ADDITIONAL SYSTEM INFORMATION :
MacBookPro 2020 M1, Monterey, Oracle JDK 17.0.6, Zulu 17.0.6
MacBookPro 2022, Ventura, M2 Max, Temurin JDK 17.0.6


A DESCRIPTION OF THE PROBLEM :
Calling Unsafe::getLongVolatile() on an allocateDirect ByteBuffer causes SIGBUS on Apple Silicon CPU:s.

The issue occurs from idk 15 and onwards.
Running a heap allocated buffer works.
We did manage to get a dump file from Zulu 15 that shows a SIGBUS.

REGRESSION : Last worked in version 11

STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
Run the test code,.

EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
Not crash
ACTUAL -
InternalError

---------- BEGIN SOURCE ----------
import java.nio.ByteBuffer;
import java.util.concurrent.ThreadLocalRandom;

import sun.misc.Unsafe;


public class UnsafeVolatileLong
{
    private static final int OFFSET =28;

    public static void main(String[] args) throws ReflectiveOperationException
    {
        var f = Unsafe.class.getDeclaredField("theUnsafe");
        f.setAccessible(true);
        var u = (Unsafe)f.get(null);
        byte[] bs = new byte[OFFSET*2];
        ThreadLocalRandom.current().nextBytes(bs);
        ByteBuffer bbd = ByteBuffer.allocateDirect(bs.length);
        bbd.put(bs);
        var val = u.getLongVolatile(bbd, OFFSET);
        System.out.println(val);
    }
}

---------- END SOURCE ----------

CUSTOMER SUBMITTED WORKAROUND :
Possible workaround is to install a x86-64 version and trust Rosetta. We have not tried that.
Otherwise, stay onJDK 11

FREQUENCY : always