X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Fvport-lisp.c;h=a1e2b2b69caa678249e3ceb259f83f2bb797719b;hb=HEAD;hp=0f013953025b3f4983752a618e8e985e10c1a5fc;hpb=a6ae068b7b3ea1e039279e35ffae6ceddfb73866;p=sliver-openvswitch.git

diff --git a/datapath/vport-lisp.c b/datapath/vport-lisp.c
index 0f0139530..a1e2b2b69 100644
--- a/datapath/vport-lisp.c
+++ b/datapath/vport-lisp.c
@@ -20,23 +20,23 @@
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
 #include <linux/version.h>
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,26)
 
 #include <linux/in.h>
 #include <linux/ip.h>
-#include <linux/list.h>
 #include <linux/net.h>
+#include <linux/rculist.h>
 #include <linux/udp.h>
 
 #include <net/icmp.h>
 #include <net/ip.h>
+#include <net/route.h>
 #include <net/udp.h>
+#include <net/xfrm.h>
 
 #include "datapath.h"
-#include "tunnel.h"
+#include "gso.h"
 #include "vport.h"
 
-
 /*
  *  LISP encapsulation header:
  *
@@ -94,34 +94,33 @@ struct lisphdr {
 
 #define LISP_HLEN (sizeof(struct udphdr) + sizeof(struct lisphdr))
 
-static inline int lisp_hdr_len(const struct tnl_mutable_config *mutable,
-			       const struct ovs_key_ipv4_tunnel *tun_key)
-{
-	return LISP_HLEN;
-}
-
 /**
  * struct lisp_port - Keeps track of open UDP ports
- * @list: list element.
- * @port: The UDP port number in network byte order.
- * @socket: The socket created for this port number.
- * @count: How many ports are using this socket/port.
+ * @dst_port: lisp UDP port no.
+ * @list: list element in @lisp_ports.
+ * @lisp_rcv_socket: The socket created for this port number.
+ * @name: vport name.
  */
 struct lisp_port {
+	__be16 dst_port;
 	struct list_head list;
-	__be16 port;
 	struct socket *lisp_rcv_socket;
-	int count;
+	char name[IFNAMSIZ];
 };
 
 static LIST_HEAD(lisp_ports);
 
-static struct lisp_port *lisp_port_exists(struct net *net, __be16 port)
+static inline struct lisp_port *lisp_vport(const struct vport *vport)
+{
+	return vport_priv(vport);
+}
+
+static struct lisp_port *lisp_find_port(struct net *net, __be16 port)
 {
 	struct lisp_port *lisp_port;
 
-	list_for_each_entry(lisp_port, &lisp_ports, list) {
-		if (lisp_port->port == port &&
+	list_for_each_entry_rcu(lisp_port, &lisp_ports, list) {
+		if (lisp_port->dst_port == port &&
 			net_eq(sock_net(lisp_port->lisp_rcv_socket->sk), net))
 			return lisp_port;
 	}
@@ -134,25 +133,6 @@ static inline struct lisphdr *lisp_hdr(const struct sk_buff *skb)
 	return (struct lisphdr *)(udp_hdr(skb) + 1);
 }
 
-static int lisp_tnl_send(struct vport *vport, struct sk_buff *skb)
-{
-	int tnl_len;
-	int network_offset = skb_network_offset(skb);
-
-	/* We only encapsulate IPv4 and IPv6 packets */
-	switch (skb->protocol) {
-	case htons(ETH_P_IP):
-	case htons(ETH_P_IPV6):
-		/* Pop off "inner" Ethernet header */
-		skb_pull(skb, network_offset);
-		tnl_len = ovs_tnl_send(vport, skb);
-		return tnl_len > 0 ? tnl_len + network_offset : tnl_len;
-	default:
-		kfree_skb(skb);
-		return 0;
-	}
-}
-
 /* Convert 64 bit tunnel ID to 24 bit Instance ID. */
 static void tunnel_id_to_instance_id(__be64 tun_id, __u8 *iid)
 {
@@ -180,22 +160,39 @@ static __be64 instance_id_to_tunnel_id(__u8 *iid)
 #endif
 }
 
-static struct sk_buff *lisp_build_header(const struct vport *vport,
-					 const struct tnl_mutable_config *mutable,
-					 struct dst_entry *dst,
-					 struct sk_buff *skb,
-					 int tunnel_hlen)
+/* Compute source UDP port for outgoing packet.
+ * Currently we use the flow hash.
+ */
+static u16 get_src_port(struct net *net, struct sk_buff *skb)
+{
+	u32 hash = skb_get_hash(skb);
+	unsigned int range;
+	int high;
+	int low;
+
+	if (!hash) {
+		struct sw_flow_key *pkt_key = OVS_CB(skb)->pkt_key;
+
+		hash = jhash2((const u32 *)pkt_key,
+			    sizeof(*pkt_key) / sizeof(u32), 0);
+	}
+
+	inet_get_local_port_range(net, &low, &high);
+	range = (high - low) + 1;
+	return (((u64) hash * range) >> 32) + low;
+}
+
+static void lisp_build_header(const struct vport *vport,
+			      struct sk_buff *skb)
 {
+	struct net *net = ovs_dp_get_net(vport->dp);
+	struct lisp_port *lisp_port = lisp_vport(vport);
 	struct udphdr *udph = udp_hdr(skb);
 	struct lisphdr *lisph = (struct lisphdr *)(udph + 1);
 	const struct ovs_key_ipv4_tunnel *tun_key = OVS_CB(skb)->tun_key;
-	__be64 out_key;
-	u32 flags;
-
-	tnl_get_param(mutable, tun_key, &flags, &out_key);
 
-	udph->dest = mutable->dst_port;
-	udph->source = htons(ovs_tnl_get_src_port(skb));
+	udph->dest = lisp_port->dst_port;
+	udph->source = htons(get_src_port(net, skb));
 	udph->check = 0;
 	udph->len = htons(skb->len - skb_transport_offset(skb));
 
@@ -210,60 +207,38 @@ static struct sk_buff *lisp_build_header(const struct vport *vport,
 	lisph->u1.nonce[1] = 0;
 	lisph->u1.nonce[2] = 0;
 
-	tunnel_id_to_instance_id(out_key, &lisph->u2.word2.instance_id[0]);
+	tunnel_id_to_instance_id(tun_key->tun_id, &lisph->u2.word2.instance_id[0]);
 	lisph->u2.word2.locator_status_bits = 1;
-
-	/*
-	 * Allow our local IP stack to fragment the outer packet even if the
-	 * DF bit is set as a last resort.  We also need to force selection of
-	 * an IP ID here because Linux will otherwise leave it at 0 if the
-	 * packet originally had DF set.
-	 */
-	skb->local_df = 1;
-	__ip_select_ident(ip_hdr(skb), dst, 0);
-
-	return skb;
 }
 
 /* Called with rcu_read_lock and BH disabled. */
 static int lisp_rcv(struct sock *sk, struct sk_buff *skb)
 {
-	struct vport *vport;
+	struct lisp_port *lisp_port;
 	struct lisphdr *lisph;
-	const struct tnl_mutable_config *mutable;
 	struct iphdr *iph, *inner_iph;
 	struct ovs_key_ipv4_tunnel tun_key;
 	__be64 key;
-	u32 tunnel_flags = 0;
 	struct ethhdr *ethh;
 	__be16 protocol;
 
-	if (unlikely(!pskb_may_pull(skb, LISP_HLEN)))
+	lisp_port = lisp_find_port(dev_net(skb->dev), udp_hdr(skb)->dest);
+	if (unlikely(!lisp_port))
 		goto error;
 
-	lisph = lisp_hdr(skb);
+	if (iptunnel_pull_header(skb, LISP_HLEN, 0))
+		goto error;
 
-	skb_pull_rcsum(skb, LISP_HLEN);
+	lisph = lisp_hdr(skb);
 
 	if (lisph->instance_id_present != 1)
 		key = 0;
 	else
 		key = instance_id_to_tunnel_id(&lisph->u2.word2.instance_id[0]);
 
-	iph = ip_hdr(skb);
-	vport = ovs_tnl_find_port(dev_net(skb->dev), iph->daddr, iph->saddr,
-		key, TNL_T_PROTO_LISP, &mutable);
-	if (unlikely(!vport))
-		goto error;
-
-	if (mutable->flags & TNL_F_IN_KEY_MATCH || !mutable->key.daddr)
-		tunnel_flags = OVS_TNL_F_KEY;
-	else
-		key = 0;
-
 	/* Save outer tunnel values */
-	tnl_tun_key_init(&tun_key, iph, key, tunnel_flags);
-	OVS_CB(skb)->tun_key = &tun_key;
+	iph = ip_hdr(skb);
+	ovs_flow_tun_key_init(&tun_key, iph, key, TUNNEL_KEY);
 
 	/* Drop non-IP inner packets */
 	inner_iph = (struct iphdr *)(lisph + 1);
@@ -277,6 +252,7 @@ static int lisp_rcv(struct sock *sk, struct sk_buff *skb)
 	default:
 		goto error;
 	}
+	skb->protocol = protocol;
 
 	/* Add Ethernet header */
 	ethh = (struct ethhdr *)skb_push(skb, ETH_HLEN);
@@ -285,7 +261,9 @@ static int lisp_rcv(struct sock *sk, struct sk_buff *skb)
 	ethh->h_source[0] = 0x02;
 	ethh->h_proto = protocol;
 
-	ovs_tnl_rcv(vport, skb);
+	ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
+
+	ovs_vport_receive(vport_from_priv(lisp_port), skb, &tun_key);
 	goto out;
 
 error:
@@ -298,8 +276,8 @@ out:
 #define UDP_ENCAP_LISP 1
 static int lisp_socket_init(struct lisp_port *lisp_port, struct net *net)
 {
-	int err;
 	struct sockaddr_in sin;
+	int err;
 
 	err = sock_create_kern(AF_INET, SOCK_DGRAM, 0,
 			       &lisp_port->lisp_rcv_socket);
@@ -311,7 +289,7 @@ static int lisp_socket_init(struct lisp_port *lisp_port, struct net *net)
 
 	sin.sin_family = AF_INET;
 	sin.sin_addr.s_addr = htonl(INADDR_ANY);
-	sin.sin_port = lisp_port->port;
+	sin.sin_port = lisp_port->dst_port;
 
 	err = kernel_bind(lisp_port->lisp_rcv_socket, (struct sockaddr *)&sin,
 			  sizeof(struct sockaddr_in));
@@ -332,31 +310,39 @@ error:
 	return err;
 }
 
-static void lisp_tunnel_release(struct lisp_port *lisp_port)
+static int lisp_get_options(const struct vport *vport, struct sk_buff *skb)
 {
-	lisp_port->count--;
+	struct lisp_port *lisp_port = lisp_vport(vport);
 
-	if (lisp_port->count == 0) {
-		/* Release old socket */
-		sk_release_kernel(lisp_port->lisp_rcv_socket->sk);
-		list_del(&lisp_port->list);
-		kfree(lisp_port);
-	}
+	if (nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, ntohs(lisp_port->dst_port)))
+		return -EMSGSIZE;
+	return 0;
+}
+
+static void lisp_tnl_destroy(struct vport *vport)
+{
+	struct lisp_port *lisp_port = lisp_vport(vport);
+
+	list_del_rcu(&lisp_port->list);
+	/* Release socket */
+	sk_release_kernel(lisp_port->lisp_rcv_socket->sk);
+
+	ovs_vport_deferred_free(vport);
 }
 
-static int lisp_tunnel_setup(struct net *net, struct nlattr *options,
-			     struct lisp_port **lport)
+static struct vport *lisp_tnl_create(const struct vport_parms *parms)
 {
+	struct net *net = ovs_dp_get_net(parms->dp);
+	struct nlattr *options = parms->options;
+	struct lisp_port *lisp_port;
+	struct vport *vport;
 	struct nlattr *a;
 	int err;
 	u16 dst_port;
-	struct lisp_port *lisp_port = NULL;
-
-	*lport = NULL;
 
 	if (!options) {
 		err = -EINVAL;
-		goto out;
+		goto error;
 	}
 
 	a = nla_find_nested(options, OVS_TUNNEL_ATTR_DST_PORT);
@@ -365,124 +351,162 @@ static int lisp_tunnel_setup(struct net *net, struct nlattr *options,
 	} else {
 		/* Require destination port from userspace. */
 		err = -EINVAL;
-		goto out;
+		goto error;
 	}
 
 	/* Verify if we already have a socket created for this port */
-	lisp_port = lisp_port_exists(net, htons(dst_port));
-	if (lisp_port) {
-		lisp_port->count++;
-		err = 0;
-		*lport = lisp_port;
-		goto out;
+	if (lisp_find_port(net, htons(dst_port))) {
+		err = -EEXIST;
+		goto error;
 	}
 
-	/* Add a new socket for this port */
-	lisp_port = kzalloc(sizeof(struct lisp_port), GFP_KERNEL);
-	if (!lisp_port) {
-		err = -ENOMEM;
-		goto out;
-	}
+	vport = ovs_vport_alloc(sizeof(struct lisp_port),
+				&ovs_lisp_vport_ops, parms);
+	if (IS_ERR(vport))
+		return vport;
 
-	lisp_port->port = htons(dst_port);
-	lisp_port->count = 1;
-	list_add_tail(&lisp_port->list, &lisp_ports);
+	lisp_port = lisp_vport(vport);
+	lisp_port->dst_port = htons(dst_port);
+	strncpy(lisp_port->name, parms->name, IFNAMSIZ);
 
 	err = lisp_socket_init(lisp_port, net);
 	if (err)
-		goto error;
+		goto error_free;
 
-	*lport = lisp_port;
-	goto out;
+	list_add_tail_rcu(&lisp_port->list, &lisp_ports);
+	return vport;
 
+error_free:
+	ovs_vport_free(vport);
 error:
-	list_del(&lisp_port->list);
-	kfree(lisp_port);
-out:
-	return err;
+	return ERR_PTR(err);
 }
 
-static int lisp_tnl_set_options(struct vport *vport, struct nlattr *options)
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3,12,0)
+
+static void lisp_fix_segment(struct sk_buff *skb)
 {
-	int err;
-	struct net *net = ovs_dp_get_net(vport->dp);
-	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
-	struct tnl_mutable_config *config;
-	struct lisp_port *old_port = NULL;
-	struct lisp_port *lisp_port = NULL;
+	struct udphdr *udph = udp_hdr(skb);
 
-	config = rtnl_dereference(tnl_vport->mutable);
+	udph->len = htons(skb->len - skb_transport_offset(skb));
+}
 
-	old_port = lisp_port_exists(net, config->dst_port);
+static int handle_offloads(struct sk_buff *skb)
+{
+	if (skb_is_gso(skb))
+		OVS_GSO_CB(skb)->fix_segment = lisp_fix_segment;
+	else if (skb->ip_summed != CHECKSUM_PARTIAL)
+		skb->ip_summed = CHECKSUM_NONE;
+	return 0;
+}
+#else
+static int handle_offloads(struct sk_buff *skb)
+{
+	if (skb_is_gso(skb)) {
+		int err = skb_unclone(skb, GFP_ATOMIC);
+		if (unlikely(err))
+			return err;
 
-	err = lisp_tunnel_setup(net, options, &lisp_port);
-	if (err)
-		goto out;
+		skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
+	} else if (skb->ip_summed != CHECKSUM_PARTIAL)
+		skb->ip_summed = CHECKSUM_NONE;
+
+	skb->encapsulation = 1;
+	return 0;
+}
+#endif
 
-	err = ovs_tnl_set_options(vport, options);
+static int lisp_send(struct vport *vport, struct sk_buff *skb)
+{
+	int network_offset = skb_network_offset(skb);
+	struct rtable *rt;
+	int min_headroom;
+	__be32 saddr;
+	__be16 df;
+	int sent_len;
+	int err;
 
-	if (err)
-		lisp_tunnel_release(lisp_port);
-	else {
-		/* Release old socket */
-		lisp_tunnel_release(old_port);
+	if (unlikely(!OVS_CB(skb)->tun_key))
+		return -EINVAL;
+
+	if (skb->protocol != htons(ETH_P_IP) &&
+	    skb->protocol != htons(ETH_P_IPV6)) {
+		kfree_skb(skb);
+		return 0;
 	}
-out:
-	return err;
-}
 
-static const struct tnl_ops ovs_lisp_tnl_ops = {
-	.tunnel_type	= TNL_T_PROTO_LISP,
-	.ipproto	= IPPROTO_UDP,
-	.hdr_len	= lisp_hdr_len,
-	.build_header	= lisp_build_header,
-};
+	/* Route lookup */
+	saddr = OVS_CB(skb)->tun_key->ipv4_src;
+	rt = find_route(ovs_dp_get_net(vport->dp),
+			&saddr,
+			OVS_CB(skb)->tun_key->ipv4_dst,
+			IPPROTO_UDP,
+			OVS_CB(skb)->tun_key->ipv4_tos,
+			skb->mark);
+	if (IS_ERR(rt)) {
+		err = PTR_ERR(rt);
+		goto error;
+	}
 
-static void lisp_tnl_destroy(struct vport *vport)
-{
-	struct lisp_port *lisp_port;
-	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
-	struct tnl_mutable_config *config;
+	min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
+			+ sizeof(struct iphdr) + LISP_HLEN;
 
-	config = rtnl_dereference(tnl_vport->mutable);
+	if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
+		int head_delta = SKB_DATA_ALIGN(min_headroom -
+						skb_headroom(skb) +
+						16);
 
-	lisp_port = lisp_port_exists(ovs_dp_get_net(vport->dp),
-				     config->dst_port);
+		err = pskb_expand_head(skb, max_t(int, head_delta, 0),
+					0, GFP_ATOMIC);
+		if (unlikely(err))
+			goto err_free_rt;
+	}
 
-	lisp_tunnel_release(lisp_port);
+	/* Reset l2 headers. */
+	skb_pull(skb, network_offset);
+	skb_reset_mac_header(skb);
+	vlan_set_tci(skb, 0);
 
-	ovs_tnl_destroy(vport);
-}
+	skb_reset_inner_headers(skb);
 
-static struct vport *lisp_tnl_create(const struct vport_parms *parms)
-{
-	int err;
-	struct vport *vport;
-	struct lisp_port *lisp_port = NULL;
+	__skb_push(skb, LISP_HLEN);
+	skb_reset_transport_header(skb);
+
+	lisp_build_header(vport, skb);
 
-	err = lisp_tunnel_setup(ovs_dp_get_net(parms->dp), parms->options,
-				&lisp_port);
+	/* Offloading */
+	err = handle_offloads(skb);
 	if (err)
-		return ERR_PTR(err);
+		goto err_free_rt;
 
-	vport = ovs_tnl_create(parms, &ovs_lisp_vport_ops, &ovs_lisp_tnl_ops);
+	skb->local_df = 1;
 
-	if (IS_ERR(vport))
-		lisp_tunnel_release(lisp_port);
+	df = OVS_CB(skb)->tun_key->tun_flags &
+				  TUNNEL_DONT_FRAGMENT ?  htons(IP_DF) : 0;
+	sent_len = iptunnel_xmit(rt, skb,
+			     saddr, OVS_CB(skb)->tun_key->ipv4_dst,
+			     IPPROTO_UDP, OVS_CB(skb)->tun_key->ipv4_tos,
+			     OVS_CB(skb)->tun_key->ipv4_ttl, df, false);
 
-	return vport;
+	return sent_len > 0 ? sent_len + network_offset : sent_len;
+
+err_free_rt:
+	ip_rt_put(rt);
+error:
+	return err;
+}
+
+static const char *lisp_get_name(const struct vport *vport)
+{
+	struct lisp_port *lisp_port = lisp_vport(vport);
+	return lisp_port->name;
 }
 
 const struct vport_ops ovs_lisp_vport_ops = {
 	.type		= OVS_VPORT_TYPE_LISP,
-	.flags		= VPORT_F_TUN_ID,
 	.create		= lisp_tnl_create,
 	.destroy	= lisp_tnl_destroy,
-	.get_name	= ovs_tnl_get_name,
-	.get_options	= ovs_tnl_get_options,
-	.set_options	= lisp_tnl_set_options,
-	.send		= lisp_tnl_send,
+	.get_name	= lisp_get_name,
+	.get_options	= lisp_get_options,
+	.send		= lisp_send,
 };
-#else
-#warning LISP tunneling will not be available on kernels before 2.6.26
-#endif /* Linux kernel < 2.6.26 */