On Fri, Nov 06, 2026 at 11:06:22AM -0800, E Shattow wrote:
> Hi Conor,
> 
> On 11/6/26 10:54, Conor Dooley wrote:
> > From: Conor Dooley <conor.dooley@microchip.com>
> > 
> > On Fri, 06 Nov 2026 18:40:55 +0000, Owen O'Hehir wrote:
> >> The firmware turns the status power LED on, but the node has no
> >> default-state, so the LED switches off when leds-gpio probes and the
> >> board looks powered down.
> >>
> >> Default the LED to on.
> >>
> >> Tested on a VisionFive 2 v1.3B.
> >>
> >> [...]
> > 
> > Applied to riscv-dt-fixes, thanks!
> > 
> > [1/1] riscv: dts: starfive: jh7110-common: keep the status power LED on
> >       https://git.kernel.org/conor/c/68dbe2550bd4
> > 
> > Thanks,
> > Conor.
> 
> I replied this a moment earlier:
> https://lore.kernel.org/lkml/179fe9e8-322b-4745-8734-d5de5f1e8487@freeshell.de/
> 
> This disagrees with VisionFive 2 Lite schematic; if when this node is
> consumed by a bootloader it will be evident we are not describing the
> hardware correctly, no?

I am not sure if that's actually relevant though? If the visionfive2lite
needs this gpio then it shouldn't be listed in a leds node. Changing the
default status of it doesn't affect the fact that it is wrong to begin
with, wrong before the patch and wrong afterwards. If there are some
boards that include this common.dtsi but don't have these leds, they're
not common and should be moved to the relevant dts files. If you know
what those are, please send a patch doing the move.

> If you want the light on and it wasn't on at reset or due to boot ROM,
> then that is userland stuff and not hardware description... but that is
> your decision to make as maintainer.

It's a power status led, I think it's pretty reasonable for it to
actually represent power status without intervention in userspace, the
alternative I guess could be something like setting the function
property to "power" - but I'm not sure if that does anything for
gpio-leds.

Cheers,
Conor.