[URGENT]: [11u] RFR 8244287: JFR: Methods samples have line number 0

[Jaroslav Bachorík]

Hi Goetz,

I support rolling back the commit. Will you do it or should I initiate
the process?

Regards,

-JB-

On Tue, Sep 29, 2020 at 5:01 PM Thomas Stüfe <thomas.stuefe at gmail.com> wrote:
>
> Hi Goetz,
>
> not a JFR expert, but I had a look at the code both in head and in 11u.
>
> While the fix proposed by Martin seems sufficient, I would rather roll back the initial commit. The code in question in head had been rewritten in jdk14 for https://bugs.openjdk.java.net/browse/JDK-8236743 which fixes a problem with tracing code via JFR when classes are redefined. As a side effect it also fixes the original issue fixed by 8244287 by passing the line number to the constructor. I would rather go back to a stable point and later integrate JDK-8236743.
>
> Cheers Thomas
>
>
>
> On Tue, Sep 29, 2020 at 4:18 PM Lindenmaier, Goetz <goetz.lindenmaier at sap.com> wrote:
>>
>> Hi,
>>
>> Any opinion on this?
>> Planned freeze of jd11u is in 2 hours ...
>>
>> Best regards,
>>   Goetz.
>>
>> > -----Original Message-----
>> > From: Doerr, Martin <martin.doerr at sap.com>
>> > Sent: Tuesday, September 29, 2020 3:11 PM
>> > To: Lindenmaier, Goetz <goetz.lindenmaier at sap.com>; Jaroslav Bachorík
>> > <jaroslav.bachorik at datadoghq.com>; jdk-updates-dev <jdk-updates-
>> > dev at openjdk.java.net>; hotspot-jfr-dev <hotspot-jfr-dev at openjdk.java.net>
>> > Subject: RE: [11u] RFR 8244287: JFR: Methods samples have line number 0
>> >
>> > Hi all,
>> >
>> > we have seen several crashes on several platforms after this change was
>> > pushed.
>> >
>> > Aarch64 example:
>> > Stack: [0x0000ffff90437000,0x0000ffff90637000],  sp=0x0000ffff90635520,
>> > free space=2041k
>> > Native frames: (J=compiled Java code, A=aot compiled Java code,
>> > j=interpreted, Vv=VM code, C=native code)
>> > V  [libjvm.so+0x9fb0f8]
>> > ObjectSampleCheckpoint::write_stacktrace(JfrStackTrace const*,
>> > JfrCheckpointWriter&)+0x208
>> > V  [libjvm.so+0x9fb770]
>> > StackTraceBlobInstaller::install(ObjectSample*)+0x1b8
>> > V  [libjvm.so+0x9fba40]
>> > ObjectSampleCheckpoint::on_rotation(ObjectSampler const*,
>> > JfrStackTraceRepository&)+0x150
>> > V  [libjvm.so+0x72df6c]  JfrRecorderService::pre_safepoint_write()+0x3a4
>> > V  [libjvm.so+0x72e828]  JfrRecorderService::write()+0x58
>> > V  [libjvm.so+0x72fd2c]  JfrRecorderService::rotate(int)+0x3b4
>> > V  [libjvm.so+0x730628]  recorderthread_entry(JavaThread*,
>> > Thread*)+0x160
>> > V  [libjvm.so+0xbb2dd0]  JavaThread::thread_main_inner()+0xd8
>> > V  [libjvm.so+0xbb08fc]  Thread::call_run()+0x94
>> > V  [libjvm.so+0xa21718]  thread_native_entry(Thread*)+0x108
>> > C  [libpthread.so.0+0x7088]  start_thread+0xb0
>> >
>> > siginfo: si_signo: 11 (SIGSEGV), si_code: 1 (SEGV_MAPERR), si_addr:
>> > 0x0000000000000008
>> >
>> >
>> > This backport uses a different JfrStackFrame constructor (which was
>> > probably unused before):
>> > JfrStackFrame::JfrStackFrame(const traceid& id, int bci, int type, int lineno) :
>> >   _method(NULL), _methodid(id), _line(lineno), _bci(bci), _type(type) {}
>> >
>> > It sets _method to NULL!
>> >
>> > Should probably be:
>> > JfrStackFrame::JfrStackFrame(const traceid& id, int bci, int type, int lineno,
>> > const Method* method) :
>> >   _method(method), _methodid(id), _line(lineno), _bci(bci), _type(type) {}
>> >
>> > and:
>> > _frames[count] = JfrStackFrame(mid, bci, type, lineno, method);
>> >
>> > Do you think we can fix this quickly or should we better revert this backport?
>> >
>> > Best regards,
>> > Martin
>> >
>> >
>> > > -----Original Message-----
>> > > From: jdk-updates-dev <jdk-updates-dev-bounces at openjdk.java.net> On
>> > > Behalf Of Lindenmaier, Goetz
>> > > Sent: Montag, 22. Juni 2020 14:42
>> > > To: Jaroslav Bachorík <jaroslav.bachorik at datadoghq.com>; jdk-updates-
>> > dev
>> > > <jdk-updates-dev at openjdk.java.net>; hotspot-jfr-dev <hotspot-jfr-
>> > > dev at openjdk.java.net>
>> > > Subject: RE: [11u] RFR 8244287: JFR: Methods samples have line
>> > > number 0
>> > >
>> > > Hi Jaroslav,
>> > >
>> > > The fix looks good to me.  Reviewed.
>> > > Please flag the change with jdk11u-fix-request and
>> > > add corresponding comment.
>> > >
>> > > Best regards,
>> > >   Goetz
>> > >
>> > >
>> > > -----Original Message-----
>> > > From: jdk-updates-dev <jdk-updates-dev-bounces at openjdk.java.net> On
>> > > Behalf Of Jaroslav Bachorík
>> > > Sent: Donnerstag, 18. Juni 2020 17:27
>> > > To: jdk-updates-dev <jdk-updates-dev at openjdk.java.net>; hotspot-jfr-dev
>> > > <hotspot-jfr-dev at openjdk.java.net>
>> > > Subject: [11u] RFR 8244287: JFR: Methods samples have line number 0
>> > >
>> > > Hi all,
>> > >
>> > > Could I get this simple fix reviewed?
>> > > It is fixing a regression introduced while backporting JDK-823790 and
>> > > adding a new test to prevent such regressions in the future.
>> > >
>> > > JIRA      :  https://bugs.openjdk.java.net/browse/JDK-8244287
>> > > Webrev : http://cr.openjdk.java.net/~jbachorik/8244287/webrev/
>> > >
>> > > Thanks,
>> > >
>> > > -JB-