C2 compiler gets SIGILL/ILL_ILLOPC

John Neffenger john at status6.com
Thu Feb 18 17:05:57 UTC 2021


On 2/18/21 4:20 AM, Matthias Klose wrote:
> this came up at https://bugs.debian.org/960153, citing
> https://www.gitmemory.com/issue/AdoptOpenJDK/openjdk-build/1070/489695028

I guess there could be more than one bug that's avoided by passing 
'-XX:-AssumeMP' to the Java launcher. It would be related to the bug I 
encountered only if you're still getting a SIGILL fatal error when 
trying to execute the PLDW instruction.

For details on the issue I reported, the full conversation can be found 
here on GitHub:

Add support for ARM Cortex-A8 cores with Client VM
https://github.com/AdoptOpenJDK/openjdk-build/issues/1070

That issue was closed by the following pull request:

Add Client VM to builds for Linux arm32 platform
https://github.com/AdoptOpenJDK/openjdk-build/pull/1078

That pull request was just to work around the following bug:

JDK-8222825: ARM32 SIGILL issue on single core CPU (not supported PLDW 
instruction)
https://bugs.openjdk.java.net/browse/JDK-8222825

And that bug was fixed by Boris Ulasevich in JDK 13 with this commit:

8222825: ARM32 SIGILL issue on single core CPU (not supported PLDW 
instruction)
https://github.com/openjdk/jdk/commit/989fa190a601654d203b8af533de9e3b81bb317e

Both the AdoptOpenJDK fix (adding the Client VM) and the OpenJDK fix 
(avoiding the PLDW instruction) are in the 32-bit ARM builds, the last 
time I checked.

Let me know if you need any further details.

John

-- 
Status Six Communications <https://www.status6.com>


More information about the aarch32-port-dev mailing list