Milter-greylist

Hi,

>>>>> On Fri, 15 May 2009 16:13:22 +0200
>>>>> attila.bruncsak@... <attila.bruncsak@...> said:

attila> I still feel it is an overhead to try
attila> to open a new socket for every new recipient,
attila> even if there is no error message about the
attila> bind failure because we hide it.

attila> What about the following:
attila> When we once get the message
attila> "cannot set IPV6_V6ONLY"
attila> consider it as serious
attila> (actually no real support for IPv6 in the kernel)
attila> so close the socket do not start
attila> the IPv6 sync_master thread
attila> but set the sync_master6.runs equals to -1
attila> without associated thread running.
attila> The mg will not try to start it again later.

attila> That way the IPv4 thread sync_master thread will start
attila> happily no need the SO_REUSEPORT.

attila> Please see the attached patch for an implementation.
attila> Also based on the vanilla 4.3.2.

It seems your idea is good to me.  However, we cannot use IPv6 where
IPV6_V6ONLY is defined but is not usable.
I think sync_listen() should set -1 to sms->runs for all errors,
instead.  The candidate patch against fresh 4.3.2 is attached.

Sincerely,