Re: [PATCH v3 1/2] dt-bindings: nvmem: add byte-mask property for non-contiguous byte selection

From: Hongmei Xu

Date: Tue Nov 17 2026 - 22:21:09 EST


On Tue, Nov 17, 2026 at 11:29 PM Srinivas Kandagatla <srini@xxxxxxxxxx> wrote:
>
>
>
> On 11/17/26 6:04 AM, Hongmei Xu wrote:
> > Add the byte-mask property to the fixed-cell binding. This allows
> > selecting non-contiguous physical bytes from an NVMEM region to form
> > a logical cell value.
> >
> > Some SoCs (e.g. Qualcomm efuse controllers) store a logical value such
> > as a MAC address across non-contiguous physical bytes, with one or more
> > bytes in between reserved or unused. byte-mask lets the DT describe this
> > directly in the fixed-cell binding without requiring a device-specific
> > provider for a common pattern.
>
> Pl take a look at layouts, which does exactly what you are trying to do.
>
> --srini
>
Thanks for the review. Do you mean there is already a driver in
nvmem/layout that implements the same functionality as byte-mask?
I looked at all the existing layout drivers. onie-tlv and sl28vpd
parse proprietary binary formats at runtime — the cell structure is
unknown until the device is read. fixed-layout just wraps
nvmem_add_cells_from_dt() for standard cells. None of them handle
statically-described non-contiguous byte selection.

Or do you mean I should implement byte-mask as a new layout driver? If
so, the concern with this approach is that byte-mask is
hardware-agnostic — any NVMEM provider with scrambled byte order could
benefit from it, just as any provider can use bits for contiguous bit
selection. A per-platform layout driver would duplicate the same
byte-compaction logic across every platform that needs it, while
implementing it in core (similar to bits) makes it universally
available.

Could you clarify if I'm missing something, or share any concern with
the core approach?

Best Regards,
Hongmei Xu