kpowersave stuck at battery charging

This is did not happen before; I am not sure right now what caused this (i.e. battery aging or some software change) nor whether this is kernel/HAL/kpowersave issue.

kpowersave is stuck at assuming battery is loading and at 94%. Sysfs displays battery state as Full:

UEVENT[1199264702.345795]
change /devices/LNXSYSTM:00/device:00/PNP0C0A:00/power_supply/BAT1
(power_supply)
ACTION=change DEVPATH=/devices/LNXSYSTM:00/device:00/PNP0C0A:00/power_supply/BAT1 SUBSYSTEM=power_supply
POWER_SUPPLY_NAME=BAT1
POWER_SUPPLY_TYPE=Battery
POWER_SUPPLY_STATUS=Full
POWER_SUPPLY_PRESENT=1
POWER_SUPPLY_TECHNOLOGY=Li-ion POWER_SUPPLY_VOLTAGE_MIN_DESIGN=10800000 POWER_SUPPLY_VOLTAGE_NOW=11340000 POWER_SUPPLY_CURRENT_NOW=0 POWER_SUPPLY_ENERGY_FULL_DESIGN=38880000 POWER_SUPPLY_ENERGY_FULL=37530000 POWER_SUPPLY_ENERGY_NOW=35640000 POWER_SUPPLY_MODEL_NAME=XM2038P04 POWER_SUPPLY_MANUFACTURER=
SEQNUM=6472

HAL did not set (dis-)charging either:

{pts/0}% lshal -l -u /org/freedesktop/Hal/devices/computer_power_supply_0 udi = '/org/freedesktop/Hal/devices/computer_power_supply_0'
battery.charge_level.current = 35640 (0x8b38) (int) battery.charge_level.last_full = 37530 (0x929a) (int) battery.charge_level.percentage = 94 (0x5e) (int) battery.charge_level.rate = 0 (0x0) (int)
battery.current = 0 (0x0) (int)
battery.present = true (bool)
battery.rechargeable.is_charging = false (bool) battery.rechargeable.is_discharging = false (bool) battery.reporting.current = 35640 (0x8b38) (int) battery.reporting.design = 38880 (0x97e0) (int) battery.reporting.last_full = 37530 (0x929a) (int) battery.reporting.technology = 'Li-ion' (string) battery.reporting.unit = 'mWh' (string)
battery.technology = 'lithium-ion' (string)
battery.type = 'primary' (string)
battery.vendor = '' (string)
battery.voltage.current = 11340 (0x2c4c) (int) info.capabilities = {'battery'} (string list)
info.category = 'battery' (string)
info.parent = '/org/freedesktop/Hal/devices/computer' (string) info.product = 'Li-ion' (string)
info.udi = '/org/freedesktop/Hal/devices/computer_power_supply_0' (string) linux.hotplug_type = 2 (0x2) (int)
linux.subsystem = 'power_supply' (string)
linux.sysfs_path
= '/sys/devices/LNXSYSTM:00/device:00/PNP0C0A:00/power_supply/BAT1' (string)

Kernel 2.6.24-rc6, HAL 0.5.10, kpowersave 0.7.3.

I suspect this is kpowersave which ignores is_(dis)charging attribute and just compares energy_full with energy_now. According to ACPI spec, energy_full gives "predicted full charge capacity" - actually it is called Last Full Charge Capacity. At least this is what ACPI battery code implements. As such this is purely informational and cannot be relied upon for deciding when charging is complete. Actually HAL /proc/acpi/battery code did not even use this property.

So

- does ACPI battery code misuse POWER_SUPPLY_PROP_ENERGY_FULL?
- does HAL misuse .../energy_full?
- does kpowersave misuse battery.charge_level.last_full?

Frequent battery charging shortens lifetime of the battery, so some (may be all now) notebook manufacturers do not start charging battery until it discharge to some degree (~90%). It could be your case. Please try to discharge battery to, say, 89% and then check if it charges to the 100%.

(i.e.

kernel/HAL/kpowersave

displays

I thought Li-Ion batteries do not have memory effect. Actually I remember to have read recommendation to avoid deep discharges of Li-Ion battery, it was adviced to charge it as often as possible.

That is exactly the question - how do you compute 100%? As far as I can tell the only possibility is - when battery stops charging. At this point you have to assume battery is fully charged.

I tried to discharge battery (it was around 78%) and plug AC in again. It went on Charging until the same limit after that state changed to Full (well, in case of ACPI battery we really only can state - not (dis-)charging, there is no special Full state flag); kpowersave still believes battery is not fully charged. Main interface shows 84% (no Charging) - tooltip states it is being charged.

POWER_SUPPLY_NAME=BAT1
POWER_SUPPLY_TYPE=Battery
POWER_SUPPLY_STATUS=Full
POWER_SUPPLY_PRESENT=1
POWER_SUPPLY_TECHNOLOGY=Li-ion
POWER_SUPPLY_VOLTAGE_MIN_DESIGN=10800000
POWER_SUPPLY_VOLTAGE_NOW=11340000
POWER_SUPPLY_CURRENT_NOW=0
POWER_SUPPLY_ENERGY_FULL_DESIGN=38880000
POWER_SUPPLY_ENERGY_FULL=37530000
POWER_SUPPLY_ENERGY_NOW=35640000
POWER_SUPPLY_MODEL_NAME=XM2038P04
POWER_SUPPLY_MANUFACTURER=

This would indicate that my battery lost 6% of capacity in new year, but question still remains - how should user tools properly calculate charge level?

IBM has advice to not re-charge battery until it discharged to 95%. It was implemented in their batteries/notebooks.
There is no memory effect in Li-Ion batteries, there is only limited number of charge cycles -- about 1000 times, thus "as often as possible" is advice of the battery seller :)

Last charge would be good reference. Think of the battery as constantly degrading resource.

So, that is the state of your battery. If you buy new one, it will go high again.

Regards,
Alex.