[PATCH 03/15] arm64: dts: rockchip: add status LED to opi5max

[Jimmy Hon]

On Sat, Oct 26, 2024 at 12:59 PM Krzysztof Kozlowski <krzk at kernel.org> wrote:
>
> On 26/10/2024 11:48, Jimmy Hon wrote:
> > Describe the Orange Pi 5 Max 2 status LEDs in its device tree.
> >
> > Signed-off-by: Jimmy Hon <honyuenkwun at gmail.com>
> > ---
> >  .../dts/rockchip/rk3588-orangepi-5-max.dts    | 35 +++++++++++++++++++
> >  1 file changed, 35 insertions(+)
> >
> > diff --git a/arch/arm64/boot/dts/rockchip/rk3588-orangepi-5-max.dts b/arch/arm64/boot/dts/rockchip/rk3588-orangepi-5-max.dts
> > index 53a34cb37487..83a118e52bb0 100644
> > --- a/arch/arm64/boot/dts/rockchip/rk3588-orangepi-5-max.dts
> > +++ b/arch/arm64/boot/dts/rockchip/rk3588-orangepi-5-max.dts
>
> This is a new board. Why do you add incomplete board in first patch?
That's an artifact of the development process. It's easier for git to
keep track of which nodes in the DTS goes with which function.

I originally started this 6-weeks ago. But the schematic was only
published this week. So I was hoping the Orange Pi 5 Max was a subset
of the Orange Pi 5 Plus (which both use full RK3588). However, I was
wrong, a lot of pins are different. I compared the differences in the
vendor kernel between the two boards [1][2], and applied it to the
mainline Orange Pi 5 Plus dts [3]

So in the process of trimming the nodes down to the ones actually used
by the OPI5Max,I got it down to the first patch and had a working
board using microSD.

For the LED functionality in particular. I originally just used
leds-gpio like the vendor kernel [2]. However, when the schematic came
out [4] page 26, I noticed the pins were called PWM_LED1 and PWM_LED2,
so I did the same thing as the mainline OPI5Plus, and used leds-pwm.
Luckily, PWM4 and PWM5 are not available from the 40-pin GPIO header.
[5] So there's not much downside. On other boards, the PWM used to
drive the LED could have been used to drive a pin on the GPIO header.
But basically, if I was told it would be better to revert back to
using gpio-leds like the vendor kernel, it was simpler to keep track
of the nodes in the same commit.

Side Note: since it's actually an RGB led, and not three individual
leds, it's tempting to use the leds-pwm-multicolor. However, the RED
portion is hardwired on when the power is on and not connected to PWM.

I still have some more functions that I'm working on (i.e. WIFi), so
those commits are not ready. I'm afraid that the Android BCMDHD driver
may need DTS nodes that are different from the DTS nodes the bcrmfmac
driver would use.

Anyways, should I squash the commits together in the new revision? I
was able to test the majority of the functionality, I don't have the
additional hardware to test fan, rtc, and eMMC.

[1] https://github.com/orangepi-xunlong/linux-orangepi/blob/orange-pi-6.1-rk35xx/arch/arm64/boot/dts/rockchip/rk3588-orangepi-5-plus.dts#L18-L37
[2] https://github.com/orangepi-xunlong/linux-orangepi/blob/orange-pi-6.1-rk35xx/arch/arm64/boot/dts/rockchip/rk3588-orangepi-5-max.dts#L394-L413
[3] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/rockchip/rk3588-orangepi-5-plus.dts#n96
[4] http://www.orangepi.org/html/hardWare/computerAndMicrocontrollers/service-and-support/Orange-Pi-5-Max.html
[5] http://www.orangepi.org/img/pi5-max/max-15.png

>
> Best regards,
> Krzysztof
>

Jimmy