Re: [RFC next v1 0/5] stmmac: honor the GPIO flags for the PHY reset GPIO

From: Martin Blumenstingl
Date: Mon Jun 10 2019 - 11:57:27 EST

Hi Andrew,

On Mon, Jun 10, 2019 at 3:25 PM Andrew Lunn <andrew@xxxxxxx> wrote:
> > if anyone else (just like me) doesn't know about it, there are generic
> > bindings defined here: [0]
> >
> > I just tested this on my X96 Max by defining the following properties
> > inside the PHY node:
> > reset-delay-us = <10000>;
> > reset-assert-us = <10000>;
> > reset-deassert-us = <10000>;
> > reset-gpios = <&gpio GPIOZ_15 (GPIO_ACTIVE_LOW | GPIO_OPEN_DRAIN)>;
> >
> > that means I don't need any stmmac patches which seems nice.
> > instead I can submit a patch to mark the snps,reset-gpio properties in
> > the dt-bindings deprecated (and refer to the generic bindings instead)
> > what do you think?
> Hi Martin
> I know Linus wants to replace all users of old GPIO numbers with gpio
> descriptors. So your patches have value, even if you don't need them.
OK, then I will send my patches anyways

> One other things to watch out for. We have generic code at two
> levels. Either the GPIO is per PHY, and the properties should be in
> the PHY node, or the reset is for all PHYs of an MDIO bus, and then
> the properties should be in the MDIO node.
our Amlogic boards only have one PHY and all schematics I'm aware of
route the SoC's GPIO line directly to the PHY's reset line.
so in my opinion defining the resets for the PHY is the right thing to do