Hi Simon,
2015-12-28 23:20 GMT+09:00 Simon Glass sjg@chromium.org:
index 371784a..1efbaf2 100644 --- a/include/clk.h +++ b/include/clk.h @@ -49,6 +49,16 @@ struct clk_ops { * @return new clock rate in Hz, or -ve error code */ ulong (*set_periph_rate)(struct udevice *dev, int periph, ulong rate);
/*** get_id() - Get peripheral ID** @dev: clock provider* @args_count: number of arguments* @args: arguments. The meaning is driver specific.* @return peripheral ID, or -ve error code*/int (*get_id)(struct udevice *dev, int args_count, uint32_t *args);What are the arguments for? Who will use them? I wonder if the simple case you have is the only useful case for now?
I think it would be better for the device tree parsing to happen within the uclass rather than requiring the caller to understand how it works. Here, we must obtain the arguments after the phandle and pass them to this function.
My intention was to make the interface as generic as possible.
It is true that most of drivers have '#clock-cells = <1>', but I found some have '#clock-cells = <2>'.
Under linux/Documentation:
devicetree/bindings/clock/qoriq-clock.txt:When the clockgen node is a clock provider, #clock-cells = <2>. devicetree/bindings/clock/qoriq-clock.txt: #clock-cells = <2>; devicetree/bindings/clock/renesas,cpg-mssr.txt: #clock-cells = <2>; devicetree/bindings/clock/st,stm32-rcc.txt: #clock-cells = <2> devicetree/bindings/clock/ux500.txt: #clock-cells = <2>; devicetree/bindings/clock/ux500.txt: #clock-cells = <2>;
So, my idea was to cover the simple case with clk_get_id_simple(), but still allow such rare-case drivers to have their own .get_id() callback.
However, U-Boot might never be hit by such a case, and your simple one meets our demand well enough.