Hi Andrew,

On Wed, 2026-07-08 at 16:35 +0200, Andrew Lunn wrote:

quoted

On Tue, 2026-07-07 at 14:42 +0200, Andrew Lunn wrote:

quoted

quoted

+  - if:
+      properties:
+        compatible:
+          contains:
+            enum:
+              - mediatek,mt8189-gmac
+    then:
+      properties:
+        clocks:
+          items:
+            - description: MAC Main clock
+            - description: PTP clock
+            - description: RMII reference clock provided by
MAC

Since this is a MAC, it sounds like it is consuming its own
clock?

In the driver ([1]), this clock is described as being only used and
needed in RMII when MAC provides the reference clock, and useless
otherwise (RGMII/MII or RMII when PHY provides the reference
clock).

So it sounds like this is a clock output, going to the PHY, as its
reference clock input. So ideally, the PHY should consume this clock,
not the MAC.

quoted

Its use and configuration also depends on the "mediatek,rmii-clk-
from-
mac" vendor property ([2]) presence in devicetree.

This makes it sounds like it is historically wrong, and the patch is
just extending this to the new device.

Do you have a board using RMII? Can you list the clock in the PHY
node, not the MAC, and see if it still works?

I don't. 
All the MTK Genio boards (510, 520, 720, 1200 EVK) I have are using
RGMII only.

Ideally, for a new device, we should not repeat past errors.

So, do you want I remove this clock from mt8189 clocks list for v2?

Regards,
Louis-Alexis

	 Andrew