Update the FFI specific code in JDK21 for JEP442

[ChengJin01]

The modifications are based on the latest APIs in JDK21
to support JEP442 as mentioned at #16951.

Signed-off-by: ChengJin01 jincheng@ca.ibm.com

[ChengJin01]

Note:
[1] The code changes have been verified with the updated test suites in JDK21 on JDK21/JDKnext.
[2] The changes include all code/tests related to the Arena/AddressLayout(introduced in JDK21 to replace OfAddress for pointer) as required at #16951.
[3] There is no updated code in OpenJDK as everything was addressed via ibmruntimes/openj9-openjdk-jdk#598 previously.
[4] There is no code required for the fallback native linker that is literally a libffi-based solution for new architectures without the FFI support which should be enabled by default.
[5] VaList related test cases are removed as there is no support from OpenJDK since JDK21.
[6] Other features (e.g. isTrivial() the union type, etc) will be handled separately via the Jtreg test suites.

[ChengJin01]

Reviewer: @tajila
FYI: @keithc-ca, @pshipton

[ChengJin01]

I will need to double-check the code here to see whether it still works with the new changes in JDK21 at ibmruntimes/openj9-openjdk-jdk21#4 (just merged).

[ChengJin01]

Probably need to wait till the changes at ibmruntimes/openj9-openjdk-jdk21#4 is merged to the openj9 branch.

[ChengJin01]

As talked to Jason offline, the follow changes in OpenJDK should be merged altogether at first to get code related compiled correctly in JDK21/next:
[1] ibmruntimes/openj9-openjdk-jdk21#5 (depends on #17514 & #17125)
[2] ibmruntimes/openj9-openjdk-jdk#608

[ChengJin01]

There are a bunch of test failures with padding/alignment detected in OpenJ9 FFI test suites with the newly merged changes (with check on the padding/alignment) in JDK21. I will need to check case by case to see whether it is simply a test issue or anything related to our code that needs to be addressed.

[ChengJin01]

I've went through all failing test cases which was caused by https://github.com/ibmruntimes/openj9-openjdk-jdk21/blob/0a869c9c5496e25f7f89805471071640a4457d68/src/java.base/share/classes/jdk/internal/foreign/abi/AbstractLinker.java#L222C1-L229C6

    // check for trailing padding
    private static void checkGroupSize(GroupLayout gl, long maxUnpaddedOffset) {
        long expectedSize = Utils.alignUp(maxUnpaddedOffset, gl.byteAlignment());
        if (gl.byteSize() != expectedSize) {
            throw new IllegalArgumentException("Layout '" + gl + "' has unexpected size: "
                    + gl.byteSize() + " != " + expectedSize); <------
        }
    }

which the alignment/padding related code has been adjusted in OpenJDK in which case the trailing padding is no longer required in struct except a few cases on AIX (e.g. [double, int]/[double, float]). So there is no more code specific changes in OpenJ9 as padding is ignored by default in downcall.

[ChengJin01]

The PR has been verified on all supported platforms (no code required in OpenJDK) with the updated test cases and is ready for review.

[ChengJin01]

@tajila, any more comment for this PR?

[tajila]

jenkins sanity test xlinux,plinux jdk21

[tajila]

jenkins test sanity xlinux,plinux jdk21

[ChengJin01]

The test failures at https://openj9-jenkins.osuosl.org/job/Test_openjdk21_j9_sanity.functional_x86-64_linux_Personal/2/tapTestReport/ & https://openj9-jenkins.osuosl.org/job/Test_openjdk21_j9_sanity.functional_ppc64le_linux_Personal/2/tapTestReport/ are related to Jep425Tests_testVirtualThread (irrelevant to FFI) which was fixed by #17510 (not yet merged).

[ChengJin01]

@tajila, please help run Jenkins for this PR to see whether there is no other issue as #17510 is merged.

[tajila]

jenkins test sanity xlinux,plinux jdk21