PING: Check PerfCounter in core image through JSnap

[Yasumasa Suenaga]

Ping: What do you think about this issue?

Yasumasa

On 2016/02/08 19:17, Yasumasa Suenaga wrote:
> Ping: Can I file it to JBS as new issue?
>
> Yasumasa
>
> 2016/01/30 23:08 "Yasumasa Suenaga" <yasuenag at gmail.com <mailto:yasuenag at gmail.com>>:
>
>     Hi all,
>
>     I want to check PerfCounter in core image to analyze state of crashed JVM.
>     When I use JSnap tool, I encountered NPE on Fedora 23 x86_64.
>
>     I checked core image, and I found PerfMemory::_start, _top, etc are set to NULL,
>     however, PerfMemory::_initialized is set to 1.
>
>     PerfMemory::_start is set to NULL in PerfMemory::destroy(), and this function
>     is called at before aborting VM.
>
>     I fixed as below, and it works fine with JSnap:
>     -------------------
>     diff -r edfc9e2316e3 src/share/vm/runtime/perfMemory.cpp
>     --- a/src/share/vm/runtime/perfMemory.cpp       Sun Jan 24 23:20:12 2016 +0900
>     +++ b/src/share/vm/runtime/perfMemory.cpp       Sat Jan 30 22:41:05 2016 +0900
>     @@ -196,12 +196,6 @@
>           //
>           delete_memory_region();
>         }
>     -
>     -  _start = NULL;
>     -  _end = NULL;
>     -  _top = NULL;
>     -  _prologue = NULL;
>     -  _capacity = 0;
>       }
>
>       // allocate an aligned block of memory from the PerfData memory
>     -------------------
>
>     I found JDK-6224040: enhance jstat to extract counter information from core files .
>     At least, we can control shared memory through coredump_filter in modern Linux kernel.
>
>     Should I work this fix in JDK-6224040 ? or another issue ?
>     Please cooperate.
>
>
>     Thanks,
>
>     Yasumasa
>
>