selftests/bpf: expand test_tc_tunnel with SIT encap

So far, all BPF tc tunnel testcases encapsulate in the same network
protocol. Add an encap testcase that requires updating skb->protocol.

The 6in4 tunnel encapsulates an IPv6 packet inside an IPv4 tunnel.
Verify that bpf_skb_net_grow correctly updates skb->protocol to
select the right protocol handler in __netif_receive_skb_core.

The BPF program should also manually update the link layer header to
encode the right network protocol.

Changes v1->v2
  - improve documentation of non-obvious logic
Signed-off-by: Willem de Bruijn <willemb@google.com>
Tested-by: Alan Maguire <alan.maguire@oracle.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

tools/testing/selftests/bpf/config

@@ -33,3 +33,4 @@ CONFIG_MPLS=y
 CONFIG_NET_MPLS_GSO=m
 CONFIG_MPLS_ROUTING=m
 CONFIG_MPLS_IPTUNNEL=m
+CONFIG_IPV6_SIT=m

tools/testing/selftests/bpf/progs/test_tc_tunnel.c

@@ -77,17 +77,52 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, __u8 encap_proto,
 	struct v4hdr h_outer;
 	struct tcphdr tcph;
 	int olen, l2_len;
+	int tcp_off;
 	__u64 flags;
 
-	if (bpf_skb_load_bytes(skb, ETH_HLEN, &iph_inner,
-			       sizeof(iph_inner)) < 0)
-		return TC_ACT_OK;
+	/* Most tests encapsulate a packet into a tunnel with the same
+	 * network protocol, and derive the outer header fields from
+	 * the inner header.
+	 *
+	 * The 6in4 case tests different inner and outer protocols. As
+	 * the inner is ipv6, but the outer expects an ipv4 header as
+	 * input, manually build a struct iphdr based on the ipv6hdr.
+	 */
+	if (encap_proto == IPPROTO_IPV6) {
+		const __u32 saddr = (192 << 24) | (168 << 16) | (1 << 8) | 1;
+		const __u32 daddr = (192 << 24) | (168 << 16) | (1 << 8) | 2;
+		struct ipv6hdr iph6_inner;
+
+		/* Read the IPv6 header */
+		if (bpf_skb_load_bytes(skb, ETH_HLEN, &iph6_inner,
+				       sizeof(iph6_inner)) < 0)
+			return TC_ACT_OK;
+
+		/* Derive the IPv4 header fields from the IPv6 header */
+		memset(&iph_inner, 0, sizeof(iph_inner));
+		iph_inner.version = 4;
+		iph_inner.ihl = 5;
+		iph_inner.tot_len = bpf_htons(sizeof(iph6_inner) +
+				    bpf_ntohs(iph6_inner.payload_len));
+		iph_inner.ttl = iph6_inner.hop_limit - 1;
+		iph_inner.protocol = iph6_inner.nexthdr;
+		iph_inner.saddr = __bpf_constant_htonl(saddr);
+		iph_inner.daddr = __bpf_constant_htonl(daddr);
+
+		tcp_off = sizeof(iph6_inner);
+	} else {
+		if (bpf_skb_load_bytes(skb, ETH_HLEN, &iph_inner,
+				       sizeof(iph_inner)) < 0)
+			return TC_ACT_OK;
+
+		tcp_off = sizeof(iph_inner);
+	}
 
 	/* filter only packets we want */
 	if (iph_inner.ihl != 5 || iph_inner.protocol != IPPROTO_TCP)
 		return TC_ACT_OK;
 
-	if (bpf_skb_load_bytes(skb, ETH_HLEN + sizeof(iph_inner),
+	if (bpf_skb_load_bytes(skb, ETH_HLEN + tcp_off,
 			       &tcph, sizeof(tcph)) < 0)
 		return TC_ACT_OK;
 
@@ -129,6 +164,7 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, __u8 encap_proto,
 						  l2_len);
 		break;
 	case IPPROTO_IPIP:
+	case IPPROTO_IPV6:
 		break;
 	default:
 		return TC_ACT_OK;
@@ -164,6 +200,17 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, __u8 encap_proto,
 				BPF_F_INVALIDATE_HASH) < 0)
 		return TC_ACT_SHOT;
 
+	/* if changing outer proto type, update eth->h_proto */
+	if (encap_proto == IPPROTO_IPV6) {
+		struct ethhdr eth;
+
+		if (bpf_skb_load_bytes(skb, 0, &eth, sizeof(eth)) < 0)
+			return TC_ACT_SHOT;
+		eth.h_proto = bpf_htons(ETH_P_IP);
+		if (bpf_skb_store_bytes(skb, 0, &eth, sizeof(eth), 0) < 0)
+			return TC_ACT_SHOT;
+	}
+
 	return TC_ACT_OK;
 }
 
@@ -325,6 +372,15 @@ int __encap_udp_eth(struct __sk_buff *skb)
 		return TC_ACT_OK;
 }
 
+SEC("encap_sit_none")
+int __encap_sit_none(struct __sk_buff *skb)
+{
+	if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6))
+		return encap_ipv4(skb, IPPROTO_IPV6, ETH_P_IP);
+	else
+		return TC_ACT_OK;
+}
+
 SEC("encap_ip6tnl_none")
 int __encap_ip6tnl_none(struct __sk_buff *skb)
 {

tools/testing/selftests/bpf/test_tc_tunnel.sh

@@ -97,6 +97,9 @@ if [[ "$#" -eq "0" ]]; then
 	echo "ip6ip6"
 	$0 ipv6 ip6tnl none 100
 
+	echo "sit"
+	$0 ipv6 sit none 100
+
 	for mac in none mpls eth ; do
 		echo "ip gre $mac"
 		$0 ipv4 gre $mac 100
@@ -211,11 +214,20 @@ else
 	targs=""
 fi
 
+# tunnel address family differs from inner for SIT
+if [[ "${tuntype}" == "sit" ]]; then
+	link_addr1="${ns1_v4}"
+	link_addr2="${ns2_v4}"
+else
+	link_addr1="${addr1}"
+	link_addr2="${addr2}"
+fi
+
 # serverside, insert decap module
 # server is still running
 # client can connect again
 ip netns exec "${ns2}" ip link add name testtun0 type "${ttype}" \
-	${tmode} remote "${addr1}" local "${addr2}" $targs
+	${tmode} remote "${link_addr1}" local "${link_addr2}" $targs
 
 expect_tun_fail=0
 
@@ -260,6 +272,12 @@ else
 	server_listen
 fi
 
+# bpf_skb_net_shrink does not take tunnel flags yet, cannot update L3.
+if [[ "${tuntype}" == "sit" ]]; then
+	echo OK
+	exit 0
+fi
+
 # serverside, use BPF for decap
 ip netns exec "${ns2}" ip link del dev testtun0
 ip netns exec "${ns2}" tc qdisc add dev veth2 clsact