Native call framework prototype issues

[Chan, Sunny]

Hello,

I am trying to compile and run the Native call framework prototype that has been uploaded by Vladimir in February [1] but I seems to hit a few issues when I try this on Linux x86_64:


1)      In RTLD_DEFAULT in Linux is 0 not -2 [2]  - -2 crashes on Linux and replacing it with 0 (grep from linux header) fixed this

2)      When the JIT generate the native method wrapper, it causes this assertion:


#  Internal Error (/local/scratch/chasun/panama/hotspot/src/cpu/x86/vm/register_x86.hpp:64), pid=23604, tid=140592728340224

#  assert(is_valid()) failed: invalid register



The stack trace looks like this:



V  [libjvm.so+0x421f5b]  Assembler::testq(RegisterImpl*, RegisterImpl*)+0x3b

V  [libjvm.so+0xd9434e]  MethodHandles::verify_klass(MacroAssembler*, RegisterImpl*, SystemDictionary::WKID, char const*)+0xce

V  [libjvm.so+0xd94ae1]  MethodHandles::generate_native_call(MacroAssembler*, RegisterImpl*, bool)+0x111

V  [libjvm.so+0xf8e9e2]  SharedRuntime::generate_native_wrapper(MacroAssembler*, methodHandle, int, BasicType*, VMRegPair*, BasicType)+0x5482

V  [libjvm.so+0xf6fb3d]  AdapterHandlerLibrary::create_native_wrapper(methodHandle)+0x77d

V  [libjvm.so+0x102641c]  SystemDictionary::find_method_handle_intrinsic(vmIntrinsics::ID, Symbol*, Thread*)+0x63c

V  [libjvm.so+0xc5907e]  LinkResolver::lookup_polymorphic_method(methodHandle&, KlassHandle, Symbol*, Symbol*, KlassHandle, Handle*, Handle*, Thread*)+0x5de

V  [libjvm.so+0xc59ef3]  LinkResolver::resolve_method(methodHandle&, KlassHandle, Symbol*, Symbol*, KlassHandle, bool, bool, Thread*)+0x5a3

V  [libjvm.so+0xc5d5fa]  LinkResolver::linktime_resolve_static_method(methodHandle&, KlassHandle, Symbol*, Symbol*, KlassHandle, bool, Thread*)+0x8a

V  [libjvm.so+0xc5dab0]  LinkResolver::resolve_static_call(CallInfo&, KlassHandle&, Symbol*, Symbol*, KlassHandle, bool, bool, Thread*)+0xa0

V  [libjvm.so+0xd8d23b]  MethodHandles::resolve_MemberName(Handle, KlassHandle, Thread*)+0xecb

V  [libjvm.so+0xd92620]  MHN_resolve_Mem+0x310


                It looks like in generate_native_call, the member_reg is set to "no_reg" and when it try to verify_klass it seems to reference to the no_reg value and hence invalid register in the testq() assembler

                Is that a known issue or caused by recent change in hotspot?

[1] http://mail.openjdk.java.net/pipermail/panama-dev/2015-February/000095.html
[2] http://hg.openjdk.java.net/panama/panama/hotspot/file/298f5197762b/src/share/vm/prims/unsafe.cpp#l776


Sunny Chan, Executive Director, Technology
Goldman Sachs (Asia) L.L.C. | 68th Floor | Cheung Kong Center | 2 Queens Road Central | Hong Kong
email:  sunny.chan at gs.com | Tel: +852 2978 6481 | Fax: +852 2978 0633

This message may contain information that is confidential or privileged.  If you are not the intended recipient, please advise the sender immediately and delete this message.  See http://www.gs.com/disclaimer/email for further information on confidentiality and the risks inherent in electronic communication.