diff --git a/include/linux/netdev_features.h b/include/linux/netdev_features.h
index c26d0ec2ef3a..f1338e0f9866 100644
--- a/include/linux/netdev_features.h
+++ b/include/linux/netdev_features.h
@@ -45,6 +45,7 @@ enum {
 	NETIF_F_GSO_IPIP_BIT,		/* ... IPIP tunnel with TSO */
 	NETIF_F_GSO_SIT_BIT,		/* ... SIT tunnel with TSO */
 	NETIF_F_GSO_UDP_TUNNEL_BIT,	/* ... UDP TUNNEL with TSO */
+	NETIF_F_GSO_UDP_TUNNEL_CSUM_BIT,/* ... UDP TUNNEL with TSO & CSUM */
 	NETIF_F_GSO_MPLS_BIT,		/* ... MPLS segmentation */
 	/**/NETIF_F_GSO_LAST =		/* last bit, see GSO_MASK */
 		NETIF_F_GSO_MPLS_BIT,
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index d8d397acb52c..5a6d10a538f5 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -345,6 +345,8 @@ enum {
 	SKB_GSO_UDP_TUNNEL = 1 << 9,
 
 	SKB_GSO_MPLS = 1 << 10,
+
+	SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
 };
 
 #if BITS_PER_LONG > 32
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 0e9bb08a91e4..0070ab87109b 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -1258,6 +1258,7 @@ static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
 		       SKB_GSO_SIT |
 		       SKB_GSO_TCPV6 |
 		       SKB_GSO_UDP_TUNNEL |
+		       SKB_GSO_UDP_TUNNEL_CSUM |
 		       SKB_GSO_MPLS |
 		       0)))
 		goto out;
diff --git a/net/ipv4/tcp_offload.c b/net/ipv4/tcp_offload.c
index d8de7b9e0720..c02f2d2e7bab 100644
--- a/net/ipv4/tcp_offload.c
+++ b/net/ipv4/tcp_offload.c
@@ -61,6 +61,7 @@ struct sk_buff *tcp_gso_segment(struct sk_buff *skb,
 			       SKB_GSO_SIT |
 			       SKB_GSO_MPLS |
 			       SKB_GSO_UDP_TUNNEL |
+			       SKB_GSO_UDP_TUNNEL_CSUM |
 			       0) ||
 			     !(type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6))))
 			goto out;
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index a84f6762ea9e..8d8c33d84c9a 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -2528,7 +2528,11 @@ struct sk_buff *skb_udp_tunnel_segment(struct sk_buff *skb,
 	int tnl_hlen = skb_inner_mac_header(skb) - skb_transport_header(skb);
 	__be16 protocol = skb->protocol;
 	netdev_features_t enc_features;
-	int outer_hlen;
+	int udp_offset, outer_hlen;
+	unsigned int oldlen;
+	bool need_csum;
+
+	oldlen = (u16)~skb->len;
 
 	if (unlikely(!pskb_may_pull(skb, tnl_hlen)))
 		goto out;
@@ -2540,6 +2544,10 @@ struct sk_buff *skb_udp_tunnel_segment(struct sk_buff *skb,
 	skb->mac_len = skb_inner_network_offset(skb);
 	skb->protocol = htons(ETH_P_TEB);
 
+	need_csum = !!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL_CSUM);
+	if (need_csum)
+		skb->encap_hdr_csum = 1;
+
 	/* segment inner packet. */
 	enc_features = skb->dev->hw_enc_features & netif_skb_features(skb);
 	segs = skb_mac_gso_segment(skb, enc_features);
@@ -2550,10 +2558,11 @@ struct sk_buff *skb_udp_tunnel_segment(struct sk_buff *skb,
 	}
 
 	outer_hlen = skb_tnl_header_len(skb);
+	udp_offset = outer_hlen - tnl_hlen;
 	skb = segs;
 	do {
 		struct udphdr *uh;
-		int udp_offset = outer_hlen - tnl_hlen;
+		int len;
 
 		skb_reset_inner_headers(skb);
 		skb->encapsulation = 1;
@@ -2564,31 +2573,20 @@ struct sk_buff *skb_udp_tunnel_segment(struct sk_buff *skb,
 		skb_reset_mac_header(skb);
 		skb_set_network_header(skb, mac_len);
 		skb_set_transport_header(skb, udp_offset);
+		len = skb->len - udp_offset;
 		uh = udp_hdr(skb);
-		uh->len = htons(skb->len - udp_offset);
+		uh->len = htons(len);
 
-		/* csum segment if tunnel sets skb with csum. */
-		if (protocol == htons(ETH_P_IP) && unlikely(uh->check)) {
-			struct iphdr *iph = ip_hdr(skb);
+		if (need_csum) {
+			__be32 delta = htonl(oldlen + len);
+
+			uh->check = ~csum_fold((__force __wsum)
+					       ((__force u32)uh->check +
+						(__force u32)delta));
+			uh->check = gso_make_checksum(skb, ~uh->check);
 
-			uh->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
-						       skb->len - udp_offset,
-						       IPPROTO_UDP, 0);
-			uh->check = csum_fold(skb_checksum(skb, udp_offset,
-							   skb->len - udp_offset, 0));
 			if (uh->check == 0)
 				uh->check = CSUM_MANGLED_0;
-
-		} else if (protocol == htons(ETH_P_IPV6)) {
-			struct ipv6hdr *ipv6h = ipv6_hdr(skb);
-			u32 len = skb->len - udp_offset;
-
-			uh->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr,
-						     len, IPPROTO_UDP, 0);
-			uh->check = csum_fold(skb_checksum(skb, udp_offset, len, 0));
-			if (uh->check == 0)
-				uh->check = CSUM_MANGLED_0;
-			skb->ip_summed = CHECKSUM_NONE;
 		}
 
 		skb->protocol = protocol;
diff --git a/net/ipv4/udp_offload.c b/net/ipv4/udp_offload.c
index 88b4023ecfcf..5c23f4765af9 100644
--- a/net/ipv4/udp_offload.c
+++ b/net/ipv4/udp_offload.c
@@ -56,7 +56,8 @@ static struct sk_buff *udp4_ufo_fragment(struct sk_buff *skb,
 	__wsum csum;
 
 	if (skb->encapsulation &&
-	    skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL) {
+	    (skb_shinfo(skb)->gso_type &
+	     (SKB_GSO_UDP_TUNNEL|SKB_GSO_UDP_TUNNEL_CSUM))) {
 		segs = skb_udp_tunnel_segment(skb, features);
 		goto out;
 	}
@@ -71,6 +72,7 @@ static struct sk_buff *udp4_ufo_fragment(struct sk_buff *skb,
 
 		if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY |
 				      SKB_GSO_UDP_TUNNEL |
+				      SKB_GSO_UDP_TUNNEL_CSUM |
 				      SKB_GSO_IPIP |
 				      SKB_GSO_GRE | SKB_GSO_MPLS) ||
 			     !(type & (SKB_GSO_UDP))))
diff --git a/net/ipv6/ip6_offload.c b/net/ipv6/ip6_offload.c
index b2f091566f88..d54c5744e3db 100644
--- a/net/ipv6/ip6_offload.c
+++ b/net/ipv6/ip6_offload.c
@@ -100,6 +100,7 @@ static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb,
 		       SKB_GSO_IPIP |
 		       SKB_GSO_SIT |
 		       SKB_GSO_UDP_TUNNEL |
+		       SKB_GSO_UDP_TUNNEL_CSUM |
 		       SKB_GSO_MPLS |
 		       SKB_GSO_TCPV6 |
 		       0)))
diff --git a/net/ipv6/udp_offload.c b/net/ipv6/udp_offload.c
index b261ee8b83fc..79da8b305ced 100644
--- a/net/ipv6/udp_offload.c
+++ b/net/ipv6/udp_offload.c
@@ -63,6 +63,7 @@ static struct sk_buff *udp6_ufo_fragment(struct sk_buff *skb,
 		if (unlikely(type & ~(SKB_GSO_UDP |
 				      SKB_GSO_DODGY |
 				      SKB_GSO_UDP_TUNNEL |
+				      SKB_GSO_UDP_TUNNEL_CSUM |
 				      SKB_GSO_GRE |
 				      SKB_GSO_IPIP |
 				      SKB_GSO_SIT |
@@ -76,7 +77,8 @@ static struct sk_buff *udp6_ufo_fragment(struct sk_buff *skb,
 		goto out;
 	}
 
-	if (skb->encapsulation && skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL)
+	if (skb->encapsulation && skb_shinfo(skb)->gso_type &
+	    (SKB_GSO_UDP_TUNNEL|SKB_GSO_UDP_TUNNEL_CSUM))
 		segs = skb_udp_tunnel_segment(skb, features);
 	else {
 		/* Do software UFO. Complete and fill in the UDP checksum as HW cannot