RFR (S) 8245487: Potential double-free of interfaces array

coleen.phillimore at oracle.com coleen.phillimore at oracle.com
Thu Jun 4 15:00:40 UTC 2020

On 6/4/20 10:34 AM, Lois Foltan wrote:
> On 6/4/2020 7:36 AM, coleen.phillimore at oracle.com wrote:
>> Summary: free _local_interfaces when we free _transitive_interfaces 
>> since they can share memory.
>> Tested with tier1-3, also test case with fastdebug and product 
>> builds.  ClassLoaderDataGraph::verify is enabled in product for 
>> -XX:+VerifyBeforeExit.
>> open webrev at 
>> http://cr.openjdk.java.net/~coleenp/2020/8245487.01/webrev
>> bug link https://bugs.openjdk.java.net/browse/JDK-8245487
>> Thanks to Fred for finding the problem and testing on valhalla where 
>> it was first observed, and to Ioi for the test case (and finding the 
>> problem too!)
>> Thanks,
>> Coleen
> Looks good Coleen!
> Minor comment on the summary description within the test, 
> classFileParserBug/TestBadPackageWithInterface.java.  The test 
> runtime/modules/JVMDefineModule.java does already test for a resulting 
> IAE when a package named "java" is defined to a class loader other 
> than the boot or platform class loader.  So my understanding is that 
> TestBadPackageWithInterface.java is really testing that if an 
> exception is raised when a class is defined, then 
> InstanceKlass::verify_on() will exit cleanly, correct?

Thank you for reviewing.  How about this for the comment:

  * @summary Check that if the VM rejects classes from packages starting 
with "java/", it will exit
  *          cleanly after InstanceKlass::verify_on(), and not leave 
freed memory in _local_interfaces.

> Thanks,
> Lois

More information about the hotspot-runtime-dev mailing list