[1.1.9.28] heavy battery drain even in stand by

answered 2015-09-24 10:03:28 +0200

spiiroin
13237 ●151 ●224 ●199

  moderator

Reaching the lowest possible standby battery consumption requires that the device manages to enter late suspend and no peripherals stay erroneously powered on. Note that if the device does suspend, it does not matter if some ill behaved application would try to use cpu -> tracking just cpu usage is usually useless.

Failure to suspend makes power consumption jump approximately from 4mA to 13mA (or so and ignoring led, wlan and cellular activity), even if there is no cpu activity. As this also allows misbehaving apps to consume cpu resources, the impact could be higher too -> tracking cpu usage could reveal problems.

Check that autosuspend policy is not disabled - you should see:

mcetool | grep ^Autosu
Autosuspend policy:                  enabled

If it says "early" or "disabled", do:

mcetool --set-suspend-policy=enabled

Check if there are active wakelocks that would block system from entering late suspend:

cat /sys/power/wake_lock

• "mce_display_on" - On Jolla phone this shows up always even if it is not active due to kernel side issues -> ignore
• "mce_cpu_keepalive" - Some application is blocking late suspend via nemo-keepalive/lower level interfaces. Unless there is a valid reason for this (audio playback, ongoing synchronization) it indicates a bug somewhere.
• Anything else that stays listed for extended periods of time is suspicious.

How to identify processes that are potentially misusing cpu-keepalive:

journalctl -u mce | grep keepalive| grep 'long session'
Sep 20 22:16:52 Jolla mce[407]: modules/cpu-keepalive.c: cka_session_renew(): long session active after 180020 ms; id=34/undefined name=:1.197 pid=16244 cmd=/usr/libexec/packagekitd
Sep 20 22:22:03 Jolla mce[407]: modules/cpu-keepalive.c: cka_session_finish(): long session lasted 490707 ms; id=34/undefined name=:1.197 pid=16244 cmd=/usr/libexec/packagekitd

In the above example packagekitd was blocking suspend for fairly long time, but it is a) more or less expected b) it eventually finished -> If there is an unexpected process holding a long keepalive session without finishing it, there is a bug somewhere.

If none of the above reveals possible problems, it is something more exotic - like erroneously powered up peripheral device (the tohd bug left the nfc chip powered on causing IIRC ~30mA power drain).

answered 2015-09-24 08:45:14 +0200

Upp15
2741 ●44 ●61 ●61

http://trellet.net/

Obvious answer: install Lighthouse and see who's using a lot of cpu.

I have 1.1.9.28 and battery consumption is perhaps even slightly better than what it was before update. With light usage easily 48 hours between charging.