NetBSD-Bugs archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

port-xen/49075: OpenJDK7 hangs in lwp_park50 in Xen PVs



>Number:         49075
>Category:       port-xen
>Synopsis:       OpenJDK7 hangs in lwp_park50 in Xen PVs
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    port-xen-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Aug 05 18:35:00 +0000 2014
>Originator:     Brad Spencer
>Release:        NetBSD 6.1.4_PATCH
>Organization:
>Environment:
NetBSD bradrun2 6.1.4_PATCH NetBSD 6.1.4_PATCH (XEN3_DOMU) #0: Mon Jun 30 
22:24:35 EDT 2014  brad@bradnetbsd:/usr/obj/sys/arch/amd64/compile/XEN3_DOMU 
amd64

>Description:
It appears that some Java applications can get stuck in the lwp_park50 system 
call in DOMUs.  A ktrace of a hung application shows the following repeating 
forever:

   781     10 java     CALL  
___lwp_park50(0x7f7fef7ffaf0,0,0x7f7ff7b4f440,0x7f7ff7b4f440)
   781     10 java     RET   ___lwp_park50 -1 errno 60 Connection timed out
   781     10 java     CALL  __clock_gettime50(3,0x7f7fef7ffbc0)
   781     10 java     RET   __clock_gettime50 0
   781     10 java     CALL  __gettimeofday50(0x7f7fef7ffb20,0)
   781     10 java     RET   __gettimeofday50 0
   781     10 java     CALL  __clock_gettime50(3,0x7f7fef7ffbc0)
   781     10 java     RET   __clock_gettime50 0
   781     10 java     CALL  __gettimeofday50(0x7f7fef7ffae0,0)
   781     10 java     RET   __gettimeofday50 0
   781     10 java     CALL  
___lwp_park50(0x7f7fef7ffaf0,0,0x7f7ff7b4f440,0x7f7ff7b4f440)
   781     10 java     RET   ___lwp_park50 -1 errno 60 Connection timed out
   781     10 java     CALL  __clock_gettime50(3,0x7f7fef7ffbc0)
   781     10 java     RET   __clock_gettime50 0
   781     10 java     CALL  __gettimeofday50(0x7f7fef7ffb20,0)
   781     10 java     RET   __gettimeofday50 0
   781     10 java     CALL  __clock_gettime50(3,0x7f7fef7ffbc0)
   781     10 java     RET   __clock_gettime50 0
   781     10 java     CALL  __gettimeofday50(0x7f7fef7ffae0,0)
   781     10 java     RET   __gettimeofday50 0

Further, if a CTRL-C is attempted against the java process, something goes into 
a tight loop such that the PV guest becomes completely unresponsive and has to 
be destroyed.

Also note that if I run this particular Java application in DOM0 using the same 
JVM it does not hang at all.

>How-To-Repeat:
I can't provide the application that trips this for me, but the same basic 
problem can be seen with the jar command.  Executing the following:

openjdk7-jar t /usr/pkg/java/openjdk7/jre/lib/rt.jar

will produce the same output in ktrace as the stuck java application.  The 
difference is that a CTRL-C will kill the jar process and not cause it to run 
away and DOM0 in this case exhibits the same problem as a DOMU.

Also:

openjdk7-1.7.51nb2  Open-source implementation of the Java Platform, Standard 
Edition

>Fix:
Not known.  I tried a DOMU kernel without any of the COMPAT stuff complied in 
just to see and tried a DOMU without MULTIPROCESSOR enabled.  Neither helped.



Home | Main Index | Thread Index | Old Index