Request for review - 7199349

Jon Masamitsu jon.masamitsu at oracle.com
Mon Oct 1 14:51:33 PDT 2012


Thanks.

On 10/01/12 13:55, John Cuthbertson wrote:
> Hi Jon,
>
> Looks good to me.
>
> JohnC
>
> On 09/28/12 16:59, Jon Masamitsu wrote:
>> 7199349: NPG: PS: Crash seen in jprt
>>
>> The error is caused by the calculation in scavenge_contents_parallel()
>> of start_card and end_card where end_card was incorrectly set beyond
>> start_card.  The incorrect code had been introduced to fix a problem
>> exposed when the card table for perm gen was removed.  That first
>> problem was an assertion error resulting when byte_for() was passed
>> an addressed not covered by the card table.   I reverted to
>> the previous code and do not invoke this code when the old gen
>> is empty.  scavenge_contents_parallel() processes the card table
>> to find old-to-young pointers and so is not needed in that case.
>>
>> Added a delay in two places in the young collection to widen the
>> opportunity for these circumstances to occur.
>>
>> http://cr.openjdk.java.net/~jmasa/7199349
>>
>> Thanks to StefanK for his analysis of the problem.
>


More information about the hotspot-dev mailing list