request for review (S): 6814943: getcpool001 catches more than one JvmtiThreadState problem
Daniel D. Daugherty
daniel.daugherty at oracle.com
Mon Jan 10 14:13:08 PST 2011
On 1/10/2011 10:36 AM, Keith McGuigan wrote:
> This closes a race condition hole between
> JvmtiThreadState::state_for_while_locked() and ~JavaThread(). Without
> this, the state_for_while_locked() could see a value of false for
> thread->is_exiting(), then the entirety of ~JavaThread() could run,
> the state_for_while_locked() could then finish leaving the
> JvmtiThreadState referring to a zombie thread.
> webrev: http://cr.openjdk.java.net/~kamg/6814943/webrev.00/
> Thanks for any review!
Nicely done and thumbs up!
line 675: I would have used the infamous "sanity check" for
the assert message, but what you have is fine by me.
The JvmtiEnv::environments_might_exist() call is the right
way to optimize this code path.
More information about the hotspot-runtime-dev