On 21 August 2012 20:33, Rob Herring [off-list ref] wrote:

On 08/21/2012 09:48 AM, Chris Ball wrote:

quoted

Hi, adding Shawn and Wolfram,

snip...

quoted

Rob Herring said:

quoted

This makes the most sense to me. However, I prefer broken-cd over
cd-internal. The binding should add properties for exceptions, not SDHCI
spec compliant implementations.

Agreed, I was going to say the same thing.  Putting it all together, it
sounds like we want:

no extra properties:  the CD pin on the host just works.
broken-cd:            the CD pin on the host is broken; use polling.
cd-gpios:             the GPIO listed is the CD pin on the host being
                      brought out directly to a GPIO.
cd-external:          when used with cd-gpios, specifies that the GPIO
                      in cd-gpios is external to the CD pin on the host.

cd-gpios and cd-external can be present on the same node.  if broken-cd
is present, it must be the only one of these nodes used.

I don't see the point of cd-external. Either you just use the CD
interrupt defined within the SDHCI or you have a gpio line independent
of the SDHCI and use cd-gpios.

There should be way to distinguish between the two types of 'cd-gpios' value.

(A) a 'cd-gpios' line that connects the card-detect pin of the mmc
physical slot to the card-detect pad to the mmc host controller.

(B) a 'cd-gpios' line that is connected to the card-detect pin of the
slot but not to the card-detect pad of the mmc host controller (used
either as gpio interrupt or polled gpio line).

So the binding 'cd-external' can act as an additional qualifier to
value of 'cd-gpios' to represent the second type. Hence, the bindings
proposed by Chris does support all the possible combinations of
card-detection used by the Samsung sdhci controller driver.

Thanks,
Thomas.

Rob