Re: [Linux-stm32] [PATCH v5 1/2] dt-bindings: arm: stm32: Document st, stm32mp23/25-syscfg subnodes and cells
From: Marek Vasut
Date: Wed Sep 02 2026 - 08:24:14 EST
On 9/2/26 10:55 AM, Fabrice Gasnier wrote:
Hello Fabrice,
@@ -57,10 +68,17 @@ allOf:
- if:
properties:
compatible:
- const: st,stm32mp25-syscfg
+ contains:
+ enum:
+ - st,stm32mp23-syscfg
+ - st,stm32mp25-syscfg
then:
required:
- "#clock-cells"
+ - "#address-cells"
+ - "#size-cells"
+ - phy@2400
+ - phy@2800
Hi Marek,
Is just saw the discussing on V4 with Conor. That's mainly a thought:
I'm not sure USB is required, e.g. mandatory for a product, even if not
so likely (I don't know if this could exist out there).
I assume you have two use cases in mind:
A) A wall-mount HMI device with only display/touchscreen (No USB plug)
B) A RPi form evaluation kit with all kinds of plugs (Yes USB plug)
?
Still, the nodesDT is a hardware description. The PHY IP is not removed from the chip silicon only because the USB signals are not used on that specific product. Therefore, for device type:
could be tweaked (deleted) in a board, to reduce dtb size in such case.
I'd relax, e.g. let these 4 props as optional. Do you see any drawback ?
That can also be relaxed later, if the case is seen.
A) without USB plug - no action is necessary, the PHYs are already disabled by default in SoC DTSI
B) with USB plug - board DT sets status = "okay" on the PHYs and controllers.
As for reducing the DT size, removing DT nodes for hardware that is present in the silicon negatively impacts the ability to apply DTOs outside of the kernel build process, so I would argue against preemptively removing any DT nodes from DTs to reduce size.
One exception is U-Boot, which adds additional generic bootph-* properties to a subset of DT nodes which should be available early on, and also contains a mechanism which reduces DT size for its U-Boot SPL which operates in severely size constrained environment. However, this is a special case, where removal of nodes from the DT is justified, because in that case, it is certain that those nodes would not and could not be used, and would only waste time.
On the contrary, the DTs that are used to boot Linux should be complete, because the user might enable an IP in there by setting status = "okay" after reaching U-Boot shell, and before booting Linux.