RFR(S): 8246676: monitor list lock operations need more fencing and 8247280
Daniel D. Daugherty
daniel.daugherty at oracle.com
Wed Jul 15 02:30:34 UTC 2020
These fixes are targeted for JDK15 and I would like to push both of these
fixes before the RDP2 cutoff on Thursday.
I have a JDK15 fix ready for a couple of related ObjectMonitor bug fixes:
JDK-8246676 monitor list lock operations need more fencing
JDK-8247280 more fencing needed in async deflation for non-TSO machines
The fix for JDK-8246676 has been through three rounds of preliminary
code review with David H.; Erik O. and Robbin participated in the first
preliminary code review round. Mostly comment changes or backouts of
code changes to return to the baseline code were done in the second and
third preliminary rounds.
The fix for JDK-8247280 has been through two rounds of preliminary
code review with David H. The bug fix itself was suggested by Erik O. so
he's likely on-board with my implementation of the fix. :-)
Many thanks to David H., Erik O. and Robbin for their many emails on
these topics and for reviewing the preliminary webrevs.
Here are the two webrevs:
The project is currently baselined on jdk-15+30 and has gone through
Mach5 Tier[1-3],4,5,6,7,8 testing with no regressions. I've also run
my inflation stress kit on Linux-X64 and macOSX without any regressions.
Thanks, in advance, for any comments, questions or suggestions.
More information about the hotspot-runtime-dev