Re: [PATCH v2 4/5] dt-bindings: soc: ti: omap-iva: Make ti,hwmods optional

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Globally removing 'ti,hwmods' from the 'required' list silently breaks schema validation for legacy OMAP generations, and the commit message justifies this using a fundamental misunderstanding of the driver architecture.
--

commit b2b7440e95e8ab91078767b3f9e9a47ec5186538
Author: Eduard Bostina <egbostina@gmail.com>

dt-bindings: soc: ti: omap-iva: Make ti,hwmods optional

This commit removes "ti,hwmods" from the required list in the ti,iva
binding schema. The commit message indicates this is safe because the
ti-sysc driver does not strictly require it for probing.

> A review of the ti-sysc driver (drivers/bus/ti-sysc.c) shows that the
> property is not strictly required for probing. The driver only reads
> it to populate the 'ddata->legacy_mode' flag. If the property is absent,
> the probe sequence continues normally without error.

[Severity: Medium]
Is this justification accurate?

Looking at sysc_check_one_child() in drivers/bus/ti-sysc.c, the driver
actually warns if it finds this property on a child node like ti,iva:

drivers/bus/ti-sysc.c:sysc_check_one_child() {
    ...
	name = of_get_property(np, "ti,hwmods", NULL);
	if (name && !of_device_is_compatible(np, "ti,sysc"))
		dev_warn(ddata->dev, "really a child ti,hwmods property?");
    ...
}

> diff --git a/Documentation/devicetree/bindings/soc/ti/ti,iva.yaml b/Documentation/devicetree/bindings/soc/ti/ti,iva.yaml
> index 5724b8367a158..799461a20b049 100644
> --- a/Documentation/devicetree/bindings/soc/ti/ti,iva.yaml
> +++ b/Documentation/devicetree/bindings/soc/ti/ti,iva.yaml
> @@ -24,7 +24,6 @@ properties:
>  
>  required:
>    - compatible
> -  - ti,hwmods

[Severity: Medium]
Does removing this requirement globally break schema validation for older OMAP
generations?

Legacy platforms (like OMAP2/OMAP3) rely on arch/arm/mach-omap2/omap_device.c,
which strictly requires this property to initialize the device:

arch/arm/mach-omap2/omap_device.c:omap_device_build_from_dt() {
    ...
	oh_cnt = of_property_count_strings(node, "ti,hwmods");
	if (oh_cnt <= 0) {
		dev_dbg(&pdev->dev, "No 'hwmods' to build omap_device\n");
		return -ENODEV;
	}
    ...
}

If "ti,hwmods" is globally optional, won't invalid device trees for these
legacy platforms silently pass dtbs_check, only to fail at runtime with
-ENODEV?

>  additionalProperties: false

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260706085917.835875-1-egbostina@gmail.com?part=4