[PATCH RFC] ARM: dts: add support for Turris Omnia

[tomas.hlavacek at nic.cz]

Hello Uwe and all!

On Sat, Nov 5, 2016 at 9:38 PM, Uwe Kleine-König 
<uwe at kleine-koenig.org> wrote:
> This machine is an open hardware router by cz.nic driven by a
> Marvell Armada 385.
> 
> Signed-off-by: Uwe Kleine-König <uwe at kleine-koenig.org>
> ---
> 
> Hello,
> 
> the following components are working:
> 
>  - WAN port
>  - eMMC

But I not not sure about DDR50 mode. At least with kernel 4.4, that we 
use in production, we had to limit to SDR50 to overcome I/O errors and 
communication instability, if I can remember it correctly. So it might 
need more testing with the current kernel.

> 
>  - UART0
>  - USB

It is worth noting that the USB 2.0 interface of Armada 385 is wired to 
USB pinout of the last (right-most) PCIe connector. The two USB3 
interfaces routed through SERDES 1 and 3 go directly to the external 
USB connectors.

> 
> Still missing is support for the switch. Wireless fails to probe, 
> didn't
> debug this up to now. SFP is untested as is UART1.

Actually SFP is connected to SGMII interface of eth1, which is routed 
through SERDES 5. The SGMII line is shared between the SFP and metallic 
PHY 88E1514. There is a autonomous high-speed switch connected to the 
SFPDET signal from SFP cage. It disconnects the metallic SFP and 
connects SGMII to SFP once the module is connected.

The SFP is also connected to the I2C mux port 4 and to GPIO expander 
for reading/driving SFPDET, LOS, TXFLT, TXDIS signals:

&i2c0 {
        pinctrl-names = "default";
        pinctrl-0 = <&i2c0_pins>;
        status = "okay";
        clock-frequency = <100000>;

        i2cmux at 70 {
                compatible = "nxp,pca9547";
                #address-cells = <1>;
                #size-cells = <0>;
                reg = <0x70>;
                status = "okay";

...

                i2c at 7 {
                        /* SFP+ GPIO expander */
                        #address-cells = <1>;
                        #size-cells = <0>;
                        reg = <7>;

                        sfpgpio: gpio at 71 {
                                compatible = "nxp,pca9538";
                                reg = <0x71>;
                                interrupt-parent = <&gpio1>;
                                interrupts = <14 IRQ_TYPE_LEVEL_LOW>;
                                gpio-controller;
                                #gpio-cells = <2>;
                        };
                };
	};
};

We have our proprietary support hacked onto mvneta driver for 
disconnecting PHY on the fly. It is a bit nasty, so I suggest to ignore 
SFP in this DTS altogether and let's wait till "phylink based SFP 
module support" or something alike hits upstream, so we can base the 
SFP support on solid code; unless somebody has a better idea, of course.

> 
> 
> diff --git a/arch/arm/boot/dts/armada-385-turris-omnia.dts 
> b/arch/arm/boot/dts/armada-385-turris-omnia.dts
> new file mode 100644
> index 000000000000..d3cd8a4d713d
> --- /dev/null
> +++ b/arch/arm/boot/dts/armada-385-turris-omnia.dts
> @@ -0,0 +1,246 @@
...
> +
> +			/* USB part of the eSATA/USB 2.0 port */

This comment is perhaps some error inherited from my development DTS. 
We do not have any eSATA, perhaps PCIe/USB 2.0 slot.

> 
> +			usb at 58000 {
> +				status = "okay";
> +			};
> +
> +
> +&eth0 {
> +	pinctrl-names = "default";
> +	pinctrl-0 = <&ge0_rgmii_pins>;
> +	status = "okay";
> +	phy-mode = "rgmii-id";
> +
> +	fixed-link {
> +		speed = <1000>;
> +		full-duplex;
> +	};
> +};
> +
> +&eth1 {
> +	pinctrl-names = "default";
> +	pinctrl-0 = <&ge1_rgmii_pins>;
> +	status = "okay";
> +	phy-mode = "rgmii-id";
> +
> +	fixed-link {
> +		speed = <1000>;
> +		full-duplex;
> +	};
> +};
> +

Actually eth0 and eth1 (both are RGMII) are connected to the 88E6176 
switch. The problem is that from what I have read so far the switch can 
not operate in DSA mode with two CPU ports. We currently operate the 
switch in "normal mode" with the eth0 and eth1 set to fixed-link 
1000/full and with proprietary driver (derived from OpenWRT switch 
drivers). I would say that these records for eth0 and eth1 are 
therefore redundant, because it does nothing without the switch support 
and it would most likely change once we have DSA driver (using only 
eth0).

Tomas