[concurrency-interest] Nested synchronized
Dr Heinz M. Kabutz
heinz at javaspecialists.eu
Fri Jan 6 12:17:07 EST 2012
Hi Nathan,
I would bet my bottom dollar that when the system hangs, it won't
respond to JConsole anymore either. This is very similar to another
test that we wrote.
Incidentally, I discovered a condition that puts the JVM in an
unbreakable hard spin. I reported this at least a year ago. However,
it seems to still not be fixed. I don't want to report it to publicly
as it would put any modern Java system at risk. I think all JVM's
after 1.6.0_16 are affected. It is easy to reproduce on Windows,
Linux and Mac. Anybody here with the clout to get this one fixed?
Please contact me directly.
Heinz
On 06/01/2012, Nathan Reynolds <nathan.reynolds at oracle.com> wrote:
> Try connecting with JConsole before the hang and then dump call stacks
> after the hang via JConsole.
>
> Also, please send the output of java -version so that all can match JVM
> versions. If you could provide the download link, that would help.
>
> Nathan Reynolds
> <http://psr.us.oracle.com/wiki/index.php/User:Nathan_Reynolds> |
> Consulting Member of Technical Staff | 602.333.9091
> Oracle PSR Engineering <http://psr.us.oracle.com/> | Server Technology
>
> On 1/5/2012 7:05 PM, Howard Lovatt wrote:
>> Hi Vitaly,
>>
>> jstack didn't work.
>>
>> gar-ln:~ lov080$ ps
>> PID TTY TIME CMD
>> 62924 ttys000 0:00.02 -bash
>> 62955 ttys000 0:08.15 /usr/bin/java
>> nestedsynchronizedproblem.NestedSynchron
>> 62967 ttys001 0:00.01 -bash
>> gar-ln:~ lov080$ jstack -l -F 62955
>> Attaching to process ID 62955, please wait...
>> attach: task_for_pid(62955) failed (5)
>> Error attaching to process: Error attaching to process, or no such process
>>
>> Good sugestion though.
>>
>> Thanks,
>>
>> -- Howard.
>>
>> On 6 January 2012 12:44, Vitaly Davidovich <vitalyd at gmail.com
>> <mailto:vitalyd at gmail.com>> wrote:
>>
>> a bit crude but if its somewhat easy to repro try increasing the
>> await period to something like 10 mins and then jstack it a few
>> times and see what it shows.
>>
>> Also can you repro in interpreter (i.e. -Xint)?
>>
>> On Jan 5, 2012 8:27 PM, "Howard Lovatt" <howard.lovatt at gmail.com
>> <mailto:howard.lovatt at gmail.com>> wrote:
>>
>> Hi Vitaly,
>>
>> If I increase the awaitTermination to 2 minutes I can run the
>> program in the debugger, however it works then! Is there
>> another away, without using the debugger, of generating a
>> stack trace for a hanging program?
>>
>> Thanks,
>>
>> -- Howard.
>>
>> On 6 January 2012 12:18, Vitaly Davidovich <vitalyd at gmail.com
>> <mailto:vitalyd at gmail.com>> wrote:
>>
>> Howard,
>>
>> Have you captured call stacks when it hangs?
>>
>> Vitaly
>>
>> On Jan 5, 2012 8:14 PM, "Howard Lovatt"
>> <howard.lovatt at gmail.com <mailto:howard.lovatt at gmail.com>>
>> wrote:
>>
>> Hi David,
>>
>> There is only one sum object shared between the two
>> threads (1st line of main) and hence all
>> synchronization is on the same object. Therefore I
>> think the code should work (even though the second
>> synchronization is redundant). As a double check on my
>> understanding I just added a specific mutex object to
>> the code and synchronized on that and got the same result.
>>
>> Have I understood you comment correctly?
>>
>> Thanks,
>>
>> -- Howard.
>>
>> On 6 January 2012 12:00, David Holmes
>> <davidcholmes at aapt.net.au
>> <mailto:davidcholmes at aapt.net.au>> wrote:
>>
>> That's not nested synchronization as you are using
>> two different objects. It is a classic deadlock:
>> - sync method on Obj A calls sync method on Obj B
>> - sync method on Obj B calls sync method on Objj A
>> Thread 1 does the call to ObjA
>> Thread 2 does the call to Obj B
>> David
>> ------
>>
>> -----Original Message-----
>> *From:*
>> concurrency-interest-bounces at cs.oswego.edu
>>
>> <mailto:concurrency-interest-bounces at cs.oswego.edu>
>> [mailto:concurrency-interest-bounces at cs.oswego.edu
>>
>> <mailto:concurrency-interest-bounces at cs.oswego.edu>]*On
>> Behalf Of *Howard Lovatt
>> *Sent:* Friday, 6 January 2012 10:53 AM
>> *To:* concurrency-interest at cs.oswego.edu
>> <mailto:concurrency-interest at cs.oswego.edu>
>> *Subject:* [concurrency-interest] Nested
>> synchronized
>>
>> Hi,
>>
>> I have seen something I think is a JVM bug but
>> would like to check my understanding before
>> reporting a problem. The following program
>> normally hangs, i.e. the problem is
>> intermittent, on my computer, MacBook Pro,
>> Java 6 or 7, 4 core processor. The problem is
>> that there are synchronized methods, isSetA1
>> and isSetA2 (near end of listing below), that
>> call another synchronized method,
>> conditionallySumArguments (at end of listing
>> below). The second synchronized
>> is unnecessary since the method is always
>> called within an already synchronized method
>> and if the second synchronized is removed the
>> program works as expected. However I think an
>> extra synchronized should be redundant, not a
>> problem?
>>
>> package nestedsynchronizedproblem;
>>
>> import java.util.concurrent.Callable;
>> import java.util.concurrent.ExecutorService;
>> import java.util.concurrent.Executors;
>> import java.util.concurrent.TimeUnit;
>>
>> import static java.lang.System.*;
>>
>> /**
>> * Test of nested synchronized. Mimics
>> calling a parallel sum method.
>> *
>> * @author Howard Lovatt
>> */
>> public class NestedSynchronizedProblem {
>> private static final int loops = 10 *
>> 1000 * 1000; // This needs to be large for
>> hanging!
>>
>> public static void main( final String...
>> notUsed ) throws InterruptedException {
>> final ParrallelSumMethod sum = new
>> ParrallelSumMethod();
>> final Callable<Void> setA1 = new
>> Callable<Void>() {
>> @Override public Void call() throws
>> Exception {
>> for ( int l = 0; l < loops; l++ )
>> { sum.setA1( l ); }
>> return null;
>> }
>> };
>> final Callable<Void> setA2 = new
>> Callable<Void>() {
>> @Override public Void call() throws
>> Exception {
>> for ( int l = 0; l < loops; l++ )
>> { sum.setA2( l ); }
>> return null;
>> }
>> };
>> final ExecutorService pool =
>> Executors.newCachedThreadPool();
>> pool.submit( setA1 );
>> pool.submit( setA2 );
>> pool.shutdown();
>> final boolean ok =
>> pool.awaitTermination( 1, TimeUnit.MINUTES );
>> out.println( sum.getSum() + (ok ? ",
>> terminated ok" : ", failed to terminate") );
>> pool.shutdownNow();
>> }
>> }
>>
>>
>> final class ParrallelSumMethod {
>> private long sum = 0;
>> private Long a1 = null;
>> private Long a2 = null;
>>
>> public void setA1( final long a1Arg )
>> throws InterruptedException {
>> for ( ;; ) {
>> if ( isSetA1( a1Arg ) ) { return; }
>> checkForInterrupt();
>> }
>> }
>>
>> public void setA2( final long a2Arg )
>> throws InterruptedException {
>> for ( ;; ) {
>> if ( isSetA2( a2Arg ) ) { return; }
>> checkForInterrupt();
>> }
>> }
>>
>> public Long getSum() { return sum; }
>>
>> private static void checkForInterrupt()
>> throws InterruptedException {
>> if ( Thread.interrupted() ) { throw
>> new InterruptedException(); }
>> }
>>
>> private synchronized boolean isSetA1(
>> final long a1Arg ) {
>> if ( a1 == null ) {
>> a1 = a1Arg;
>> conditionallySumArguments();
>> return true;
>> }
>> return false;
>> }
>>
>> private synchronized boolean isSetA2(
>> final long a2Arg ) {
>> if ( a2 == null ) {
>> a2 = a2Arg;
>> conditionallySumArguments();
>> return true;
>> }
>> return false;
>> }
>>
>> private synchronized void
>> conditionallySumArguments() { // Works if
>> not synchronized!!!
>> if ( ( a1 == null ) || ( a2 == null )
>> ) { return; }
>> sum += a1 + a2;
>> a1 = a2 = null;
>> }
>> }
>>
>>
>> Thanks in advance for any comments,
>>
>> -- Howard.
>>
>>
>>
>>
>> --
>> -- Howard.
>>
>>
>> _______________________________________________
>> Concurrency-interest mailing list
>> Concurrency-interest at cs.oswego.edu
>> <mailto:Concurrency-interest at cs.oswego.edu>
>> http://cs.oswego.edu/mailman/listinfo/concurrency-interest
>>
>>
>>
>>
>> --
>> -- Howard.
>>
>>
>>
>>
>> --
>> -- Howard.
>>
>>
>>
>> _______________________________________________
>> Concurrency-interest mailing list
>> Concurrency-interest at cs.oswego.edu
>> http://cs.oswego.edu/mailman/listinfo/concurrency-interest
>
--
Dr Heinz M. Kabutz (PhD CompSci)
Author of "The Java(tm) Specialists' Newsletter"
Sun Java Champion
IEEE Certified Software Development Professional
http://www.javaspecialists.eu
Tel: +30 69 72 850 460
Skype: kabutz
More information about the Concurrency-interest
mailing list