El dia 14/05/2014 19:46, "Sören Brinkmann" [off-list ref] va
escriure:

Hi Matthias,

On Wed, 2014-05-14 at 07:12PM +0200, Matthias Brugger wrote:

quoted

Add binding documentation for the General Porpose Timer driver of
the Mediatek SoCs.

Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
---
 .../devicetree/bindings/timer/mediatek,mtk-timer.txt   | 18

++++++++++++++++++

quoted

 1 file changed, 18 insertions(+)
 create mode 100644

Documentation/devicetree/bindings/timer/mediatek,mtk-timer.txt

quoted

diff --git

a/Documentation/devicetree/bindings/timer/mediatek,mtk-timer.txt
b/Documentation/devicetree/bindings/timer/mediatek,mtk-timer.txt

quoted

new file mode 100644
index 0000000..938adf6

--- /dev/null
+++ b/Documentation/devicetree/bindings/timer/mediatek,mtk-timer.txt

@@ -0,0 +1,18 @@
+Mediatek MT6589, MT6577 and MT6572 Timers
+---------------------------------------
+
+Required properties:
+- compatible: Should be "mediatek,mtk6589-timer"

Don't you want compatible strings for the other supported SOCs too?

I'm not sure if it makes sense to define the compatible strings for all the
SoCs that have this timer as it's actually totally the same IP.

In other cases e.g. the arm gic, not all compatible strings are defined,
that's why I use cortex-a15-gic in the dtsi although I'm using a cortex A7
SoC.

I would instead say, that the SoC version of the compatible string, mt6589,
is something arbitrary. So I propose to use the first known SoC which uses
the timer. That would be AFAIK mt6577.

quoted

+- reg: Should contain location and length for timers register.
+- clocks: Clocks driving the timer hardware. This list shoud include

two

quoted

+     clocks. The order is system clock and as second clock the RTC

clock.

Sorry for being pedantic, but we had so many ugly bindings in our vendor
tree. I'm still getting caught up on the clocks. I have the feeling the

SOC

has two clocks (system and rtc) which are used to drive this timer,
while the timer probably refers differently to its clock inputs.
If this was the case, the names 'rtc' and 'system' clock are SOC specific
names/implementation details. From the timer perspective it is probably
just important that it receives two clock inputs. Since the timer is
described here and not the SOC, these SOC specific names should
disappear and be replaced with the names the timer IP uses for its input
clocks.

The timer is exactly identically described in the three SoC datasheets. In
the description the clocks are called system and RTC, so I think we are on
the safe side here. :)

The clock-names should be listed too, probably as optional properties.

This will lead to a problem, as the clock-names description includes the
order of the clocks. If we make this property optional, it can be easily
missunderstood. So I propose to do it the way other drivers in the kernel
(e.g. arm,sp804) do it and leave the clock-names property as it is.

quoted

+
+Examples:
+
+     timer {
+             compatible = "mediatek,mtk6589-timer";
+             reg = <0x10008000 0x80>;
+             interrupts = <GIC_SPI 113 IRQ_TYPE_LEVEL_LOW>;
+             clocks = <&system_clk>, <&rtc_clk>;
+             clock-names = "system-clk", "rtc-clk";

        Sören