RFR: 8273695: Safepoint deadlock on VMOperation_lock

[Daniel D.Daugherty]

On Tue, 21 Sep 2021 20:04:03 GMT, Robbin Ehn <rehn at openjdk.org> wrote:

>> Hmmm... good point about `should_process()` not having side effects.
>
> This case could notice the update:
> bool q1 = should_process(jt, true);
> bool q_that_updates_poll_value = should_process(jt, false);
> bool q2 = should_process(jt, true);
> 
> So q2 can be false while q1 is true. I do not think this is a problem, but worth considering the implications.
> 
> I think we should update the poll value even if I do not like it. (running some tests)

If you decide to add the call to `update_poll_values(thread)`, then
please add a comment with a short explanation about the reason
for this "side effect".

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

PR: https://git.openjdk.java.net/jdk/pull/5613