Fatal Error on PPC64 and PPC64LE for both JDK7 and JDK8

Sat Apr 18 18:22:40 UTC 2015

Volker and Goetz,

RHEL 7.1's spec file for java-1.8.0-openjdk had a clause for ppc64le
that enabled the CC_INTERP=true flag. I opened a bug with RedHat and
attached a small patch just to remove that entry.

Ubuntu seems safe and is using the template interpreter for both 7 & 8.
Later on I try to get a machine running SLES to check how they are
building openjdk.

Meanwhile, please keep me posted if you figure out something about the
bug. I'm still investigating it myself, but unless I am really lucky it
will take some time to understand what is going on with that part of the
hotspot - assuming I'm even looking into the right place. =)

Regards,
Tiago

On Sat, 2015-04-18 at 08:22 +0200, Volker Simonis wrote:
> On Fri, Apr 17, 2015 at 9:42 PM, Tiago Sturmer Daitx
> <tdaitx at linux.vnet.ibm.com> wrote:
> > Goetz,
> >
> > You are right, it seems RHEL 7.1 JDK 8 is using the C++ interpreter:
> >
> > $ /usr/lib/jvm/java-1.8.0-openjdk/bin/java -XX:
> > +UnlockDiagnosticVMOptions -XX:+PrintInterpreter -version
> > ----------------------------------------------------------------------
> > Interpreter
> >
> > code size        =      5K bytes
> > total space      =     11K bytes
> > wasted space     =      6K bytes
> >
> > # of codelets    =      3
> > avg codelet size =   2005 bytes
> >
> >
> > ----------------------------------------------------------------------
> > slow signature handler  [0x00003fff98005f80, 0x00003fff98006400]  1152
> > bytes
> >
> > Could not load hsdis-ppc64le.so; library not loadable; PrintAssembly is
> > disabled
> >
> > ----------------------------------------------------------------------
> > result handlers for native calls  [0x00003fff98006480,
> > 0x00003fff98006680]  512 bytes
> >
> >
> > ----------------------------------------------------------------------
> > (kind = frame_manager)  [0x00003fff98006700, 0x00003fff98007680]  3968
> > bytes
> >
> > ----------------------------------------------------------------------
> >
> > openjdk version "1.8.0_45"
> > OpenJDK Runtime Environment (build 1.8.0_45-b13)
> > OpenJDK 64-Bit Server VM (build 25.45-b02, mixed mode)
> >
> >
> > Running the same flags against RHEL 7.1 JDK 7 or the custom build jdk7u
> > shows a muuuuch bigger output, which I assume to be the template.
> >
> > Anyway, would you still be able to help with the JDK 7? That one is
> > running the Template as far as I can tell.
> >
> > Meanwhile I will take a look at source RPM for JDK8 and check what they
> > are doing there.
> >
> > I'm curious: why would the C++ interpreter still be enabled/available if
> > it is not maintained?
> >
> 
> The decision template vs. c++ interpreter is a build time decision
> (--with-jvm-interpreter=template|cpp with 'template' being the
> default). It's probably this configure option which is still lurking
> around in the build scripts of some of the Linux distros and which we
> strongly recommend to remove for Linux/ppc64.
> 
> Regards,
> Volker
> 
> > Regards,
> > Tiago
> >
> > On Fri, 2015-04-17 at 13:49 +0000, Lindenmaier, Goetz wrote:
> >> Hi Tiago,
> >>
> >> from the hs_err-pid24587-1.8.0.45.log it looks like you are using the bytecode interpreter (C++ interpreter).
> >> Why that?
> >> Is this the case for all the VMs you tested?
> >> I can’t tell whether the bug is related to that, but we didn’t test it
> >> for about a year now.
> >> Could you please try to reproduce the problem with the template interpreter?
> >> That would be great!
> >>
> >> Best regards,
> >>   Goetz.
> >>
> >> -----Original Message-----
> >> From: Tiago Sturmer Daitx [mailto:tdaitx at linux.vnet.ibm.com]
> >> Sent: Freitag, 17. April 2015 15:17
> >> To: Lindenmaier, Goetz
> >> Subject: Re: Fatal Error on PPC64 and PPC64LE for both JDK7 and JDK8
> >>
> >> Goetz,
> >>
> >> Thanks for your help!
> >>
> >> I have attached a zip file containing a few hs_err reports. They have:
> >> - the pid number, in case you want additional debugging information,
> >> just refer to it
> >> - which of the 3 JVMs was used to generate them, between RHEL 7.1
> >> OpenJDK 7 and 8 and a custom jdk7u build from ppc-aix-port repo
> >>
> >> You will see 2 files for each jvm above, they come from 2 components:
> >> oozie/core and oozie/sharedlib/pig.
> >>
> >> I do have more if those are not enough.
> >>
> >> If at all possible (I completely understand if you are unable to), I
> >> would really appreciate if you could share what kind of information you
> >> were able to find and how (no need to be very detailed, just enough so I
> >> can take a peek and try to understand it myself). I tried to get some
> >> information out of the hs_err files but didn't amount to much, that's
> >> why I decided to install the debug symbols and poke around.
> >>
> >> Regards,
> >> Tiago
> >>
> >> On Fri, 2015-04-17 at 10:03 +0000, Lindenmaier, Goetz wrote:
> >> > Hi Tiago
> >> >
> >> >
> >> >
> >> > Can you send us the full hs_err file?   What’s the Java args?
> >> >
> >> >
> >> >
> >> > Best regards,
> >> >
> >> >   Goetz.
> >> >
> >> >
> >> >
> >> > PS: I included your power 8 recognition in a larger change I
> >> >
> >> > want to submit into the ppc directory.  Will come right after
> >> >
> >> > my change to adlc was pushed, which is a prerequisite.
> >> >
> >> >
> >> >
> >> > From: ppc-aix-port-dev
> >> > [mailto:ppc-aix-port-dev-bounces at openjdk.java.net] On Behalf Of Tiago
> >> > Sturmer Daitx
> >> > Sent: Freitag, 17. April 2015 01:07
> >> > To: ppc-aix-port-dev at openjdk.java.net
> >> > Subject: Re: Fatal Error on PPC64 and PPC64LE for both JDK7 and JDK8
> >> >
> >> >
> >> >
> >> >
> >> > On Thu, 2015-04-16 at 20:00 -0300, Tiago Sturmer Daitx wrote:
> >> >
> >> >
> >> >
> >> >
> >> > JDK7u from PPC-AIX-Port repository
> >> > ----------------------------------------------
> >> >
> >> >
> >> > $ grep -e "Core dump" -e "hs_error" oozie-ppc-aix-port.stdout
> >> > # Core dump written. Default location: /home/tdaitx/oozie/client/target/site/apidocs/core or core.26317
> >> > # Core dump written. Default location: /home/tdaitx/oozie/core/core or core.29406
> >> > # Core dump written. Default location: /home/tdaitx/oozie/sharelib/pig/core or core.30071
> >> > # Core dump written. Default location: /home/tdaitx/oozie/sharelib/hive/core or core.30108
> >> >
> >> >
> >> >
> >> > Sorry, I failed earlier to include a hs_err file just as example for
> >> > the custom ppc-aix-port jdk7u build.
> >> >
> >> >
> >> > # A fatal error has been detected by the Java Runtime Environment:
> >> > #
> >> > #  Internal Error (/home/tdaitx/ppc-aix-port-jdk7u/hotspot/src/share/vm/gc_implementation/parallelScavenge/parMarkBitMap.cpp:263), pid=30108, tid=70365572231632
> >> > #  assert(*p == 0) failed: bitmap not clear
> >> > #
> >> > # JRE version: OpenJDK Runtime Environment (7.0) (build 1.7.0-internal-fastdebug-tdaitx_2015_04_16_17_13-b00)
> >> > # Java VM: OpenJDK 64-Bit Server VM (24.75-b04-fastdebug mixed mode linux-ppc64 compressed oops)
> >> > # Core dump written. Default location: /home/tdaitx/oozie/sharelib/hive/core or core.30108
> >> > #
> >> > # If you would like to submit a bug report, please visit:
> >> > #   http://bugreport.sun.com/bugreport/crash.jsp
> >> > #
> >> >
> >> > ---------------  T H R E A D  ---------------
> >> >
> >> > Current thread (0x00003fff88341000):  VMThread [stack: 0x00003fff42d00000,0x00003fff42f00000] [id=30123]
> >> >
> >> > Stack: [0x00003fff42d00000,0x00003fff42f00000],  sp=0x00003fff42efda00,  free space=2038k
> >> > Native frames: (J=compiled Java code, j=interpreted, Vv=VM code, C=native code)
> >> > V  [libjvm.so+0xe55120]  VMError::report_and_die()+0x240
> >> > V  [libjvm.so+0x548ef4]  report_vm_error(char const*, int, char const*, char const*)+0xb4
> >> > V  [libjvm.so+0xbb7bbc]  ParMarkBitMap::verify_clear() const+0x8c
> >> > V  [libjvm.so+0xc4c930]  PSParallelCompact::pre_compact(PreGCValues*)+0x290
> >> > V  [libjvm.so+0xc5b668]  PSParallelCompact::invoke_no_policy(bool)+0x258
> >> > V  [libjvm.so+0xc6c444]  PSScavenge::invoke()+0x434
> >> > V  [libjvm.so+0xbc8d58]  ParallelScavengeHeap::failed_mem_allocate(unsigned long)+0x1a8
> >> > V  [libjvm.so+0xe581b4]  VM_ParallelGCFailedAllocation::doit()+0x174
> >> > V  [libjvm.so+0xe7fb78]  VM_Operation::evaluate()+0xb8
> >> > V  [libjvm.so+0xe7ca10]  VMThread::evaluate_operation(VM_Operation*)+0x410
> >> > V  [libjvm.so+0xe7d6fc]  VMThread::loop()+0x6bc
> >> > V  [libjvm.so+0xe7d9d4]  VMThread::run()+0xf4
> >> > V  [libjvm.so+0xb801b0]  java_start(Thread*)+0x170
> >> > C  [libpthread.so.0+0x8944]  start_thread+0xf4
> >> >
> >> > VM_Operation (0x00003ffd3b9f8440): ParallelGCFailedAllocation, mode: safepoint, requested by thread 0x00003fff11cba800
> >> >
> >> >
> >> >
> >> >
> >>
> >
> >
> 