mbuf cluster leak on CURRENT?

While fixing sk(4) driver issues I'm seeing these mbuf cluster
leaks on sparc64/i386. Both machines are SMP.

sparc64(sk(4))
386/1946/2332 mbufs in use (current/cache/total)
1710/290/2000/2000 mbuf clusters in use (current/cache/total/max)
^^^^^^^^^^^^^^^^^^
0/5/4528 sfbufs in use (current/peak/max)
3516K/1066K/4583K bytes allocated to network (current/cache/total)
0 requests for sfbufs denied
0 requests for sfbufs delayed
0 requests for I/O initiated by sendfile
0 calls to protocol drain routines

i386(em(4))
388/1929/2317 mbufs in use (current/cache/total)
1727/273/2000/2000 mbuf clusters in use (current/cache/total/max)
^^^^^^^^^^^^^^^^^^
0/7/3504 sfbufs in use (current/peak/max)
3551K/1028K/4579K bytes allocated to network (current/cache/total)
0 requests for sfbufs denied
0 requests for sfbufs delayed
0 requests for I/O initiated by sendfile
0 calls to protocol drain routines

Related information from vm.zone:
mbuf_ext_ref:      4,        0,      0,      0,        0
mbuf_jumbo_1:  16384,        0,      0,      0,        0
mbuf_jumbo_9:   9216,        0,      0,      0,        0
mbuf_jumbo_4:   4096,        0,      0,      0,        0
mbuf_cluster:   2048,     2000,   2000,      0,     2000
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
mbuf:            256,        0,    788,   1657, 87811655
mbuf_packet:     256,        0,   2160,    285, 50505646

To confirm the mbuf leak I lowered the cluter limit. netperf is
blocked on a "zonelimit".
Any ideas?

-- 
Regards,
Pyun YongHyeon