In the case of FreeBSD, there was a change made to its tcp_input()
implementation in October 1996 which probably has the side effect of
protecting against this attack. This change was removed in early October
1997 because it caused problems if spoofed SYN's with the source addresses
of legitimate hosts (other than the victim) were sent to it.
It looks to me like FreeBSD 2.2.2 should not be vulnerable unless it has
an updated version of tcp_input.c. I believe FreeBSD 2.2.5 is vulnerable.
A single attack packet may or may not cause the problem to occur, depending
on the TCP sequence numbers.