RFR(xxs): 8213236: A partial removed/deleted JavaThread cannot transition

[Robbin Ehn]

Hi Dan,
>> I don't think we should grab Heap_lock at all and we should just remove it.
>> If there is a problem with "unparseable heap", taking Heap_lock seems to the 
>> wrong solution.
> 
> I don't think you want to drob the grab of the Heap_lock from
> this code path as part of this fix. That should be done separately.
> Of course, you have GC folks close to you so you might be able to
> resolve the issue (theoretically) quickly. However, I think that
> will mean that this fix needs more testing.

I was wrong, we can do as you say just skip safepoint checks on Heap_lock.

The incremental is:
diff -r f5f6ed30bb13 src/hotspot/share/runtime/thread.cpp
--- a/src/hotspot/share/runtime/thread.cpp	Thu Nov 01 20:10:18 2018 +0100
+++ b/src/hotspot/share/runtime/thread.cpp	Thu Nov 01 20:17:58 2018 +0100
@@ -4289,1 +4289,1 @@
-    MutexLocker ml(Heap_lock);
+    MutexLockerEx ml(Heap_lock, Mutex::_no_safepoint_check_flag);
diff -r f5f6ed30bb13 src/hotspot/share/runtime/vmThread.cpp
--- a/src/hotspot/share/runtime/vmThread.cpp	Thu Nov 01 20:10:18 2018 +0100
+++ b/src/hotspot/share/runtime/vmThread.cpp	Thu Nov 01 20:17:58 2018 +0100
@@ -323,1 +323,1 @@
-  assert(((JavaThread*)Thread::current())->is_terminated(), "Should be off 
threadslist");
+  assert(((JavaThread*)Thread::current())->is_terminated(), "Should be 
terminated");

Re-running t1-2, I expect no problems.

Thanks!

/Robbin

> 
> Dan
> 
> 
>>
>> /Robbin
>>
>>>
>>> Dan
>>>
>>>
>>>
>>>> _should_terminate = true;
>>>>      VMOperationQueue_lock->notify();
>>>>    }
>>>>
>>>> Passes t1-2.
>>>>
>>>> Thanks Robbin
>>>
>