Details
Bug
Resolved
Blocker
Description
I am seeing responses that contain fragments from previous response buffers prepended to the correct response. I set a breakpoint in HttpGenerator.completeHeader() to see if the result of _buffers.getBuffer() was a non-empty, and it intermittently returns a buffer that has a non-zero length(). It turns out that something is causing the same buffer to be returned to the buffer pool multiple times. I added some logging to getBuffer/returnBuffer to look for mismatched calls. In my application I observed the following sequence of returnBuffer calls without an intervening getBuffer call. It appears the HttpGenerator.reset() is being called in a way that is not thread-safe.
INFO | jvm 1 | 2007/04/18 18:12:12 | 2007-04-18 18:12:12.203::WARN: returnBuffer returns 2356633: len=0 INFO | jvm 1 | 2007/04/18 18:12:12 | java.lang.Exception INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.AbstractBuffers.returnBuffer(AbstractBuffers.java:150) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.AbstractGenerator.reset(AbstractGenerator.java:153) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.HttpGenerator.reset(HttpGenerator.java:78) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.HttpConnection.reset(HttpConnection.java:435) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.HttpConnection.handle(HttpConnection.java:416) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.io.nio.SelectChannelEndPoint.run(SelectChannelEndPoint.java:365) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.thread.BoundedThreadPool$PoolThread.run(BoundedThreadPool.java:442) INFO | jvm 1 | 2007/04/18 18:12:12 | INFO | jvm 1 | 2007/04/18 18:12:12 | 2007-04-18 18:12:12.203::WARN: returnBuffer returns 2356633: len=0 INFO | jvm 1 | 2007/04/18 18:12:12 | java.lang.Exception INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.AbstractBuffers.returnBuffer(AbstractBuffers.java:150) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.AbstractGenerator.reset(AbstractGenerator.java:153) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.HttpGenerator.reset(HttpGenerator.java:78) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.HttpConnection.destroy(HttpConnection.java:125) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.AbstractConnector.connectionClosed(AbstractConnector.java:737) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.nio.SelectChannelConnector.access$100(SelectChannelConnector.java:64) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.nio.SelectChannelConnector$1.endPointClosed(SelectChannelConnector.java:92) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.io.nio.SelectChannelEndPoint.doUpdateKey(SelectChannelEndPoint.java:352) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.io.nio.SelectorManager$SelectSet.doSelect(SelectorManager.java:347) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.io.nio.SelectorManager.doSelect(SelectorManager.java:175) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.nio.SelectChannelConnector.accept(SelectChannelConnector.java:124) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.jetty.AbstractConnector$Acceptor.run(AbstractConnector.java:500) INFO | jvm 1 | 2007/04/18 18:12:12 | at org.mortbay.thread.BoundedThreadPool$PoolThread.run(BoundedThreadPool.java:442)
Ah yes - I saw this issue some time ago when you get sockets being closed at the same time as the response is sent.
you end up with double resets.
I protected the HttpParser.reset() method but stupidly left HttpGenerator unprotected!
I believe it is fixed in SVN head now.... but I don't have a test harness that can reproduce this....