jdk24 OpenJDK java/foreign/TestLinker crash vmState=0x00000000

[pshipton]

Likely all platforms, I didn't check them all.
https://openj9-jenkins.osuosl.org/job/Test_openjdk24_j9_sanity.openjdk_x86-64_linux_Nightly_testList_0/2
java/foreign/TestLinker.java

04:31:05  Type=Segmentation error vmState=0x00000000
04:31:05  J9Generic_Signal_Number=00000018 Signal_Number=0000000b Error_Value=00000000 Signal_Code=00000001
04:31:05  Handler1=00007FC5ADD80DF0 Handler2=00007FC5AF34E3C0 InaccessibleAddress=0000000000000003
04:31:05  RDI=0000000000552BA8 RSI=00007FC565B9600F RAX=0000000000000000 RBX=00000000003AD600
04:31:05  RCX=0000000000000006 RDX=0000000000000020 R8=0000000000000001 R9=00000000005527F8
04:31:05  R10=0000000000552748 R11=0000000000000002 R12=000000000056F778 R13=000000001C0C0100
04:31:05  R14=00007FC5ADF8F760 R15=0000000000000000
04:31:05  RIP=00007FC5ADDC7A59 GS=0000 FS=0000 RSP=00007FC5971158C0
04:31:05  EFlags=0000000000010216 CS=0033 RBP=00000000003AD300 ERR=0000000000000004
04:31:05  TRAPNO=000000000000000E OLDMASK=0000000000000000 CR2=0000000000000003
04:31:05  xmm0=0000000000000000 (f: 0.000000, d: 0.000000e+00)
04:31:05  xmm1=0000000000552720 (f: 5580576.000000, d: 2.757171e-317)
04:31:05  xmm2=00007fc5adf5bd70 (f: 2918563072.000000, d: 6.940980e-310)
04:31:05  xmm3=ffffffffffffffff (f: 4294967296.000000, d: -nan)
04:31:05  xmm4=00007fc5adf38540 (f: 2918417664.000000, d: 6.940980e-310)
04:31:05  xmm5=0000003000000020 (f: 32.000000, d: 1.018558e-312)
04:31:05  xmm6=00007fc5141f0580 (f: 337577344.000000, d: 6.940853e-310)
04:31:05  xmm7=00000000ffd00528 (f: 4291822848.000000, d: 2.120442e-314)
04:31:05  xmm8=000000000000000a (f: 10.000000, d: 4.940656e-323)
04:31:05  xmm9=0000000040400000 (f: 1077936128.000000, d: 5.325712e-315)
04:31:05  xmm10=000000003e8c05e2 (f: 1049363968.000000, d: 5.184547e-315)
04:31:05  xmm11=000000003c075284 (f: 1007112832.000000, d: 4.975799e-315)
04:31:05  xmm12=0000000040400000 (f: 1077936128.000000, d: 5.325712e-315)
04:31:05  xmm13=000001380000013f (f: 319.000000, d: 6.620627e-312)
04:31:05  xmm14=0000000008001800 (f: 134223872.000000, d: 6.631540e-316)
04:31:05  xmm15=000001420000013b (f: 315.000000, d: 6.832826e-312)
04:31:05  Module=/home/jenkins/workspace/Test_openjdk24_j9_sanity.openjdk_x86-64_linux_Nightly_testList_0/jdkbinary/j2sdk-image/lib/default/libj9vm29.so
04:31:05  Module_base_address=00007FC5ADD40000
04:31:05  Target=2_90_20250125_2 (Linux 4.4.0-210-generic)
04:31:05  CPU=amd64 (4 logical CPUs) (0x5e2f20000 RAM)
04:31:05  ----------- Stack Backtrace -----------
04:31:05  walkStackFrames+0x909 (0x00007FC5ADDC7A59 [libj9vm29.so+0x87a59])
04:31:05  _ZN32VM_BytecodeInterpreterCompressed3runEP10J9VMThread+0xec4d (0x00007FC5ADE00BCD [libj9vm29.so+0xc0bcd])
04:31:05  bytecodeLoopCompressed+0xca (0x00007FC5ADDF1F6A [libj9vm29.so+0xb1f6a])
04:31:05   (0x00007FC5ADEBDDE2 [libj9vm29.so+0x17dde2])

[pshipton]

Exclude it via adoptium/aqa-tests#5912

[babsingh]

@h3110n3rv3 Please take a look at this issue.

[babsingh]

@h3110n3rv3 i was looking at this failure earlier this week. here are a few insights about the failure.

gdb stack trace:

#13 walkBytecodeFrame (walkState=0x3de900) at /root/openj9-openjdk-jdk24/openj9/runtime/vm/swalk.c:999
#14 walkStackFrames (currentThread=<optimized out>, walkState=0x3de900) at /root/openj9-openjdk-jdk24/openj9/runtime/vm/swalk.c:353
#15 0x00007f6e18ef0fe1 in VM_BytecodeInterpreterCompressed::inlThrowableFillInStackTrace (_pc=<optimized out>, _sp=<optimized out>, this=<optimized out>) at /root/openj9-openjdk-jdk24/openj9/runtime/vm/BytecodeInterpreter.hpp:3242
#16 VM_BytecodeInterpreterCompressed::run (this=0x6, this@entry=0x7f6e0028ae30, vmThread=0x0) at /root/openj9-openjdk-jdk24/openj9/runtime/vm/BytecodeInterpreter.hpp:10729
#17 0x00007f6e18ee5039 in bytecodeLoopCompressed (currentThread=<optimized out>) at /root/openj9-openjdk-jdk24/openj9/runtime/vm/BytecodeInterpreter.inc:112
#18 0x00007f6e18fb16f2 in c_cInterpreter () at /root/openj9-openjdk-jdk24/build/linux-x86_64-server-release/vm/runtime/vm/xcinterp.s:157
#19 0x00007f6e18e4e99a in internalSendExceptionConstructor (currentThread=currentThread@entry=0x3de600, exceptionClass=0x7f6e190799d0 <throwableNameAndSig>, exception=0xffe38098, detailMessage=<optimized out>, detailMessage@entry=0x0, constructorIndex=<optimized out>,
    constructorIndex@entry=0) at /root/openj9-openjdk-jdk24/openj9/runtime/vm/callin.cpp:639
#20 0x00007f6e18e673b7 in internalSetCurrentExceptionWithCause (currentThread=0x3de600, exceptionNumber=<optimized out>, detailMessage=0x0, utfMessage=<optimized out>, cause=<optimized out>) at /root/openj9-openjdk-jdk24/openj9/runtime/vm/exceptionsupport.c:725
#21 0x00007f6e18e67967 in setCurrentExceptionWithUtfCause (currentThread=<optimized out>, exceptionNumber=<optimized out>, detailMessage=<optimized out>, utfMessage=<optimized out>, cause=<optimized out>) at /root/openj9-openjdk-jdk24/openj9/runtime/vm/exceptionsupport.c:807
#22 0x00007f6e18ea53ec in OutOfLineINL_openj9_internal_foreign_abi_InternalDowncallHandler_initCifNativeThunkData (currentThread=0x3de600, method=<optimized out>)
    at /root/openj9-openjdk-jdk24/openj9/runtime/vm/OutOfLineINL_openj9_internal_foreign_abi_InternalDowncallHandler.cpp:191
#23 0x00007f6e18ee7a49 in VM_BytecodeInterpreterCompressed::outOfLineINL (_pc=<optimized out>, _sp=<optimized out>, this=<optimized out>) at /root/openj9-openjdk-jdk24/openj9/runtime/vm/BytecodeInterpreter.hpp:5613
#24 VM_BytecodeInterpreterCompressed::run (this=0x6, this@entry=0x7f6e0028b8f0, vmThread=0x0) at /root/openj9-openjdk-jdk24/openj9/runtime/vm/BytecodeInterpreter.hpp:10931
...

The segfault happens while throwing an InternalError from

openj9/runtime/vm/OutOfLineINL_openj9_internal_foreign_abi_InternalDowncallHandler.cpp

Lines 189 to 193 in e88bbde

	if (FFI_OK != status) {
	rc = GOTO_THROW_CURRENT_EXCEPTION;
	setCurrentException(currentThread, J9VMCONSTANTPOOL_JAVALANGINTERNALERROR, NULL);
	goto freeAllMemoryThenExit;
	}

The following test code crashes: TestLinker.java#L254-L260

        if (linker.getClass().equals(FallbackLinker.class)) {
            // The fallback linker does not support empty layouts (FFI_BAD_TYPEDEF)
            var iae = expectThrows(IllegalArgumentException.class, () -> linker.downcallHandle(fd));
            assertTrue(iae.getMessage().contains("is empty"));
        } else {
            linker.downcallHandle(fd);
        }

The test expects an IllegalArgumentException but OpenJ9 throws an InternalError.

For the above test case, we can throw an IllegalArgumentException in OpenJ9 by changing the below code:

 	// InternalError changed to IllegalArgumentException
 	setCurrentException(currentThread, J9VMCONSTANTPOOL_JAVALANGILLEGALARGUMENTEXCEPTION, NULL); 

The segfault will still happen while throwing the exception.

To resolve the segfault, my initial premise was that we need to buildInternalNativeStackFrame before throwing an error/exception similar to other functions in *InternalDowncallHandler.cpp:

openj9/runtime/vm/OutOfLineINL_openj9_internal_foreign_abi_InternalDowncallHandler.cpp

Lines 269 to 280 in e88bbde

	VM_OutOfLineINL_Helpers::buildInternalNativeStackFrame(currentThread, method);
	VM_VMHelpers::pushObjectInSpecialFrame(currentThread, nativeMH);
	invokeCache = resolveFfiCallInvokeHandle(currentThread, nativeMH);
	nativeMH = VM_VMHelpers::popObjectInSpecialFrame(currentThread);
	if (VM_VMHelpers::exceptionPending(currentThread)) {
	rc = GOTO_THROW_CURRENT_EXCEPTION;
	goto done;
	}
	VM_AtomicSupport::writeBarrier();
	J9VMJAVALANGINVOKENATIVEMETHODHANDLE_SET_INVOKECACHE(currentThread, nativeMH, invokeCache);
	VM_OutOfLineINL_Helpers::restoreInternalNativeStackFrame(currentThread);

Building an internal native stack frame in *InternalDowncallHandler_initCifNativeThunkData led to a different segfault. We are probably missing some preliminary setup before throwing the exception. You will need to figure out the missing piece.

[babsingh]

@h3110n3rv3 Can you open a PR to re-enable TestLinker? See adoptium/aqa-tests#5967 for an example.

[babsingh]

Test re-enabled via adoptium/aqa-tests#5974.

@h3110n3rv3 After enabling the test, the next steps are to remove the test excluded label and then close the issue. For future issues, I encourage you to try executing these steps independently. If you don’t have the necessary permissions to add or remove labels, you can request access through @AdamBrousseau.