Deadlock in 0.15.x between MultipleAssignmentSubscription and CompositeSubscription

[mattrjacobs]

Here's the relevant thread dump:

"RxComputationThreadPool-8":
at rx.subscriptions.CompositeSubscription.unsubscribe(CompositeSubscription.java:100)
- waiting to lock <0x00007f2a2846be00> (a rx.subscriptions.CompositeSubscription)
at rx.subscriptions.MultipleAssignmentSubscription.unsubscribe(MultipleAssignmentSubscription.java:43)
- locked <0x00007f2a26de2a80> (a rx.subscriptions.MultipleAssignmentSubscription)
at rx.subscriptions.CompositeSubscription.add(CompositeSubscription.java:92)
- locked <0x00007f2a2846be60> (a rx.subscriptions.CompositeSubscription)
at rx.concurrency.ExecutorScheduler$4.run(ExecutorScheduler.java:128)
at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:471)
at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:334)
at java.util.concurrent.FutureTask.run(FutureTask.java:166)
at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$201(ScheduledThreadPoolExecutor.java:178)
at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:292)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
at java.lang.Thread.run(Thread.java:724)
"RxComputationThreadPool-7":
at rx.subscriptions.MultipleAssignmentSubscription.unsubscribe(MultipleAssignmentSubscription.java:40)
- waiting to lock <0x00007f2a26de2a80> (a rx.subscriptions.MultipleAssignmentSubscription)
at rx.subscriptions.CompositeSubscription.add(CompositeSubscription.java:92)
- locked <0x00007f2a2846be00> (a rx.subscriptions.CompositeSubscription)
at rx.concurrency.ExecutorScheduler$4.run(ExecutorScheduler.java:128)
at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:471)
at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:334)
at java.util.concurrent.FutureTask.run(FutureTask.java:166)
at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$201(ScheduledThreadPoolExecutor.java:178)
at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:292)

[mattrjacobs]

In English:

2 CompositeSubscriptions (denoted using last 2 letters of memory address): 00 and 60
1 MultipleAssignmentSubscription : 80

Thread 7:
Composite-00.add
MultipleAssignment-80.unsubscribe

Thread 8:
Composite-60.add
MultipleAssignment-80.unsubscribe
Composite-00.unsubscribe

So I believe there are 2 bugs here - a circular reference between subscriptions and a deadlock when unsubscribes happen on circular subscriptions

[mattrjacobs]

Also, based on our codebase, rolling back to 0.14.10 did not exhibit the problem (yet). We also believe that the Scheduler.schedule(Action1) method is the one creating the MultipleAssignmentSubscription (The only other places are OperationRetry and OperationSwitch).

[akarnokd]

Generally, unsubscribing should not happen while holding locks, especially inside these subscription instances. I'll take a look at them.

[mattrjacobs]

Thanks for the really quick work, @akarnokd. Once we've got the next release of RxJava into our codebase, we'll let you know if the issue is resolved.

[benjchristensen]

Should be fixed in #593