Re: [PATCH 01/14] qcom: mtd: nand: Add driver data for QPIC DMA

[Abhishek Sahu <absahu@codeaurora.org>]

On 2017-07-17 12:52, Boris Brezillon wrote:
> On Mon, 17 Jul 2017 11:41:01 +0530
> Abhishek Sahu <absahu@codeaurora.org> wrote:
> 
>> >> > +
>> >> > +nand@79b0000 {
>> >
>> > nand-controller@xxxx {
>> >
>> > BTW, glad to see another driver moving to the new DT representation
>> > :-).
>> >
>> >> > +	compatible = "qcom,qpic-nandc-v1.4.0";
>> >> > +	reg = <0x79b0000 0x1000>;
>> >> > +
>> >> > +	clocks = <&gcc GCC_QPIC_CLK>,
>> >> > +		<&gcc GCC_QPIC_AHB_CLK>;
>> >> > +	clock-names = "core", "aon";
>> >> > +
>> >> > +	#address-cells = <1>;
>> >> > +	#size-cells = <0>;
>> >> > +
>> >> > +	nandcs@0 {
>> >
>> > 	nand@0 {
>> >
>> >> > +		compatible = "qcom,nandcs";
>> >
>> > Why do you need a compatible here?
>> It is the part of original driver. We can connect multiple
>> NAND devices in the same bus and qcom,nandcs is being used
>> for each connected NAND device. Each NAND device can use
>> different  chip select, ecc strength etc which we can specify
>> under this sub node.
> 
> 
> Still don't understand why you need a compatible? Is this a memory bus
> where you can connect other kind of memories (parallel NORs,
> SRAMs, ...)?
> 
> If that's not the case, then considering all subnodes of the
> nand-controller node containing a reg property as NAND devices is fine,
> you don't need this compatible = "nand,cs" (see sunxi-nand bindings
> [1]).
> 

  Thanks Boris for giving detailed references.

  We can connect other parallel devices also but we have
  different hardware wrappers over generic EBI2/QPIC which
  will MUX/arbitrate the device access from the hardware itself.
  So this NAND driver will only control multiple NAND devices.

  We can remove this compatible and use the bindings
  similar to sunxi-nand. I will do the required changes and
  will post in v2 of the same patch series.

> If the bus is generic and can be attached non-NAND devices, I'd
> recommend looking at atmel's binding [2], because you're likely to
> have one instance of the NAND controller logic for all NAND devices
> connected on this bus.
> And more importantly, if the bus a generic, the node should not be
> named nand or nand-controller, and the compatible should not contain
> 'nandc' in it.
> 
> [1]http://elixir.free-electrons.com/linux/latest/source/Documentation/devicetree/bindings/mtd/sunxi-nand.txt#L34
> [2]http://elixir.free-electrons.com/linux/latest/source/Documentation/devicetree/bindings/mtd/atmel-nand.txt#L70

-- 
Abhishek Sahu