Hello Krzysztof, On 06/09/22 12:33, Krzysztof Kozlowski wrote: > On 06/09/2022 07:02, Siddharth Vadapalli wrote: >>>> >>>> Please let me know if the above description is fine. >>> >>> Hm, but that's a phy node, not address of register... Isn't this a phy >>> node representing the phy of the CPSW MAC ports? >> >> Despite it being a phy node, the phy-gmii-sel driver actually uses this >> node to obtain the address of the CTRLMMR_ENETx_CTRL registers which >> correspond to the CPSW MAC configuration and are therefore unrelated to >> the PHY. Please let me know if my suggested description would be fine. > > Either I miss some more pieces or this is wrong design. The phy node > should not be used to pass some addresses somewhere. It is used to > define a device which will be instantiated (as parent is simple-mfd). If > you use it only to obtain some address, not to describe child device, > then this is wrong property type. Sorry for describing it incompletely, and at some places incorrectly. Yes, you were right when you initially mentioned that the phy node corresponds to the phy of the ethernet MAC ports. I had incorrectly understood the term "phy" there as the external Layer-1 ethernet phy.The phy node corresponds to the phy used by the ethernet MAC, based on the phy-mode configured. The am65-cpsw-nuss driver which is responsible for ethernet MAC, requires the ethernet MAC's phy to be configured by the phy-gmii-sel driver. Thus, the phy node corresponds to an actual phy (ethernet MAC's PHY). I plan on updating the description for the phy pattern property to the following: "The phy node corresponding to the ethernet MAC." Regards, Siddharth.
