Request for review: 7059899: Stack overflows in Java code cause 64-bit JVMs to exit due to SIGSEGV

Chris Hegarty chris.hegarty at oracle.com
Mon Feb 13 07:39:35 PST 2012


Karen,

One of the issues that provokes this problem is the native networking 
socket read/write code on unix. When running a 64bit JDK the read/write 
implementation can try to create an on stack buffer of up to 128K. I 
believe this may have been greater than the old StackShadowPages size.

-Chris.

On 13/02/2012 15:11, Karen Kinnear wrote:
> Coleen,
>
> That was quick. Code change looks good.
> Is there a reason the larger size isn't needed on all platforms?
>
> thanks,
> Karen
>
> On Feb 13, 2012, at 7:23 AM, Coleen Phillimore wrote:
>
>> Summary: Increase StackShadowPages to accomodate the JDK changes to increase buffer size in socketWrite
>>
>> open webrev at http://cr.openjdk.java.net/~coleenp/7059899/
>> bug link at http://bugs.sun.com/view_bug.do?bug_id=7059899
>>
>> Tested with test case and internal stack overflow tests.
>>
>> Thanks,
>> Coleen
>>
>>
>


More information about the hotspot-runtime-dev mailing list