Index: nf_nat_core.c
===================================================================
--- nf_nat_core.c	(mainline version)
+++ nf_nat_core.c	(working copy)
@@ -358,7 +358,8 @@
 		  enum nf_nat_manip_type maniptype)
 {
 	struct net *net = nf_ct_net(ct);
-	struct nf_conntrack_tuple curr_tuple, new_tuple;
+	struct nf_conntrack_tuple orig_tuple, new_tuple;
+	struct nf_conntrack_tuple *curr_tuple = &orig_tuple;
 	struct nf_conn_nat *nat;
 
 	/* nat helper or nfctnetlink also setup binding */
@@ -378,14 +379,18 @@
 	/* What we've got will look like inverse of reply. Normally
 	   this is what is in the conntrack, except for prior
 	   manipulations (future optimization: if num_manips == 0,
-	   orig_tp =
-	   conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple) */
-	nf_ct_invert_tuplepr(&curr_tuple,
+	   curr_tp = orig_tp =
+	   conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple) 
+	*/
+	if (!(ct->status & IPS_DST_NAT)) /* we do the optimization, as mentioned above */
+		curr_tuple = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple;
+	else 
+		nf_ct_invert_tuplepr(curr_tuple,
 			     &ct->tuplehash[IP_CT_DIR_REPLY].tuple);
 
-	get_unique_tuple(&new_tuple, &curr_tuple, ranges, ct, maniptype);
+	get_unique_tuple(&new_tuple, curr_tuple, ranges, ct, maniptype);
 
-	if (!nf_ct_tuple_equal(&new_tuple, &curr_tuple)) {
+	if (!nf_ct_tuple_equal(&new_tuple, curr_tuple)) {
 		struct nf_conntrack_tuple reply;
 
 		/* Alter conntrack table so will recognize replies. */
@@ -405,19 +410,20 @@
 		srchash = nat_hash_by_src(net, 0,
 				      &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple);
 		spin_lock_bh(&nf_nat_lock);