JDK-6855809 : Unaligned memory access on Solaris within java.io.BufferedInputStream
  • Type: Bug
  • Component: hotspot
  • Sub-Component: compiler
  • Affected Version: 6u14
  • Priority: P2
  • Status: Closed
  • Resolution: Duplicate
  • OS: solaris_10
  • CPU: sparc
  • Submitted: 2009-06-29
  • Updated: 2011-02-16
  • Resolved: 2010-02-11
Related Reports
Duplicate :  
JDK-6840775 - Multiple JVM crashes seen with 1.6.0_10 through early access of 1.6.0_14 - possibly related to GC
Description
FULL PRODUCT VERSION :
java version "1.6.0_12"
Java(TM) SE Runtime Environment (build 1.6.0_12-b04)
Java HotSpot(TM) Server VM (build 11.2-b01, mixed mode)

(Error also occurs in 1.6.0_14)

FULL OS VERSION :
SunOS 5.10 Generic_137137-09 sun4v sparc SUNW,SPARC-Enterprise-T5220

A DESCRIPTION OF THE PROBLEM :
During the normal operation of our application we read/write lots of data. Our code uses the java.io.BufferedInputStream class to read this data. Intermittently the JVM will crash and produce a Hot Spot error, stating the problematic frame as java.io.BufferedInputStream.read().

Initial research into the error has drawn our attention to Unaligned Memory Access issues that can occur in the Solaris SPARC architecture. We have the same set up on a RedHat Linux environment (on an HP server) and no Hot Spot errors occur, though research indicates x86 architectures can handled unaligned memory access.

Another update is that the VMs did not have the -d64 flag set, though the version of java used to launch them was the 64-bit version, as detailed in the bug report. I'm not certain what effect, if any, this would have had.

THE PROBLEM WAS REPRODUCIBLE WITH -Xint FLAG: The testing we have done when the -Xint flag is set on the VMs shows that the issue does not occur.

THE PROBLEM WAS REPRODUCIBLE WITH -server FLAG: Did not try

ERROR MESSAGES/STACK TRACES THAT OCCUR :
Attached seperatly


REPRODUCIBILITY :
This bug can be reproduced occasionally.
Comments
EVALUATION waiting for the release of 6u18 and customer's confirm of the fix.
19-12-2009
EVALUATION According to the hs_err log, this is most likely to be a duplicate of 6840775. I have asked the customer to try -XX:-ReduceBulkZeroing as a workaround, and the customer has confirmed the problem did go away with this Vm option. The fix of 6840775 has been integrated into hotspot 16, which is in 6u18. The customer has been informed about this and is waiting for the official release of 6u18.
14-12-2009
EVALUATION Customer provided the core file (with 6u17). I put the core file in /home/cf231476/bug6855809.
14-12-2009
EVALUATION Not enough information to triage the problem. We need at least a core file to work on. From the attached hs_err_decoded file, it looks like an object pointer was not set up correctly. I have suggested the customer to try 6u18 (http://download.java.net/jdk6/index.html), and if it still fails, send us the core file (pending response).
18-11-2009