Re: [patch/rfc 2.6.19-rc5] arch-neutral GPIO calls

From: David Brownell <david-b@pacbell.net>
To: Bill Gatliff <bgat@billgatliff.com>
Cc: Paul Mundt <lethal@linux-sh.org>,
	Linux Kernel list <linux-kernel@vger.kernel.org>,
	Andrew Morton <akpm@osdl.org>,
	Andrew Victor <andrew@sanpeople.com>,
	Haavard Skinnemoen <hskinnemoen@atmel.com>,
	jamey.hicks@hp.com, Kevin Hilman <khilman@mvista.com>,
	Nicolas Pitre <nico@cam.org>, Russell King <rmk@arm.linux.org.uk>,
	Tony Lindgren <tony@atomide.com>
Subject: Re: [patch/rfc 2.6.19-rc5] arch-neutral GPIO calls
Date: Mon, 13 Nov 2006 12:53:00 -0800	[thread overview]
Message-ID: <200611131253.00828.david-b@pacbell.net> (raw)
In-Reply-To: <4558D4E6.4020601@billgatliff.com>

On Monday 13 November 2006 12:26 pm, Bill Gatliff wrote:

> >Nah; look at arch/arm/plat-omap/gpio.c and ignore the mess, but observe
> >that what you see there is essentially a bunch of "gpio controller"
> >classes using the ugly "switch(type)" dispatch scheme instead of the
> >prettier "type->op()" dispatch scheme.  All that stuff needs to be
> >cleaner, but for now it'd suffice to add a new FPGA typecode.
> 
> Agreed.  But if we add to the machine descriptor, then not only do you 
> not need to touch arch-omap/gpio.c, but you can take that switch 
> statement out, too.  Just one less chunk of code to tweak when a new 
> platform is supported.

Do non-ARM platforms have board/machine descriptors on Linux, though?
I thought most didn't ...

One could come up with an implementation that uses GPIO numbers
as indices into a descriptor array, and using board-specific
initialization of that array ... just like with IRQs and irq_chip.

That could lead to heavier weight implementations than I'd prefer
to see (since GPIOs are a very light weight notion!), but it'd
certainly provide a more reusable way to add GPIO controllers.

All behind the API I proposed, note -- no changes needed.

- Dave