Phone reboots after removing and installing scope

[Marcus Tomlinson]

Once I’ve finished running / debugging my scope (i.e. I hit the stop button), the phone returns to the home scope, then after approx. 10s, the phone reboots (I mean, back to the boot screen, not just a unity crash).

[Zoltan Balogh (bzoltan)]

With all good intention to fix this problem I wonder what the SDK has to do with this problem. What happens on the SDK side is not more then removing the scope in a conventional way.

[Benjamin Zeller (zeller-benjamin)]

I can reproduce this behaviour. It does not happen everytime though.
However the SDK is not doing anything special when stopping a scope. It does uninstall the click package and sends
a DBUS message to show the homescreen again.

This must be a problem of a underlying system (scopes registry, unity, ...), a quick look through the logfiles
did not help much. But I wonder what could trigger a full reboot.

[XiaoGuo, Liu (liu-xiao-guo)]

I can reproduce the problem almost 100 percent in the latest SDK release. Please help to fix it. Thanks!

[Marcus Tomlinson (marcustomlinson)]

It seems this issue can be reproduced with the following script without the qtc involved: http://pastebin.ubuntu.com/11761687/

[Marcus Tomlinson (marcustomlinson)]

Looks like the culprit is the last line of the script: pkcon remove "untitled.zeller-benjamin;0.1;armhf;local:click"

[Benjamin Zeller (zeller-benjamin)]

Retargeting to unity-scopes-api. I was able to reproduce the problem with this script: http://pastebin.ubuntu.com/11761687/ as well as uninstalling a scope from the store (imgur scope). It is not directly related to the qtcreator plugin.

[Marcus Tomlinson (marcustomlinson)]

I've attached a .zip containing a scope click package and a install_loop.sh script that installs and uninstalls the click every 15s. After a few installs, the phone reboots. (Tested on mako 234)

[Paweł Stołowski (stolowski)]

Added 'sudo dmesg' for every step performed by Marcus' script. The last thing reported by dmesg before it reboots is apparmor related:

[ 573.731268] type=1400 audit(1435068499.118:64): apparmor="STATUS" operation="profile_replace" profile="unconfined" name="com.ubuntu.clock_clock_3.3.280.1" pid=9132 comm="apparmor_parser"
[ 573.749397] type=1400 audit(1435068499.138:65): apparmor="STATUS" operation="profile_load" profile="unconfined" name="newsagg.marcustomlinson_newsagg_0.1" pid=9132 comm="apparmor_parser"
[ 590.941583] type=1400 audit(1435068516.345:66): apparmor="STATUS" operation="profile_replace" profile="unconfined" name="com.ubuntu.clock_clock_3.3.280.1" pid=9470 comm="apparmor_parser"

[Alejandro J. Cura (alecu)]

I'm running Marcus' script on krillin (wily r119), no crashes after 15 install/uninstall loops.
Seems to be an issue only with mako.

[Marcus Tomlinson (marcustomlinson)]

Hmmm, just tried the same thing with an app click package, and no crash. (zip attached)

[Marcus Tomlinson (marcustomlinson)]

Adding 'sudo dmesg' to the app install loop script now produces:

[ 1823.580344] type=1400 audit(1435070740.169:104): apparmor="STATUS" operation="profile_replace" profile="unconfined" name="com.viclog.hodor_hodor_0.2" pid=22736 comm="apparmor_parser"
[ 1859.959133] type=1400 audit(1435070776.548:105): apparmor="STATUS" operation="profile_replace" profile="unconfined" name="com.viclog.hodor_hodor_0.2" pid=23161 comm="apparmor_parser"
[ 1896.309873] type=1400 audit(1435070812.897:106): apparmor="STATUS" operation="profile_replace" profile="unconfined" name="com.viclog.hodor_hodor_0.2" pid=23571 comm="apparmor_parser"
[ 1932.570364] type=1400 audit(1435070849.156:107): apparmor="STATUS" operation="profile_replace" profile="unconfined" name="com.viclog.hodor_hodor_0.2" pid=24012 comm="apparmor_parser"
[ 1968.891988] type=1400 audit(1435070885.474:108): apparmor="STATUS" operation="profile_replace" profile="unconfined" name="com.viclog.hodor_hodor_0.2" pid=24403 comm="apparmor_parser"
[ 2005.267205] type=1400 audit(1435070921.853:109): apparmor="STATUS" operation="profile_replace" profile="unconfined" name="com.viclog.hodor_hodor_0.2" pid=24849 comm="apparmor_parser"
[ 2041.485609] type=1400 audit(1435070958.072:110): apparmor="STATUS" operation="profile_replace" profile="unconfined" name="com.viclog.hodor_hodor_0.2" pid=25284 comm="apparmor_parser"

Firstly, why is profile_replace called so often for one package. Secondly, looking at the dmesg output from the scope click install loop, why does "com.ubuntu.clock_clock_3.3.280.1" get updated? When installing/uninstalling the app we only see profile_replace calls against "com.viclog.hodor_hodor_0.2"

[Marcus Tomlinson (marcustomlinson)]

Hmmm, ok so it looks every time the app installs, another profile_replace call shows up in dmesg.

So after the first install: 1 profile_replace call is shown.
Then when its uninstalled: still 1 profile_replace call is shown.
Then it is installed again: now 2 profile_replace calls are shown.
etc...

[Marcus Tomlinson (marcustomlinson)]

Looking further into this issue, it actually looks related to this bug: https://bugs.launchpad.net/canonical-devices-system-image/+bug/1470750

This reboot issue seems to occur on the nexus but not the bq. The bug above describes behaviour experienced on the arale and bq when uninstalling a scope, so that leads me to believe the reboot does not occur on arale either.

Basically, upon uninstalling a scope we see the scope registry process CPU usage shoot up. On the nexus, after a short time of eating the CPU like this, the phone reboots. If however, I stop the registry process somewhere between the uninstall and where I'd expect a reboot, the phone does not reboot.

Therefore, I believe if we fix bug #1470750, we fix this one.

[Marcus Tomlinson (marcustomlinson)]

I have just confirmed that this bug is directly related to bug #1470750. The proposed fix for that bug fixes this one. Marking as duplicate.