Closed Bug 1154629 Opened 7 years ago Closed 7 years ago

[MTBF][Nexus5-L] adb crashed after running MTBF for a while

Categories

(Firefox OS Graveyard :: Stability, defect)

ARM
Gonk (Firefox OS)
defect
Not set
normal

Tracking

(Not tracked)

RESOLVED WORKSFORME

People

(Reporter: ashiue, Unassigned)

References

Details

2 run of Nexus 5 L have ever encountered adb daemon crashed issue when running MTBF a while. (4 hours and 9 hours for each device)

Version info:
application_buildid: 20150413040332
application_display_name: B2G
application_id: {3c2e2abc-06d4-11e1-ac3b-374f68613e61}
application_name: B2G
application_remotingname: b2g
application_vendor: Mozilla
application_version: 37.0
device_firmware_date: 1428868011
device_firmware_version_incremental: eng..20150413.034606
device_firmware_version_release: 5.1
device_id: hammerhead
gaia_changeset: cec00d643f517ffd96cde559cd3bbd43ab85816c
gaia_date: 1428702072
platform_buildid: 20150413040332
platform_version: 37.0
How often does it happen? only once or always happen on the same device?
Flags: needinfo?(ashiue)
Until now, hit rate is 2/18, and this issue happened at different devices.
Flags: needinfo?(ashiue)
Hit this issue again.
We found that before adb crashed, devices all tried to execute test_mtbf_camera_multiple_shots.py.

Here are logs for refer: https://drive.google.com/a/mozilla.com/folderview?id=0B_lQ-lNDWe2ofnNHZFBuNDZMc045VHRTVEFQU0NwX3pMS0UyaW9IdHZ3QUhHcDl5ZW5BQXc&usp=sharing
blocking-b2g: --- → 2.2?
QA Whiteboard: [COM=MTBF-Nexus5L]
Summary: [MTBF][Nexus5-L] adb daemon crashed after running MTBF for a while → [MTBF][Nexus5-L] adb crashed after running MTBF for a while
we met this few times in Nexus 5 and thanks for Alison's help to pass the failed device to me and see if we can have more analysis. 

All fail symptoms are same: adb crashed but gecko still work. we can still take picture or play music on those devices. Here are the analysis we can have:
1. there's no device shows on lsusb
2. I try to unplug/plug the USB cable and looks like USB controller didn't reset.
3. disable/enable adb can not recover this fail symptom.

I also check the log in comment 3 but there's no log when issue happened.
Two proposals I have for this bug:
1. remove the test for test_mtbf_camera_multiple_shots.py and see if we can keep MTBF running.
2. keep running test_mtbf_camera_multiple_shots.py test case and see what we get in kernel log. keep recording kernel log by "adb shell cat /proc/kmsg" should be fine.
Blocks: MTBF-B2G
Hi Viral,
Could you help to check whether log per comment 5 helps?
Thanks!
Flags: needinfo?(vwang)
so far we met several times about adb not working case.
All failed devices can not recovery after several times plug/unplug usb cable.

It looks like the failed symptom is that the usb interface is crashed.
The best way to check this kind of issue should be using usb analyzer to check the usb signal, but it more likely the low level issue that we can not handle.

The only thing we can try to do now is try to avoid this kind of failed happened again.
Since the failed case all stop in camera, I would like to suggest to use "adb wait-for-device" to make sure all commands can be run after adb is online.
Flags: needinfo?(vwang)
Hi Alison,
Can you try the suggestion from comment 8?
Thanks!
Flags: needinfo?(ashiue)
Yes, we already added "adb wait-for-device" in framework last Friday. 
We are wait and see if this issue still occur or not.
Flags: needinfo?(ashiue)
Close this bug since until now we do not hit this issue.
We would renominate if we observe the issue again.
Status: NEW → RESOLVED
blocking-b2g: 2.2? → ---
Closed: 7 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.