RFR: 8009204 : [dtrace] signatures returned by Java 7 jstack() are corrupted on Solaris

serguei.spitsyn at oracle.com serguei.spitsyn at oracle.com
Sun Jun 30 01:06:09 PDT 2013


Hi Tomas,

The fix looks good.
Thank you for discovering, tracking down and fixing the issue!

Thanks,
Serguei

On 6/28/13 6:29 AM, Tomas Hurka wrote:
> Hi All,
> I would like to ask for review of JDK-8009204 <http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=8009204>
>
> Description of the problem:
> jstack() returns corrupted method signatures on Solaris. The issue can be easily reproduce using the following DTrace script:
>
> dtrace -x jstackstrsize=2048 -Z \
> 	-n 'hotspot_jni$target::: /0/{}' \
> 	-n 'syscall::write:entry /pid==$target/{jstack(1024)}' \
> 	-c 'java -version'
>
> With Java 6, the output is roughly:
>
>               java/io/FileOutputStream.writeBytes([BII)V
>               java/io/FileOutputStream.write([BII)V
>               java/io/BufferedOutputStream.flushBuffer()V
>               java/io/BufferedOutputStream.flush()V
>               java/io/PrintStream.write([BII)V
>
> With Java 7 and Java 8, the output is roughly:
>
>               java/io/FileOutputStream.riteBytes
>               java/io/FileOutputStream.rite
>               java/io/BufferedOutputStream.lushBuffer
>               java/io/BufferedOutputStream.lush
>               java/io/PrintStream.rite
>
> Evaluation:
> The problem is caused by SymbolTable changes JDK-6990754. jhelper.d was never updated with changes for CPSlot, so the low bit of the address of the Symbol is set, which causes the off by 1-ness of the output. The klass name uses a untagged constant pool entry, which is why it doesn't have the problem in the output. There is a very similar bug JDK-7019165 reported against pstack output and the proposed fix is basically backport of JDK-7019165 to jhelper.d.
>
> Webrev:
> <http://cr.openjdk.java.net/~thurka/8009204/webrev.00/>
>
> I tested the fix using above D-script on latest JDK 8 and JDK 7u sources.
>   
> Thanks,
> --
> Tomas Hurka   <mailto:tomas.hurka at oracle.com>
> NetBeans Profiler http://profiler.netbeans.org
> VisualVM http://visualvm.java.net
> Software Developer
> Oracle, Praha Czech Republic
>



More information about the serviceability-dev mailing list