RFR: 8330027: Identity hashes of archived objects must be based on a reproducible random seed [v5]

[Ioi Lam]

On Fri, 3 May 2024 07:54:40 GMT, Thomas Stuefe <stuefe at openjdk.org> wrote:

>> src/hotspot/share/runtime/synchronizer.cpp line 952:
>> 
>>> 950:   if (t == nullptr) {
>>> 951:     Atomic::cmpxchg(&cds_dump_java_thread, t, cur);
>>> 952:     return true;
>> 
>> I think this is not thread safe, as I got few intermittent asserts.
>
> Hmm, it should not really need to be thread safe, since it should only ever be called from a single thread. Do the asserts always happen in the one and only java thread?

I found the problem. A new `JavaThread` is created during VM exit. This thread is attached to the main OS thread:


Thread 2 "java" hit Breakpoint 3, JavaThread::JavaThread (this=0x7ffff002c4e0, is_attaching_via_jni=true) at javaThread.cpp:532
532	JavaThread::JavaThread(bool is_attaching_via_jni) : JavaThread() {
(gdb) where
#0  JavaThread::JavaThread (this=0x7ffff002c4e0, is_attaching_via_jni=true)
#1  attach_current_thread (vm=0x7ffff7c283d8 <main_vm>, penv=0x7ffff5358d50, _args=0x7ffff5358d60, daemon=false)
#2  jni_AttachCurrentThread (vm=0x7ffff7c283d8 <main_vm>, penv=0x7ffff5358d50, _args=0x7ffff5358d60)
#3  JavaVM_::AttachCurrentThread (this=0x7ffff7c283d8 <main_vm>, penv=0x7ffff5358d50, args=0x7ffff5358d60)
#4  jni_DestroyJavaVM_inner (vm=0x7ffff7c283d8 <main_vm>)
#5  jni_DestroyJavaVM (vm=0x7ffff7c283d8 <main_vm>)
#6  JavaMain ()
#7  ThreadJavaMain ()
#8  start_thread (arg=<optimized out>)
#9  clone3 ()


This thread can execute Java code (shutdown hooks) which can call `Object::identityHashcode()`.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/18735#discussion_r1590208765