Re: [PATCH 2/2] power: supply: leds: add a not-charging trigger

Sebastian Reichel <[email protected]> Sat, 25 Jul 2026 04:40:09 +0200
Newsgroups	org.kernel.vger.linux-pm,org.kernel.vger.linux-kernel
Message-ID	<amQWViWt8ovRisp0@venus>
Hi,

On Tue, Jul 21, 2026 at 08:56:47AM +0200, Steffen Dirkwinkel wrote:
> On Tue, 2026-07-21 at 00:23 +0200, Sebastian Reichel wrote:
> > On Thu, Jun 25, 2026 at 07:07:35PM +0200, Steffen Dirkwinkel wrote:
> > > From: Steffen Dirkwinkel <[email protected]>
> > > 
> > > We intend to use this with a gpio-charger ups device that reports
> > > charging or not-charging based on a gpio to set a "power failure" led.
> > > 
> > > Signed-off-by: Steffen Dirkwinkel <[email protected]>
> > > ---
> > >  drivers/power/supply/power_supply_leds.c | 7 +++++++
> > >  include/linux/power_supply.h             | 1 +
> > >  2 files changed, 8 insertions(+)
> > 
> > POWER_SUPPLY_STATUS_NOT_CHARGING means, that a battery is neither
> > charged **nor discharged**. I think the NOT_CHARGING status has a
> > bad name, but it's ABI and cannot be changed easily. But I certainly
> > don't want it to spread further. Let's find a better name for this
> > trigger. Maybe idle?
> 
> Hm, I guess I was also mislead by the name, as it's somewhat correct for my
> usecase, our gpio connected ups is indicating "not-charging" via a gpio and I'm
> using it as a gpio-charger charge-status-gpio.
> 
> I guess it would be more correct if I could map the gpio-charger charge status
> to switch between charging and discharging instead of indicating not-charging.
> If you agree I'd add that as an option via device tree binding there and add a
> discharging trigger instead of the not charging one to the led triggers:
> 
>   charge-status-discharging:
>       type: boolean
>       description:
>         Interpret a deasserted charge-status-gpio as "discharging" instead of
> 	"not charging".

For a normal charger your gpio would be the ONLINE state effectively
(i.e. "gpios" in the binding). For most devices that's basically the
only thing a charger reports.

Looking at UPS support in the mainline kernel, it does not look like
anyone used it so far (not just with gpio-charger; I cannot finy *any*
user for POWER_SUPPLY_TYPE_UPS). I do see some issues with some of
the power-supply framework's properties as a UPS is basically a
combination of a charger and a battery. I wonder if it's better to
simply expose them separately to avoid these problems.

Can you share a bit more details about your platform? Would it be
possible to expose your UPS like this?

ups_charger: charger {
    compatible = "gpio-charger";
    charger-type = "mains";
    gpios = <your_status_gpio>;
};

ups_battery: battery {
    compatible = "gpio-battery";
    power-supplies = <&ups_charger>;

    /*
     * does not yet exist, any GPIOs are optional. In the simple
     * case it creates a TYPE_BATTERY device, which sets its own
     * status based on the ONLINE state of the charger it is being
     * supplied from. Potentially could have a gpio to notify
     * critical battery status.
     */
};

Greetings,

-- Sebastian
signature.asc (application/pgp-signature, 833 B)