datapath: Add Stateless TCP Tunneling protocol.

The Stateless TCP Tunnel (STT) protocol encapsulates traffic in IPv4/TCP packets. STT uses TCP segmentation offload available in most of NIC. On packet xmit STT driver appends STT header along with TCP header to the packet. For GSO packet GSO parameters are set according to tunnel configuration and packet is handed over to networking stack. This allows use of segmentation offload available in NICs The protocol is documented at http://www.ietf.org/archive/id/draft-davie-stt-06.txt Signed-off-by: Pravin B Shelar <[email protected]> Signed-off-by: Jesse Gross <[email protected]> Acked-by: Jesse Gross <[email protected]>
JerryChiao · Apr 29, 2015 · 4237026 · 4237026
1 parent a51a508
commit 4237026
Show file tree

Hide file tree

Showing 14 changed files with 1,920 additions and 12 deletions.
diff --git a/FAQ.md b/FAQ.md
@@ -218,6 +218,7 @@ A: Support for tunnels was added to the upstream Linux kernel module
 | VXLAN    |    3.12
 | Geneve   |    3.18
 | LISP     | <not upstream>
+| STT      | <not upstream>
 
    If you are using a version of the kernel that is older than the one
    listed above, it is still possible to use that tunnel protocol. However,

diff --git a/NEWS b/NEWS
@@ -80,6 +80,7 @@ Post-v2.3.0
    - The kernel vports with dependencies are no longer part of the overall
      openvswitch.ko but built and loaded automatically as individual kernel
      modules (vport-*.ko).
+   - Support for STT tunneling.
 
 
 v2.3.0 - 14 Aug 2014

diff --git a/datapath/Modules.mk b/datapath/Modules.mk
@@ -9,6 +9,7 @@ both_modules = \
 	vport_geneve \
 	vport_gre \
 	vport_lisp \
+	vport_stt \
 	vport_vxlan
 # When changing the name of 'build_modules', please also update the
 # print-build-modules in Makefile.am.
@@ -30,6 +31,7 @@ vport_geneve_sources = vport-geneve.c
 vport_vxlan_sources = vport-vxlan.c
 vport_gre_sources = vport-gre.c
 vport_lisp_sources = vport-lisp.c
+vport_stt_sources = vport-stt.c
 
 openvswitch_headers = \
 	compat.h \

diff --git a/datapath/linux/.gitignore b/datapath/linux/.gitignore
@@ -35,6 +35,7 @@
 /random32.c
 /reciprocal_div.c
 /skbuff-openvswitch.c
+/stt.c
 /table.c
 /time.c
 /tmp
@@ -50,6 +51,7 @@
 /vport-lisp.c
 /vport-netdev.c
 /vport-patch.c
+/vport-stt.c
 /vport-vxlan.c
 /vport.c
 /vxlan.c

diff --git a/datapath/linux/Modules.mk b/datapath/linux/Modules.mk
@@ -12,6 +12,7 @@ openvswitch_sources += \
 	linux/compat/net_namespace.c \
 	linux/compat/reciprocal_div.c \
 	linux/compat/skbuff-openvswitch.c \
+	linux/compat/stt.c \
 	linux/compat/udp.c \
 	linux/compat/udp_tunnel.c \
 	linux/compat/vxlan.c	\
@@ -75,6 +76,7 @@ openvswitch_headers += \
 	linux/compat/include/net/udp.h \
 	linux/compat/include/net/udp_tunnel.h \
 	linux/compat/include/net/sock.h \
+	linux/compat/include/net/stt.h \
 	linux/compat/include/net/vxlan.h \
 	linux/compat/include/net/sctp/checksum.h
 EXTRA_DIST += linux/compat/build-aux/export-check-whitelist
diff --git a/datapath/linux/compat/gso.h b/datapath/linux/compat/gso.h
@@ -26,6 +26,15 @@ struct ovs_gso_cb {
 };
 #define OVS_GSO_CB(skb) ((struct ovs_gso_cb *)(skb)->cb)
 
+static inline void skb_clear_ovs_gso_cb(struct sk_buff *skb)
+{
+	OVS_GSO_CB(skb)->fix_segment = NULL;
+}
+#else
+static inline void skb_clear_ovs_gso_cb(struct sk_buff *skb)
+{
+
+}
 #endif
 
 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,10,0)

diff --git a/datapath/linux/compat/include/linux/openvswitch.h b/datapath/linux/compat/include/linux/openvswitch.h
@@ -229,6 +229,7 @@ enum ovs_vport_type {
 	OVS_VPORT_TYPE_GENEVE,	 /* Geneve tunnel. */
 	OVS_VPORT_TYPE_GRE64 = 104, /* GRE tunnel with 64-bit keys */
 	OVS_VPORT_TYPE_LISP = 105,  /* LISP tunnel */
+	OVS_VPORT_TYPE_STT = 106, /* STT tunnel */
 	__OVS_VPORT_TYPE_MAX
 };
 

diff --git a/datapath/linux/compat/include/net/stt.h b/datapath/linux/compat/include/net/stt.h
@@ -0,0 +1,71 @@
+#ifndef __NET_STT_H
+#define __NET_STT_H  1
+
+#include <linux/kconfig.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3,5,0) && IS_ENABLED(CONFIG_NETFILTER)
+#include <net/ip_tunnels.h>
+#define OVS_STT
+
+struct stthdr {
+	__u8		version;
+	__u8		flags;
+	__u8		l4_offset;
+	__u8		reserved;
+	__be16		mss;
+	__be16		vlan_tci;
+	__be64		key;
+};
+
+/* Padding after the end of the tunnel headers to provide alignment
+ * for inner packet IP header after 14 byte Ethernet header.
+ */
+#define STT_ETH_PAD 2
+
+#define STT_BASE_HLEN   (sizeof(struct stthdr) + STT_ETH_PAD)
+#define STT_HEADER_LEN	(sizeof(struct tcphdr) + STT_BASE_HLEN)
+
+static inline struct stthdr *stt_hdr(const struct sk_buff *skb)
+{
+	return (struct stthdr *)(skb_transport_header(skb) +
+				 sizeof(struct tcphdr));
+}
+
+struct stt_sock;
+typedef void (stt_rcv_t)(struct stt_sock *stt_sock, struct sk_buff *skb);
+
+/* @list: Per-net list of STT ports.
+ * @rcv: The callback is called on STT packet recv, STT reassembly can generate
+ * multiple packets, in this case first packet has tunnel outer header, rest
+ * of the packets are inner packet segments with no stt header.
+ * @rcv_data: user data.
+ * @sock: Fake TCP socket for the STT port.
+ */
+struct stt_sock {
+	struct list_head	list;
+	stt_rcv_t		*rcv;
+	void			*rcv_data;
+	struct socket		*sock;
+	struct rcu_head		rcu;
+};
+
+#define stt_sock_add rpl_stt_sock_add
+struct stt_sock *rpl_stt_sock_add(struct net *net, __be16 port,
+			      stt_rcv_t *rcv, void *data);
+
+#define stt_sock_release rpl_stt_sock_release
+void rpl_stt_sock_release(struct stt_sock *stt_sock);
+
+#define stt_xmit_skb rpl_stt_xmit_skb
+int rpl_stt_xmit_skb(struct sk_buff *skb, struct rtable *rt,
+		 __be32 src, __be32 dst, __u8 tos,
+		 __u8 ttl, __be16 df, __be16 src_port, __be16 dst_port,
+		 __be64 tun_id);
+
+#define stt_init_module ovs_stt_init_module
+int ovs_stt_init_module(void);
+
+#define stt_cleanup_module ovs_stt_cleanup_module
+void ovs_stt_cleanup_module(void);
+
+#endif
+#endif /*ifdef__NET_STT_H */