Prevent overload of incomming s2s connections

Three changes were introduced:

1) ejabberd_s2s_in now uses p1_fsm instead of gen_fsm.  And uses
   the {max_queue, N} option to kill the process if its input
   queue grows too much.
2) If a ejabberd_s2s_in process is overload and killed, the server
   that originated that connection is not allowed to connect back
   to us for X seconds (set to 60seconds on the source)
3) The list of blocked (both statically and dynamically by the above
   method) host is now also checked for hosts authenticating by
   starttls+sasl. Previusly it was only used during dialback.

0 files changed, 0 insertions, 0 deletions