RFR: 8260366: ExtendedSocketOptions <clinit> can deadlock in some circumstances [v4]
Jaikiran Pai
jpai at openjdk.java.net
Fri Feb 19 04:29:56 UTC 2021
On Thu, 18 Feb 2021 13:06:44 GMT, Jaikiran Pai <jpai at openjdk.org> wrote:
>> Changes requested by dfuchs (Reviewer).
>
> I'm closing this for now, for the reason stated here https://github.com/openjdk/jdk/pull/2601#discussion_r578400669.
After thinking a bit more about this issue and the patch I had proposed, I realized what I did wrong in this patch. The `synchronized` block in the `sun.net.ext.ExtendedSocketOptions#getInstance()` was scoped one statement too many. We don't (and shouldn't) need the `Class.forName("jdk.net.ExtendedSocketOptions")` to be part of that synchronized block. I've now modified and fixed that part to have the synchronized block only at the place where it should be needed. I've also enhanced the test to introduce a couple of additional concurrent tasks which now call `sun.net.ext.ExtendedSocketOptions#getInstance()`. So in total, the test now fires off 4 tasks, 2 of which load the respective classes and the other 2 call the getInstance() method, all concurrently. This should make the test a bit more robust and should catch any potential deadlocks. With this fix, the test now passes.
I've now updated and reopened this PR for further reviews.
-------------
PR: https://git.openjdk.java.net/jdk/pull/2601
More information about the net-dev
mailing list