mautz | # 471 | 2017-01-23, 11:26 | Report

Sailfish 2.0.5.6. OTA will be hopefully available after release, but for OTA i need the camera working.

I already talked t nh1402, he'll help getting sfdroid running on cm13 base, but same as above, camera first.

rinigus | # 472 | 2017-01-23, 11:45 | Report

That's really low CPU usage. I wonder if the phone is entering CPU sleep? I have seen so low usage only when the sleep is zero(ish).

mautz | # 473 | 2017-01-23, 12:12 | Report

Most of the times it is in c3 state.

rinigus | # 474 | 2017-01-23, 14:16 | Report

C3 is good, but its still awake as far as I understand the states.

What about "CPU sleep %"? When all is good, that should be 90%+ while your phone is in your pocket. For me, its 95% with idle states taking in total less than 1% (i.e. phone is either asleep or doing something while awake to get into the sleep ASAP). Its different hardware + kernel, though.

mautz | # 475 | 2017-01-23, 17:25 | Report

Is there any other way to check the suspend time, in Systemdatascope this never showed anything, it does not work on 12.1port either, or it really never gets to sleep.

/sys/kernel/debugsuspend_stats shows 0, but its the same on 12.1.

---

Last edited by mautz; 2017-01-23 at 17:31.

rinigus | # 476 | 2017-01-23, 18:47 | Report

Yes, you could use mce-tools package provided mcetool:

mcetool --get-suspend-stats
uptime: 522012.239
suspend_time: 406547.645


The collectd plugin reporting CPU sleep is very simple (merged into upstream at https://github.com/collectd/collectd...src/cpusleep.c ) and is based on following two kernel clocks, as mcetool.

I presume that you would get suspend_time = 0 in mcetool output. If not, then run mcetool several times and see how the times are increasing. By following the increase in times, you could easily calculate CPU sleep percentage. Just let the phone "rest" for a bit (10 minutes) between measurements. Please, if it turns out that SystemDataScope/collectd reports sleep wrongly, let me know - github issue at SystemDataScope repo would probably be simplest (or via this thread, I'll follow it).

Usually, absence of CPU sleep means that some wakelock is not released. Maybe dmesg has something to say?

mautz | # 477 | 2017-01-23, 20:49 | Report

Not looking good:

mcetool --get-suspend-stats
uptime: 3253.005
suspend_time: 0.000

The wakelocks who keep the device from getting into sleep mode are from msm_serial_hs_dma. Seems to be bluetooth related. Strange thing is that i don't use bluetooth.

EDIT: After disabling all bluetooth services, the device does enter suspend mode.

Maybe someone can try this also on cm12.1?

---

Last edited by mautz; 2017-01-23 at 21:24.

claustn | # 478 | 2017-01-23, 21:39 | Report

If you say me how to disable bluetooth services i'll try!
btw, i'm satisfied with the battery usage with cm12 and your kernel. But if with cm13 is better i will give a try!

minimec | # 479 | 2017-01-23, 23:44 | Report

Also confirmed for cm12.1 with Sailfish 2.0.5.6!
Result...

mautz | # 480 | 2017-01-24, 05:08 | Report

Thanks minimec.

@rinigus

With bluetooth disabled, suspend works as it should, but the power consumption in idle does not improve.

Btw, it is also shown in systemdatascope now.