On Mon, Feb 07, 2011 at 01:24:12PM -0600, Rob Herring wrote:

David, Thomas,

On 02/06/2011 06:04 PM, David Gibson wrote:

quoted

On Sun, Feb 06, 2011 at 06:47:28PM +0530, Thomas Abraham wrote:

quoted

This patch adds a basic dts file for Samsung's SMDKV310 machine.

Signed-off-by: Thomas Abraham<redacted>
---
 arch/arm/mach-s5pv310/mach-smdkv310.dts |   38 +++++++++++++++++++++++++++++++
 1 files changed, 38 insertions(+), 0 deletions(-)
 create mode 100755 arch/arm/mach-s5pv310/mach-smdkv310.dts

diff --git a/arch/arm/mach-s5pv310/mach-smdkv310.dts b/arch/arm/mach-s5pv310/mach-smdkv310.dts
new file mode 100755
index 0000000..74d80bf
--- /dev/null
+++ b/arch/arm/mach-s5pv310/mach-smdkv310.dts

@@ -0,0 +1,38 @@
+/dts-v1/;
+
+/ {
+	model = "smdkv310";
+	compatible = "samsung,smdkv310";
+	#address-cells =<1>;
+	#size-cells =<1>;
+
+	memory {
+		device_type = "memory";
+		reg =<0x40000000 0x08000000>;
+	};

Uh.. where are the cpus?

But for ARM, all the details of the cpu are probe-able. So what
would we gain by putting cpu info in the DTS?

You can detect what cpus are present, but you cannot necessarily
detect if all cpus are available (for instance in a multicore AMP
system where some cores run a different operating environment.  It is
also always good practice to have the structure of the device tree
fully populated even if the OS doesn't currently use it because it
provides points that (for instance) cpu affinity or non-probeable
configuration can be hung off of.