[oe] OpenJDK 8 Segfault on i.MX6

[Attie Grande]

Hi again,

As a follow-up (under Aleksey's guidance) I applied the following patches:

- JDK-8202600 [1][2] had no effect, but I'll be keeping it applied.
- JDK-8194739 [3][4] and things have sprung into life.

I'll prepare a patch and submit tomorrow.

Attie

[1]: https://bugs.openjdk.java.net/browse/JDK-8202600
[2]: http://hg.openjdk.java.net/jdk8u/jdk8u/hotspot/rev/ca4663e58916
[3]: https://bugs.openjdk.java.net/browse/JDK-8194739
[4]: http://hg.openjdk.java.net/jdk/jdk10/rev/69d1a1590485
On Wed, 28 Nov 2018 at 17:30, Attie Grande
<attie.grande at argentum-systems.co.uk> wrote:
>
> Hi all,
>
> I've been trying to get OpenJDK 8 162-b12 running on Linux on an ARM
> platform (i.MX6), using the Yocto Sumo recipes from meta-java [1].
>
> Unfortunately, the result is a java executable that cannot startup
> without hitting a SIGSEGV. Even running "simple" things like `java` or
> `java -help` result in a segmentation fault.
>
> I've tried to investigate it, but I don't have the insight into the
> project to really be able to debug the issue properly myself.
>
> I've been in contact with jdk-dev [2], who have been helpful so far,
> but we're starting to wonder about the patches applied as part of the
> Yocto build.
>
> After enabling fastdebug, I still get a segfault with little insight
> into the issue - a fatal error log is available here [3].
>
> Investigating with GDB shows that the segfault is in libjvm.so, but
> there are no symbols in this region of memory, and the stack has been
> destroyed (I suspect as part of the VM's operation).
>
> We've confirmed that another build of OpenJDK 8 runs on this platform
> [4] (currently jdk8u191-b12).
>
> I've spent some time trying to update to 191-b12, but a number of the
> patches appear to be necessary for Yocto's build environment, while
> others require major rework to apply cleanly.
>
> I see that Thud is using 172b11 [5] which is somewhere in between...
> so I could give those recipes a try - though there is a Sumo / Thud
> dependency issue with xorg-proto.
>
> Has anyone seen this before, or have any insight into the patches / Yocto build?
>
> Aleksey (from jdk-dev) mentioned [6] a user on IRC called hitendra had
> a similar sounding issue with Yocto Sumo / i.MX6 / OpenJDK 8 162-b12,
> and I've also come across a brief / frutiless IRC log in #yocto [7].
>
> Regards,
> Attie
>
> [1]: https://git.yoctoproject.org/cgit/cgit.cgi/meta-java/tree/recipes-core/openjdk?h=sumo
> [2]: http://mail.openjdk.java.net/pipermail/jdk-dev/2018-November/002246.html
> [3]: https://pastebin.com/raw/iXTz8v9z
> [4]: https://builds.shipilev.net/openjdk-jdk8-redhat/openjdk-jdk8-redhat-latest-linux-arm32-hflt-zero-fastdebug.tar.xz
> [5]: https://git.yoctoproject.org/cgit/cgit.cgi/meta-java/tree/recipes-core/openjdk?h=master-next
> [6]: http://mail.openjdk.java.net/pipermail/jdk-dev/2018-November/002251.html
> [7]: https://www.yoctoproject.org/irc/%23yocto.2018-10-23.log.html#t2018-10-23T09:46:21