On 3/26/19 1:08 PM, Marcel Ziswiler wrote:

Hi Hannes

Hi Marcel, thanks for helping me working on this.

On Tue, 2019-03-26 at 12:57 +0100, Hannes Schmelzer wrote:

...

Hi marcel,

Okay. Behaviour is now clear to me. You don't have a phy node within your dts

We never ever had any such anywhere but it still used to work just fine so far.

Before that you didn't operate the FEC with DM i think, so there was no description at all.

...

nor you have MXC_PHY_ADDR set in your board header.

I actually once even tried setting that without success but maybe I did it wrong.

Sorry, the correct name of the #define is 'CONFIG_FEC_MXC_PHYADDR'. Prior converting your board to DM you had here: #define CONFIG_FEC_MXC_PHYADDR 0

...

The implicit fallback mechanism for searching the whole mdio bus for st least one phy has been gone with my patch.

OK, maybe I just missed noticing such behavioural change having been advertised anywhere.

There was no explicit notice, it did happen ;-) I didn't expect that there is somebody who neither defines a phy-adress within dts nor board-header.

...

Could you try describing your phy within dts to get sure that iam right here.

Sure, let me try that.

Did you have success in this?

...

Later on we have to discuss if the 'fallback' should be re implemented.

Yeah, I assume such behavioural change breaks other boards as well.

Mhm, you're right - i will provide a patch to the mailinglist for catching this. But in general, a precise description of the hardware would be the better way.

cheers, Hannes