Re: [PATCH 02/10] mach-u300: rewrite gpio driver, move to drivers/gpio

From: Kyungmin Park <kmpark@infradead.org>
To: Shawn Guo <shawn.guo@freescale.com>
Cc: Nicolas Pitre <nicolas.pitre@linaro.org>,
	Linus Walleij <linus.walleij@stericsson.com>,
	Jonas Aaberg <jonas.aberg@stericsson.com>,
	Sascha Hauer <s.hauer@pengutronix.de>,
	linux-kernel@vger.kernel.org,
	Grant Likely <grant.likely@secretlab.ca>,
	Lee Jones <lee.jones@linaro.org>,
	Linus Walleij <linus.walleij@linaro.org>,
	linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH 02/10] mach-u300: rewrite gpio driver, move to drivers/gpio
Date: Fri, 20 May 2011 12:43:46 +0900	[thread overview]
Message-ID: <BANLkTikonewi-fxCxoxx=ARsTs-CPvfoTg@mail.gmail.com> (raw)
In-Reply-To: <20110520031818.GB27251@S2100-06.ap.freescale.net>

On Fri, May 20, 2011 at 12:18 PM, Shawn Guo <shawn.guo@freescale.com> wrote:
> On Thu, May 19, 2011 at 03:30:36PM -0400, Nicolas Pitre wrote:
>> On Thu, 19 May 2011, Sascha Hauer wrote:
>>
>> > On Thu, May 19, 2011 at 09:56:32PM +0800, Shawn Guo wrote:
>> > > On Thu, May 19, 2011 at 02:21:27PM +0200, Linus Walleij wrote:
>> > > > On Thu, May 19, 2011 at 10:56 AM, Shawn Guo <shawn.guo@freescale.com> wrote:
>> > > >
>> > > > > I start working on moving mxs gpio (arch/arm/mach-mxs/gpio.c) into
>> > > > > driver/gpio, and I see the possibility to go a different approach
>> > > > > from U300 one posted here.
>> > > >
>> > > > I've tried to figure out what relation the mail has to the U300 driver
>> > > > but cannot find any, more than that it's moving a driver... Please
>> > > > start a new mail thread.
>> > > >
>> > > I will post mxs-gpio driver once I get it done.  Then please review
>> > > the code and see the difference between mxs-gpio and u300-gpio,
>> > > though these hardwares have something in common.
>> >
>> > I'm pretty sure they have something in common and even more that *all*
>> > gpio drivers have something in common. I wonder if it really makes sense
>> > to move the gpio driver to drivers/gpio without creating a common
>> > mmio_gpio_chip beforehand. This can't be very hard.
>>
>> I do think that performing the move first will make a subsequent
>> conversion easier.  And since a move is a no-op from a functional point
>> of view, it is the safest thing to do first.
>>
> That's also what I heard on UDS week.  Move stuff into driver/gpio
> first, then try to find the pattern in those drivers and come up with
> some framework.

I agree, I reviewed the mmio_gpio_chip, but it's too simple to cover
current implementation.
So first move to the drivers/gpio and make more common mmio_gpio
instead of basic.

BTW, how do you think the GPIO interrupt routine. If it uses the
drivers/gpio drivers, it also cover the gpio interrupt codes at
drivers/gpio? or separate it?

Thank you,
Kyungmin Park