Handle the case of m == NULL in m_copym().

This case should not happen, but actually does happen for the
userland stack. See #160.

The real fix for #160 is
to figure out why there is an entry in the asoc->send_queue with
chk->data == NULL. But I was not able to reproduce the problem
locally. So this change will still panic(), if the code is compiled
with INVARIANTS, but will not crash if build without it (for
production).

usrsctplib/user_mbuf.c

@@ -987,7 +987,13 @@ m_copym(struct mbuf *m, int off0, int len, int wait)
 
 	KASSERT(off >= 0, ("m_copym, negative off %d", off));
 	KASSERT(len >= 0, ("m_copym, negative len %d", len));
+	KASSERT(m != NULL, ("m_copym, m is NULL"));
 
+#if !defined(INVARIANTS)
+	if (m == NULL) {
+		return (NULL);
+	}
+#endif
 	if (off == 0 && m->m_flags & M_PKTHDR)
 		copyhdr = 1;
 	while (off > 0) {