RFR: 8346123: [REDO] NMT should not use ThreadCritical

[David Holmes]

On Mon, 16 Dec 2024 22:15:45 GMT, Kim Barrett <kbarrett at openjdk.org> wrote:

>> This is a redo of [JDK-8304824](https://bugs.openjdk.org/browse/JDK-8304824) which was backed out by [JDK-8343726](https://bugs.openjdk.org/browse/JDK-8343726) due to problems documented in [JDK-8343244](https://bugs.openjdk.org/browse/JDK-8343244).
>> 
>> The problem was that `NmtVirtualMemoryLocker` was not locking when the current thread is not attached by checking `Thread::current_or_null_safe() != nullptr`. This is necessary during VM init, but should not be allowed afterward. NMT may be used in `attach_current_thread` before the current thread is set. The lock was not being acquired in that case, which intermittently caused NMT accounting to become corrupted, triggering various assertions when future NMT operations are done.  To fix this, I've adopted [Thomas' suggestion](https://github.com/openjdk/jdk/pull/21928#issuecomment-2460238057) to reverse the order of 
>> 
>> 
>> thread->register_thread_stack_with_NMT();
>> thread->initialize_thread_current();
>> 
>> 
>> in `attach_current_thread`.  This allows `NmtVirtualMemoryLocker` to be locked after current thread is set. 
>> 
>> To allow for `NmtVirtualMemoryLocker` to be used during VM init, I've replaced the `ConditionalMutexLocker` check `Thread::current_or_null_safe() != nullptr` with a new flag: `_done_bootstrap`. This flag prevents the lock from being used during VM init, at which point we are single threaded anyway. This avoids errors due to Hotspot mutexes and current thread not yet being ready. 
>> 
>> I also added new asserts in `virtualMemoryTracker.cpp` to catch future bugs like this where the lock is not held when it should be. I updated the appropriate VMT tests to also lock (there were a few cases where locking was being bypassed) so they can pass the new asserts.
>> 
>> I also removed the unused `_query_lock` variable in `MemTracker`.
>> 
>> Testing: 
>> 
>> - On Linux amd64, I was able to consistently reproduce the errors described in [JDK-8343244](https://bugs.openjdk.org/browse/JDK-8343244) by increasing the number of test threads in `java/lang/Thread/jni/AttachCurrentThread/AttachTest.java`. The test consistently passes with the new changes in this PR.
>> - hotspot_nmt , gtest:VirtualSpace, tier1
>
> src/hotspot/share/nmt/memTracker.hpp line 281:
> 
>> 279:   // Same as MutexLocker but can be used during VM init while single threaded and before mutexes are ready or current thread has been assigned.
>> 280:   // Performs no action during VM init.
>> 281:   class NmtVirtualMemoryLocker: public ConditionalMutexLocker {
> 
> Should not derive from ConditionalMutexLocker - it's not designed to be a base class.
> Either use has-a relationship, or just derive directly from MutexLockerImpl, since we don't need
> the assert in ConditionalMutexLocker anyway.  And I'm surprised there isn't a constructor taking
> the current thread, like all the other locker variants.

You should be able to use a ConditionalMutexLocker directly to handle this situation - it is one of the usecases that caused CML to be introduced.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/22745#discussion_r1887738017