8076110: VM crash when class is redefined with Instrumentation.redefineClasses

[Andreas Eriksson]

Hi,

Deleting the resolution errors from InstanceKlass::deallocate contents 
works, but I'm unsure if the deletion code should be placed inside the 
is_shared() if-block or not.
I don't think that CDS should affect our decision to delete the cached 
resolution errors, but I might be missing something.
Coleen, do you know if it affects anything?

New webrev:
http://cr.openjdk.java.net/~aeriksso/8076110/webrev.01/

Regards,
Andreas

On 2015-06-09 01:49, serguei.spitsyn at oracle.com wrote:
> Hi Andreas,
>
> I agree, it is a good idea to delete the resolution errors from 
> ConstantPool::deallocate_contents.
> So, I'm waiting for next webrev.
>
> Thanks,
> Serguei
>
> On 6/8/15 9:53 AM, Andreas Eriksson wrote:
>> Hi Coleen, thanks for looking at this.
>>
>> That might work, I'll make a build to test it.
>>
>> Just FYI:
>> My first attempt to fix this was actually to delete the resolution 
>> errors from ConstantPool::deallocate_contents, but that was sometimes 
>> called when not at safepoint, which made me do this change instead.
>>
>> Regards,
>> Andreas
>>
>>
>> On 2015-06-06 00:07, Coleen Phillimore wrote:
>>>
>>> Thank you so much for fixing this, Andreas!
>>>
>>> I was wondering if you could delete the resolution errors in 
>>> InstanceKlass::deallocate_contents instead at the point where you 
>>> free the constant pool?   This code is at a safepoint at this time, 
>>> even with the CMS collector.
>>>
>>> Thanks,
>>> Coleen
>>>
>>> On 6/4/15 5:23 PM, serguei.spitsyn at oracle.com wrote:
>>>> Added the SVC list as it impacts the Serviceability as well.
>>>>
>>>> Thanks,
>>>> Serguei
>>>>
>>>> On 6/4/15 4:56 AM, Andreas Eriksson wrote:
>>>>> Hi,
>>>>>
>>>>> Please review this fix of a crash because of a class redefinition 
>>>>> race.
>>>>>
>>>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8076110
>>>>> Webrev: http://cr.openjdk.java.net/~aeriksso/8076110/webrev.00/
>>>>>
>>>>> The crash happens when the redefine occurs while the redefined 
>>>>> method is still running.
>>>>> This is because VM_RedefineClasses::redefine_single_class deletes 
>>>>> the resolution errors at redefine time, but the running method 
>>>>> might still need them.
>>>>>
>>>>> This is fixed by moving the deletion of the cached resolution 
>>>>> errors to when the {add,purge}_previous_version functions have 
>>>>> come to the conclusion that no running methods need the old class 
>>>>> any more (and therefore no one needs the constant pool and its 
>>>>> associated resolution errors either).
>>>>>
>>>>> A jtreg regression test that reproduces the problem using 
>>>>> redefineclasses and objectweb.asm is included in the change.
>>>>>
>>>>> Thanks,
>>>>> Andreas
>>>>
>>>
>>
>