On Sun, Jan 21, 2018 at 02:08:40AM +0100, Andrew Lunn wrote:

quoted

I'm not familiar with MDIO bus but an alternative to GeneriSerialBus
would be to follow what SDIO is doing, e.g have the PHY devices listed
below the MDIO controller and use _ADR to describe their "address" on
that bus. You can see how _ADR applies to SDIO bus from ACPI spec.

Hi Mika

SDIO is not a serial bus, well it can be in its simplest form, but
high speed implementations have 4 data lines. So i can understand them
not using GenericSerialBus.

Well, I think SDIO is more of a serial bus pretty much the same way than
SPI but it can support more data lines as needed (in the same way than
SPI). But I'm not an expert in SDIO.

However, that's not the point here :-) SATA (which is definitely a serial
bus) uses the same mechanism and not GenericSerialBus.

MDIO is a serial bus, very similar to SPI, I2C, and UART.

quoted

If you go with the SDIO way then each PHY is described as normal ACPI
device and you can use ACPI _HID/_CID to match the device to the
corresponding driver.

Just some background here. If you have a plain PHY as a device on an
MDIO bus, you don't need to match it to a driver within ACPI.
Registers 2 and 3 contain a vendor and product ID. That is what it
used to match the device to the driver.

What you might need to know is the protocol to talk on the bus. Most
devices use clause 22 protocol. A few devices are clause 45. 22 is the
default in Linux, and you need to indicate if 45 should be used. You
can also indicate 22.

It gets more complex when the device on the bus is not a PHY. It is a
generic bus, you can connect anything to it. Ethernet switches can be
on the bus. They generally cannot be identified using registers 2 and
3. So you do need to match the device to the driver. Most do have ID
registers, so the driver can work out what specific device is on the
bus. However, Marvell moved the ID registers on there newer generation
of devices, so we need to give the driver a hint where to look. So in
device tree, we have two different compatible string.

Broadcom really do use it as a generic bus. They have their USB PHYs
and PCIE PHYs on an MDIO bus. In DT, they have compatible strings to
match the device to the driver, as normal.

OK, thanks for the explanation.

We need to ensure what we define for ACPI has the same level of
flexibility.

Right. So if you need to have some additional "parameters" with the
connection, then I suppose you may want to go with the GenericSerialBus
route. However, looking at the sample device tree description:

        davinci_mdio: ethernet at 5c030000 {
                compatible = "ti,davinci_mdio";
                reg = <0x5c030000 0x1000>;
                #address-cells = <1>;
                #size-cells = <0>;

                reset-gpios = <&gpio2 5 GPIO_ACTIVE_LOW>;
                reset-delay-us = <2>;

                ethphy0: ethernet-phy at 1 {
                        reg = <1>;
                };

                ethphy1: ethernet-phy at 3 {
                        reg = <3>;
                };
        };

would pretty much translate directly to this in ACPI if you don't need
any additional attributes:

	Device (ETH0) {
		Name (_ADR, /* PCI address of the NIC */)

		Device (PHY0) {
			Name (_ADR, 1)
			...
		} 

		Device (PHY1) {
			Name (_ADR, 3)
			...
		} 
	}

which looks pretty simple to me. You can also use _DSM and _DSD here to
pass information (like the protocol number) for the PHY devices to Linux.