Segfault

[Tobias Hartmann]

Hi Volker,

On 20.01.2017 19:30, Volker Simonis wrote:
> Sorry for the previous mail - it was intended as a private answer to
> Rafael but I accidently also posted it to the list.
> 
> But once it's public, here are the details:
> 
> Rafael sent me hs_err file (together with replay) which looks exactly
> like https://bugs.openjdk.java.net/browse/JDK-8081379 which is a
> duplicate of https://bugs.openjdk.java.net/browse/JDK-8078262 which
> should be fixed with https://bugs.openjdk.java.net/browse/JDK-6675699
> (in 9 but also in  8u111b01).

In the past, many loop optimization problems manifested as crashes in PhaseIdealLoop::build_loop_late_post(), including JDK-6675699 which I fixed some time ago in 9 and 8u.

I would suggest to file a bug and attach the replay compilation and hs_err file. Were you able to reproduce this?

Thanks,
Tobias

> But the crash happened in 8.0_111-b14 (see below). If somebody is
> interested, I can open a new bug and attach the hs_err and replay
> file.
> 
> Regards,
> Volker
> 
> #
> # A fatal error has been detected by the Java Runtime Environment:
> #
> #  SIGSEGV (0xb) at pc=0x00007fe6aa553804, pid=6871, tid=0x00007fe685302700
> #
> # JRE version: Java(TM) SE Runtime Environment (8.0_111-b14) (build
> 1.8.0_111-b14)
> # Java VM: Java HotSpot(TM) 64-Bit Server VM (25.111-b14 mixed mode
> linux-amd64 compressed oops)
> # Problematic frame:
> # V  [libjvm.so+0x818804]  PhaseIdealLoop::build_loop_late_post(Node*)+0x144
> #
> # Failed to write core dump. Core dumps have been disabled. To enable
> core dumping, try "ulimit -c unlimited" before starting Java
>  again
> #
> # If you would like to submit a bug report, please visit:
> #   http://bugreport.java.com/bugreport/crash.jsp
> #
> 
> Stack: [0x00007fe685202000,0x00007fe685303000],
> sp=0x00007fe6852fd110,  free space=1004k
> Native frames: (J=compiled Java code, j=interpreted, Vv=VM code, C=native code)
> V  [libjvm.so+0x818804]  PhaseIdealLoop::build_loop_late_post(Node*)+0x144
> V  [libjvm.so+0x818dbc]  PhaseIdealLoop::build_loop_late(VectorSet&,
> Node_List&, Node_Stack&)+0x13c
> V  [libjvm.so+0x81b9d5]  PhaseIdealLoop::build_and_optimize(bool, bool)+0x835
> V  [libjvm.so+0x4a2ad0]  Compile::Optimize()+0xbe0
> V  [libjvm.so+0x4a5559]  Compile::Compile(ciEnv*, C2Compiler*,
> ciMethod*, int, bool, bool, bool)+0x13c9
> V  [libjvm.so+0x3f24f5]  C2Compiler::compile_method(ciEnv*, ciMethod*,
> int)+0x1f5
> V  [libjvm.so+0x4afb8a]
> CompileBroker::invoke_compiler_on_method(CompileTask*)+0xc9a
> V  [libjvm.so+0x4b0b36]  CompileBroker::compiler_thread_loop()+0x5d6
> V  [libjvm.so+0xa742b3]  JavaThread::thread_main_inner()+0x103
> V  [libjvm.so+0xa743fc]  JavaThread::run()+0x11c
> V  [libjvm.so+0x924ea8]  java_start(Thread*)+0x108
> C  [libpthread.so.0+0x7aa1]
> 
> 
> 2017-01-20 19:16 GMT+01:00 Volker Simonis <volker.simonis at gmail.com>:
>> Ja, das ist ganz klar ein C2 crash!
>>
>> Scheint das gleiche zu sein wie:
>> https://bugs.openjdk.java.net/browse/JDK-8081379 was wiederum ein
>> Duplikat von https://bugs.openjdk.java.net/browse/JDK-8078262 ist und
>> angeblich mit https://bugs.openjdk.java.net/browse/JDK-6675699 gefixt
>> wurde (auch in 8u111).
>>
>> Ich würde das man an hotspot-compiler-dev at openjdk.java.net schicken.
>>
>>
>> Zu unserem ursprünglichen Problem: ich glaube mittlerweile nicht mehr,
>> dass das Problem direkt mit Classentransormation zusammenhängt, da
>> diese an einem safepoint passiert. Weisst Du, ob parallel zur
>> Transformation auch viele Klassen geladen werden (entweder neue oder
>> generierte oder alte in einem neuen Classloader)?
>>
>>
>>
>> 2017-01-20 16:40 GMT+01:00 Rafael Winterhalter <rafael.wth at gmail.com>:
>>> Sorry, falsche Datei.
>>>
>>> ---------- Weitergeleitete Nachricht ----------
>>> Von: Rafael Winterhalter <rafael.wth at gmail.com>
>>> Datum: 20. Januar 2017 um 16:31
>>> Betreff: Segfault
>>> An: Volker Simonis <volker.simonis at gmail.com>
>>>
>>>
>>>
>>> Hei Volker,
>>> ich weiß nicht, ob das mit dem besprochenen Bug zusammenhängt aber die
>>> gleiche Routine verursacht mittlerweile ab und an den angehängten Fehler.
>>>
>>> Schaut mir aber eher aus als ob der JIT compiler (c2) zusammenfliegt. Ist
>>> das richtig? Schaut mir jedenfalls nach nem HotSpot bug aus.
>>>
>>> Vielleicht hilft's! Viele Grüße, Rafael
>>>