[8u-dev] Request for approval: 8162766: Unsafe_DefineClass0 accesses raw oops while in _thread_in_native

[Coleen Phillimore]

Thanks!  Thanks for pointing out that JVM_GetClassLoader wasn't deleted.
Coleen

On 10/7/16 5:22 PM, Leela Mohan wrote:
> Looks good.
>
> Thanks,
> Leela
>
> On Fri, Oct 7, 2016 at 2:20 PM, Coleen Phillimore 
> <coleen.phillimore at oracle.com <mailto:coleen.phillimore at oracle.com>> 
> wrote:
>
>
>     Here's the (retested) fix using JVM_GetClassLoader.
>
>     open webrev at
>     http://cr.openjdk.java.net/~coleenp/8162766.02/webrev
>     <http://cr.openjdk.java.net/%7Ecoleenp/8162766.02/webrev>
>
>     From the history it seems that this part was backported with the
>     Class.getClassLoader() speedup change, and the code was corrected
>     in jdk9.
>
>     Thanks,
>     Coleen
>
>
>     On 10/7/16 4:19 PM, Leela Mohan wrote:
>>     Hi Coleen,
>>
>>     Sorry i never came back after we found the issue. We are
>>     certainly doing many unnecessary transitions from native to vm.
>>     It seems like, we can certainly avoid them if we have UNSAFE_LEAF
>>     (equivalent to JVM_LEAF) . But that might be larger change.
>>
>>     Coming back to your change, I am wondering if we can use
>>     "JVM_GetClassLoader" instead.
>>
>>     Thanks,
>>     Leela
>>
>>     On Fri, Oct 7, 2016 at 1:01 PM, Coleen Phillimore
>>     <coleen.phillimore at oracle.com
>>     <mailto:coleen.phillimore at oracle.com>> wrote:
>>
>>         Summary: Go to thread_in_vm to get the class_loader oop.
>>
>>         This error was found by Leela Venati.  Tested with jtreg
>>         hotspot/test.
>>
>>         open webrev at
>>         http://cr.openjdk.java.net/~coleenp/8162766.01/webrev
>>         <http://cr.openjdk.java.net/%7Ecoleenp/8162766.01/webrev>
>>         bug link https://bugs.openjdk.java.net/browse/JDK-8162766
>>         <https://bugs.openjdk.java.net/browse/JDK-8162766>
>>
>>         Thanks,
>>         Coleen
>>
>>
>
>