[RFC PATCH 11/11] arm/versatile: Add device tree support

[Grant Likely]

On Thu, Jun 16, 2011 at 04:20:37PM +0200, Arnd Bergmann wrote:
> On Thursday 16 June 2011, Grant Likely wrote:
> > This patch adds a new versatile platform for when using the device
> > tree.  Add platform and amba devices are discovered and registered by
> > parsing the device tree.  Clocks and initial io mappings are still
> > configured statically.
> 
> Hi Grant,
> 
> Two small questions about the device tree contents:
> 
> > +       i2c at 10002000 {
> > +               #address-cells = <1>;
> > +               #size-cells = <0>;
> > +               compatible = "arm,versatile-i2c";
> > +               reg = <0x10002000 0x1000>;
> > +
> > +               rtc at 68 {
> > +                       compatible = "dallas,ds1338";
> > +                       reg = <0x68>;
> > +               };
> > +       };
> > +
> > +       net at 10010000 {
> > +               compatible = "smsc,lan91c111";
> > +               reg = <0x10010000 0x10000>;
> > +               interrupts = <25>;
> > +       };
> > +
> > +       lcd at 10008000 {
> > +               compatible = "arm,versatile-lcd";
> > +               reg = <0x10008000 0x1000>;
> > +       };
> 
> Why are these devices on the top level, rather than on the AMBA bus or
> the FPGA? From the documentation, it seems that they are implemented
> in the FPGA, which would also match the address layout.

Mostly because I 'faked' this device tree based on what is currently
in the kernel for the Versatile platform.  I need to look at the
documentation and make it reflect reality, including setting up ranges
correctly (as you commented on below).

> 
> > +       amba {
> > +               compatible = "arm,amba-bus";
> > +               #address-cells = <1>;
> > +               #size-cells = <1>;
> > +               ranges;
> > +
> > +               vic: intc at 10140000 {
> > +                       compatible = "arm,versatile-vic", "arm,vic";
> > +                       interrupt-controller;
> > +                       #interrupt-cells = <1>;
> > +                       reg = <0x10140000 0x1000>;
> > +               };
> > +
> 
> Why the empty ranges property? All device registers are based on 0x10000000,
> so I'd expect this to be described like that here.
> 
> 	Arnd