1072155 - Constant 40-80% b2g main process CPU load while idling

Status: RESOLVED WORKSFORME

(Firefox OS Graveyard :: Gaia::System, defect)

Description

[:gerard-majax]

Still reproducing this kind of bad behavior, after a while of use on my Xperia ZR, running a JB base system and a gecko/gaia from yesterday.

Logcat non conclusive, apart a lot of Geolocation messages like:
> 09-24 10:05:29.313  1605  1605 E GeckoConsole: geo: Debug: GPS ignored 0

I'm wondering if it may still be related to Settings API.

Not able to profile for now, SIGUSR2 is being handled by libdiag on my device.

Comment 1

[:gerard-majax]

FYI, the logcat message comes from: http://hg.mozilla.org/mozilla-central/annotate/5e704397529b/dom/system/gonk/GonkGPSGeolocationProvider.cpp#l1028 which is triggered by a mozsetting-changed matching the kSettingDebugGpsIgnored name.

Comment 2

[:gerard-majax]

This may be related/triggered by FMD

Comment 3

[Guilherme Gonçalves [:ggp]]

(In reply to Alexandre LISSY :gerard-majax from comment #2)
> This may be related/triggered by FMD

What makes you think that? Was the FMD app running in the background while this happened? If so, does killing it help in any way? And can I see the logcat to look for FMD-related messages?

Comment 4

[:gerard-majax]

(In reply to Guilherme Gonçalves [:ggp] from comment #3)
> (In reply to Alexandre LISSY :gerard-majax from comment #2)
> > This may be related/triggered by FMD
> 
> What makes you think that? Was the FMD app running in the background while
> this happened? If so, does killing it help in any way? And can I see the
> logcat to look for FMD-related messages?

Because I saw logs showing FMD state changes:
> 09-24 13:12:22.977  1186  1186 I GeckoDump: [findmydevice] got wake up request
> 09-24 13:12:22.977  1186  1186 I GeckoDump: [findmydevice] unlocked

And yes, FMD is up:
> alex@portable-alex:~/codaz/Mozilla/b2g/devices/XperiaZR/B2G$ adb shell b2g-ps
> APPLICATION    SEC USER     PID   PPID  VSIZE  RSS     WCHAN    PC         NAME
> b2g              0 root      250   1     255836 91672 ffffffff b6eb467c S /system/b2g/b2g
> (Nuwa)           0 root      494   250   71928  13820 ffffffff b6eb467c S /system/b2g/b2g
> Homescreen       0 u0_a938   938   494   134964 36836 ffffffff b6eb467c S /system/b2g/b2g
> Built-in Keyboa  0 u0_a1063  1063  250   94276  30464 ffffffff b6e6b67c S /system/b2g/plugin-container
> Smart Collectio  0 u0_a1132  1132  250   95784  31944 ffffffff b6f2967c S /system/b2g/plugin-container
> Find My Device   0 u0_a1186  1186  494   83236  21648 ffffffff b6eb467c S /system/b2g/b2g
> (Preallocated a  0 u0_a1380  1380  494   80080  19340 ffffffff b6eb467c S /system/b2g/b2g
> alex@portable-alex:~/codaz/Mozilla/b2g/devices/XperiaZR/B2G$

Comment 5

[Guilherme Gonçalves [:ggp]]

After discussing this on IRC with Alexandre, what we can infer from his logs is that FMD was launched by the lockscreen being unlocked (even though it was disabled, this is bug 1062558), but did nothing else -- it's just sitting in the background and holding no wakelocks.

If geolocation is indeed the root cause of this, the only way we could see FMD being involved is if launching the FMD app caused some faulty geolocation code to run, despite the fact that FMD never used the Geolocation API. I don't know how realistic that is though.

Comment 6

[:gerard-majax]

Attachment: memory-reports

Kyle, this is what I got after roughly one day of use on my Nexus S

Comment 7

[Kyle Machulis [:qdot] [:kmachulis] (INACTIVE)]

Hmm. I'm not seeing anything WAY out of the ordinary I think, but I'm also not quite sure how to parse that log. How'd you get that?

Comment 8

[:gerard-majax]

Attachment: adb shell top -m 10 -d 1 -t

Comment 9

[:gerard-majax]

This is still reproducing, even after applying the WIP patch for bug 1076597.

Comment 10

[:gerard-majax]

Attachment: adb shell cat /proc/meminfo

Comment 11

[:gerard-majax]

Attachment: adb shell b2g-procrank

Comment 12

[:gerard-majax]

I suspect this is because of bug 1089619. Since I hacked it on my Xperia ZR, I could not reproduce this issue.

Comment 13

[:gerard-majax]

(In reply to Alexandre LISSY :gerard-majax from comment #12)
> I suspect this is because of bug 1089619. Since I hacked it on my Xperia ZR,
> I could not reproduce this issue.

Closing since no more manifestation since bug 1089619 landed.