On 10/8/26 21:02, Andrew Lunn wrote:
quoted
Would clarifying that a package node is required for unaligned RTL8224
layouts be sufficient to address the medium finding, while preserving
the fallback for existing aligned layouts?
Probably. This is the AI being a bit picky.
quoted
For the schema-coverage concern, would adding a dedicated RTL8224
package binding with a compatible such as "realtek,rtl8224-package"
be the preferred approach? This could follow the QCA807x package-binding
pattern and provide explicit validation of the package and child PHYs.
A dedicated realtek,rtl8224-package binding probably does make
sense. You need some documentation anyway.
Thanks, Andrew. A dedicated RTL8224 package binding seems a clearer
approach. I would retain the aligned-address fallback for existing
device trees and explicitly document that unaligned layouts require
a package node.