I'm having problem with really poor NFS performance between NetBSD and
   Solaris 2.6.

What *specific* symptoms are you seeing?  slow reads?  slow writes?

One possibility here..  I believe that recent versions of Solaris send
larger NFS packets by default (16k fragmented udp rather than the more
typical 8k used by most other clients); by default, the NetBSD NFS
server only allocates a ~32k receive buffer size, so back-to-back ~16k
requests would appear to stand a significant chance of getting
dropped..

Can you see if the problem gets any better if you change solaris to
mount the NetBSD fileserver with rsize=8192,wsize=8192?

						- Bill