Re: [PATCH v1 3/3] ARM64 LPC: update binding doc
From: Benjamin Herrenschmidt
Date: Wed Jan 13 2016 - 19:12:43 EST
On Tue, 2015-12-29 at 21:33 +0800, Rongrong Zou wrote:
> Signed-off-by: Rongrong Zou <zourongrong@xxxxxxxxx>
> ---
> Â.../devicetree/bindings/arm64/low-pin-count.txtÂÂÂÂÂÂ| 20 ++++++++++++++++++++
> Â1 file changed, 20 insertions(+)
> Âcreate mode 100644 Documentation/devicetree/bindings/arm64/low-pin-count.txt
>
> diff --git a/Documentation/devicetree/bindings/arm64/low-pin-count.txt b/Documentation/devicetree/bindings/arm64/low-pin-count.txt
> new file mode 100644
> index 0000000..215f2c4
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/arm64/low-pin-count.txt
> @@ -0,0 +1,20 @@
> +Low Pin Count bus driver
> +
> +Usually LPC controller is part of PCI host bridge, so the legacy ISA
> +port locate on LPC bus can be accessed directly. But some SoC have
> +independent LPC controller, and we can access the legacy port by specifying
> +LPC address cycle. Thus, LPC driver is introduced.
> +
> +Required properties:
> +- compatible: "low-pin-count"
I'm not sure about the above. I'd rather just make it "isa" or maybe
isa-lpc. The LPC bus is fundamentally an ISA bus with the 3 cycle
types of ISA etc... I would also allow the node to be named "isa".
> +- reg: specifies low pin count address range
> +
> +
> +Example:
> +
> +ÂÂÂÂÂÂÂÂlpc_0: lpc@a01b0000 {
> + #address-cells = <1>;
> + #size-cells = <1>;
As discussed earlier, address-cells should be 2 with the first cell
indicating the address space type (0 = mem, 1 = IO, possibly 2 =
firmware but that remains somewhat TBD).
Â
> +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂcompatible = "low-pin-count";
> +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂreg = <0x0 0xa01b0000 0x0 0x10000>;
And also as discussed, this is the business of the "ranges" property so
that children devices can be properly expressed.
> +ÂÂÂÂÂÂÂÂ};
Also, this being a bus binding, it should describe the format for
children (for example, PNP related properties).
That leads to the obvious question: Why not just reference the existing
Open Firmware ISA binding ?
Cheers,
Ben.