Re: [osmosis-dev] calling DataPostbox.release() multiple times
On 19 January 2012 10:17, Richard Hansen wrote: > > release() and outputRelease() have been designed to support re-use. I >> don't think I've tested it, but I believe it will work. The two >> variables you mention are initialized near the start of those methods >> and only read after the Released flags have been set so their >> initial state isn't important. However I can see that is confusing so >> I've modified their initial state to match their final state of true >> (I'll check it in shortly). >> > > I agree with that change. The only other thing I would suggest is a > comment indicating that the class is designed to support reuse. (Or are > all tasks supposed to be reusable?) These days I try to support re-use for all tasks, but I wouldn't be surprised if many of the older tasks don't work. I've updated the comments to describe the re-use support. Let me know if it doesn't make sense. Brett ___ osmosis-dev mailing list osmosis-dev@openstreetmap.org http://lists.openstreetmap.org/listinfo/osmosis-dev
Re: [osmosis-dev] calling DataPostbox.release() multiple times
On 2012-01-18 07:23, Brett Henderson wrote: On 16 January 2012 07:22, Richard Hansen wrote: On 2012-01-15 07:34, Brett Henderson wrote: You're correct in saying that Releasable classes can be released multiple times, however that is not typically how they're used in the Osmosis pipeline. Does it make sense to change the Releasable API definition to say that release can only be called once? Or might that break multiple-sink stages like merge? I've updated the Releasable API definition to say that implementations should support release being called multiple times but that it isn't mandatory and cannot be relied on by clients. Cool. With that change, no other changes should be necessary. release() and outputRelease() have been designed to support re-use. I don't think I've tested it, but I believe it will work. The two variables you mention are initialized near the start of those methods and only read after the Released flags have been set so their initial state isn't important. However I can see that is confusing so I've modified their initial state to match their final state of true (I'll check it in shortly). I agree with that change. The only other thing I would suggest is a comment indicating that the class is designed to support reuse. (Or are all tasks supposed to be reusable?) Your patch would probably work but I'd prefer to continue to support re-use rather than to use that variable for a second purpose. Good point. With your Releasable contract change, the patch is unnecessary anyway. An alternative is to increase the burden on the users of DataPostbox. They could be required to either ensure that the output thread restarts or avoid calling DataPostbox.release() once the output thread exits. For example, EntityBuffer.run() could set an 'outputThreadIsRunning' boolean at the top and clear it at the end. Then EntityBuffer.release() would be modified to only call buffer.release() if outputThreadIsRunning is true. Race conditions between threads are tough to avoid, and I believe the above change would introduce one. If the output thread hasn't entered EntityBuffer.run yet and the outputThreadIsRunning flag hasn't been set, but the calling thread encounters an early exception and calls EntityBuffer.release, the calling thread will think that the output thread isn't running and there is no need to perform release functionality. This would then lead to the output thread blocking when it finally does start. Yes, you are correct -- there is a race condition in that scheme. Concurrency is fun, isn't it? :) The way the class is constructed now, each thread will synchronise at the initialize, complete and release methods. If either thread skips straight to release Oh, I didn't realize that was possible. That makes it hard if not impossible to support both reuse and multiple calls to release(). then the other thread detects that as an error condition and aborts appropriately, but in all cases both threads will synchronise around release before exiting. The only reason to synchronise around release is to support re-use which may never be used in practice but it should work well. We can't even rely on the initialize methods being called because it is possible for a thread to fail before that is called. I'd rather leave the class as it is designed now. It's a tradeoff between supporting re-use and incurring the limitation that every call to release must be matched by a call to outputRelease in another thread. I don't think it's possible to have it both ways. I agree with your analysis. Thanks for making the changes! -Richard ___ osmosis-dev mailing list osmosis-dev@openstreetmap.org http://lists.openstreetmap.org/listinfo/osmosis-dev
Re: [osmosis-dev] calling DataPostbox.release() multiple times
On 16 January 2012 07:22, Richard Hansen wrote: > On 2012-01-15 07:34, Brett Henderson wrote: > >> You're correct in saying that Releasable classes can be released >> multiple times, however that is not typically how they're used in the >> Osmosis pipeline. >> > > Does it make sense to change the Releasable API definition to say that > release can only be called once? Or might that break multiple-sink stages > like merge? I've updated the Releasable API definition to say that implementations should support release being called multiple times but that it isn't mandatory and cannot be relied on by clients. Multiple sink stages like merge expose two separate sink instances to the two upstream pipes (through MultiSink interface which allows multiple Sink instances to be obtained) which feed into separate DataPostbox instances internally. Each sink instance only has its release method called once. So a single release call limitation shouldn't break anything. > > Looking at the code a bit more I think it might only need minor changes. >> I had thought that release() and outputRelease() changed the state to >> equal that of a freshly constructed DataPostbox (to support reuse?), but >> that's not true: inputExit and outputExit are false after construction but >> true after release(). So simply checking inputExit at the beginning of >> release() should take care of it. See the attached (untested) patch. >> > release() and outputRelease() have been designed to support re-use. I don't think I've tested it, but I believe it will work. The two variables you mention are initialized near the start of those methods and only read after the Released flags have been set so their initial state isn't important. However I can see that is confusing so I've modified their initial state to match their final state of true (I'll check it in shortly). Your patch would probably work but I'd prefer to continue to support re-use rather than to use that variable for a second purpose. > > An alternative is to increase the burden on the users of DataPostbox. They > could be required to either ensure that the output thread restarts or avoid > calling DataPostbox.release() once the output thread exits. For example, > EntityBuffer.run() could set an 'outputThreadIsRunning' boolean at the top > and clear it at the end. Then EntityBuffer.release() would be modified to > only call buffer.release() if outputThreadIsRunning is true. Race conditions between threads are tough to avoid, and I believe the above change would introduce one. If the output thread hasn't entered EntityBuffer.run yet and the outputThreadIsRunning flag hasn't been set, but the calling thread encounters an early exception and calls EntityBuffer.release, the calling thread will think that the output thread isn't running and there is no need to perform release functionality. This would then lead to the output thread blocking when it finally does start. The way the class is constructed now, each thread will synchronise at the initialize, complete and release methods. If either thread skips straight to release then the other thread detects that as an error condition and aborts appropriately, but in all cases both threads will synchronise around release before exiting. The only reason to synchronise around release is to support re-use which may never be used in practice but it should work well. We can't even rely on the initialize methods being called because it is possible for a thread to fail before that is called. I'd rather leave the class as it is designed now. It's a tradeoff between supporting re-use and incurring the limitation that every call to release must be matched by a call to outputRelease in another thread. I don't think it's possible to have it both ways. Brett ___ osmosis-dev mailing list osmosis-dev@openstreetmap.org http://lists.openstreetmap.org/listinfo/osmosis-dev
Re: [osmosis-dev] calling DataPostbox.release() multiple times
On 2012-01-15 07:34, Brett Henderson wrote: You're correct in saying that Releasable classes can be released multiple times, however that is not typically how they're used in the Osmosis pipeline. Does it make sense to change the Releasable API definition to say that release can only be called once? Or might that break multiple-sink stages like merge? DataPostbox is a bit tricky because it is the main synchronisation point between multiple threads. Two threads communicating via a DataPostbox perform handshakes during the initialize, complete, and release methods (the process methods run mostly independently). It is the release method where each threads tells the other thread that they are about to exit. DataPostbox does technically support multiple calls to release, but it would require each thread to restart after the release calls and this isn't supported (or even desirable) by the Osmosis pipeline. ... What changes would you suggest? I believe it's working as I originally intended, but that's not to say there isn't a better way :-) Looking at the code a bit more I think it might only need minor changes. I had thought that release() and outputRelease() changed the state to equal that of a freshly constructed DataPostbox (to support reuse?), but that's not true: inputExit and outputExit are false after construction but true after release(). So simply checking inputExit at the beginning of release() should take care of it. See the attached (untested) patch. An alternative is to increase the burden on the users of DataPostbox. They could be required to either ensure that the output thread restarts or avoid calling DataPostbox.release() once the output thread exits. For example, EntityBuffer.run() could set an 'outputThreadIsRunning' boolean at the top and clear it at the end. Then EntityBuffer.release() would be modified to only call buffer.release() if outputThreadIsRunning is true. -Richard >From 86c93d88e011ab76b79737c6185bd138f2eced76 Mon Sep 17 00:00:00 2001 From: Richard Hansen Date: Sun, 15 Jan 2012 15:13:01 -0500 Subject: [PATCH] Do nothing in DataPostbox.release() if already released The Releasable interface specification says that release() may be called multiple times. Before, if DataPostbox.release() was called a second time, it would block forever waiting for the output thread (which would no longer exist) to release. Now subsequent calls to release() are a no-op, preventing the lockup. --- .../osmosis/core/store/DataPostbox.java|5 + 1 files changed, 5 insertions(+), 0 deletions(-) diff --git a/core/src/main/java/org/openstreetmap/osmosis/core/store/DataPostbox.java b/core/src/main/java/org/openstreetmap/osmosis/core/store/DataPostbox.java index 372de7f..415ee40 100644 --- a/core/src/main/java/org/openstreetmap/osmosis/core/store/DataPostbox.java +++ b/core/src/main/java/org/openstreetmap/osmosis/core/store/DataPostbox.java @@ -335,6 +335,11 @@ public class DataPostbox implements Initializable { lock.lock(); try { + // support release() being called multiple + // times (do nothing if already released) + if (inputExit) +return; + // If release is being called without having completed successfully, // it is an error condition. if (!inputComplete) { -- 1.7.8.2 ___ osmosis-dev mailing list osmosis-dev@openstreetmap.org http://lists.openstreetmap.org/listinfo/osmosis-dev
Re: [osmosis-dev] calling DataPostbox.release() multiple times
On 15 January 2012 14:09, Richard Hansen wrote: > Hi all, > > I believe I have found a bug in DataPostbox.release(). (I'm using git > master, currently 43089ed.) > > According to the javadoc comments, Releaseable.release() can be called > multiple times. However, DataPostbox.release() can only be called once. > If it is called a second time, it blocks forever waiting for the output > thread (which no longer exists) to release. > Good find, thanks for the report. You're correct in saying that Releasable classes can be released multiple times, however that is not typically how they're used in the Osmosis pipeline. DataPostbox is a bit tricky because it is the main synchronisation point between multiple threads. Two threads communicating via a DataPostbox perform handshakes during the initialize, complete, and release methods (the process methods run mostly independently). It is the release method where each threads tells the other thread that they are about to exit. DataPostbox does technically support multiple calls to release, but it would require each thread to restart after the release calls and this isn't supported (or even desirable) by the Osmosis pipeline. > > I noticed this because the pbf reader can call release() twice, triggering > the lockup. To recreate, you can run the following: > >osmosis --read-pbf /dev/null --buffer --write-null > > The two calls to DataPostbox.release() have the following stack traces: > > java.lang.Exception: Stack trace >at java.lang.Thread.dumpStack(**Thread.java:1266) >at org.openstreetmap.osmosis.**core.store.DataPostbox.** > release(DataPostbox.java:335) >at org.openstreetmap.osmosis.**core.buffer.v0_6.EntityBuffer.** > release(EntityBuffer.java:64) >at crosby.binary.osmosis.**OsmosisBinaryParser.complete(** > OsmosisBinaryParser.java:36) >at crosby.binary.file.**BlockInputStream.process(** > BlockInputStream.java:37) >at crosby.binary.osmosis.**OsmosisReader.run(** > OsmosisReader.java:45) >at java.lang.Thread.run(Thread.**java:679) > java.lang.Exception: Stack trace >at java.lang.Thread.dumpStack(**Thread.java:1266) >at org.openstreetmap.osmosis.**core.store.DataPostbox.** > release(DataPostbox.java:335) >at org.openstreetmap.osmosis.**core.buffer.v0_6.EntityBuffer.** > release(EntityBuffer.java:64) >at crosby.binary.osmosis.**OsmosisReader.run(** > OsmosisReader.java:50) >at java.lang.Thread.run(Thread.**java:679) > > Deleting line 36 of > pbf/src/main/java/crosby/**binary/osmosis/**OsmosisBinaryParser.java > will cause DataPostbox.release() to be called only once, but that just > avoids the bug. > I've removed line 36 as you've suggested because it shouldn't be there. Release should only be called in a finally block that gets called regardless of success or failure, the complete method only gets called on success. I think it was me that added the other release in OsmosisReader recently when I added initialize, but apparently I didn't make a corresponding change in OsmosisBinaryParser. Please let me know if you see any further issues. > > I believe a proper fix to DataPostbox.release() will require a minor > redesign of the DataPostbox lifetime management code. I'm not familiar > enough with the code to feel comfortable making such a change. I can take > a crack at it, but I'm hoping an expert has the time to take a look. > What changes would you suggest? I believe it's working as I originally intended, but that's not to say there isn't a better way :-) Brett ___ osmosis-dev mailing list osmosis-dev@openstreetmap.org http://lists.openstreetmap.org/listinfo/osmosis-dev