Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752520AbdHIWmS (ORCPT ); Wed, 9 Aug 2017 18:42:18 -0400 Received: from mail-pf0-f196.google.com ([209.85.192.196]:38572 "EHLO mail-pf0-f196.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752403AbdHIWmO (ORCPT ); Wed, 9 Aug 2017 18:42:14 -0400 Date: Wed, 09 Aug 2017 15:42:05 -0700 User-Agent: K-9 Mail for Android In-Reply-To: <20170809084541.GB7137@Red> References: <1501654546-17292-1-git-send-email-david.wu@rock-chips.com> <1501654888-19514-1-git-send-email-david.wu@rock-chips.com> <6b1ebf64-6903-6d35-b1fc-92ec47334425@gmail.com> <20170809084541.GB7137@Red> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Subject: Re: [PATCH v3 05/11] net: stmmac: dwmac-rk: Add internal phy support To: Corentin Labbe , andrew@lunn.ch, davem@davemloft.net CC: David Wu , =?ISO-8859-1?Q?Heiko_St=FCbner?= , Rob Herring , Mark Rutland , Catalin Marinas , Will Deacon , Olof Johansson , Russell King , Arnd Bergmann , Tao Huang , hwg@rock-chips.com, alexandre.torgue@st.com, devicetree , netdev , linux-kernel , "open list:ARM/Rockchip SoC..." , Giuseppe Cavallaro , linux-arm-kernel , wens@csie.org From: Florian Fainelli Message-ID: <77E3C546-E3A8-4EE5-8505-2C4B2252E08C@gmail.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by nfs id v79MgPJV026520 Content-Length: 4920 Lines: 114 On August 9, 2017 1:45:41 AM PDT, Corentin Labbe wrote: >On Thu, Aug 03, 2017 at 07:06:33PM +0800, Chen-Yu Tsai wrote: >> On Thu, Aug 3, 2017 at 1:38 AM, Florian Fainelli > wrote: >> > On 08/01/2017 11:21 PM, David Wu wrote: >> >> To make internal phy work, need to configure the phy_clock, >> >> phy cru_reset and related registers. >> >> >> >> Signed-off-by: David Wu >> >> --- >> >> .../devicetree/bindings/net/rockchip-dwmac.txt | 6 +- >> >> drivers/net/ethernet/stmicro/stmmac/dwmac-rk.c | 81 >++++++++++++++++++++++ >> >> 2 files changed, 86 insertions(+), 1 deletion(-) >> >> >> >> diff --git >a/Documentation/devicetree/bindings/net/rockchip-dwmac.txt >b/Documentation/devicetree/bindings/net/rockchip-dwmac.txt >> >> index 8f42755..ec39b31 100644 >> >> --- a/Documentation/devicetree/bindings/net/rockchip-dwmac.txt >> >> +++ b/Documentation/devicetree/bindings/net/rockchip-dwmac.txt >> >> @@ -25,7 +25,8 @@ Required properties: >> >> - clock-names: One name for each entry in the clocks property. >> >> - phy-mode: See ethernet.txt file in the same directory. >> >> - pinctrl-names: Names corresponding to the numbered pinctrl >states. >> >> - - pinctrl-0: pin-control mode. can be <&rgmii_pins> or ><&rmii_pins>. >> >> + - pinctrl-0: pin-control mode. can be <&rgmii_pins>, ><&rmii_pins> or led pins >> >> + for internal phy mode. >> >> - clock_in_out: For RGMII, it must be "input", means main >clock(125MHz) >> >> is not sourced from SoC's PLL, but input from PHY; For RMII, >"input" means >> >> PHY provides the reference clock(50MHz), "output" means GMAC >provides the >> >> @@ -40,6 +41,9 @@ Optional properties: >> >> - tx_delay: Delay value for TXD timing. Range value is 0~0x7F, >0x30 as default. >> >> - rx_delay: Delay value for RXD timing. Range value is 0~0x7F, >0x10 as default. >> >> - phy-supply: phandle to a regulator if the PHY needs one >> >> + - clocks: <&cru MAC_PHY>: Clock selector for internal macphy >> >> + - phy-is-internal: A boolean property allows us to know that MAC >will connect to >> >> + internal phy. >> > >> > This is incorrect in two ways: >> > >> > - this is a property of the PHY, so it should be documented as such >in >> > Documentation/devicetree/bindings/net/phy.txt so other bindings can >> > re-use it >> > >> > - if it was specific to your MAC you would expect a vendor prefix >to >> > this property, which is not there >> > >> > An alternative way to implement the external/internal logic >selection >> > would be mandate that your Ethernet PHY node have a compatible >string >> > like this: >> > >> > phy@0 { >> > compatible = "ethernet-phy-id1234.d400", >"ethernet-phy-802.3-c22"; >> > }; >> > >> > Then you don't need this phy-is-internal property, because you can >> > locate the PHY node by the phy-handle (see more about that in a >reply to >> > patch 10) and you can determine ahead of time whether this PHY is >> > internal or not based on its compatible string. >> >> This doesn't really work for us (sunxi). The "internal" PHY of the H3 >> is also available in the X-Powers AC200 combo chip, which would be >> external. Same ID. So if someone were to be stupid and put the two >> together, you wouldn't know which one it actually is. Generically >> it doesn't make sense to match against the ID either. The PHY is >> only integrated or inlined into the SoC, meaning it could be moved >> elsewhere or even be a discreet part. It actually makes a lot of sense to differentiate on the PHY ID because you are supposed to allocate an unique ID based on how the integration of the PHY is done. Not doing that is making a sloppy job at integrating HW blocks, but such is life and there is no shortage of creativity amongst HW engineers when they are not given feedback from SW people. >> >> The way I see it is more like a reversed pinmux. The system should >> select either the internal set or external set of xMII pins to use. >> >> A phy-is-internal property in the PHY node would work for us. We >> already need a PHY node to describe its clocks and resets. >> >> A side note to this solution is that, since the internal PHY is >defined >> at the .dtsi level, any external PHYs at the same address would need >to >> make sure to delete the property, which is kind of counterintuitive, >but >> it is how device tree works. One would want to put the internal PHY's >> address, assuming it is configurable, on something that is rarely >used. >> > >Hello David, Florian, Andrew > >Could someone ack this ? or nack if you think that the chance for >having two PHY id both internal and external is too low. >Anyway, we need a choice. Let's move forward with the 'phy-is-integrated' Boolean property ('phy-is-internal' is too close from the proprietary "internal" phy-mode IMHO). Andrew, does that also work for you? -- Florian