[Bug 66574] Deadlock in websocket code

https://bz.apache.org/bugzilla/show_bug.cgi?id=66574 Mark Thomas changed: What|Removed |Added Resolution|--- |FIXED Status|NEW

[Bug 66574] Deadlock in websocket code

https://bz.apache.org/bugzilla/show_bug.cgi?id=66574 --- Comment #10 from Boris Petrov --- I tested for some time and all seems fine. 9.0.74 blocked often but with this build it didn't happen so it looks good. -- You are receiving this mail because: You are the assignee for the bug. ---

[Bug 66574] Deadlock in websocket code

https://bz.apache.org/bugzilla/show_bug.cgi?id=66574 --- Comment #9 from Mark Thomas --- Great. 9.0.75-dev build that includes the proposed fix for this deadlock is available at: https://people.apache.org/~markt/dev/v9.0.75-dev/ Usual caveats apply. This is not an official release. It is only in

[Bug 66574] Deadlock in websocket code

https://bz.apache.org/bugzilla/show_bug.cgi?id=66574 --- Comment #8 from Boris Petrov --- Great, thanks a lot for the quick fix! I'll try to test with whatever you give me, yes. 9.0.x, correct. Thanks! -- You are receiving this mail because: You are the assignee for the bug. ---

[Bug 66574] Deadlock in websocket code

https://bz.apache.org/bugzilla/show_bug.cgi?id=66574 --- Comment #7 from Mark Thomas --- I have a patch that addresses this issue by refactoring the session state so it uses an AtomicReference (with a couple of extra states) rather than a lock. Initial testing is positive but I am currently runn

[Bug 66574] Deadlock in websocket code

https://bz.apache.org/bugzilla/show_bug.cgi?id=66574 --- Comment #6 from Mark Thomas --- This is related to one of the fixes for bug 66508. Specifically this change: https://github.com/apache/tomcat/commit/dccc2644ce701e88b152563473a350ec33a29a81 In short, the above change replaced the internal

[Bug 66574] Deadlock in websocket code

https://bz.apache.org/bugzilla/show_bug.cgi?id=66574 --- Comment #5 from Remy Maucherat --- Well, it seems like it is triggered by your code. "thread" #824 decides to wait in your code, blocking inside the main Servlet request processing (so it is still holding the socket wrapper lock). "thread"

[Bug 66574] Deadlock in websocket code

https://bz.apache.org/bugzilla/show_bug.cgi?id=66574 --- Comment #4 from Boris Petrov --- Created attachment 38543 --> https://bz.apache.org/bugzilla/attachment.cgi?id=38543&action=edit Thread dump -- You are receiving this mail because: You are the assignee for the bug. -

[Bug 66574] Deadlock in websocket code

https://bz.apache.org/bugzilla/show_bug.cgi?id=66574 Boris Petrov changed: What|Removed |Added Status|NEEDINFO|NEW --- Comment #3 from Boris Petrov -

[Bug 66574] Deadlock in websocket code

https://bz.apache.org/bugzilla/show_bug.cgi?id=66574 Christopher Schultz changed: What|Removed |Added Status|NEW |NEEDINFO --- Comment #2 from Chr

[Bug 66574] Deadlock in websocket code

https://bz.apache.org/bugzilla/show_bug.cgi?id=66574 --- Comment #1 from Boris Petrov --- I want to mention something that I'm noticing which might not be true but so far that's always been the case. Whenever this deadlock happens, there's only ONE such pair of threads with the stacktraces I post