Re: [PATCH 2/2] dt-bindings: arm: omap: Convert IVA to DT schema
From: Eddie
Date: Thu Jul 02 2026 - 11:18:30 EST
On Thu, Jul 2, 2026 at 10:27 AM, Krzysztof Kozlowski <krzk@xxxxxxxxxx> wrote:
>
> On Tue, Jun 30, 2026 at 01:28:56PM +0000, Eduard Bostina wrote:
> > Convert the Texas Instruments IVA bindings to DT schema.
> >
> > During the conversion, several updates were made to reflect actual hardware
> > usage and resolve dtbs_check warnings:
> > - Added a dsp sub node that references the new ti,omap3-c64.yaml
> > schema to support OMAP3 configurations where the DSP is defined as a
> > child of the IVA node.
> > - Updated the compatible property to allow ti,ivahd as a standalone
> > string.
>
> Heh? Old binding already said that, so you did not implement such
> change. Instead you invented ti,iva without a reason...
I copied ti,iva from the old example and failed to check if it's
actually used anymore.
I will drop it in v2.
> > - ti,hwmods has been made optional.
>
> Why?
I will add the justification to the commit message in v2: the TI hwmod
framework is legacy and is being phased out, so new device tree nodes
should not be required to include it.
> > +properties:
> > + compatible:
> > + oneOf:
> > + - items:
> > + - const: ti,ivahd
> > + - const: ti,iva
>
> From where did you get such list (ti,iva fallback)? ti,iva is completely
> undocumented and completely unused (not existing), so why it did come up
> with it?
As mentioned above, this was a mistake carried over from the old .txt
example. I will remove it.
> > + - enum:
> > + - ti,iva1
> > + - ti,iva2.1
> > + - ti,iva2.2
> > + - ti,ivahd
>
> This is wrong. ti,ivahd cannot be both: compatible and not compatible
> with ti,iva
Will clean this up by removing the items list and just using a single
enum containing the valid strings in v2.
> > +
> > + ti,hwmods:
> > + description: Name of the hwmod associated to the iva
> > + $ref: /schemas/types.yaml#/definitions/string
>
> And here it is a string. Confusing.
I will change the DSP ti,hwmods to string too.
> > + const: iva
> > +
> > + dsp:
> > + type: object
> > + $ref: /schemas/arm/omap/ti,omap3-c64.yaml#
>
> Missing unevaluatedProperties: false.
>
> Actually the other binding could be squashed/folded into this place...
> but split is fine as well.
I will add unevaluatedProperties: false in v2.
I will keep them split for now, but I will also move this IVA binding
to bindings/soc/ti/ in v2 so that both live in the appropriate
directory.
Thanks for the review.
Best regards,
Eduard