On Fri, Nov 28, 2014 at 01:22:29PM +0000, Laszlo Ersek wrote:
> On 11/28/14 13:59, Arnd Bergmann wrote:
> > On Friday 28 November 2014 13:26:44 Laszlo Ersek wrote:
> >> +Example:
> >> +
> >> +/ {
> >> + #size-cells = <0x2>;
> >> + #address-cells = <0x2>;
> >> +
> >> + fw-cfg@9020000 {
> >> + reg = <0x0 0x9020000 0x0 0x2 0x0 0x9020002 0x0 0x1>;
> >> + compatible = "fw-cfg,mmio";
> >> + };
> >> +};
> >>
> >
> > "fw-cfg" is not a valid vendor string. Are you emulating an actual piece
> > of hardware here or is this something that comes from qemu?
>
> It's a made up name.
>
> > How about using "qemu,fwcfg" as the compatible string, and documenting
> > "qemu" as an official vendor string in Documentation/devicetree/bindings/vendor-prefixes.txt
>
> I can do that, certainly, but I'm not sure if that implies other QEMU
> devices should follow suit.
>
> For example, the virtio-mmio transports that qemu advertises have the
> "compatible" property "virtio,mmio". "virtio" is not a vendor prefix
> either (according to the above text file).
True, but it's a documented standard. It's obvious that 'virtio' has
something to do with the virtio standard, while it's not obvious that
'fw-cfg' as a prefix is for a qemu-specific device.
We should probably add 'virtio' as a vendor prefix.
>
> Here's the full list of compatible strings from the generated DTB:
>
> compatible = "arm,cortex-a15";
> compatible = "arm,armv7-timer";
> compatible = "arm,cortex-a15-gic";
> compatible = "arm,pl011", "arm,primecell";
> compatible = "arm,pl031", "arm,primecell";
> compatible = "arm,psci-0.2", "arm,psci";
All of these are emulated real devices or standards from ARM, so that
makes sense.
> compatible = "cfi-flash";
> compatible = "fixed-clock";
These are generic bindings, arguably they should have been 'linux,'
prefixed but it's too late now.
> compatible = "fw-cfg,mmio";
> compatible = "virtio,mmio";
> compatible = "linux,dummy-virt";
>
> According to
> <http://devicetree.org/Device_Tree_Usage#Understanding_the_compatible_Property>,
> "cfi-flash" and "fixed-clock" don't even have the correct format
> (there's no comma separating <manufacturer> from <model>).
>
> > We don't normally list contiguous registers separately. Maybe just round
> > up to one page and make the register property
> >
> > reg = <0x0 0x9020000 0x0 0x1000>;
>
> The registers are not necessarily placed shoulder to shoulder (it's not
> a requirement in the QEMU source).
It might not be at the moment, but if no-one separates them it could be
a requirement for using this binding. Is there any benefit to having
them separated?
Thanks,
Mark.
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
