Re: [PATCH V2 2/2] mfd: stmpe: Extend DT support in stmpe driver

From: Lee Jones <lee.jones@linaro.org>
To: Shiraz Hashim <shiraz.hashim@st.com>
Cc: Viresh Kumar <viresh.kumar@linaro.org>,
	sameo@linux.intel.com, devicetree-discuss@lists.ozlabs.org,
	linux-kernel@vger.kernel.org, spear-devel@list.st.com,
	Vipul Kumar Samar <vipulkumar.samar@st.com>
Subject: Re: [PATCH V2 2/2] mfd: stmpe: Extend DT support in stmpe driver
Date: Fri, 23 Nov 2012 09:33:52 +0000	[thread overview]
Message-ID: <20121123093352.GC17471@gmail.com> (raw)
In-Reply-To: <20121123034557.GB5384@localhost.localdomain>

On Fri, 23 Nov 2012, Shiraz Hashim wrote:

> On Thu, Nov 22, 2012 at 10:31:48PM +0530, Viresh Kumar wrote:
> > On 22 November 2012 21:16, Lee Jones <lee.jones@linaro.org> wrote:
> > > The STMPE GPIO controller can't be used by Device Tree yet in
> > > any case, because it doesn't have an IRQ domain. This is
> > > compulsory, or it won't work. Have you tried to test this
> > > functionality yet?
> > 
> > I don't have SPEAr board to test it anymore. I have moved out of
> > ST now and working in linaro as ARM asignee. Just pushing these
> > as an part time activity.
> > 
> > Though ST guys would have tested stmpe, but stmpe-gpio, i am not
> > sure about.
> 
> Let me bring some more information here. I totally understand
> Jones concerns, but the way stmpe (and may be other mfd devices)
> are handled is this that the parent block (i.e. stmpe) decides on
> the variants (say by probing device itself) and then prepares
> associated data for the (probed) variant and creates a platform
> device for the same.

I realise this, but now we're using Device Tree, there's no need
to stuff pdata in the parent driver now. It's better that the
child devices are self sufficient.

> For the interrupts case also, it is stmpe which registers the
> irq domain. This is because, stmpe driver probes variant and
> populates its platform data and stmpe-gpio may not be aware of the
> variant it serves. At the same time, it (stmpe) needs few of the
> (virtual) interrupts for its internal purpose also.

I know. I wrote the IRQ domain for STMPE. ;)

STMPE needs its own one too, which I will work on now.

STMPE-GPIO doesn't need to be aware of anything, the device
which wishes to use its GPIOs/IRQs will reference it from
Device Tree in the manor previously explained.

> Hence stmpe passes irq_base to the stmpe-gpio driver while
> allocating and registering irq domain by itself.

Not anymore it doesn't. There are no irq_base:s with DT. All
IRQs are dynamic, hence why STMPE requires its own domain to
play with. I can fix that.

> With this approach we have tested the functionality on SPEAr
> platform.

You'll need to test it again with the new DT approach too. :)

-- 
Lee Jones
Linaro ST-Ericsson Landing Team Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog