[PATCH 04/14] ARM: dts: armada-375: Fixup bootrom DT warning

Hi Thomas,
 
 On jeu., nov. 10 2016, Thomas Petazzoni [off-list ref] wrote:

[...]

quoted

quoted

A good example of why I'm worried is the sa-sram case:

+		crypto_sram0: sa-sram0 at 0 {
 			compatible = "mmio-sram";
 			reg = <MBUS_ID(0x09, 0x09) 0 0x800>;

+		crypto_sram1: sa-sram1 at 0 {
 			compatible = "mmio-sram";
 			reg = <MBUS_ID(0x09, 0x05) 0 0x800>;

The node names should be just "sram" without a number. Indeed for UARTs
for example, you use uart at XYZ, uart at ABC and not uart0 at XYZ and
uart1 at ABC. But then, if you do that, with your scheme, you end up with
both nodes named sa-sram at 0.

Which clearly shows that the way you set this unit-address is not
correct: those two devices are mapped at completely different
locations, but you end up with an identical unit address.

I have no idea what is the rule for setting the unit address in this
case, but I'm pretty sure the rule you've chosen is not good.  

I don't know if there is an existing rules for this case. But I see your
concern. What I propose then is to expose the memory windows ID by
adding the target and the attributes like this:

		crypto_sram0: sa-sram at 09_09_0 {
 			compatible = "mmio-sram";
  			reg = <MBUS_ID(0x09, 0x09) 0 0x800>;


		crypto_sram1: sa-sram at 09_05_0 {
  			compatible = "mmio-sram";
  			reg = <MBUS_ID(0x09, 0x05) 0 0x800>;

I have no idea if 09_05_0 is considered a valid unit address. Indeed
the _ character in an address looks a bit weird.

I guess we need guidance from the DT binding maintainers on this, since
it's really a matter of interpreting what "unit address" means in
relation to the value of the "reg" property.

So I looked for in the reference: Power_ePAPR_APPROVED_v1.0, and in
paragraph "2.2.1.1 Node Name Requirements" we have:

"The unit-address component of the name is specific to the bus type on
which the node sits. It consists of one or more ASCII characters from
the set of characters in Table 2-1. The fundamental requirement is that
at any level of the device tree the unit-address be unique in order to
differentiate nodes with the same name at the same level in the
tree. The binding for a particular bus may specify additional, more
specific requirements for the format of a unit-address."

In Table 2-1 we have the following characters:
0-9
a-z
A-z
,
.
_
+
-

So the underscore is valid. And by adding information about the memory
windows we match the fundamental requirement :"at any level of the
device tree the unit-address be unique".

Gregory

-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

-- 
Gregory Clement, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com