[PATCH 2/3] ARM: shmobile: r8a7790: Configure R-Car GPIO for IRQ_TYPE_EDGE_BOTH

[laurent.pinchart@ideasonboard.com (Laurent Pinchart)]

Hi Magnus,

On Tuesday 14 May 2013 15:01:01 Magnus Damm wrote:
> On Tue, May 14, 2013 at 11:40 AM, Simon Horman <horms@verge.net.au> wrote:
> > On Tue, May 14, 2013 at 12:33:42AM +0200, Laurent Pinchart wrote:
> >> Hi Simon,
> >> 
> >> On Monday 13 May 2013 17:53:52 Simon Horman wrote:
> >> > "gpio-rcar: Support IRQ_TYPE_EDGE_BOTH" adds support to the R-Car GPIO
> >> > driver for IRQ_TYPE_EDGE_BOTH. As hardware support for this feature is
> >> > not universal for all SoCs a flag, has_both_edge_trigger, has been
> >> > added to the platform data of the driver to allow this feature to be
> >> > enabled.
> >> 
> >> What about moving this information to a platform ID table in the
> >> gpio-rcar driver ?
> > 
> > Magnus, do you have an opinion on this?
> 
> I don't mind so much as long as we:
> a) can control things freely per driver instance
> and
> b) don't have to update all drivers for every new SoC
> 
> The current approach allows for that, so as an example if we can have
> two slightly different GPIO hardware blocks then we can setup two GPIO
> controller driver instances where the configuration can be selected
> per-instance. Perhaps Laurent's proposal would allow for this too, I'm
> not sure.

I agree that parameters related to the instantiation of a GPIO IP core in a 
SoC should come from DT. Parameters related to the IP core version (usch as 
the ability to trigger on both edges) should in my opinion be computed by the 
GPIO driver based on the IP core version number.

> On top of that I'd like to avoid doing per-SoC string matching in drivers.
> Those would force us to update every darn driver for every new SoC. It is
> better to try to use a per-hardware device version in the drivers and then
> configure each SoC to use the appropriate version.

Agreed. Are the version numbers for the GPIO IP core available ?

-- 
Regards,

Laurent Pinchart