river-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gregg Wonderly <gr...@wonderly.org>
Subject Re: The importance of safe publication
Date Tue, 04 Jun 2013 17:21:05 GMT
Did you use  a concurrent Set implementation?  The "synchronized" nature of 
"Hashtable" and "Vector" will, unfortunately, "fix" a number of concurrency 
issues by causing constant cache line updates to occur.  This of course is a 
huge impact to non-concurrent code, and hence "HashMap", "HashSet" and 
"ArrayList" make things go a lot faster there.

Perhaps your change in classes inadvertently removed synchronization and thus 
Happens Before which was keeping the JIT out of the mix?

Gregg Wonderly

On 6/3/2013 5:13 PM, Peter Firmstone wrote:
> Found a beaut bug, this time it relates to com.sun.jini.outrigger.EntryRep, 
> this is what I think's occurring on the client side.
>
> During construction arrays were created, written to volatile variables, then 
> populated with values.
>
> Now EntryRep uses default serialization, it isn't synchronized if marshalled 
> by a different thread, and an EntryRep is created for every Entry written into 
> the space.
>
> Previously I'd only seen similar test failures on Arm, but now I could observe 
> it on Windows, the platform so far least affected by concurrency issues.
>
> com/sun/jini/test/impl/outrigger/leasing/UseNotifyLeaseTest.td
>
> How did I find it?
>
> An unrelated class com.sun.jini.outrigger.TypeTree used the data structure, 
> Hashtree<String,Vector<String>> internally, to cache all subclasses, I 
> replaced the data structure with
> ConcurrentMap<String,Set<String>>, which simplified the code somewhat, this

> also allowed the unrelated EntryRep to fail on Window's where previously it 
> wasn't evident.
>
> The good news is, it even failed while being observed with visualvm.  It 
> appears that the test was running well until hotspot optimised reflective 
> method invocation, after that, the EntryRep array contents went missing and 
> the test subsequently failed because the Watchers no longer matched the 
> EntryRep and didn't send any more event notifications.
>
> I've just committed the fix, feel free to reverse the changes to EntryRep and 
> play around with unsafe publication.
>
> Anyone seen any strange behaviour writing Entry's to the space in deployment?  
> Eg entry's going missing, not matching, or update notifications not 
> occurring?  It's likely this could have been confused with network failure, 
> which the Jini infrastructure handles quite well.
>
> Regards,
>
> Peter.
>
>
>
>
>


Mime
View raw message