Re: [PATCH] dt-bindings: net: wireless: brcm,bcm4329-fmac: add nvmem-consumer reference

[Rob Herring]

On Sun, Aug 02, 2026 at 10:39:04AM +0200, Arend van Spriel wrote:
> On Mon, 27 Jul 2026 15:53:11 -0700, Rosen Penev <rosenp@xxxxxxxxx> wrote:
> > The brcm,bcm4329-fmac binding has unevaluatedProperties set to false
> > but does not allow nvmem-cells or nvmem-cell-names, which are used by
> > DTS files (e.g. bcm4709-linksys-ea9200.dts) to supply MAC addresses.
> >
> > The driver already supports nvmem-based MAC address retrieval via
> > of_get_mac_address(), so add the nvmem-consumer schema reference to
> > document and allow these properties.
> > ---
> >  .../devicetree/bindings/net/wireless/brcm,bcm4329-fmac.yaml | 1 +
> 
> The problem is real but I think the fix belongs in network-class.yaml rather
> than the brcmfmac binding specifically. ethernet-controller.yaml already
> defines nvmem-cells and nvmem-cell-names for wired NICs, but the wireless
> binding hierarchy (wireless-controller.yaml -> network-class.yaml) is missing
> them. Fixing it in network-class.yaml would cover all network nodes at once
> instead of requiring each wireless driver to add the same reference
> individually.

It is wrong because nvmem-consumer.yaml does not define how many entries 
and what they are. IOW, the valid names in nvmem-cell-names must be 
defined.

However, expecting all network class devices to use the same 
nvmem-cells/nvmem-cell-names will probably not work long term. Some 
device may use it, but not use "mac-address" entry. Unfortunately, 
there's not a way to override a default in json-schema. So we'll 
probably have to just define this in every user rather than centrally. 

So this should just add the definitions to the brcm,bcm4329-fmac 
binding:

nvmem-cells:
  maxItems: 1

nvmem-cell-names:
  const: mac-address

Rob