On 07/26/2017 09:20 AM, Suman Anna wrote:

On 07/26/2017 08:36 AM, Keerthy wrote:

quoted

On Wednesday 26 July 2017 06:57 PM, Franklin S Cooper Jr wrote:

quoted

On 07/26/2017 08:00 AM, Suman Anna wrote:

quoted

Hi Keerthy,

On 07/26/2017 01:45 AM, Keerthy wrote:

quoted

The patch adds keystone-k2g compatible, specific properties and
an example.

Seems we are adding information regarding several Keystone 2 SoCs. So
the commit and subject should be tweaked to reflect this.

Okay i can add that as well.

quoted

quoted

Please update patch header to "dt-bindings: gpio: davinci: ...."

quoted

Signed-off-by: Keerthy <redacted>
---

Changes in v3:

  * Added details about family of SoCs corresponding to compatibles.

 .../devicetree/bindings/gpio/gpio-davinci.txt      | 40 +++++++++++++++++++++-
 1 file changed, 39 insertions(+), 1 deletion(-)

diff --git a/Documentation/devicetree/bindings/gpio/gpio-davinci.txt b/Documentation/devicetree/bindings/gpio/gpio-davinci.txt
index 5079ba7..fb9efee 100644
--- a/Documentation/devicetree/bindings/gpio/gpio-davinci.txt
+++ b/Documentation/devicetree/bindings/gpio/gpio-davinci.txt

@@ -1,7 +1,10 @@
 Davinci/Keystone GPIO controller bindings
 
 Required Properties:
-- compatible: should be "ti,dm6441-gpio", "ti,keystone-gpio"
+- compatible: should be "ti,dm6441-gpio": for Davinci da850 SoCs 
+			"ti,keystone-gpio": for Keystone 2 66AK2H/K, 66AK2L,
+						66AK2E SoCs
+			"ti,keystone-k2g-gpio", "ti,keystone-gpio": for 66AK2G
 
 - reg: Physical base address of the controller and the size of memory mapped
        registers.

@@ -26,6 +29,17 @@ The GPIO controller also acts as an interrupt controller. It uses the default
 two cells specifier as described in Documentation/devicetree/bindings/
 interrupt-controller/interrupts.txt.
 
+Required Properties specific to keystone-k2g

Thanks for updating the binding for the clocks, but clocks are not
specific to K2G. They also apply to other K2 SoCs. Davinci platforms do
not have DT clocks atm, so the Davinci portion can be updated later if
and when they get added.

What about power-domain property?

The correct name is "power-domains".

quoted

Driver has no pm_runtime implemented yet.

True, not yet, but this is in general a required property on K2G SoCs to
automatically enable clocks through runtime_pm. Clock properties on K2G
nodes should only be truly required if a driver is using clk API
(ideally to control optional clocks or for adjusting clock frequencies).
When the gpio-davinci driver gets updated to use pm_runtime, the clock
properties will be rendered obsolete for K2G.

Now I understand better when we do or do not need this property after
some offline discussion. If the driver doesn't support pm_runtime then
no point in adding this property. Binding should discuss how to use the
current driver. Not how a feature that may never exist may possibly be used.

Rob,
Any suggestions on how we need to handle this? Should we be adding the
property now or later when we adapt the driver for runtime_pm? This
would be a common theme for K2G nodes that are reusing Davinci drivers.

My take on this would be to add the property now, and mark the clock
properties obsolete when the driver gets converted.

regards
Suman

quoted

quoted

quoted

quoted

+
+- clocks: Should contain devices input clock. The first parameter
+           is a handle to k2g_clks. The second parameter is the
+           device ID and the third parameter is the clock ID. One can
+           refer: http://processors.wiki.ti.com/index.php/TISCI#66AK2G02_Data

No need for this link here, just refer to the appropriate clock
bindings. Some thing like the following would be better

This information is helpful especially with the macros being dropped.
However, I agree that this is not the place for this information.
Probably should be linked to in the ti,sci-clk.txt and
sci-pm-domain.txt. However, both of these are outdated since it is
referring to macros and includes that don't exist or will no longer exist.

Hence included here but yes i can point to something like what Suman
asked me to.

quoted

quoted

- clocks:               Should contain the device's input clock, and
                        should be defined as per the appropriate clock
bindings consumer usage in,

Documentation/devicetree/bindings/clock/keystone-gate.txt
                            for 66AK2HK/66AK2L/66AK2E SoCs or,

Documentation/devicetree/bindings/clock/ti,sci-clk.txt
                            for 66AK2G SoCs

quoted

+
+           Example: <&k2g_clks 0x001c 0x0>;

This can be dropped as well, below example already demonstrates it.

quoted

+
+- clock-names: The driver expects the clock name to be "gpio";

Just say, Should be "gpio". No need of mentioning about the driver.

quoted

+
 Example:
 
 gpio: gpio@1e26000 {

@@ -60,3 +74,27 @@ leds {
 		...
 	};
 };
+
+Example for keystone-k2g:

s/keystone-k2g/66AK2G/

quoted

+
+gpio0: gpio@2603000 {
+	compatible = "ti,keystone-k2g-gpio", "ti,keystone-gpio";
+	reg = <0x02603000 0x100>;
+	gpio-controller;
+	#gpio-cells = <2>;
+	interrupts = <GIC_SPI 432 IRQ_TYPE_EDGE_RISING>,
+			<GIC_SPI 433 IRQ_TYPE_EDGE_RISING>,
+			<GIC_SPI 434 IRQ_TYPE_EDGE_RISING>,
+			<GIC_SPI 435 IRQ_TYPE_EDGE_RISING>,
+			<GIC_SPI 436 IRQ_TYPE_EDGE_RISING>,
+			<GIC_SPI 437 IRQ_TYPE_EDGE_RISING>,
+			<GIC_SPI 438 IRQ_TYPE_EDGE_RISING>,
+			<GIC_SPI 439 IRQ_TYPE_EDGE_RISING>,
+			<GIC_SPI 440 IRQ_TYPE_EDGE_RISING>;
+	interrupt-controller;
+	#interrupt-cells = <2>;
+	ti,ngpio = <144>;
+	ti,davinci-gpio-unbanked = <0>;
+	clocks = <&k2g_clks 0x001b 0x0>;
+	clock-names = "gpio";
+};

If your going to talk about other Keystone 2 devices it would be helpful
to include an example for one of them since they have slightly different
properties.

Sure.

quoted

quoted

quoted

regards
Suman

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html