Devices lose adb connection after phablet-flash loop

[Francis Ginther]

Devices will eventually lose their adb connection to the host computer after running them through a phablet-flash loop. This has been seen in our jenkins upstream merger infrastructure for a while and I've been able to isolate it and reproduce it on my own device (a maguro) using the attached reflash-loop.sh script.

This is seen at a much higher rate on our maguros and not as much on our makos. We have swapped hosts and swapped hubs, neither change helped with the failure rate. On the maguros, it's usually seen within 10 runs. It my test at home, I was able to reproduce multiple times after 4, 6, and 4 iterations.

adb kill-server/start-server on the host have no effect. The device itself completes flash and is sitting at the Intro screen (waiting for me to try the right edge). The device appears to be fully functional. With the terminal app, I was able to determine that adbd was running on the device (via "ps -ef | grep adbd").

The MTP connection is working, I'm able to browse files on the device.

The device adb connection can be restored by pulling the battery.

[Test script - reflash-loop.sh]
http://paste.ubuntu.com/6379760/

[Sergio Schvezov (sergiusens)]

In any case this is an android-tools bug.

[Sergio Schvezov (sergiusens)]

"The device adb connection can be restored by pulling the battery."

Can you explain what this means?

[Francis Ginther (fginther)]

> "The device adb connection can be restored by pulling the battery."
>
> Can you explain what this means?

The adb connection is lost. The only way I've been able to restore it so that my device is once again shown under "adb devices" is to remove the battery for 10-15 seconds or so, then to power it on.

[Francis Ginther (fginther)]

Restarting adbd on the device when in the failed state results in a partial recovery:
fginther@blackened:~$ adb devices
List of devices attached
???????????? no permissions

To get all the way back I had to also restart the adb server:
fginther@blackened:~$ adb kill-server
fginther@blackened:~$ adb start-server
* daemon not running. starting it now on port 5037 *
* daemon started successfully *
fginther@blackened:~$ adb devices
List of devices attached
01498AC41901F014 device

I collected logs from the device before restarting adbd, via 'sudo pkill -9 adbd':
dmesg.pre, syslog.pre

and after:
dmesg.post, syslog.post

[Alexander Sack (asac)]

replugging USB combined with restarting server doesnt help recover from this state either?

[Francis Ginther (fginther)]

> replugging USB combined with restarting server doesnt help recover from this state either?

I think I tried that but not certain. Can give it another try and report back.

[Sergio Schvezov (sergiusens)]

On Wed, Nov 20, 2013 at 8:12 PM, Alexander Sack <email address hidden> wrote:

> replugging USB combined with restarting server doesnt help recover from
> this state either?
>

That's missing in the bug feedback; yes it does.

[Sergio Schvezov (sergiusens)]

On Wed, Nov 20, 2013 at 8:15 PM, Francis Ginther <
<email address hidden>> wrote:

> > replugging USB combined with restarting server doesnt help recover
> from this state either?
>
> I think I tried that but not certain. Can give it another try and report
> back.
>

I actually do remember :-)
Told you to start ssh from the terminal app and use that to restart it or
restart from that same terminal app ;-)

You added that after that you needed to also restart adbd on the host.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Ursula Junque (ursinha)]

Inheriting status/importance from duplicate