ipvs-toa.patch

diff --git a/include/net/ip_vs.h b/include/net/ip_vs.h
index 5d08c19..27df4d8 100644
--- a/include/net/ip_vs.h
+++ b/include/net/ip_vs.h
@@ -348,6 +348,22 @@ enum ip_vs_sctp_states {
 	IP_VS_SCTP_S_LAST
 };
 
+#define TCPOPT_ADDR  254
+#define TCPOLEN_ADDR 8		/* |opcode|size|ip+port| = 1 + 1 + 6 */
+
+/*
+ * insert client ip in tcp option, now only support IPV4,
+ * must be 4 bytes alignment.
+ */
+struct ip_vs_tcpo_addr {
+	__u8 opcode;
+	__u8 opsize;
+	__u16 port;
+	__u32 addr;
+};
+
+extern int sysctl_ip_vs_toa_entry;
+
 /* Delta sequence info structure
  * Each ip_vs_conn has 2 (output AND input seq. changes).
  * Only used in the VS/NAT.
diff --git a/net/netfilter/ipvs/ip_vs_ctl.c b/net/netfilter/ipvs/ip_vs_ctl.c
index e8f1556..3a5bbdd 100644
--- a/net/netfilter/ipvs/ip_vs_ctl.c
+++ b/net/netfilter/ipvs/ip_vs_ctl.c
@@ -66,6 +66,8 @@ int ip_vs_get_debug_level(void)
 }
 #endif
 
+int sysctl_ip_vs_toa_entry = 1;
+
 
 /*  Protos */
 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
@@ -1637,6 +1639,7 @@ static int ip_vs_zero_all(struct netns_ipvs *ipvs)
 #ifdef CONFIG_SYSCTL
 
 static int zero;
+static int one = 1;
 static int three = 3;
 
 static int
@@ -1884,6 +1887,15 @@ static int ip_vs_zero_all(struct netns_ipvs *ipvs)
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec,
 	},
+	{
+	 	.procname = "nat_toa_entry",
+	 	.data = &sysctl_ip_vs_toa_entry,
+	 	.maxlen = sizeof(int),
+	 	.mode = 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &zero,
+		.extra2		= &one,
+	 },
 #ifdef CONFIG_IP_VS_DEBUG
 	{
 		.procname	= "debug_level",
diff --git a/net/netfilter/ipvs/ip_vs_proto_tcp.c b/net/netfilter/ipvs/ip_vs_proto_tcp.c
index 121a321..cfa842c 100644
--- a/net/netfilter/ipvs/ip_vs_proto_tcp.c
+++ b/net/netfilter/ipvs/ip_vs_proto_tcp.c
@@ -142,6 +142,110 @@
 						csum_unfold(tcph->check))));
 }
 
+/*
+ * add client (ip and port) in tcp option
+ * return 0 if success
+ */
+static int tcp_opt_add_toa(struct ip_vs_conn *cp,
+				       struct sk_buff *skb,
+				       struct tcphdr **tcph)
+{
+	__u32 mtu;
+	struct ip_vs_tcpo_addr *toa;
+	unsigned int tcphoff;
+	struct tcphdr *th;
+	__u8 *p, *q;
+
+	/* now only process IPV4 */
+	if (cp->af != AF_INET) {
+		return 1;
+	}
+
+	/* skb length and tcp option length checking */
+	if (skb_dst(skb))
+		mtu = dst_mtu(skb_dst(skb));
+	else		 /* fast_xmit can reach here */
+		mtu = sizeof(struct ip_vs_tcpo_addr);
+
+	if (skb->len > (mtu - sizeof(struct ip_vs_tcpo_addr))) {
+		return 1;
+	}
+
+	/* the maximum length of TCP head is 60 bytes, so only 40 bytes for options */
+	if ((60 - ((*tcph)->doff << 2)) < sizeof(struct ip_vs_tcpo_addr)) {
+		return 1;
+	}
+
+	/* expand skb if needed */
+	if ((sizeof(struct ip_vs_tcpo_addr) > skb_tailroom(skb)) &&
+			pskb_expand_head(skb, 0,
+			sizeof(struct ip_vs_tcpo_addr), GFP_ATOMIC)){
+		return 1;
+	}
+
+	/*
+	 * add client ip
+	 */
+	tcphoff = ip_hdrlen(skb);
+	/* get new tcp header */
+	*tcph = th =
+	    (struct tcphdr *)((void *)skb_network_header(skb) + tcphoff);
+
+	/* ptr to old opts */
+	p = skb_tail_pointer(skb) - 1;
+	q = p + sizeof(struct ip_vs_tcpo_addr);
+
+	/* move data down, offset is sizeof(struct ip_vs_tcpo_addr) */
+	while (p >= ((__u8 *) th + sizeof(struct tcphdr))) {
+		*q = *p;
+		p--;
+		q--;
+	}
+
+	/* move tail to new postion */
+	skb->tail += sizeof(struct ip_vs_tcpo_addr);
+
+	/* put client ip opt , ptr point to opts */
+	toa = (struct ip_vs_tcpo_addr *)(th + 1);
+	toa->opcode = TCPOPT_ADDR;
+	toa->opsize = TCPOLEN_ADDR;
+	toa->port = cp->cport;
+	toa->addr = cp->caddr.ip;
+
+	/* reset tcp header length */
+	th->doff += sizeof(struct ip_vs_tcpo_addr) / 4;
+	/* reset ip header totoal length */
+	ip_hdr(skb)->tot_len =
+	    htons(ntohs(ip_hdr(skb)->tot_len) +
+		  sizeof(struct ip_vs_tcpo_addr));
+	/* reset skb length */
+	skb->len += sizeof(struct ip_vs_tcpo_addr);
+
+	/* re-calculate tcp csum, if no csum_offload */
+        bool sysctl_ip_vs_csum_offload = false;
+	if (!sysctl_ip_vs_csum_offload) {
+		th->check = 0;
+		skb->csum = skb_checksum(skb, tcphoff,
+					skb->len - tcphoff, 0);
+		th->check = csum_tcpudp_magic(cp->caddr.ip,
+					cp->vaddr.ip,
+					skb->len - tcphoff,
+					cp->protocol, skb->csum);
+
+		if (skb->ip_summed == CHECKSUM_PARTIAL) {
+			skb->ip_summed = CHECKSUM_COMPLETE;
+			skb_shinfo(skb)->gso_size = 0;
+		}
+	}
+
+	/* re-calculate ip head csum, tot_len has been adjusted */
+	ip_send_check(ip_hdr(skb));
+
+	return 0;
+}
+
+
+
 
 static int
 tcp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
@@ -225,6 +329,7 @@
 tcp_dnat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
 		 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph)
 {
+
 	struct tcphdr *tcph;
 	unsigned int tcphoff = iph->len;
 	int oldlen;
@@ -261,6 +366,24 @@
 	}
 
 	tcph = (void *)skb_network_header(skb) + tcphoff;
+
+
+	/*
+	 * for syn packet
+	 * add toa
+	 */
+	if (tcph->syn & !tcph->ack) {
+		tcp_opt_add_toa(cp, skb, &tcph);
+	}
+
+	/* TOA: add client ip */
+	if ((sysctl_ip_vs_toa_entry == 1)
+	    // && (ntohl(tcph->ack_seq) == cp->in_seq.init_seq)
+	    && !tcph->syn && !tcph->rst && !tcph->fin) {
+			tcp_opt_add_toa(cp, skb, &tcph);
+	}
+
+
 	tcph->dest = cp->dport;
 
 	/*
@@ -299,7 +422,6 @@
 	return 1;
 }
 
-
 static int
 tcp_csum_check(int af, struct sk_buff *skb, struct ip_vs_protocol *pp)
 {