OpenJDK 8 FreeBSD 10.1 SIGSEGV issues
Seán Coffey
sean.coffey at oracle.com
Wed Jun 17 15:07:08 UTC 2015
If this is a FreeBSD only issue, you might have more experts in that
area on the bsd-port-dev mailing list :
http://mail.openjdk.java.net/pipermail/bsd-port-dev/
Regards,
Sean.
On 17/06/15 15:52, Paul wrote:
> Hello David
>
> Thanks for the reply.
> From my understanding the problem is over a year old.
> I am testing ports to OpenServer X (XinuOS Freebsd 10.1)
> At this point OSRX is plain vanilla FreeBSD 10.1.
> We need a Java to work on our platform and OpenNMS to work.
> We have a small ISP business.
> If the FreeBSD patches will work I can probably get 8u60 to work.
> Currently FreeBSD Java group doesn't have 8u60 ported.
> The problematic frame as you showed below happens but not always in
> the same place
> and doesn't always involve SocketTimeoutException. Always involves
> java/lang/String.
>
> I will keep at it until I understand better of what is going on.
> I am still analyzing jdb and truss traces.OpenNMS is a bear to debug
> for me.
>
> Thanks again
> Paul McNary
> McNary Computer Services
> Cameron, MO
> pmcnary at cameron.net
>
>
> On 6/17/2015 3:22 AM, David Holmes wrote:
>> Hi Paul,
>>
>> On 17/06/2015 3:14 PM, Paul wrote:
>>> Hello
>>>
>>> This is a thread from FreeBSD Java list.
>>> Issue has been reported to OpenNMS and FreeBSD Java.
>>>
>>> It has been advised to send it further upstream to OpenJDK
>>>
>>>
>>> Nobody seems to want to take ownership of this problem!
>>
>> Well it looks like a JVM problem to me, but on an unsupported
>> platform so ...
>>
>> The primary crash is in the compiler thread:
>>
>> Problematic frame:
>> # J 17481 C1
>> java.net.SocketTimeoutException.<init>(Ljava/lang/String;)V (6 bytes)
>> @ 0x0000000806f71e40 [0x0000000806f71e00+0x40]
>>
>> The SEGV you see trying to debug are likely due to the internal
>> mechanisms the JVM uses for safepoint polling (mprotect then segv)
>> and other things (like null reference dereferencing).
>>
>> Is it possible to test with a version of 8u60 ?
>>
>> Is this a new issue - what caused it to start happening?
>>
>> David
>>
>>> Please Help?
>>>
>>> Thanks
>>> Paul McNary
>>> pmcnary at cameron.net
>>> --------------------------------------------------------------------
>>>
>>>
>>> Another Followup -
>>>
>>> hotspot/src/os/bsd/vm/os_bsd.cpp is where mprotect is used and
>>> some of the BSD patches indicate that there might be problem
>>> with the patches.
>>> Has this been looked at as a possible source of the SIGSEGV problems?
>>>
>>> ----------------------------------------------------------------------------
>>>
>>>
>>>
>>>
>>> In tracing system calls ....
>>> Multiple SIGSEGV's are occuring
>>> Many are near: system call:
>>> 26326: mprotect(0x800659000,4096,PROT_NONE) = 0 (0x0)
>>> 26326: SIGNAL 11 (SIGSEGV)
>>> Several others occurred:
>>> 26326:
>>> stat("/usr/local/openjdk8/jre/lib/jaxp.properties",0x7ffffebec078)
>>> ERR#2
>>> 'No such file or directory'
>>> 26326: SIGNAL 11 (SIGSEGV)
>>> A couple more near:
>>> 26326:
>>> madvise(0x8df4cb000,0xa0000,0x5,0xaaaaaaaaaaaaaaab,0x0,0x80100bbe0) = 0
>>> (0x0)
>>> 26326: SIGNAL 11 (SIGSEGV)
>>>
>>> So something in the OpenJDK is having at least a problem with mprotect
>>> and madvise.
>>> These probably have POSIX and non-POSIX versions on FreeBSD.
>>>
>>> So problem is not 1 SIGSEGV but many.
>>> ---------------------------------------------------------------------------------------------------
>>>
>>>
>>> OpenNMS Jira Trouble Ticket
>>>
>>>
>>> [http://issues.opennms.org/browse/NMS-7687?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=40259#comment-40259
>>>
>>> ]
>>>
>>> Jesse White commented on NMS-7687:
>>> ----------------------------------
>>>
>>> We would need help from FreeBSD's OpenJDK maintainers to move this
>>> forward.
>>>
>>> The problem is easy to reproduce, but I haven't had any success in
>>> isolating it.
>>>
>>>
>>>> SegFault on FreeBSD 10.1 OpenNMS 16.0.0 OpenJDK 8
>>>> -------------------------------------------------
>>>>
>>>> Key: NMS-7687
>>>> URL:http://issues.opennms.org/browse/NMS-7687
>>>> Project: OpenNMS
>>>> Issue Type: Bug
>>>> Security Level: Default(Default Security Scheme)
>>>> Components: Startup / Shutdown / Status
>>>> Affects Versions: 16.0.0
>>>> Environment: FreeBSD 10.1-RELEASE-p9(amd64) OpenJDK 1.8.0-40
>>>> Reporter: Paul McNary
>>>> Fix For: Backlog
>>>>
>>>> Attachments: logs0001.zip, logs-20150527.zip
>>>>
>>>>
>>>> OpenNMS crashes after a few minutes
>>>> Have attached logs and crash dumps
>>>> as recommended on mailing lists
>>>> This has been reported by several people on the various mailing lists
>>>> I don't see an issue reported to add this to.
>>>> Please help?
>>>
>>>
>>>
>>>
>>>
>>>
>>> On 6/16/2015 3:27 PM, Paul wrote:
>>>> Hello Jean-Yves
>>>>
>>>> Has anyone taken ownership of this problem (FreeBSD and/or OpenJDK?
>>>> Has OpenJDK responded to anyone about these issues?
>>>> Is there an OpenJDK trouble ticket for this?
>>>>
>>>> Thanks
>>>> Paul
>>>> pmcnary at cameron.net
>>>>
>>>>
>>>> On 6/16/2015 5:07 AM, Jean-Yves Moulin wrote:
>>>>> Hi,
>>>>>
>>>>>
>>>>>> On 11 Jun 2015, at 23:45 , Paul <pmcnary at cameron.net> wrote:
>>>>>>
>>>>>> Bug report is on OpenNMS JIRA as replies and links show below.
>>>>>> Dumps and crash reports are attached to the JIRA bug report.
>>>>>>
>>>>>> Can the FreeBSD JAVA team please assist with this problem?
>>>>>>
>>>>>> Problem only seems to exist on FreeBSD OpenJDK!
>>>>>
>>>>> This OpenJDK issue doesn’t concern only OpenNMS: we have the same
>>>>> kind of segfault in our internal application, with both OpenJDK-7 and
>>>>> OpenJDK-8. Segfault which doesn’t happen when using linux binary JVM.
>>>>>
>>>>> Unfortunately, we are unable to reproduce the issue with a simple
>>>>> code.
>>>>>
>>>>>
>>>>> Regards,
>>>>> jym
>>>>> _______________________________________________
>>>>> freebsd-java at freebsd.org mailing list
>>>>> http://lists.freebsd.org/mailman/listinfo/freebsd-java
>>>>> To unsubscribe, send any mail to
>>>>> "freebsd-java-unsubscribe at freebsd.org"
>>>>
>>>
>>
>>
>
More information about the jdk8u-dev
mailing list