From mboxrd@z Thu Jan  1 00:00:00 1970
From: jbrunet@baylibre.com (Jerome Brunet)
Date: Thu, 02 Mar 2017 11:22:27 +0100
Subject: [PATCH] ARM64: dts: meson-gxbb-odroidc2: switch ethernet phy
 mode to rgmii-txid
In-Reply-To: <5baa94a2-f45c-2ab3-0085-422436e12a84@gmail.com>
References: <5baa94a2-f45c-2ab3-0085-422436e12a84@gmail.com>
Message-ID: <1488450147.2420.2.camel@baylibre.com>
To: linus-amlogic@lists.infradead.org
List-Id: linus-amlogic.lists.infradead.org

On Fri, 2017-02-10 at 00:09 +0100, Heiner Kallweit wrote:
> Documentation/networking/phy.txt states:
> Whenever possible,use the PHY side RGMII delay.
> 
> The board uses a RTl8211F phy which supports creating the needed
> tx delay internally. Therefore switch to rgmii-txid mode.
> 
> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
> ---
> ?arch/arm64/boot/dts/amlogic/meson-gxbb-odroidc2.dts | 1 +
> ?1 file changed, 1 insertion(+)
> 
> diff --git a/arch/arm64/boot/dts/amlogic/meson-gxbb-odroidc2.dts
> b/arch/arm64/boot/dts/amlogic/meson-gxbb-odroidc2.dts
> index 3877dda6..94896139 100644
> --- a/arch/arm64/boot/dts/amlogic/meson-gxbb-odroidc2.dts
> +++ b/arch/arm64/boot/dts/amlogic/meson-gxbb-odroidc2.dts
> @@ -152,6 +152,7 @@
> ?	pinctrl-0 = <&eth_rgmii_pins>;
> ?	pinctrl-names = "default";
> ?	phy-handle = <&eth_phy0>;
> +	phy-mode = "rgmii-txid";
> ?
> ?	mdio {
> ?		compatible = "snps,dwmac-mdio";

Heiner,

In principle, I agree with this patch but:

* We need the patches [0] [1] from net-next to merged first (or pulled
in kevin's tree). Without this, the proposed change would result in a
4ns Tx delay instead of 2ns. If Martin send another version of patch
[2], then "amlogic,tx-delay-ns = <2>;" should be removed for the
odroidc2 as the delay would be introduced by the PHY (the value would
ignored anyway, but it would be misleading)

* We would be changing a configuration tested by many people for a
barely tested one, without any benefits expected. These 2
configurations are supposed to be equivalent, and I'm willing to test
it again once the prerequisites are merged but I'd prefer if more
people could test this and confirm there is no regression before
merging it.

Cheers

Jerome

[0]: commit b765234e72fc ("net: stmmac: dwmac-meson8b: make the RGMII
TX delay configurable")
[1]: commit d6db61a40ee0 ("net: stmmac: don't set tx delay in RGMII_ID
and RGMII_TXID mode")
[2]: https://marc.info/?l=linux-arm-kernel&m=148754182813497&w=2