Post-Mortem on Panic/Reboot

[Doug]

Hi, I am an Android app developer, all of my work is within the Java
runtime environment. Before you point me to the android-developers
list, please consider the following:

I have had a number of users of my app report that with it installed
on their rooted 2.2 devices (primarily the Verizon Droid) it causes
the phone to reboot anywhere from every 15 minutes to every hour.

I realize that the 2.2 codebase is not released yet (don't even know
when), but the last thing I want is to be stung by 1,000's of users
who suddenly encounter this problem after the OS is released into the
wild.

As such, I am doing my best to determine the cause of this problem.
Users have reported that the phone will spontaneously reboot at pretty
much any time (on the phone, off the phone, browsing the web, idle,
asleep) with the one exception being that it seems to be stable when
it is "plugged in" (Most of this info has been gleaned from various
forum postings, so accuracy cannot be 100% confirmed)

Is there any way to access the Android equivalent of /etc/messages (or
some other crash-dump file) AFTER the phone has rebooted? (given that
the users of these devices have root access)

I *suspect* that the problem may be related to the network stack (the
app tries to poll a server for changes periodically), and I think the
problem may be somewhere in the handling of 'bad/unavailable' IP
connections.

Doug

[Dianne Hackborn]

System crashes are a bug in the platform, not in the application.  It should be up to the device manufacturer to fix such a problem...  and if these people are running some random build of who knows what, there is obviously nobody who can help them with crashes they are experiencing on their device, so they should wait until there is a real build to run to report any such errors to the manufacturer.

--
unsubscribe: android-kerne...@googlegroups.com
website: http://groups.google.com/group/android-kernel

--
Dianne Hackborn
Android framework engineer
hac...@android.com

Note: please don't send private questions to me, as I don't have time to provide private support, and so won't reply to such e-mails.  All such questions should be posted on public forums, where I and others can see and answer them.

[Mike Chan]

Do you know if its a kernel reboot or a framework restart?

If its the kernel crashing you can get a persistent logs found:

For Nexus One:
You can check /proc/last_kmsg

On Droid:
/data/dontpanic/apanic_console

Make sure to check the timestamp of the file created or in the dmesg,
sometimes if the reboot is due to the modem there will be no logs.

What build is this 2.2 Droid running?

-- Mike

[Doug]

first part for Dianne, second for Mike

Dianne - I completely agree that this is theoretically not an issue
that I should be concerned with, however, as the developer of an app
that seems to be triggering this bad behavior it is up to me to try to
determine whether or not anything can be done about it. I can point
my finger in any direction (Motorola, Verizon, even Google/AOSP) but
as you know users don't know (or even care) where the problem lies,
all they see is my app misbehaving and thats the end of it. I'd
rather do a little research to help identify/resolve the problem now
than have to deal with the user rants later.

Mike - I went to the trouble of installing what I *believe* is the
broken build on one of my test phones (a Droid) - so far I *think* the
phone has misbehaved twice, but I didn't have logcat running, so don't
know what it was up to at the time (whether it will misbehave
connected via usb is another question...).

"Build Number" = "voles-userdebug 2.2 FRFxx 33333 test-keys"

I did look in /data/dontpanic - the cupboard is bare

Doug

On Jun 14, 2:38 pm, Mike Chan <m...@android.com> wrote:
> Do you know if its a kernel reboot or a framework restart?
>
> If its the kernel crashing you can get a persistent logs found:
>
> For Nexus One:
> You can check /proc/last_kmsg
>
> On Droid:
> /data/dontpanic/apanic_console
>
> Make sure to check the timestamp of the file created or in the dmesg,
> sometimes if the reboot is due to the modem there will be no logs.
>
> What build is this 2.2 Droid running?
>
> -- Mike
>
>
>
> On Mon, Jun 14, 2010 at 2:06 PM, Dianne Hackborn <hack...@android.com> wrote:
> > System crashes are a bug in the platform, not in the application.  It should
> > be up to the device manufacturer to fix such a problem...  and if these
> > people are running some random build of who knows what, there is obviously
> > nobody who can help them with crashes they are experiencing on their device,
> > so they should wait until there is a real build to run to report any such
> > errors to the manufacturer.
>

> > hack...@android.com

[Mike Chan]

Were you able to determine if it was a kernel crash or a framework
restart? I believe you can verify with the uptime of the device.

Also what kernel came with your build? Are you on 2.6.32 for Droid?

-- Mike

[Doug Henderson]

Umm, ok

Kernel is 2.6.32.9-_____

Uptime per the settings -> About -> Status does seem to have been reset

I made a copy of the output from dmesg and noticed that the timestamps indicate it was a kernel restart about 2.5 hrs ago... 

<6>[    0.000000] boot_mode=kpanic

If that means what I think it might...

If you are interested in looking deeper into logs, etc then feel free to contact me direct - people probably don't want to see screeds of logging go by for a technically unsupported release :)

Thanks

Doug

[Dianne Hackborn]

On Mon, Jun 14, 2010 at 3:37 PM, Doug <dougf...@gmail.com> wrote:

Dianne - I completely agree that this is theoretically not an issue
that I should be concerned with, however, as the developer of an app
that seems to be triggering this bad behavior it is up to me to try to
determine whether or not anything can be done about it.  I can point
my finger in any direction (Motorola, Verizon, even Google/AOSP) but
as you know users don't know (or even care) where the problem lies,
all they see is my app misbehaving and thats the end of it.  I'd
rather do a little research to help identify/resolve the problem now
than have to deal with the user rants later.

I care much more about the platform software being fixed than applications working around such a bug.  (Of course for some unknown piece of software running on a device, I care less about it.)

In other words, if your app is crashing the platform, I would like to have a bug filed with steps to reproduce so this issue in the platform can be fixed.

--
Dianne Hackborn
Android framework engineer

hac...@android.com

[Doug]

Hence my original request for information on how to determine the
cause of (what appears to be) a panic-caused reboot.
If I can determine the problem then you'll definitely see a ticket.

In terms of unknown software - how would I go about getting access to
official pre-release (i.e. Beta or RC) versions of Android for the
purposes of dev/test? Is that even an option?

On Jun 14, 4:06 pm, Dianne Hackborn <hack...@android.com> wrote:

> On Mon, Jun 14, 2010 at 3:37 PM, Doug <dougforp...@gmail.com> wrote:
> > Dianne - I completely agree that this is theoretically not an issue
> > that I should be concerned with, however, as the developer of an app
> > that seems to be triggering this bad behavior it is up to me to try to
> > determine whether or not anything can be done about it.  I can point
> > my finger in any direction (Motorola, Verizon, even Google/AOSP) but
> > as you know users don't know (or even care) where the problem lies,
> > all they see is my app misbehaving and thats the end of it.  I'd
> > rather do a little research to help identify/resolve the problem now
> > than have to deal with the user rants later.
>
> I care much more about the platform software being fixed than applications
> working around such a bug.  (Of course for some unknown piece of software
> running on a device, I care less about it.)
>
> In other words, if your app is crashing the platform, I would like to have a
> bug filed with steps to reproduce so this issue in the platform can be
> fixed.
>
> --
> Dianne Hackborn
> Android framework engineer

> hack...@android.com

[Doug]

If anyone is still interested, I've been able to reproduce the
problem, but not a lot of logging I'm afraid.

* If the device is connected via USB the problem doesn't seem to
occur.
* If both "Mobile Data" and "WiFi" are turned OFF, the problem doesn't
seem to occur.
* If either "Mobile Data" OR "WiFi" is enabled, the problem occurs (In
my case mobile-data is 'dead' as the phone isn't registered with a
mobile network)

I've been able to capture logcat output on phone (using "logcat > /
sdcard/logcat.out" from the Android Terminal app) on three occasions,
but after a quick glance there doesn't seem to be much in common
(don't know if this is due to output-buffering/file-flushing)

Doug

[Doug]

While I know you guys aren't interested in resolving problems with
rooted devices, how about a bunch of users who are running official
builds on non-rooted devices?
(Specifically the LG Ally on the Verizon network)

How should they escalate the problem given their carrier is probably
not at all interested?