant.likely@...retlab.ca>, linuxppc-dev@...abs.org,
	David Brownell <dbrownell@...rs.sourceforge.net>,
	linux-kernel@...r.kernel.org, Pierre Ossman <drzeus-mmc@...eus.cx>
Subject: Re: [PATCH 1/3] powerpc: Add mmc-spi-slot bindings

On Fri, Oct 31, 2008 at 02:02:53AM +0300, Anton Vorontsov wrote:
> On Thu, Oct 30, 2008 at 02:37:31PM -0600, Grant Likely wrote:
> [...]
> > > +- gpios : (optional) may specify GPIOs in this order: Card-Detect GPIO,
> > > +  Write-Protect GPIO.
> > 
> > I wonder if we're following the example of irq mappings too closely
> > for the gpios property.  I like the layout of the property
> > (<controller> <specifier>), but I think the 'gpios' name is getting
> > too overloaded.  In this case a single property 'gpios' is being used
> > to encode 2 unrelated bits of information; the write protect pin and
> > the card detect pins.
> > 
> > In this particular case I think it is better to use 2 properties in
> > this case; something like 'spi-writeprotect-gpio' and
> > 'spi-carddetect-gpio' using the same specifier format.  Doing so adds
> > a bit more clarity to the purpose of the properties.
> > 
> > I my mind I differentiate this from other examples (for instance a
> > series of CS pins) based on how closely related the pin functions are.
> >  So I would say for the following examples...
> > 1) GPIO data bus (SPI, MDIO and I2C are great examples); all pins must
> > be present - single gpio property
> > 2) This MMC case (pins are optional and unrelated); separate gpio properties
> > 3) LCD with backlight and contrast control pins; one gpio property for
> > backlight pins, one for constrast pins.
> > 
> > Thoughts?
> 
> It's pretty trivial to implement (of_get_named_gpio() -- could be just
> factored out of of_get_gpio()).
> 
> Though,
> 
> 1. The idea is quite extreme. It needs discussion, and furthermore,
>    we need to define when do we use gpios = <> and when something-gpio =
>    <>; We need to be consistent, and to be consistent, the rules should
>    be clear and written.
> 
> 2. We should think about it very very carefully. Do we want to lose the
>    track of gpios? For example, there are quite defined rules when (and
>    in what properties) you may encounter memory addresses, when and
>    where you can encounter interrupt specifiers. We do the same for
>    gpios, and so far it works great. We need to think about any possible
>    drawbacks of the scheme you purpose (we would never know where to
>    expect gpios - it isn't a problem per se, but maybe it could lead
>    to some problem in future? I don't know.)
> 
> Quite honestly I don't like the idea... maybe I just used to
> interrupts = <>, reg = <>, ranges = <>, interrupt-map = <> and so
> forth, and now my subconsciousness tells me "it's wrong to do
> something-interrupt = <> stuff." ;-)

Fwiw, I agree.  The current scheme works, adding new places to look
for gpio specifiers will just complexify things.  Long lists of gpios
may be somewhat awkward to work with, but I don't think it's
sufficiently bad to warrant another scheme.

-- 
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
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists