[11u] RFR JDK-8213231 ThreadSnapshot::_threadObj can become stale

[Doerr, Martin]

Hi Jiangli,

your backport looks good. You only had to integrate manually because of unrelated changes in the context. That's fine.

Best regards,
Martin


> -----Original Message-----
> From: jdk-updates-dev <jdk-updates-dev-retn at openjdk.java.net> On
> Behalf Of Jiangli Zhou
> Sent: Sonntag, 21. März 2021 01:10
> To: jdk-updates-dev <jdk-updates-dev at openjdk.java.net>
> Subject: [11u] RFR JDK-8213231 ThreadSnapshot::_threadObj can become
> stale
> 
> Hello,
> 
> We recently noticed JDK-8213231 related GC crashes in production
> environments when running on JDK 11. The crashes were always
> associated with a stale pointer from a java.lang.management.ThreadInfo
> object to a bad j.l.String object. Applying JDK-8213231's fix resolved
> the specific crashes. I'd like to backport JDK-8213231 to JDK 11u.
> 
> Bug: https://bugs.openjdk.java.net/browse/JDK-8213231
> Original commit: http://hg.openjdk.java.net/jdk/jdk/rev/391d671f222b
> Webrev: http://cr.openjdk.java.net/~jiangli/backport-8213231/webrev.00/
> 
> The patch did not apply cleanly. The following diff in
> threadService.hpp were hand merged:
> 
> --- threadService.hpp
> +++ threadService.hpp
> @@ -213,12 +213,15 @@
>    ThreadConcurrentLocks* _concurrent_locks;
>    ThreadSnapshot*        _next;
> 
> -public:
> -  // Dummy snapshot
> +  // ThreadSnapshot instances should only be created via
> +  // ThreadDumpResult::add_thread_snapshot.
> +  friend class ThreadDumpResult;
>    ThreadSnapshot() : _thread(NULL), _threadObj(NULL),
>                       _blocker_object(NULL), _blocker_object_owner(NULL),
>                       _stack_trace(NULL), _concurrent_locks(NULL),
> _next(NULL) {};
> -  ThreadSnapshot(ThreadsList * t_list, JavaThread* thread);
> +  void        initialize(ThreadsList * t_list, JavaThread* thread);
> +
> +public:
>    ~ThreadSnapshot();
> 
> Testing: Linux x86_64 tier1 tests
> 
> Best,
> Jiangli