ubuntu-phone team mailing list archive

[Alexander Antimonov <alexander.antimonov.od.ua@xxxxxxxxx>]

Hi, Petr.

There is, probably, nothing wrong with ANDROID_MEMORY_BARRIER.
The code of android_atomic_acquire_store and android_atomic_release_load is
not much different from CM (http://get.cm/?device=tenderloin).
If the code was wrong, it would affect many Touchpad users.

>From the backtrace it's obvious the crash happened at the
android_atomic_add function.
The crash log you posted in the letter is not full, so I looked at
http://pastebin.com/ir3Gdk1K.
>From that one I can tell, the crash is caused by ldrex instruction, ldrex
r0, [r4].
It is a load from memory and the address was wrong at that time, r4 ==
00000105.
The function prototype is the following:
int32_t android_atomic_add(int32_t increment, volatile int32_t *ptr),
passed ptr is wrong memory address.
Also, in the logcat in stack "section" there was
android::RefBase::incStrong(void const*) call.
It must be related to memory management by reference counting.

Let sum up.
The problematic peace of software (almost obvious) is ubuntuappmanager.
Seems like it is in development and may crash.
Why it crashes is another question. Maybe ubuntuappmanager developers can
say why it crashes most often?
Just in case, re-check that software you build is up to date.

So, you should stop blame yourself and (maybe) post a bug report on
ubuntuappmanager (platform-api?), if you didn't already.
I suppose, those are the "child issues" of ubuntuappmanager and will be
resolved with time.

Cheers.


On Mon, Apr 22, 2013 at 1:10 PM, Petr Bláha <pb@xxxxxxxx> wrote:

> Hi to all again,
>
> after some time I am back with my problem connected to HP Toucpad and
> Booting in to Black Screen. I have made some investigation and ended with
> this:  1) my ubuntuappmanager dies
>


> What is wrong with ANDROID_MEMORY_BARRIER ? Am I missing some
> configuration related to kernel?
>
> Thanks in advance
> --
>
> Petr Bláha
>