Code Review request for bug fixes found by the Contended Locking project
david.holmes at oracle.com
Sun Jan 20 15:08:32 PST 2013
Thanks for all your efforts on this. Reviewed.
Minor nit: There seem to be some pre-existing indentation issues in a
couple of places that you've got caught in:
eg src/os/bsd/vm/os_bsd.cpp (and linux & solaris)
_Event = 0 ;
status = pthread_mutex_unlock(_mutex);
assert_status(status == 0, status, "mutex_unlock");
// Paranoia to ensure our locked and lock-free paths interact
// correctly with each other.
Actually there seems to be a lot of incorrect indenting in the
os_<os>.cpp files :(
On 18/01/2013 4:45 AM, Daniel D. Daugherty wrote:
> I have been working on the Contended Locking project and there are some
> bug fixes that I'd like to push into HSX-25 independently of the Contended
> Locking project.
> I'm using three different bug IDs to track these very distinct bug
> 6444286 Possible naked oop related to biased locking revocation
> safepoint in jni_exit()
> 8004902 correctness fixes motivated by contended locking work (6607129)
> 8004903 VMThread::execute() calls Thread::check_for_valid_safepoint_state()
> on concurrent VM ops
> Here is the URL for the webrev:
> These changes have been through two internal rounds of code review so I
> think they are ready for wider review. The changes for 8004902 are very
> tricky and there is long write-up attached to the 8004902 bug report that
> explains the intricate details of the "triangular race". The changes for
> 6444286 and 8004903 are more straight forward.
> These changes have been through JPRT build-and-test and have also been
> tested with the vm.parallel_class_loading and vm.quick subsuites via
> the Aurora ad-hoc testing mechanism.
> The current set of reviewers is:
> acorn, dholmes, dice
> As always, other reviewers are welcome. Comments, questions and
> suggestions are also welcome.
More information about the hotspot-runtime-dev