On Monday 10 September 2018 12:28 PM, Andre Naujoks wrote:
On 9/9/18 9:16 AM, Alan Bateman wrote:
On 07/09/2018 13:22, Andre Naujoks wrote:
:
I have not tried joining IPv4 groups on an IPv6 socket through java,
since we do not use IPv4 at all in this particular environment. I have
tried setting IP_MULTICAST_ALL to 0 in the IPv6 scenario (in a C++
project), hoping it would help, but it did not. Hence the patch for the
linux kernel.
Would it actually help, if I tried the IPv4 multicast group bind on an
IPv6 socket?
The bind to an address would be a workaround for the missing
IPV6_MULTICAST_ALL handling.
The tests that we have for the scenario of two sockets bound to the same
port but joining different multicast groups seems to be mostly using
IPv4 multicast addresses so one thing out this discussion is that we may
need to expand the tests to IPv6 multicast addresses. As the existing
tests use IPv6 sockets (when not disabled on the system or in the test
run) then it means they are exercising IP_MULTICAST_ALL=0 so I think we
can conclude that disabling IP_MULTICAST_ALL works correctly for IPv6
sockets when joining IPv4 multicast groups. If the tests were expanded
to IPv6 multicast groups then I assume we will run into the need for
IPV6_MULTICAST_ALL too.
Yes, I would assume the same. That was the whole reason for the kernel
patch.
As regards the patch to NET_InetAddressToSockaddr to set the scope_id
then it looks correct but need testing (both for bind and connect). I
see JDK-8210493 has been picked up by Vyom.
Please test it! This is my very first look into the java code-base. The
patch does what it says, but I cannot rule out, that it doesn't have any
unintended side effects.
Hi Andre,
i will apply your patch run our all network tests. I will update once i
will have result.
Thanks,
Vyom
-Alan
