I have been trying to migrate from random() to rnd(4), in the
	kernel network processing (including IPsec).
	so my question: is it okay to assume that
		rnd_extract_data(RND_EXTRACT_ANY)
	returns the specified bytes of random bytes?  what will it return
	in entropy-starvation situation?  how good is the randomness in
	entropy-starvation situation?
	(NOTE: in the kernel networking code, we can't wait till entropy
	to be gathered)

	tcp_new_iss() does not check the return value from rnd_extract_data(),
	and assumes that we can get enough randomness from the function.
	is it a safe thing to do?  if so, i'm happy.

itojun