Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210728161020.3905-1-michael.riesch@wolfvision.net>
 <20210728161020.3905-3-michael.riesch@wolfvision.net> <YQGaAFvqqc7wXrWD@lunn.ch>
In-Reply-To: <YQGaAFvqqc7wXrWD@lunn.ch>
From:   Peter Geis <pgwipeout@gmail.com>
Date:   Wed, 28 Jul 2021 14:29:50 -0400
Message-ID: <CAMdYzYo8zf0wjtAxTuYQnZQsBtw38prNuAA0j0sBEamcbzZbfA@mail.gmail.com>
Subject: Re: [PATCH 2/2] arm64: dts: rockchip: rk3568-evb1-v10: add ethernet support
To:     Andrew Lunn <andrew@lunn.ch>
Cc:     Michael Riesch <michael.riesch@wolfvision.net>,
        devicetree@vger.kernel.org,
        arm-mail-list <linux-arm-kernel@lists.infradead.org>,
        "open list:ARM/Rockchip SoC..." <linux-rockchip@lists.infradead.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Rob Herring <robh+dt@kernel.org>,
        Heiko Stuebner <heiko@sntech.de>,
        Liang Chen <cl@rock-chips.com>, Simon Xue <xxm@rock-chips.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Jul 28, 2021 at 1:55 PM Andrew Lunn <andrew@lunn.ch> wrote:
>
> On Wed, Jul 28, 2021 at 06:10:20PM +0200, Michael Riesch wrote:
> > +&gmac0 {
> > +     phy-mode = "rgmii";
>
> ...
> > +
> > +     tx_delay = <0x3c>;
> > +     rx_delay = <0x2f>;
>
> Hi Michael
>
> In general, we try to have the PHY introduce the RGMII delays, not the
> MAC. Did you try
>
> phy-mode = "rgmii-id";
>
> and remove these delay values? It is hard for me to say if that will
> work because i've no idea what 0x3c and 0x2f means? Are they
> equivalent to 2ns?

Unfortunately the driver and TRM are both rather non-specific as to
how this works.
The driver sets the tx_delay to 0x30 and rx_delay to 0x10 if these
values are not defined, or sets them both to 0 in case of rgmii_id.

Generally all rockchip boards use this value instead of the rgmii_id,
I imagine because it's more consistent to tune here than the hit or
miss support of the phy drivers.
The usual course of action is to test to find the lowest and highest
working values and take the median value to plug in here.

>
>      Andrew