On 18/02/2026 12:05, Konrad Dybcio wrote:

On 2/17/26 9:25 PM, Krzysztof Kozlowski wrote:

quoted

On Tue, Feb 17, 2026 at 02:30:31PM +0100, Konrad Dybcio wrote:

quoted

From: Konrad Dybcio <redacted>

The IP Accelerator hardware/firmware owns a sizeable region within the
IMEM, named 'modem-tables', containing various packet processing
configuration data.

It's not actually accessed by the OS, although we have to IOMMU-map it
with the IPA device, so that presumably the firmware can act upon it.

Allow it as a subnode of IMEM.

You do not have compatible, so rely on the node name as ABI, which is
fine in general but... I do not see usage of it in the driver. Why do
you need to define modem-tables child then?

I don't really *need* the node name to be an ABI. However, the current
binding for IMEM only allows a named "pil-reloc@.." subnode (which is
consumed via of_find_compatible_node() in the remoteproc subsystem) so I
figured the intention was to keep the list of allowed subnodes strictly
moderated

If you'd prefer a blanket pattern declaration with say '^[a-z]@[0-9a-z]+$'
with just a reg requirement inside, I'm fine with that too

No, the problem is that you do not use the ABI here at all. Neither
would you use the blanket pattern, so my question stays: why adding ABI
which is not used?

The pil-reloc is being used, as you pointed out.

Best regards,
Krzysztof