Fix a missing allocationFence in process_java_lang_StringUTF16_toBytes()

[klangman]

When we transform a StringUTF16.toBytes() into a newarray and a call to String.decompressedArrayCopy(), we neglected to add an allocationFence. This change will introduce an allocationFence after the decompressedArrayCopy() call to ensure that all threads see the new contents of memory on weak memory coherency machines like POWER and AArch64.

[vijaysun-omr]

Are allocation fence nodes generated in general after allocations on platforms such as X86, that do not have weak memory model ? I ask, because this insertion of an allocation fence seems to be unconditional in that sense, it will add it across platforms. This may not be incorrect, but I thought I'd check if it was unconventional (e.g. if no one code ever generated an allocation fence on non-weak ordered platforms) since that could lead to bugs in its own way (i.e. IL that is unexpectedly different).

[klangman]

Yes, good catch.. I had looked at this but somehow missed the cg()->getEnforceStoreOrder() check in the genFence() routine used in ILGen. I added that check now.

[vijaysun-omr]

jenkins test sanity all jdk17

[klangman]

Looking through the failures I believe the root of the issues are:

zLinux:

06:05:02  After 20s process did not stop
[Pipeline] sh
06:05:05  -----------------------------------
06:05:05  fibTest_0_FAILED
06:05:05  -----------------------------------

Linux x86_64:

21:21:28  Exception: org.jenkinsci.plugins.workflow.support.steps.AgentOfflineException: Unable to create live FilePath for ub18-x86-1; ub18-x86-1 was marked offline: Timed out for last 5 attempts

Mac AArch64:

19:09:33  Running on mac10-x86-3 in /Users/jenkins/workspace/Test_openjdk17_j9_sanity.functional_aarch64_mac_Personal
[Pipeline] {
[Pipeline] cleanWs
19:09:48  [WS-CLEANUP] Deleting project workspace...
19:09:48  [WS-CLEANUP] Deferred wipeout is disabled by the job configuration...
19:09:52  ERROR: Cannot delete workspace :Remote call on mac10-x86-3 failed
[Pipeline] echo
19:09:52  Exception: hudson.AbortException: Cannot delete workspace: Remote call on mac10-x86-3 failed

Linux PPC64le:

06:05:02  After 20s process did not stop
[Pipeline] sh
06:05:05  -----------------------------------
06:05:05  fibTest_0_FAILED
06:05:05  -----------------------------------

I haven't found any previous examples of these issues.

Maybe we can try to retest since some of these failure seem to stem from a failure to reach machines.
The zLinux and LinuxPPCle failure " fibTest_0 timeout" is more concerning, but for the zLinux case the change should have 0 effect on the IL given that the getEnforceStoreOrder() test should (I would think) return false for z? I'll check that.

[klangman]

As I expected, setEnforceStoreOrder() is only executed on AArch64 and PPC. So I see no way for this code change to have affected zLinux.

[pshipton]

The zlinux failure I see is cmdLineTester_criu_jitserverPostRestore_0 can't bind server address: Address already in use, which looks like a failure you can ignore.

[klangman]

The zlinux failure I see is cmdLineTester_criu_jitserverPostRestore_0 can't bind server address: Address already in use, which looks like a failure you can ignore.

Right... I copied from the wrong browser tab when I got the z error text. So ya, ignore all I said about the z failure.

[vijaysun-omr]

jenkins test sanity amac jdk17

[vijaysun-omr]

jenkins test sanity xlinux jdk17

[vijaysun-omr]

jenkins test sanity amac jdk17

[vijaysun-omr]

jenkins test sanity xlinux jdk17

[vijaysun-omr]

There appear to be some infra issues preventing tests from passing. This change is conservative in that it adds a fence operation on weakly ordered platforms and so I don't expect any issues. I am merging it now with this reasoning.