[ 
https://issues.apache.org/jira/browse/CASSANDRA-15726?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

ZhaoYang updated CASSANDRA-15726:
---------------------------------
     Bug Category: Parent values: Correctness(12982)
       Complexity: Normal
      Component/s: Legacy/Core
    Discovered By: Unit Test
    Fix Version/s: 4.0-alpha
         Severity: Normal
           Status: Open  (was: Triage Needed)

[~aleksey] [~benedict] do you mind reviewing?

> buffer pool may NPE with concurrent release
> -------------------------------------------
>
>                 Key: CASSANDRA-15726
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-15726
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Legacy/Core
>            Reporter: ZhaoYang
>            Assignee: ZhaoYang
>            Priority: Normal
>             Fix For: 4.0-alpha
>
>
> This can be reproduced by running {{LongBufferPoolTest}}, 1 out 5 runs..
> {code:java}
> java.lang.NullPointerException
>       at 
> org.apache.cassandra.utils.memory.BufferPool$Chunk.access$1300(BufferPool.java:836)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$LocalPool.lambda$remove$1(BufferPool.java:716)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$MicroQueueOfChunks.removeIf(BufferPool.java:460)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$MicroQueueOfChunks.access$1500(BufferPool.java:304)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$LocalPool.remove(BufferPool.java:716)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$LocalPool.put(BufferPool.java:590)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$LocalPool.recycle(BufferPool.java:709)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$Chunk.recycle(BufferPool.java:909)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$Chunk.tryRecycle(BufferPool.java:903)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$Chunk.release(BufferPool.java:896)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$MicroQueueOfChunks.removeIf(BufferPool.java:465)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$MicroQueueOfChunks.access$1500(BufferPool.java:304)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$LocalPool.addChunk(BufferPool.java:736)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$LocalPool.addChunkFromParent(BufferPool.java:725)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$LocalPool.tryGetInternal(BufferPool.java:691)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$LocalPool.tryGet(BufferPool.java:679)
>       at 
> org.apache.cassandra.utils.memory.BufferPool$LocalPool.access$000(BufferPool.java:518)
>       at 
> org.apache.cassandra.utils.memory.BufferPool.tryGet(BufferPool.java:120)      
>                  
>       at 
> org.apache.cassandra.utils.memory.LongBufferPoolTest$2.testOne(LongBufferPoolTest.java:497)
>       at 
> org.apache.cassandra.utils.memory.LongBufferPoolTest$TestUntil.call(LongBufferPoolTest.java:558)
>       at 
> org.apache.cassandra.utils.memory.LongBufferPoolTest$TestUntil.call(LongBufferPoolTest.java:538)
>       at java.util.concurrent.FutureTask.run(FutureTask.java:266)
>       at 
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
>       at 
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
>       at 
> io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
>       at java.lang.Thread.run(Thread.java:748)
> {code}
> The cause is that:
>  * When evicting a normal chunk from a full MicroQueueOfChunks, local pool 
> will try to remove corresponding tiny chunks, via 
> {{MicroQueueOfChunks#removeIf}}.
>  * If matching tiny chunk is found, tiny {{chunk.release()}} is called 
> immediately before moving null chunk to the back of the queue.
>  * Due to concurrent release from different threads, tiny {{chunk.release()}} 
> may cause its parent normal chunk, aka. the evicted chunk in #1, to be 
> removed from local pool and causes tiny pool to remove corresponding tiny 
> chunks again in {{LocalPool#remove()}}.
>  * In {{MicroQueueOfChunks#removeIf}}, due to previous in-progress 
> {{removeIf}}, it throws NPE as it violate MicroQueueOfChunks's assumption 
> which requires null chunks to be put at the back of queue.
>  
> | [patch|https://github.com/apache/cassandra/pull/537] | 
> [CI|https://circleci.com/workflow-run/a97317a0-ef21-4c01-9a97-82eaf28d7faf]|
> The fix to put null chunks to the back of queue before releasing any chunks.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org

Reply via email to