Re: [PATCH 10/10] spi: s3c64xx: add device tree support

Mark Brown <broonie@xxxxxxxxxxxxxxxxxxxxxxxxxxx> · Wed, 9 May 2012 15:32:00 +0100

On Wed, May 09, 2012 at 10:13:28PM +0800, Thomas Abraham wrote:
> On 9 May 2012 17:07, Mark Brown <broonie@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
> > On Wed, May 09, 2012 at 03:34:54AM +0530, Thomas Abraham wrote:

> >> +- gpios: The gpio specifier for clock, mosi and miso interface lines (in no
> >> +  particular order). The format of the gpio specifier depends on the gpio
> >> +  controller.

> > This seems odd...  This isn't a bitbanging controller, and surely the
> > driver will need to know which signal is which?  I suspect this is
> > actually for pinmux rather than to identify the signals but that should
> > at least be made clear and really should be being done using the pinmux
> > API.

> The driver retrieves the list of gpio's that it is allowed to use. The
> gpio numbers for miso, mosi and clk are mandatory but the order in
> which they are specified is not important since the driver never needs
> to which gpio is which interface line. I agree the pinmux api should
> be used here, but the call to pinmux api would be a incremental change
> here, not changing the code this patch is adding.

I'd suggest just specifying the order - someone might want to use it
later for some reason and it's not really a hardship for someone to use
it.  Avoids any "how does that work?" questions like I had.

> >> +  - samsung,spi-cs-gpio: A gpio specifier that specifies the gpio line used as
> >> +    the slave select line by the spi controller. The format of the gpio
> >> +    specifier depends on the gpio controller.

> > We should really have a binding for this at the SPI level (and ideally
> > some code to manage setting the GPIO too) - it's pretty common to use a
> > GPIO as /CS.

> The existing implementations vary in the way the nCS gpio lines are
> specified. For some controllers, the nCS gpio's are included in the
> spi device node whereas in this implementation, the nCS gpio is listed
> in the spi slave device node.

Yeah, I know.  I'm saying we should try to come up with a binding for
this that can be used by new SPI contollers going forward so things are
consistent.