hs25 review request (2nd round): 8008511 JSR 292: MemberName vmtarget refs to methods must be updated at class redefinition

serguei.spitsyn at oracle.com serguei.spitsyn at oracle.com
Fri Apr 12 19:13:43 PDT 2013

Please, review the following fix.

This is the 2nd version that has a review fixes suggested by Coleen and 
John in the 1st review round.

Now the mtClass is used in allocation of MemberNameTable since it's part 
of class metadata.
All call sites in the methodHandles.cpp that recursively invoke the 
are made safepoint-safe by replacing oop's with Handle's.
This allowed to get rid of the No_Safepoint_Verifier in the 
The MemberName's are added to MNT not only for methods but also for fields.

You also suggested to put all new non-jvmti code under #if INCLUDE_JVMTI 
I've not done it because now the MemberName's associated with the fields are
also registered in the MNT which is not needed for JVMTI purposes.
Just have some doubt if it is a right thing to do.
I'm Ok to add this modification, but could you, confirm that it is still 



The problem is that the old version of the bootstrap method is re-invoked
after a popframe from the bootstrap method execution.
It is because the MemberName keeps a stale reference to the old method 

The solution (suggested by John) is to lazily create and keep up-to-date 
a MemberNameTable
which plays a role of MemberName cache assosiated with the InstanceKlass.
Then, at class redefinition, this cache is used to replace the old method
references in the MemberName's with the new method references.

The MemberNameTable is based on the GrowableArray<jweek>.
A C_HEAP array is allocated at the first call to 
It is released in the InstanceKlass::release_C_heap_structures().

A global week reference to member name oop is stored in the MemberNameTable.
It allowed to avoid having the oops_do() in the MemberNameTable.
Also, the MemberNameTable won't hold member name oops in memory.

The MemberNameTable_lock mutex is added to serialize MemberNameTable's 

Test coverage: vm.mlvm, nsk.jvmti, nsk.jdi tests on multiple platforms 
(32 vs 64-bit too).
The testing looks good.


More information about the hotspot-dev mailing list