Virtual threads deadlock with next-in-line lock release

[Cedomir Segulja]

Hi folks,

We (Faire <https://www.faire.com/>) have recently started using virtual
threads. Due to pinning, we've run into some deadlocks, but were able to
proceed, with some code changes (our pinning was rooted in static class
initialization).

I am writing here to share our write up on this topic
<https://craft.faire.com/java-virtual-threads-increasing-search-performance-while-avoiding-deadlocks-f12fa296d521>.
I think for most part, for people here it will be yet another post on
deadlocks :), but the part that might be interesting - I haven't seen that
being discussed yet - is the part about lock release
<https://craft.faire.com/java-virtual-threads-increasing-search-performance-while-avoiding-deadlocks-f12fa296d521#0f96>
being done in virtual-thread-unaware manner, leading to difficult to
analyze deadlock situations.

Cheers,
Cedo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/loom-dev/attachments/20240325/15d4ef11/attachment.htm>