Memory leak spotted on i.MX6 with Android 8

cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

Memory leak spotted on i.MX6 with Android 8

3,120 Views
Fribes
Contributor II

Hi,

My company is using Android 8 on i.MX6 to display a GUI on a long running device. The leak is also present on Sabre light board and acknowledged by Boundary Devices. It seems to lurk in binaries provided by NXP.

Could I by any chance draw some attention from NXP to get a fix ?

Details can be found here https://github.com/boundarydevices/linux-imx6/issues/57

Thanks

 

8 Replies

2,970 Views
Fribes
Contributor II

Hi,

Thanks for your support.

Unfortunately we have no Android 8 mobile phone available to try. Does your test on Pixel (Snapdragon ?) implies the leak is not specific to i.MX6 implementation ?

Regards

 

0 Kudos

2,962 Views
Bio_TICFSL
NXP TechSupport
NXP TechSupport

Correct!

 

0 Kudos

2,902 Views
Fribes
Contributor II

Ok, so I assume we are on our own on this leak since it is not NXP binary specific right ?

Any advice on a efficient way to hunt this kind of leak ?

 

Regards

0 Kudos

2,983 Views
Bio_TICFSL
NXP TechSupport
NXP TechSupport

Hello Fribes,

 

To narrow down the issue, we tried with the application and seeing a similar leak on a mobile device. On Pixel mobile we kept it running for 4 hours and found the value of SUnreclaim 182392 kB. Have you tried running the application on a mobile device and observing the same issue?

 

Regards

 

0 Kudos

2,828 Views
RomainNaour
Contributor I

Hello,

I did the same test on a RaspberryPi3 using LineageOS (15.1, Android 8 based, kernel 4.4) and the Le Potato Board using the Android 8.0 prebuilt image (Android 8.0.0, kernel 4.9.61) [1].

I'm unable to reproduce the issue so far, even on Le Potato board using an Android image very close to the image provided by BoundaryDevice for the Sabrelite board.

The test on the Le Potato board has been running for 9h hours.

p212:/ # uptime
 00:50:42 up 50 min,  0 users,  load average: 3.36, 3.16, 3.04

p212:/ # cat /proc/meminfo | grep SUnreclaim: SUnreclaim: 39976 kB
p212:/ # uptime 00:50:53 up 50 min, 0 users, load average: 3.54, 3.20, 3.05 p212:/ # uptime 09:23:19 up 9:23, 0 users, load average: 3.11, 2.98, 2.89
p212:/ # cat /proc/meminfo | grep SUnreclaim: SUnreclaim: 40364 kB
p212:/ # uptime 09:23:23 up 9:23, 0 users, load average: 3.02, 2.96, 2.89

 

I can't verify the test on Pixel mobile is the same issue we have. I still think the issue is related to IMX firmware... please take a look at the debug from BoundaryDevice [2].


[1] https://libre.computer/2018/09/27/android-release-for-tritium-and-le-potato

[2] https://github.com/boundarydevices/linux-imx6/issues/57#issuecomment-679900825

Best regards,

Romain

3,039 Views
kleemann_j
Contributor I

+1 we are facing the same issue

3,080 Views
RomainNaour
Contributor I

Hello,

Here is the link to the AndroidKernelLeak application that is used to reproduce the issue (The code is available):

https://github.com/eove/AndroidKernelLeak/releases

When downloaded, you can install the application using adb:

 

adb install AndroidKernelLeak-1.0.0.apk

 


Then start the application.

You should see a blinking "Hello world" message.


After some time you can notice that "SUnreclaim" value is huge and continue to grow until the system run out of memory:

 

$ cat /proc/meminfo | grep SUnreclaim
SUnreclaim:       163000 kB

 

 

Best regards,

Romain

0 Kudos

3,092 Views
gary_bisson
Senior Contributor III

Hi,

Note that the issue was reproduced on SabreSD using the NXP prebuilt images for o8.0.0_1.0.0 release.

https://www.nxp.com/webapp/Download?colCode=IMX_O8.0.0_1.0.0_ANDROID_DEMO_SD&appType=license

So this is not related to any changes made by Boundary Devices.

Regards,

Gary