threadMXBeanTestSuite1 -Xgcpolicy:metronome j9vm.227 ASSERTION FAILED at vm/jnicsup.cpp:1691: ((vmThread)->publicFlags & J9_PUBLIC_FLAGS_VM_ACCESS)

[JasonFengJ9]

Failure link

From internal Test_openjdk25_j9_extended.functional_x86-64_linux_testList_1 (10)

java version "25.0.3-beta" 2026-04-21
IBM Semeru Runtime Certified Edition 25.0.3+6-202603240357 (build 25.0.3-beta+6-202603240357)
Eclipse OpenJ9 VM 25.0.3+6-202603240357 (build 25.0.3-beta+6-202603240357-master-2ffc340655, JRE 25 Linux amd64-64-Bit Compressed References 20260324_184 (JIT enabled, AOT enabled)
OpenJ9   - 2ffc340655
OMR      - 1207e4690
JCL      - 9658c7f50 based on jdk-25.0.3+6)

Rerun in Grinder - Change TARGET to run only the failed test targets

Optional info

Failure output (captured from console output)

[2026-03-24T07:54:05.439Z] Running test threadMXBeanTestSuite1_1 ...
[2026-03-24T07:54:05.439Z] ===============================================
[2026-03-24T07:54:05.439Z] threadMXBeanTestSuite1_1 Start Time: Tue Mar 24 00:54:03 2026 Epoch Time (ms): 1774338843556
[2026-03-24T07:54:05.439Z] variation: Mode301 -Xint
[2026-03-24T07:54:05.439Z] JVM_OPTIONS:  -Xgcpolicy:metronome -Xnocompressedrefs -Xint 

[2026-03-24T07:54:10.861Z] 07:54:08.997 0x7f7aa8003c00    j9vm.227    *   ** ASSERTION FAILED ** at /home/jenkins/workspace/build-scripts/jobs/jdk25/jdk25-linux-x64-openj9-IBM/workspace/build/src/openj9/runtime/vm/jnicsup.cpp:1691: ((vmThread)->publicFlags & J9_PUBLIC_FLAGS_VM_ACCESS)

This is a consistent failure.

openjdk21_j9_extended.functional_ppc64_aix

[2026-03-24T04:04:45.751Z] Running test threadMXBeanTestSuite1_2 ...
[2026-03-24T04:04:45.751Z] ===============================================
[2026-03-24T04:04:45.751Z] threadMXBeanTestSuite1_2 Start Time: Tue Mar 24 00:04:44 2026 Epoch Time (ms): 1774325084352
[2026-03-24T04:04:45.751Z] variation: Mode351 -Xint
[2026-03-24T04:04:45.751Z] JVM_OPTIONS:  -Xgcpolicy:metronome -Xcompressedrefs -Xint 

[2026-03-24T04:04:50.460Z] 04:04:50.149 0x300e5300    j9vm.227    *   ** ASSERTION FAILED ** at /home/jenkins/workspace/build-scripts/jobs/jdk21u/jdk21u-aix-ppc64-openj9-IBM/workspace/build/src/openj9/runtime/vm/jnicsup.cpp:1691: ((vmThread)->publicFlags & J9_PUBLIC_FLAGS_VM_ACCESS)

[dmitripivkine]

I will triage this

[dmitripivkine]

An assertion was triggered during class unloading:

07:54:08.996783641 *0x7F7AA8003C00 j9mm.55              Event       Mark end
07:54:08.996784391 *0x0 j9mm.55              Event       Mark end
07:54:08.996785361 *0x7F7AA8003C00 j9mm.60              Event       Class unloading start
07:54:08.996785501 *0x0 j9mm.60              Event       Class unloading start
07:54:08.996793881 *0x7F7AA8003C00 j9vm.227             Assert     * ** ASSERTION FAILED ** at /home/jenkins/workspace/build-scripts/jobs/jdk25/jdk25-linux-x64-openj9-IBM/workspace/build/src/openj9/runtime/vm/jnicsup.cpp:1691: ((vmThread)->publicFlags & J9_PUBLIC_FLAGS_VM_ACCESS)

Unfortunately I am struggling to get c-stack from gdb.

[JasonFengJ9]

#1  0x00007ffff7a3bfdb in freeClassNativeMemory (hook=0x7ffff00122a0, eventNum=71, eventData=0x7ffed42fa490, userData=0x0) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/openj9/runtime/vm/jvminit.c:8785
#2  0x00007ffff7f8c6ce in J9HookDispatch (hookInterface=0x7ffff00122a0, taggedEventNum=71, eventData=0x7ffed42fa490) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/omr/util/hookable/hookable.cpp:237
#3  0x00007ffff7177db5 in MM_MetronomeDelegate::addDyingClassesToList (this=0x7ffff000c958, env=0x7ffec8005df8, classLoader=0x7ffff0a71328, setAll=false, classUnloadListStart=0x0, classUnloadCountResult=0x7ffed42fa4f8)
    at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/openj9/runtime/gc_glue_java/MetronomeDelegate.cpp:524
#4  0x00007ffff7176ff9 in MM_MetronomeDelegate::processDyingClasses (this=0x7ffff000c958, env=0x7ffec8005df8, classUnloadCountResult=0x7ffed42fa668, anonymousClassUnloadCountResult=0x7ffed42fa670, classLoaderUnloadCountResult=0x7ffed42fa678, 
    classLoaderUnloadListResult=0x7ffed42fa680) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/openj9/runtime/gc_glue_java/MetronomeDelegate.cpp:434
#5  0x00007ffff717811c in MM_MetronomeDelegate::unloadDeadClassLoaders (this=0x7ffff000c958, envModron=0x7ffec8005df8) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/openj9/runtime/gc_glue_java/MetronomeDelegate.cpp:615
#6  0x00007ffff7176a83 in MM_MetronomeDelegate::incrementalCollect (this=0x7ffff000c958, env=0x7ffec8005df8) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/openj9/runtime/gc_glue_java/MetronomeDelegate.cpp:355
#7  0x00007ffff6ef5cd3 in MM_RealtimeGC::incrementalCollect (this=0x7ffff000c750, env=0x7ffec8005df8) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/openj9/runtime/gc_realtime/RealtimeGC.cpp:298
#8  0x00007ffff6ef6369 in MM_RealtimeGC::internalGarbageCollect (this=0x7ffff000c750, env=0x7ffec8005df8, subSpace=0x7ffff0a6c1b0, allocDescription=0x0) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/openj9/runtime/gc_realtime/RealtimeGC.cpp:413
#9  0x00007ffff7041a6a in MM_Collector::garbageCollect (this=0x7ffff000c750, env=0x7ffec8005df8, callingSubSpace=0x7ffff0a6c1b0, allocateDescription=0x0, gcCode=2, objectAllocationInterface=0x0, baseSubSpace=0x0, context=0x0)
    at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/omr/gc/base/Collector.cpp:496
#10 0x00007ffff6ef47ac in MM_MemorySubSpaceMetronome::collect (this=0x7ffff0a6c1b0, env=0x7ffec8005df8, gcCode=...) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/openj9/runtime/gc_realtime/MemorySubSpaceMetronome.cpp:166
#11 0x00007ffff6f04eca in MM_Scheduler::mainEntryPoint (this=0x7ffff0981c50, envModron=0x7ffec8005df8) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/openj9/runtime/gc_realtime/Scheduler.cpp:931
#12 0x00007ffff70802a4 in dispatcher_thread_proc2 (portLib=0x7ffff7cc2260 <j9portLibrary>, info=0x7ffff7d40930) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/omr/gc/base/ParallelDispatcher.cpp:89
#13 0x00007ffff784a57a in omrsig_protect (portLibrary=0x7ffff7cc2260 <j9portLibrary>, fn=0x7ffff7080122 <dispatcher_thread_proc2(OMRPortLibrary*, void*)>, fn_arg=0x7ffff7d40930, handler=0x7ffff7a0e767 <structuredSignalHandlerVM>, handler_arg=0x7ffff000fdf0, 
    flags=506, result=0x7ffed42fad00) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/omr/port/unix/omrsignal.c:427
#14 0x00007ffff70804aa in dispatcher_thread_proc (info=0x7ffff7d40930) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/omr/gc/base/ParallelDispatcher.cpp:131
#15 0x00007ffff79abdf6 in thread_wrapper (arg=0x7ffff0d15fa0) at /root/builds/jdk25-master-0324/openj9-openjdk-jdk25/omr/thread/common/omrthread.c:1739
#16 0x00007ffff7df5ac3 in ?? () from /lib/x86_64-linux-gnu/libc.so.6
#17 0x00007ffff7e878d0 in ?? () from /lib/x86_64-linux-gnu/libc.so.6

This is before the assertion was triggered (I was triaging a personal build failure).

[dmitripivkine]

Right, thank you. The same stack has been preserved in the javacore.
So the problem is in Metronome GC worker thread performing class unloading does not have 0x20 flag set. So hook subscribers can not call j9jni_deleteGlobalRef() as is.

[dmitripivkine]

I need to be clear class unloading is performed single threaded during GC increment under Exclusive access, so performing of j9jni_deleteGlobalRef() should be safe, but assertion is on the way. Another point we need double check deleting JNI Global Reference outside GC cycle is a subject of calling Barrier. I need to check that calling Barrier is not necessary during GC increment.

[dmitripivkine]

Ok, j9jni_deleteGlobalRef() calls barrier itself, so no extra barriers call is necessary.
So, the only problem left an assertion should be removed.
BTW I think only Standard Collectors would have VM Access to be set for thread doing class unloading. It is not correct for Metronome and I believe Balanced too.