On Thu, 6 Dec 2012 13:36:11 +0100, Peter Ujfalusi [off-list ref] wrote:

Hi Grant,

On 12/06/2012 11:00 AM, Grant Likely wrote:

quoted

On Mon, 12 Nov 2012 15:41:10 +0100, Peter Ujfalusi [off-list ref] wrote:

quoted

Support for device tree booted kernel.
When the kernel is booted with DeviceTree blob we support one led per
leds-pwm device to have cleaner integration with the PWM subsystem.

For usage see:
Documentation/devicetree/bindings/leds/leds-pwm.txt

Signed-off-by: Peter Ujfalusi <redacted>
---
 .../devicetree/bindings/leds/leds-pwm.txt          |  34 ++++++
 drivers/leds/leds-pwm.c                            | 125 +++++++++++++++------
 2 files changed, 127 insertions(+), 32 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/leds/leds-pwm.txt

diff --git a/Documentation/devicetree/bindings/leds/leds-pwm.txt b/Documentation/devicetree/bindings/leds/leds-pwm.txt
new file mode 100644
index 0000000..9fe3040
--- /dev/null
+++ b/Documentation/devicetree/bindings/leds/leds-pwm.txt

@@ -0,0 +1,34 @@
+LED connected to PWM
+
+Required properties:
+- compatible : should be "pwm-leds".
+- pwms : PWM property, please refer to: 
+  Documentation/devicetree/bindings/pwm/pwm.txt
+- pwm-names : (optional) Name to be used by the PWM subsystem for the PWM device
+- label : (optional) The label for this LED.  If omitted, the label is
+  taken from the node name (excluding the unit address).
+- max-brightness : Maximum brightness possible for the LED
+- linux,default-trigger :  (optional) This parameter, if present, is a
+  string defining the trigger assigned to the LED.  Current triggers are:
+    "backlight" - LED will act as a back-light, controlled by the framebuffer
+		  system
+    "default-on" - LED will turn on, but see "default-state" below
+    "heartbeat" - LED "double" flashes at a load average based rate
+    "ide-disk" - LED indicates disk activity
+    "timer" - LED flashes at a fixed, configurable rate

The binding mostly looks good. However, it seems to be gratuitously
different from the gpio-leds binding and it duplicates property
definitions. Please match the gpio-leds behaviour with each led defined
as a sub node of the pwm-leds node.

The GPIO and PWM bindings are substantially different. For start in pwm we do
not have of_get_pwm* helpers. To get the PWM itself we need to use wpm_get()

That's just the implementation. I'm talking about the binding. :-)

Implementation should follow binding design, not the other way around.
Both bindings use the same pattern. There isn't a of_get_pwm helper now,
but there is nothing preventing one from being created if you need it.

in the controller:
	#gpio-cells  vs. #pwm-cells
in the user:
	gpios = <[gpio specifier]> vs. pwms = <pwm specifier>

The PWM binding states that pwm-names is optional.

which uses the pwms = <>; pwm-names = <>; properties on the device's main node.
This is what I could do at the moment:

twl_pwm: pwm {
	compatible = "ti,twl4030-pwm";
	#pwm-cells = <2>;
};

twl_led: pwmled {
	compatible = "ti,twl4030-pwmled";
	#pwm-cells = <2>;
};

pwmleds {
	compatible = "pwm-leds";
	pwms = <&twl_pwm 0 7812500
		&twl_pwmled 0 7812500>;
	pwm-names = "omap4::keypad",
		    "omap4:green:chrg";
	kpad {
		label = "omap4::keypad";
		max-brightness = <127>;
	};

	charging {
		label = "omap4:green:chrg";
		max-brightness = <255>;
	};

That's just a goofy extra layer of indirection and still doesn't follow
the lead of the gpio-leds pattern. That makes it worse that your
original binding, not better.

It really needs to look like this:

pwmleds {
	compatible = "pwm-leds";
	kpad {
		label = "omap4::keypad";
		max-brightness = <127>;
		pwms = <&twl_pwm 0 7812500>;
	};
	charging {
		label = "omap4:green:chrg";
		max-brightness = <255>;
		pwms = <&twl_pwmled 0 7812500>;
	};
};

g.