Re: [PATCH v2 2/8] power: supply: core: Add state_of_health power supply property

[Fenglin Wu]

On 6/22/2025 9:17 AM, Sebastian Reichel wrote:
> Hi,
>
> On Thu, Jun 05, 2025 at 09:34:14AM +0300, Dmitry Baryshkov wrote:
> > On Thu, Jun 05, 2025 at 02:08:30PM +0800, Fenglin Wu wrote:
> > > On 6/3/2025 6:35 PM, Dmitry Baryshkov wrote:
> > > > > > > +What:		/sys/class/power_supply/<supply_name>/state_of_health
> > > > > > > +Date:		May 2025
> > > > > > > +Contact:	linux-arm-msm@xxxxxxxxxxxxxxx
> > > > > > > +Description:
> > > > > > > +		Reports battery power supply state of health in percentage.
> > > > > > > +
> > > > > > > +		Access: Read
> > > > > > > +
> > > > > > > +		Valid values: 0 - 100 (percent)
> > > > > > What does it mean that battery has 77% of health?
> > > > > I will update this to explain it better:
> > > > >
> > > > > Reports battery power supply state of health in percentage, indicating that the maximum charge capacity has degraded to that percentage of its original designed capacity.
> > > > Which basically means that we don't need it in the first place, as we
> > > > can read capacity_full and capacity_full_design (or energy_full /
> > > > energy_full_design) and divide one onto another.
> > > Hmm, it is true in general to quantify how the battery performance has
> > > degraded over time. However, estimating and calculating for battery state of
> > > health is much more complicated I think. I am not an expert, but as far as I
> > > know, different battery management systems might have different algorithms
> > > to calculate the battery health and report it in as percentage. For example,
> > > in Qcom battery management firmware, a "soh" parameter is provided as the
> > > battery health percentage based on the real-time calculations from learning
> > > capacity, resistance estimation, etc.
> > Ack, this is more than just full / full_design. Please consider
> > expanding ABI description (though in the vendor-neutral way).
> No, Dmitry was right. It is exactly the same.
>
> Estimating the battery state of health information is indeed tricky
> and complicated. The reason for that is that estimating and
> calculating POWER_SUPPLY_PROP_CHARGE_FULL/POWER_SUPPLY_PROP_ENERGY_FULL
> (i.e. not the *_FULL_DESIGN) is complicated :)
>
> Greetings,
>
> -- Sebastian
Hi Sebastian,

Thanks for the comment.

In the qcom_battmgr driver, both POWER_SUPPLY_PROP_CHARGE_FULL and 
POWER_SUPPLY_PROP_CHARGE_FULL_DESIGN properties are already supported, 
and CHARGE_FULL is used to represent the learned battery capacity 
calculated in Qcom BMS. Additionally, the Qcom BMS calculates a 
comprehensive battery SOH parameter by considering multiple factors, 
such as changes in battery impedance, learned capacity over time, and 
charging/discharging cycles. Such as, for the battery impedance change, 
it is specially important for calculating SOH in scenarios with high 
load or low temperatures, as only part of the CHARGE_FULL capacity may 
be usable due to significant voltage drops, especially in aged batteries.

Hence, we proposed adding "state_of_health" support in power supply ABI 
to expose this parameter provided in Qcom BMS which is different from 
the calculation of charge_full / charge_full_design.

Also, Android battery management code [1] is expecting "state_of_health" 
power supply property and it can use it if it's available.

[1] 
https://android.googlesource.com/platform/system/core/+/refs/heads/main/healthd/BatteryMonitor.cpp#927

Thanks

Fenglin