X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Fvport-lisp.c;h=259cc2b291bd1464bedaf8f0738a6425c946eaee;hb=3442636d01d2a73a557952ad9140de07418c28c2;hp=ad37c54b351c0431c4ffcd137c27bb29a81326fd;hpb=85c9de194b9f432f7b8a66bda980cbab52a55b72;p=sliver-openvswitch.git

diff --git a/datapath/vport-lisp.c b/datapath/vport-lisp.c
index ad37c54b3..259cc2b29 100644
--- a/datapath/vport-lisp.c
+++ b/datapath/vport-lisp.c
@@ -20,7 +20,6 @@
 #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>
@@ -30,13 +29,13 @@
 
 #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 "vport.h"
 
-
 /*
  *  LISP encapsulation header:
  *
@@ -94,34 +93,33 @@ struct lisphdr {
 
 #define LISP_HLEN (sizeof(struct udphdr) + sizeof(struct lisphdr))
 
-static inline int lisp_hdr_len(const struct ovs_key_ipv4_tunnel *tun_key)
-{
-	return LISP_HLEN;
-}
-
 /**
  * struct lisp_port - Keeps track of open UDP ports
- * @list: list element.
- * @vport: vport for the tunnel.
- * @socket: The socket created for this port number.
+ * @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;
-	struct vport *vport;
 	struct socket *lisp_rcv_socket;
-	struct rcu_head rcu;
+	char name[IFNAMSIZ];
 };
 
 static LIST_HEAD(lisp_ports);
 
+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_rcu(lisp_port, &lisp_ports, list) {
-		struct tnl_vport *tnl_vport = tnl_vport_priv(lisp_port->vport);
-
-		if (tnl_vport->dst_port == port &&
+		if (lisp_port->dst_port == port &&
 			net_eq(sock_net(lisp_port->lisp_rcv_socket->sk), net))
 			return lisp_port;
 	}
@@ -134,25 +132,6 @@ 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)
 {
@@ -180,16 +159,33 @@ static __be64 instance_id_to_tunnel_id(__u8 *iid)
 #endif
 }
 
+/* Compute source UDP port for outgoing packet.
+ * Currently we use the flow hash.
+ */
+static u16 ovs_tnl_get_src_port(struct sk_buff *skb)
+{
+	int low;
+	int high;
+	unsigned int range;
+	struct sw_flow_key *pkt_key = OVS_CB(skb)->pkt_key;
+	u32 hash = jhash2((const u32 *)pkt_key,
+			  sizeof(*pkt_key) / sizeof(u32), 0);
+
+	inet_get_local_port_range(&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,
 			      int tunnel_hlen)
 {
-	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+	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;
 
-	udph->dest = tnl_vport->dst_port;
+	udph->dest = lisp_port->dst_port;
 	udph->source = htons(ovs_tnl_get_src_port(skb));
 	udph->check = 0;
 	udph->len = htons(skb->len - skb_transport_offset(skb));
@@ -207,15 +203,42 @@ static void lisp_build_header(const struct vport *vport,
 
 	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), skb_dst(skb), 0);
+/**
+ *	ovs_tnl_rcv - ingress point for generic tunnel code
+ *
+ * @vport: port this packet was received on
+ * @skb: received packet
+ * @tun_key: tunnel that carried packet
+ *
+ * Must be called with rcu_read_lock.
+ *
+ * Packets received by this function are in the following state:
+ * - skb->data points to the inner Ethernet header.
+ * - The inner Ethernet header is in the linear data area.
+ * - The layer pointers are undefined.
+ */
+static void ovs_tnl_rcv(struct vport *vport, struct sk_buff *skb,
+			struct ovs_key_ipv4_tunnel *tun_key)
+{
+	struct ethhdr *eh;
+
+	skb_reset_mac_header(skb);
+	eh = eth_hdr(skb);
+
+	if (likely(ntohs(eh->h_proto) >= ETH_P_802_3_MIN))
+		skb->protocol = eh->h_proto;
+	else
+		skb->protocol = htons(ETH_P_802_2);
+
+	skb_dst_drop(skb);
+	nf_reset(skb);
+	skb_clear_rxhash(skb);
+	secpath_reset(skb);
+	vlan_set_tci(skb, 0);
+
+	ovs_vport_receive(vport, skb, tun_key);
 }
 
 /* Called with rcu_read_lock and BH disabled. */
@@ -247,8 +270,7 @@ static int lisp_rcv(struct sock *sk, struct sk_buff *skb)
 
 	/* Save outer tunnel values */
 	iph = ip_hdr(skb);
-	tnl_tun_key_init(&tun_key, iph, key, OVS_TNL_F_KEY);
-	OVS_CB(skb)->tun_key = &tun_key;
+	ovs_flow_tun_key_init(&tun_key, iph, key, TUNNEL_KEY);
 
 	/* Drop non-IP inner packets */
 	inner_iph = (struct iphdr *)(lisph + 1);
@@ -270,7 +292,9 @@ static int lisp_rcv(struct sock *sk, struct sk_buff *skb)
 	ethh->h_source[0] = 0x02;
 	ethh->h_proto = protocol;
 
-	ovs_tnl_rcv(lisp_port->vport, skb);
+	ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
+
+	ovs_tnl_rcv(vport_from_priv(lisp_port), skb, &tun_key);
 	goto out;
 
 error:
@@ -283,9 +307,8 @@ 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;
-	struct tnl_vport *tnl_vport = tnl_vport_priv(lisp_port->vport);
+	int err;
 
 	err = sock_create_kern(AF_INET, SOCK_DGRAM, 0,
 			       &lisp_port->lisp_rcv_socket);
@@ -297,7 +320,7 @@ 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 = tnl_vport->dst_port;
+	sin.sin_port = lisp_port->dst_port;
 
 	err = kernel_bind(lisp_port->lisp_rcv_socket, (struct sockaddr *)&sin,
 			  sizeof(struct sockaddr_in));
@@ -318,37 +341,39 @@ error:
 	return err;
 }
 
-
-static void free_port_rcu(struct rcu_head *rcu)
+static int lisp_get_options(const struct vport *vport, struct sk_buff *skb)
 {
-	struct lisp_port *lisp_port = container_of(rcu,
-			struct lisp_port, rcu);
+	struct lisp_port *lisp_port = lisp_vport(vport);
 
-	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_tunnel_release(struct lisp_port *lisp_port)
+static void lisp_tnl_destroy(struct vport *vport)
 {
-	if (!lisp_port)
-		return;
+	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);
-	call_rcu(&lisp_port->rcu, free_port_rcu);
+
+	ovs_vport_deferred_free(vport);
 }
 
-static int lisp_tunnel_setup(struct net *net, struct vport *vport,
-			     struct nlattr *options)
+static struct vport *lisp_tnl_create(const struct vport_parms *parms)
 {
-	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+	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;
 
 	if (!options) {
 		err = -EINVAL;
-		goto out;
+		goto error;
 	}
 
 	a = nla_find_nested(options, OVS_TUNNEL_ATTR_DST_PORT);
@@ -357,95 +382,263 @@ static int lisp_tunnel_setup(struct net *net, struct vport *vport,
 	} 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_find_port(net, htons(dst_port));
-	if (lisp_port) {
+	if (lisp_find_port(net, htons(dst_port))) {
 		err = -EEXIST;
-		goto out;
+		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;
 
-	tnl_vport->dst_port = htons(dst_port);
-	lisp_port->vport = vport;
-	list_add_tail_rcu(&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;
 
-	return 0;
+	list_add_tail_rcu(&lisp_port->list, &lisp_ports);
+	return vport;
 
+error_free:
+	ovs_vport_free(vport);
 error:
-	list_del_rcu(&lisp_port->list);
-	kfree(lisp_port);
-out:
-	return err;
+	return ERR_PTR(err);
 }
 
-static int lisp_get_options(const struct vport *vport, struct sk_buff *skb)
+static bool need_linearize(const struct sk_buff *skb)
 {
-	const struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+	int i;
 
-	if (nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, ntohs(tnl_vport->dst_port)))
-		return -EMSGSIZE;
-	return 0;
-}
+	if (unlikely(skb_shinfo(skb)->frag_list))
+		return true;
 
-static const struct tnl_ops ovs_lisp_tnl_ops = {
-	.ipproto	= IPPROTO_UDP,
-	.hdr_len	= lisp_hdr_len,
-	.build_header	= lisp_build_header,
-};
+	/*
+	 * Generally speaking we should linearize if there are paged frags.
+	 * However, if all of the refcounts are 1 we know nobody else can
+	 * change them from underneath us and we can skip the linearization.
+	 */
+	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++)
+		if (unlikely(page_count(skb_frag_page(&skb_shinfo(skb)->frags[i])) > 1))
+			return true;
 
-static void lisp_tnl_destroy(struct vport *vport)
+	return false;
+}
+
+static struct sk_buff *handle_offloads(struct sk_buff *skb)
 {
-	struct lisp_port *lisp_port;
-	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+	int err;
+
+	if (skb_is_gso(skb)) {
+		struct sk_buff *nskb;
+		char cb[sizeof(skb->cb)];
+
+		memcpy(cb, skb->cb, sizeof(cb));
+
+		nskb = __skb_gso_segment(skb, 0, false);
+		if (IS_ERR(nskb)) {
+			err = PTR_ERR(nskb);
+			goto error;
+		}
+
+		consume_skb(skb);
+		skb = nskb;
+		while (nskb) {
+			memcpy(nskb->cb, cb, sizeof(cb));
+			nskb = nskb->next;
+		}
+	} else if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		/* Pages aren't locked and could change at any time.
+		 * If this happens after we compute the checksum, the
+		 * checksum will be wrong.  We linearize now to avoid
+		 * this problem.
+		 */
+		if (unlikely(need_linearize(skb))) {
+			err = __skb_linearize(skb);
+			if (unlikely(err))
+				goto error;
+		}
+
+		err = skb_checksum_help(skb);
+		if (unlikely(err))
+			goto error;
+	}
 
-	lisp_port = lisp_find_port(ovs_dp_get_net(vport->dp),
-				   tnl_vport->dst_port);
+	skb->ip_summed = CHECKSUM_NONE;
+	return skb;
 
-	lisp_tunnel_release(lisp_port);
-	ovs_tnl_destroy(vport);
+error:
+	return ERR_PTR(err);
 }
 
-static struct vport *lisp_tnl_create(const struct vport_parms *parms)
+static int ovs_tnl_send(struct vport *vport, struct sk_buff *skb,
+			u8 ipproto, int tunnel_hlen,
+			void (*build_header)(const struct vport *,
+					     struct sk_buff *,
+					     int tunnel_hlen))
 {
-	struct vport *vport;
+	int min_headroom;
+	struct rtable *rt;
+	__be32 saddr;
+	int sent_len = 0;
 	int err;
+	struct sk_buff *nskb;
+
+	/* 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,
+			OVS_CB(skb)->tun_key->ipv4_tos,
+			skb->mark);
+	if (IS_ERR(rt)) {
+		err = PTR_ERR(rt);
+		goto error;
+	}
 
-	vport = ovs_tnl_create(parms, &ovs_lisp_vport_ops, &ovs_lisp_tnl_ops);
-	if (IS_ERR(vport))
-		return vport;
+	tunnel_hlen += sizeof(struct iphdr);
+
+	min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
+			+ tunnel_hlen
+			+ (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
 
-	err = lisp_tunnel_setup(ovs_dp_get_net(parms->dp), vport,
-				parms->options);
-	if (err) {
-		ovs_tnl_destroy(vport);
-		return ERR_PTR(err);
+	if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
+		int head_delta = SKB_DATA_ALIGN(min_headroom -
+						skb_headroom(skb) +
+						16);
+
+		err = pskb_expand_head(skb, max_t(int, head_delta, 0),
+					0, GFP_ATOMIC);
+		if (unlikely(err))
+			goto err_free_rt;
 	}
 
-	return vport;
+	/* Offloading */
+	nskb = handle_offloads(skb);
+	if (IS_ERR(nskb)) {
+		err = PTR_ERR(nskb);
+		goto err_free_rt;
+	}
+	skb = nskb;
+
+	/* Reset SKB */
+	nf_reset(skb);
+	secpath_reset(skb);
+	skb_dst_drop(skb);
+	skb_clear_rxhash(skb);
+
+	while (skb) {
+		struct sk_buff *next_skb = skb->next;
+		struct iphdr *iph;
+		int frag_len;
+
+		skb->next = NULL;
+
+		if (vlan_tx_tag_present(skb)) {
+			if (unlikely(!__vlan_put_tag(skb,
+							skb->vlan_proto,
+							vlan_tx_tag_get(skb))))
+				goto next;
+
+			vlan_set_tci(skb, 0);
+		}
+
+		frag_len = skb->len;
+		skb_push(skb, tunnel_hlen);
+		skb_reset_network_header(skb);
+		skb_set_transport_header(skb, sizeof(struct iphdr));
+
+		if (next_skb)
+			skb_dst_set(skb, dst_clone(&rt_dst(rt)));
+		else
+			skb_dst_set(skb, &rt_dst(rt));
+
+		/* Push Tunnel header. */
+		build_header(vport, skb, tunnel_hlen);
+
+		/* Push IP header. */
+		iph = ip_hdr(skb);
+		iph->version	= 4;
+		iph->ihl	= sizeof(struct iphdr) >> 2;
+		iph->protocol	= ipproto;
+		iph->daddr	= OVS_CB(skb)->tun_key->ipv4_dst;
+		iph->saddr	= saddr;
+		iph->tos	= OVS_CB(skb)->tun_key->ipv4_tos;
+		iph->ttl	= OVS_CB(skb)->tun_key->ipv4_ttl;
+		iph->frag_off	= OVS_CB(skb)->tun_key->tun_flags &
+				  TUNNEL_DONT_FRAGMENT ?  htons(IP_DF) : 0;
+		/*
+		 * 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 with __ip_select_ident(),
+		 * as ip_select_ident() assumes a proper ID is not needed when
+		 * when the DF bit is set.
+		 */
+		skb->local_df = 1;
+		__ip_select_ident(iph, skb_dst(skb), 0);
+
+		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
+
+		err = ip_local_out(skb);
+		if (unlikely(net_xmit_eval(err)))
+			goto next;
+
+		sent_len += frag_len;
+
+next:
+		skb = next_skb;
+	}
+
+	return sent_len;
+
+err_free_rt:
+	ip_rt_put(rt);
+error:
+	return err;
+}
+
+static int lisp_tnl_send(struct vport *vport, struct sk_buff *skb)
+{
+	int tnl_len;
+	int network_offset = skb_network_offset(skb);
+
+	if (unlikely(!OVS_CB(skb)->tun_key))
+		return -EINVAL;
+
+	/* 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, IPPROTO_UDP,
+				LISP_HLEN, lisp_build_header);
+		return tnl_len > 0 ? tnl_len + network_offset : tnl_len;
+	default:
+		kfree_skb(skb);
+		return 0;
+	}
+}
+
+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_name	= lisp_get_name,
 	.get_options	= lisp_get_options,
 	.send		= lisp_tnl_send,
 };
-#else
-#warning LISP tunneling will not be available on kernels before 2.6.26
-#endif /* Linux kernel < 2.6.26 */