RFR(XS): 8232005: [s390] More exception checks missing in interpreter

[Doerr, Martin]

Hi Richard,

thank you for reviewing and for checking other places, too.

Regarding the s390 code, I think the additional places you mentioned are no bugs.
Monitorenter/exit use JRT_ENTRY_NO_ASYNC, so no async exception will get installed.
I guess the s390 code was designed this way by intention? No idea why x86 has exception checks, here.

It would also be safe to omit the exception check at places, in which we already have a pending exception which is handled by specific code later on. Async exceptions will get postponed in this case. But this doesn't seem to be the case for the places you mentioned.

Regarding the PPC64 code, I think we should fix the places you found. They seem to be real bugs AFAICS.

Best regards,
Martin


> -----Original Message-----
> From: Reingruber, Richard <richard.reingruber at sap.com>
> Sent: Freitag, 11. Oktober 2019 11:36
> To: Doerr, Martin <martin.doerr at sap.com>; hotspot-runtime-
> dev at openjdk.java.net; Schmidt, Lutz <lutz.schmidt at sap.com>; Langer,
> Christoph <christoph.langer at sap.com>
> Subject: RE: RFR(XS): 8232005: [s390] More exception checks missing in
> interpreter
> 
> Hi Martin,
> 
> Thanks for analyzing and fixing those.
> 
> Have you looked at the following:
> 
> cpu/s390/interp_masm_s390.cpp:970:    call_VM(noreg,
> CAST_FROM_FN_PTR(address, InterpreterRuntime::monitorenter),
> cpu/s390/interp_masm_s390.cpp-971-            monitor,
> /*check_for_exceptions=*/false);
> cpu/s390/interp_masm_s390.cpp:1150:  call_VM(noreg,
> CAST_FROM_FN_PTR(address, InterpreterRuntime::monitorexit),
> cpu/s390/interp_masm_s390.cpp-1151-          monitor,
> /*check_for_exceptions=*/false);
> cpu/ppc/templateInterpreterGenerator_ppc.cpp:595:    __
> call_VM(Rexception, CAST_FROM_FN_PTR(address,
> InterpreterRuntime::create_klass_exception), false);
> cpu/ppc/templateInterpreterGenerator_ppc.cpp:598:    __
> call_VM(Rexception, CAST_FROM_FN_PTR(address,
> InterpreterRuntime::create_exception), false);
> cpu/ppc/templateInterpreterGenerator_ppc.cpp:2108:    __
> MacroAssembler::call_VM(R4_ARG2, CAST_FROM_FN_PTR(address,
> InterpreterRuntime::member_name_arg_or_null), R4_ARG2, R19_method,
> R14_bcp, false);
> 
> x86 does the exception check there. I think the check is needed, because
> async. exceptions can get installed there and if we don't check, they could be
> dropped.
> 
> Thanks, Richard.
> (Not yet Reviewer)
> 
> -----Original Message-----
> From: hotspot-runtime-dev <hotspot-runtime-dev-
> bounces at openjdk.java.net> On Behalf Of Doerr, Martin
> Sent: Dienstag, 8. Oktober 2019 16:42
> To: hotspot-runtime-dev at openjdk.java.net; Schmidt, Lutz
> <lutz.schmidt at sap.com>; Langer, Christoph <christoph.langer at sap.com>
> Subject: [CAUTION] RFR(XS): 8232005: [s390] More exception checks missing
> in interpreter
> 
> Hi,
> 
> the template interpreter on s390 misses more exception checks when calling
> VM functions which may install an async exception (like
> java.lang.ThreadDeath) at a safepoint.
> 
> Bug:
> https://bugs.openjdk.java.net/browse/JDK-8232005
> 
> These are the two places I found:
> http://cr.openjdk.java.net/~mdoerr/8232005_s390_check_exceptions/webr
> ev.00/
> 
> Please review.
> 
> Best regards,
> Martin