Thread (1 message) 1 message, 1 author, 2015-11-26

Preprocessor arithmetic in dtsi files (base + offset)

From: mark.rutland@arm.com (Mark Rutland)
Date: 2015-11-26 13:52:54
Also in: linux-devicetree

On Thu, Nov 26, 2015 at 02:16:16PM +0100, Mason wrote:
Hello,

In the device tree for my ARM platform, I have several nodes with
addresses within the SCU block:

	scu: scu at 20000000 {
		compatible = "arm,cortex-a9-scu";
		reg = <0x20000000 0x100>;

	gic: interrupt-controller at 20001000 {
		compatible = "arm,cortex-a9-gic";
		reg = <0x20001000 0x1000>, <0x20000100 0x0100>;

	twd-timer at 20000600 {
		compatible = "arm,cortex-a9-twd-timer";
		reg = <0x20000600 0x10>;

Can I use preprocessor arithmetic to abstract the base address,
as would be done in C?

#define SCU_BASE 0x20000000

	scu: scu at XXX {
		compatible = "arm,cortex-a9-scu";
		reg = <SCU_BASE 0x100>;
	
	gic: interrupt-controller at XXX {
		compatible = "arm,cortex-a9-gic";
		reg = <SCU_BASE+0x1000 0x1000>, <SCU_BASE+0x100 0x0100>;
The pre-processor would only do substitution, not arithmetic. So the '+'
would flow all the way to DTC.

DTC does have some basic integer expression support (see [1]), but it
looks like it can only work at the cell level, so generally it needs to
be avoided (in case #address-cells >1, for example).

So generally I think that such arithmetic should be avoided.
	twd-timer at XXX {
		compatible = "arm,cortex-a9-twd-timer";
		reg = <SCU_BASE+0x600 0x10>;

Are the @XXX important? Can they be removed altogether?
The bit after the @ is called the unit-address.

It's meant to be there (matching the base address of the first reg
entry) to disambiguate nodes and make it simple to figure out where a
node lives in the physical address space.

They should stay.

Thanks,
Mark.

[1] https://git.kernel.org/cgit/linux/kernel/git/jdl/dtc.git/commit/?id=5f0c3b2d6235dec65fff1628a97f45e21680b36d
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help