[8u] JDK-8191393: Random crashes during cfree+0x1c

[Andrew Dinn]

On 08/04/2020 16:50, Hohensee, Paul wrote:
> Looks good now, except one more small issue.
> 
> I forgot to say before that, given that rotate_log and write are now synchronized, the rotate_log method comment is inaccurate. It could be changed to eliminate the reference to 8191393 and read something like
> 
>     // rotate_log must be called from the VMThread at a safepoint. In case of need to do
>     // gc log rotation from a thread other than the VMThread, a subtype of VM_Operation
>     // should be created and be submitted to the VMThread's operation queue. DO NOT call this
>     // function directly. It is safe to rotate the log at a safepoint via the VMThread because
>     // no mutator threads run concurrently with the VMThread, and GC threads that run
>     // concurrently with the VMThread are synchronized in write and rotate_log via _file_lock.
>     // rotate_log can write log entries, so write does a recursive lock check on _file_lock.
Yes, that's very clear. Still ok for me assuming the above change. No
need for another webrev.

regards,


Andrew Dinn
-----------
Senior Principal Software Engineer
Red Hat UK Ltd
Registered in England and Wales under Company Registration No. 03798903
Directors: Michael Cunningham, Michael ("Mike") O'Neill