[PATCH] ramips: mt7621: add TP-Link EAP235-Wall support

[Sander Vanheule]

Hi Adrian,

Thanks for the review. I'll fix the naming issues in a v2, some extra
comments below.

On Wed, 2021-02-17 at 01:47 +0100, Adrian Schmutzler wrote:
> > Known issues with this device:
> > The MT7613BE radio is currently not well supported by the mt7615
> > driver:
> > - The EEPROM blob is not recognized, so (Tx) power levels aren't
> > useful.
> >   Patch sent to linux-wireless:
> >   https://lore.kernel.org/linux-wireless/20210202085953.9564-1-
> > sander at svanheule.net/
> > - DFS support is incomplete (known issue for MT7613)
> > - Radio stops responding after a while when idling, reboot required
> > to
> >   bring it to life. Reported by Stijn, appears to be introduced by
> >   2021-01-27 bump of mt76.
> 
> Please add the known issues to the preserved part of the commit
> message, I think it's valuable information also for others.

The first issue has been resolved by now, and maybe also the last. I'll
add an updated list to the commit message.

> 
> > +       leds {
> > +               compatible = "gpio-leds";
> > +
> > +               led_status: status {
> > +                       label = "white:status";
> > +                       color = <LED_COLOR_ID_WHITE>;
> > +                       function = LED_FUNCTION_STATUS;
> 
> Note that color/function do not work right now, so label is actually
> required.
> I'm fine with keeping both color/function and label, though.

I was aware that color and function aren't used at the moment. Since
label has been deprecated, I figured I would already include them to
'prepare' the DTS for 5.10.

> > +
> > +       gpio-export {
> > +               compatible = "gpio-export";
> > +
> > +               poe_passthrough {
> > +                       gpio-export,name = "tp-link:poe-
> > passthrough:enable";
> 
> I'd consider to drop the prefix, although we have no policy on this,
> so it's pure matter of taste.
> Actually, I'd drop this led-label-mimic entirely, and just call it
> "poe-passthrough".

Will update.
This was the same label as I used earlier for the EAP225-Wall. Maybe we
should change that one too then?

> 
> 
> state_default is missing?

It appears to work correctly without an explicit state_default. Is
there and advantage (e.g. power saving) to setting unused functions to
'gpio'?

> > +
> > +&pcie0 {
> > +       mt76 at 0,0 {
> 
> wifi at 0,0
> 
> > +               reg = <0x0000 0 0 0 0>;
> > +               mediatek,mtd-eeprom = <&radio 0x0>;
> > +               mtd-mac-address = <&info 0x8>;
> > +       };
> > +};
> > +
> > +&pcie1 {
> > +       mt76 at 0,0 {
> > +               reg = <0x0000 0 0 0 0>;
> > +               mediatek,mtd-eeprom = <&radio 0x8000>;
> > +               ieee80211-freq-limit = <5000000 6000000>;
> > +               mtd-mac-address = <&info 0x8>;
> > +               mtd-mac-address-increment = <1>;
> > +       };
> > +};

Renamed both mt76 at 0,0 nodes to wifi at 0,0.

> > +
> > +&gmac0 {
> > +       mtd-mac-address = <&info 0x8>;
> > +};
> > +
> > +&switch0 {
> > +       ports {
> > +               port at 0 {
> > +                       status = "okay";
> > +                       label = "lan0";
> 
> Is the zero-based indexing founded on some labels? If not, one-based
> would be the common way to do it.

The port on the back of the device is actually labelled 'eth0', so I
chose the lanX naming to reflect that. That's also why the order of the
other port labels is reversed compared to their HW index.

> > +               };
> > +
> > +               port at 1 {
> > +                       status = "okay";
> > +                       label = "lan3";
> > +               };
> > +
> > +               port at 2 {
> > +                       status = "okay";
> > +                       label = "lan2";
> > +               };
> > +
> > +               port at 3 {
> > +                       status = "okay";
> > +                       label = "lan1";
> > +               };
> > +       };
> > +};
> > diff --git a/target/linux/ramips/image/mt7621.mk
> > b/target/linux/ramips/image/mt7621.mk
> > index 203ca1b908..6efda9eb90 100644
> > --- a/target/linux/ramips/image/mt7621.mk
> > +++ b/target/linux/ramips/image/mt7621.mk
> > @@ -1114,6 +1114,18 @@ define Device/totolink_x5000r  endef
> > TARGET_DEVICES += totolink_x5000r
> > 
> > +define Device/tplink_eap235-wall-v1
> 
> dsa-migration needs to be added for new devices as well, so all
> mt7621
> start with compat version 1.1.

I realised this already while trying to sysupgrade the device. I
originally thought as this was named 'migration', it wouldn't be needed
since this device had always been a DSA device.


Best,
Sander