On Fri, Feb 28, 2014 at 10:37:08AM -0800, Stephen Boyd wrote:

On 02/28, Josh Cartwright wrote:

quoted

On Thu, Feb 27, 2014 at 05:55:18PM -0800, Stephen Boyd wrote:

quoted

+
+EXAMPLE
+
+	keypad {
+		compatible = "qcom,pm8921-keypad";
+		interrupt-parent = <&pmicintc>;
+		interrupts = <74 1>, <75 1>;
+		linux,keymap = <
+			MATRIX_KEY(0, 0, KEY_VOLUMEUP)
+			MATRIX_KEY(0, 1, KEY_VOLUMEDOWN)
+			MATRIX_KEY(0, 2, KEY_CAMERA_FOCUS)
+			MATRIX_KEY(0, 3, KEY_CAMERA)
+			>;
+		keypad,num-rows = <1>;
+		keypad,num-columns = <5>;
+		debounce = <15>;
+		scan-delay = <32>;
+		row-hold = <91500>;
+	};

It odd to me that these newly created bindings don't have 'reg'
properties, even though the device clearly has a register region.

I suppose it makes sense from a "port over from platform data to DT"
perspective, as these drivers have just assumed the location of their
registers to be fixed; however I suspect things will need to be changed
if/when we hope to share these drivers with pm8841/pm8941 and beyond...

Agreed. I would love it if the platform OF code would create
IORESOURCE_REG resources for any reg properties that aren't
translatable to CPU addresses. That way we don't have to pick out the
reg property from DT with special OF code (like you've done in
rtc-pm8xxx).

Yes, I agree this would be nice.  The rtc-pm8xxx register parsing magic
is misplaced/ugly.  I'll see about taking a crack at this and seeing
what it looks like.

I'll throw the reg property into the binding so that in
the future we can support the register moving around (although at the
moment the driver will ignore it).

Great, I think this sounds good.

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation