On 1/14/20 3:36 PM, Marek Vasut wrote: > On 1/14/20 8:20 AM, Christophe ROULLIER wrote: >> On 1/13/20 6:30 PM, Marek Vasut wrote: >>> On 1/13/20 5:14 PM, Christophe ROULLIER wrote: >>> [...] >>>>>> Christophe you can complete my answer if it is not enough clear. >>>>> I would expect that you should describe _all_ the clock which are routed >>>>> into the ethernet IP in the DT and if the driver doesn't need some of >>>>> those clock, the driver should just not enable such clock. >>>> Hi Marek, >>> Hi, >>> >>>> You are right, we will add by default all clocks used by stm32mp157, and >>>> modify driver to decide if we need to enable it or not. >>>> >>>> So ACK to add ETHCK_K in stm32mp157c.dtsi >>> Do you want me to send the kernel patch ? I have it somewhere here too, >>> I can roll it out tomorrow-ish. >> Hi Marek, >> >> I have not changed yet, the kernel and uboot driver code, to manage this >> behavior. I will implement and push new code soon. > The kernel code already does what is needed. You only need similar patch > to the DT, cfr attachment.
Yes, I'm aligned Marek (I've also done the glue driver in kernel :-)) But what I need to do in uboot and kernel driver it is to add by default all clocks in dtsi file and modified driver to manage them in my driver (enable or disable it in function of phy mode, if with or without quartz in PHY etc..) Regards, Christophe.