Looks like if bgpsim don't read a Update in single read,
it tries to interpret it's consecutive read as a new packet
even thought it has read few bytes of previous packet
from the socket. So now when it consider i't consecutive
read as a new Update packet, it gets garbage values
from there and it send a header-length notification.
from my tcpdump I can see that my Update pkt been
split
into three write because of txing TCP queue is getting
filled.