RFR (S) 8224151: Test vmTestbase/nsk/sysdict/vm/stress/chain/chain007/chain007.java might hang with release binaries

[coleen.phillimore at oracle.com]

On 6/6/19 1:56 PM, Erik Österlund wrote:
> Hi Coleen,
>
> Could you please add Atomic::load() on the CLD::next getter used by 
> the iterator? No need for new webrev. Should remove the lock on the 
> reader side in the future by making the unloading list use a different 
> next pointer. Then the reader side locking won't really do anything 
> for us.

Okay, I'll do it but there are more places we'd have to fix to make all 
the readers lock free.  To be continued ...

Thanks!
Coleen

>
> Thanks,
> /Erik
>
> On 2019-06-06 03:53, coleen.phillimore at oracle.com wrote:
>> Summary: Fix deadlock on ClassLoaderDataGraph_lock and g1 
>> clear_claimed_marks.
>>
>> Thank you to Stefan for finding the deadlock and Thomas for 
>> suggestions on the fix, plus text for comments.
>>
>> Tested with this test 100s of times, tier1-3 and tier6 almost complete.
>>
>> open webrev at 
>> http://cr.openjdk.java.net/~coleenp/2019/8224151.01/webrev
>> bug link https://bugs.openjdk.java.net/browse/JDK-8224151
>>
>> Thanks,
>> Coleen