8207404: MulticastSocket tests failing on Aix

David Holmes david.holmes at oracle.com
Sun Dec 2 23:14:07 UTC 2018


This is probably best discussed on net-dev rather than core-libs-dev.

Cheers,
David

On 1/12/2018 2:27 am, Volker Simonis wrote:
> On Fri, Nov 30, 2018 at 12:20 PM Steve Groeger <GROEGES at uk.ibm.com> wrote:
>>
>> Hi Goetz,
>>
>> I disabled IPv6 on my xLinux box and the MulticastSocket tests pass, as
>> they detected that IPv6 is not enabled and then didnt try and set the IPv6
>> options.
> 
> Not being a network expert either, I think there's two things to
> consider here. It seems that IPv6 can be disabled globally (e.g. in
> the Linux kernel) or IPv6 can be enabled globally, but not all network
> devices supporting it. A quick search revealed that it seems to be
> fairly easy to disable IPv6 on a single interface on Linux while still
> having IPv6 globally enabled (see
> https://superuser.com/questions/575684/how-to-disable-ipv6-on-a-specific-interface-in-linux).
> @Steve Groeger  can you please check if the test still succeeds on
> Linux if you only disable IPv6 on the specific interface used by the
> test? If it still succeeds, there's probably something to fix on AIX.
> If it fails on Linux as well, there may be something to fix in the
> test (i.e. to detect that the corresponding interface doesn't support
> IPv6) or even in the Java network implementation (i.e. to handle both,
> IPv6 and IPv4-only interfaces on an IPv6 enabled system correctly).
> 
> Thank you and best regards,
> Volker
> 
>> I am currently unable to disable IPv6 on my AIX box to test this but would
>> assume it would also work as the code checks to see whether IPv6 is
>> enabled and doesnt try and set the IPV6 options.
>> It seems the issue with the tests failing seems to be, if IPv6 is enabled
>> but the interface doesn't have an IPv6 address associated with it, as in
>> your case here:
>>
>> bash-4.3$ ifconfig -a
>> en0:
>> flags=1e084863,480<UP,BROADCAST,NOTRAILERS,RUNNING,SIMPLEX,MULTICAST,GROUPRT,64BIT,CHECKSUM_OFFLOAD(ACTIVE),CHAIN>
>>          inet 10.xx.xxx.xxx netmask 0xfffffe00 broadcast 10.xx.xxx.xxx
>>           tcp_sendspace 262144 tcp_recvspace 262144 rfc1323 1
>> lo0:
>> flags=e08084b,c0<UP,BROADCAST,LOOPBACK,RUNNING,SIMPLEX,MULTICAST,GROUPRT,64BIT,LARGESEND,CHAIN>
>>          inet 127.x.x.x netmask 0xff000000 broadcast 127.xxx.xxx.xxx
>>          inet6 ::1%1/0
>>           tcp_sendspace 131072 tcp_recvspace 131072 rfc1323 1
>>
>> where you can see that the en0 interface has an inet address associated
>> but doesnt have an inet6 address associated.
>> I can see that IPv6 is enabled as the lo0 interface has an inet6 address
>> (pointing to localhost)
>>
>> As I stated below, my AIX system was configured in a similar way to yours
>> and the tests failed. When I was able to add an inet6 address to en0,
>> everything works correctly.
>> NOTE: Not sure this was correct but I ran 'sudo autoconf6' which seemed to
>> add the inet6 address to my en0 interface. After that all tests worked.
>>
>> So I think the tests are working as expected, but the AIX system is not
>> configured correctly, but I would like someone else to confirm this.
>>
>> Thanks
>> Steve Groeger
>> IBM Runtime Technologies
>> Hursley, Winchester
>> Tel: (44) 1962 816911  Mobex: 279990  Mobile: 07718 517 129
>> Fax (44) 1962 816800
>> Lotus Notes: Steve Groeger/UK/IBM
>> Internet: groeges at uk.ibm.com
>>
>> Unless stated otherwise above:
>> IBM United Kingdom Limited - Registered in England and Wales with number
>> 741598.
>> Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU
>>
>>
>>
>> From:   "Lindenmaier, Goetz" <goetz.lindenmaier at sap.com>
>> To:     Steve Groeger <GROEGES at uk.ibm.com>, ppc-aix-port-dev
>> <ppc-aix-port-dev at openjdk.java.net>
>> Date:   30/11/2018 10:18
>> Subject:        RE: 8207404: MulticastSocket tests failing on Aix
>>
>>
>>
>> i Steve,
>>
>> thanks for looking at this issue.
>>
>>> Should these tests work without having IPV6 configured on the interface?
>> In general, tests should not show errors if a system does not meet
>> the requirements for the test. In such cases the test should
>> be skipped somehow. Maybe only a test fix is needed.
>>
>> Can you check on a non-aix machine without ipv6 whether the test
>> passes?
>>
>>> Goetz, do you have IPV6 configured on the machine you were running these
>>> tests on?
>> This is our test machine:
>>
>> bash-4.3$ ifconfig -a
>> en0:
>> flags=1e084863,480<UP,BROADCAST,NOTRAILERS,RUNNING,SIMPLEX,MULTICAST,GROUPRT,64BIT,CHECKSUM_OFFLOAD(ACTIVE),CHAIN>
>>          inet 10.xx.xxx.xxx netmask 0xfffffe00 broadcast 10.xx.xxx.xxx
>>           tcp_sendspace 262144 tcp_recvspace 262144 rfc1323 1
>> lo0:
>> flags=e08084b,c0<UP,BROADCAST,LOOPBACK,RUNNING,SIMPLEX,MULTICAST,GROUPRT,64BIT,LARGESEND,CHAIN>
>>          inet 127.x.x.x netmask 0xff000000 broadcast 127.xxx.xxx.xxx
>>          inet6 ::1%1/0
>>           tcp_sendspace 131072 tcp_recvspace 131072 rfc1323 1
>>
>> I think this message should be posted to core-libs-dev, too. That's also
>> the list where a potential fix needs to be reviewed.
>>
>> Best regards,
>>    Goetz.
>>
>>
>>> -----Original Message-----
>>> From: Steve Groeger <GROEGES at uk.ibm.com>
>>> Sent: Donnerstag, 29. November 2018 17:07
>>> To: ppc-aix-port-dev <ppc-aix-port-dev at openjdk.java.net>
>>> Cc: Lindenmaier, Goetz <goetz.lindenmaier at sap.com>
>>> Subject: 8207404: MulticastSocket tests failing on Aix
>>>
>>> Hi All,
>>>
>>> I have been having a look at this issue:
>>>
>> https://bugs.openjdk.java.net/browse/JDK-8207404
>>
>>> <
>> https://bugs.openjdk.java.net/browse/JDK-8207404
>>>
>>> which was raised by  Goetz Lindenmaier.
>>>
>>> When I ran the tests on my AIX system I got the same results as Goetz
>> listed
>>> in the issue.
>>>
>>> Looking at the failure I noticed that it was failing on a setsockopt
>> call in the
>>> native code in
>>> the PlainDatagramSocketImpl.c file which returned a EADDRNOTAVAIL (Can't
>>> assign requested address)
>>>
>>> Looking at my network configuration using "ifconfig -a", I noticed that
>> my
>>> interface "en0" which the test
>>> was trying to use did not have an inet6 address configured (see below -
>> some
>>> addresses have been obfuscated with x's).
>>>
>>> en0:
>>> flags=1e084863,14c0<UP,BROADCAST,NOTRAILERS,RUNNING,SIMPLEX,MULTI
>>> CAST,GROUPRT,64BIT,CHECKSUM_OFFLOAD(ACTIVE),LARGESEND,CHAIN>
>>>          inet 9.20.xx.xx netmask 0xffffff00 broadcast 9.20.xx.xxx
>>>           tcp_sendspace 262144 tcp_recvspace 262144 rfc1323 1
>>>
>>> configuring the IPV6 on my AIX box I then had a inet6 address configured
>> for
>>> en0
>>>
>>> en0:
>>> flags=1e084863,14c0<UP,BROADCAST,NOTRAILERS,RUNNING,SIMPLEX,MULTI
>>> CAST,GROUPRT,64BIT,CHECKSUM_OFFLOAD(ACTIVE),LARGESEND,CHAIN>
>>>          inet 9.20.xx.xx netmask 0xffffff00 broadcast 9.20.xx.xxx
>>>          inet6 fe80::xxxx:xxxx:xxxx:xxxx/64
>>>           tcp_sendspace 262144 tcp_recvspace 262144 rfc1323 1
>>>
>>> with this configured the MulticastSocket tests all passed.
>>>
>>> Should these tests work without having IPV6 configured on the interface?
>>>
>>> Goetz, do you have IPV6 configured on the machine you were running these
>>> tests on?
>>>
>>> Thanks
>>> Steve Groeger
>>> IBM Runtime Technologies
>>> Hursley, UK
>>> Tel: (44) 1962 816911  Mobex: 279990  Mobile: 07718 517 129
>>> Fax (44) 1962 816800
>>> Lotus Notes: Steve Groeger/UK/IBM
>>> Internet: groeges at uk.ibm.com
>>>
>>> Unless stated otherwise above:
>>> IBM United Kingdom Limited - Registered in England and Wales with number
>>> 741598.
>>> Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6
>>> 3AU
>>
>>
>>
>>
>>
>> Unless stated otherwise above:
>> IBM United Kingdom Limited - Registered in England and Wales with number
>> 741598.
>> Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU


More information about the core-libs-dev mailing list