Re: [PATCH 13/15] ARM: dts: sun8i: Add dt node for rtl8703as wifi chip on ga10h

On Mon, Jul 11, 2016 at 02:36:43PM +0200, Hans de Goede wrote:

quoted

quoted

+&mmc1 {
+	pinctrl-names = "default";
+	pinctrl-0 = <&mmc1_pins_a>;
+	vmmc-supply = <&reg_dldo1>;
+	bus-width = <4>;
+	non-removable;
+	status = "okay";
+
+	rtl8703as: sdio_wifi@1 {

Shouldn't that need a compatible ?

There is no upstream driver yet, so no devicetree binding for the specific
wifi chip used. But all we really need is to follow generic sdio / network
bindings:

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/mmc/mmc-card.txt

Describes how to create a mmc-host child-node describing an (e)mmc-card or
sdio-function, in this case we are adding a node for sdio-function 1,
hence the reg = <1>; in the child-node.

Both that file and
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/mmc/mmc.txt
list the compatible property as mandatory.

Linux implementation wise, this will make the mmc core set the dev->of_node
ptr to this node for the struct device it passes into the sdio-drivers
probe function.

And u-boot will add a local-mac-address property as documented here:
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/net/ethernet.txt

to the node pointed to by ethernet%d aliases.

To actually make use of this I've modified the out of tree driver to check
for the "local-mac-address" property:

https://github.com/jwrdegoede/rtl8723bs/commit/be5e87ea71677120ede457213208902ed9942653

So that each device actually gets a unique address rather then the
(stupid) behavior of the out of tree driver to set the same fixes
mac address on every device which does not have a mac address in
eeprom / nvram.

Since this follows the existing ethernet bindings standard I
believe the dts bits can go upstream as is, but since there
is no upstream-able driver yet, I do not believe that making up
a compatible string to add to the child-node is a good idea.

More-over on q8 tablets we will want to do the same (again for
eeprom / nvram less boards) but without pinning ourselves to
a specific sdio-wifi module / compatible, instead relying on
sdio enumeration to figure out which sdio-wifi module is present.

Feel free to raise that issue to the DT maintainers. But this patch
breaks the existing bindings, so it cannot be merged as is.

Maxime

-- 
Maxime Ripard, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com