olpc-dm/Xorg segfault only if during systemd boot

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

olpc-dm/Xorg segfault only if during systemd boot

George Hunt
Context:
XSCE server code loaded on 128GB SD card running on XO1.5. We've added lots of services, and for generic servers, disabled the GUI (systemctl disable olpc-dm).

But for aesthetic reasons, it might be more attractive to start up in sugar/mate, and return the memory, if the user really wants to serve clients. ( set it back to default.target => multi-user.target).

Observations:
Xorg starts, flashes the screen, and gives up (systemd retries about 32 times and gives up)

The /var/log/Xorg.log shows a segfault immediately after "VIA 3d engine enabled" during systemd boot. But interestingly, if ollpc-dm is started after systemd has completed, olpc-dm brings up Xorg successfully.  

So there may be a dependency, which without all the additional services, is properly sequenced.  We may need to delay olpc-dm until after that dependency is satisfied.

Does anyone have an idea of what might be happening? Or additional tests I might perform?

_______________________________________________
Devel mailing list
[hidden email]
http://lists.laptop.org/listinfo/devel
Reply | Threaded
Open this post in threaded view
|

Re: olpc-dm/Xorg segfault only if during systemd boot

James Cameron-2
I'm interested in finding the cause; any such race condition that
prevents X startup may eventually occur depending on timing, so is
there some simple reproducer that we can apply to OLPC OS?

Where are your changes relative to OLPC OS?

What does the systemd journal say about olpc-dm service?

How do the journal, olpc-dm logs, and X logs differ between working
and failing case?

Where does gdb show the segmentation fault occurs?

--
James Cameron
http://quozl.linux.org.au/
_______________________________________________
Devel mailing list
[hidden email]
http://lists.laptop.org/listinfo/devel
Reply | Threaded
Open this post in threaded view
|

Re: olpc-dm/Xorg segfault only if during systemd boot

George Hunt
I missed one obvious clue, thinking that Xorg.log was the strategic place to look.

/var/log/messages (search olpc-dm) shows 
  • "olpc-dm.service holdoff time over, scheduling restart"
  • and then "xauth: file /var/tmp/olpc-auth/.Xauthority does not exist"
googling "Xauthority create olpc" does not give a good hit, but suggests I should learn olpc-utils better.

On Mon, Sep 7, 2015 at 2:44 PM, James Cameron <[hidden email]> wrote:
I'm interested in finding the cause; any such race condition that
prevents X startup may eventually occur depending on timing, so is
there some simple reproducer that we can apply to OLPC OS?

Where are your changes relative to OLPC OS?

What does the systemd journal say about olpc-dm service?

How do the journal, olpc-dm logs, and X logs differ between working
and failing case?

Where does gdb show the segmentation fault occurs?

--
James Cameron
http://quozl.linux.org.au/


_______________________________________________
Devel mailing list
[hidden email]
http://lists.laptop.org/listinfo/devel
Reply | Threaded
Open this post in threaded view
|

Re: olpc-dm/Xorg segfault only if during systemd boot

James Cameron-2
The first message is a result of your problem, not a cause.

The second message is normal, verified in my test just now, therefore
not related to your problem.

To look at the systemd journal for olpc-dm ...

        systemctl status olpc-dm.service | cat

To look at olpc-dm client logs ...

        cat /tmp/olpc-dm*

Do both of these in the failed state as well as the working state to
identify the differences.

On Mon, Sep 07, 2015 at 03:21:10PM -0700, George Hunt wrote:

> I missed one obvious clue, thinking that Xorg.log was the strategic place to
> look.
>
> /var/log/messages (search olpc-dm) shows 
>
>   • "olpc-dm.service holdoff time over, scheduling restart"
>   • and then "xauth: file /var/tmp/olpc-auth/.Xauthority does not exist"
>
> googling "Xauthority create olpc" does not give a good hit, but suggests I
> should learn olpc-utils better.
>
> On Mon, Sep 7, 2015 at 2:44 PM, James Cameron <[1][hidden email]> wrote:
>
>     I'm interested in finding the cause; any such race condition that
>     prevents X startup may eventually occur depending on timing, so is
>     there some simple reproducer that we can apply to OLPC OS?
>
>     Where are your changes relative to OLPC OS?
>
>     What does the systemd journal say about olpc-dm service?
>
>     How do the journal, olpc-dm logs, and X logs differ between working
>     and failing case?
>
>     Where does gdb show the segmentation fault occurs?
>    
>     --
>     James Cameron
>     [2]http://quozl.linux.org.au/
>
> References:
>
> [1] mailto:[hidden email]
> [2] http://quozl.linux.org.au/

> _______________________________________________
> Devel mailing list
> [hidden email]
> http://lists.laptop.org/listinfo/devel


--
James Cameron
http://quozl.linux.org.au/
_______________________________________________
Devel mailing list
[hidden email]
http://lists.laptop.org/listinfo/devel
Reply | Threaded
Open this post in threaded view
|

Re: olpc-dm/Xorg segfault only if during systemd boot

George Hunt
The resolution to the segfault in olpc-dm turned out to be obvious. 

XSCE developers have responded to on-the-ground requests for teamviewer, and added it to the ansible playbook, without realizing that it is not compatible with olpc-dm. Teamviewer daemon was being pulled in by graphical.target.

The ansible playbook will be updated

On Mon, Sep 7, 2015 at 3:37 PM, James Cameron <[hidden email]> wrote:
The first message is a result of your problem, not a cause.

The second message is normal, verified in my test just now, therefore
not related to your problem.

To look at the systemd journal for olpc-dm ...

        systemctl status olpc-dm.service | cat

To look at olpc-dm client logs ...

        cat /tmp/olpc-dm*

Do both of these in the failed state as well as the working state to
identify the differences.

On Mon, Sep 07, 2015 at 03:21:10PM -0700, George Hunt wrote:
> I missed one obvious clue, thinking that Xorg.log was the strategic place to
> look.
>
> /var/log/messages (search olpc-dm) shows 
>
>   • "olpc-dm.service holdoff time over, scheduling restart"
>   • and then "xauth: file /var/tmp/olpc-auth/.Xauthority does not exist"
>
> googling "Xauthority create olpc" does not give a good hit, but suggests I
> should learn olpc-utils better.
>
> On Mon, Sep 7, 2015 at 2:44 PM, James Cameron <[1][hidden email]> wrote:
>
>     I'm interested in finding the cause; any such race condition that
>     prevents X startup may eventually occur depending on timing, so is
>     there some simple reproducer that we can apply to OLPC OS?
>
>     Where are your changes relative to OLPC OS?
>
>     What does the systemd journal say about olpc-dm service?
>
>     How do the journal, olpc-dm logs, and X logs differ between working
>     and failing case?
>
>     Where does gdb show the segmentation fault occurs?
>
>     --
>     James Cameron
>     [2]http://quozl.linux.org.au/
>
> References:
>
> [1] mailto:[hidden email]
> [2] http://quozl.linux.org.au/

> _______________________________________________
> Devel mailing list
> [hidden email]
> http://lists.laptop.org/listinfo/devel


--
James Cameron
http://quozl.linux.org.au/


_______________________________________________
Devel mailing list
[hidden email]
http://lists.laptop.org/listinfo/devel