X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Ftunnel.c;h=1ef81ab7893f08172d16df71861ad7bda1a7cb1f;hb=de06dce673979eb9208c539bad15fae855e1d4c5;hp=6d0e7b9c603b35e0e63753ee31a223cf7de3a049;hpb=f915f1a8ca180828983ef22cf2fd21b8f010b972;p=sliver-openvswitch.git

diff --git a/datapath/tunnel.c b/datapath/tunnel.c
index 6d0e7b9c6..1ef81ab78 100644
--- a/datapath/tunnel.c
+++ b/datapath/tunnel.c
@@ -33,6 +33,7 @@
 #include "datapath.h"
 #include "table.h"
 #include "tunnel.h"
+#include "vlan.h"
 #include "vport.h"
 #include "vport-generic.h"
 #include "vport-internal_dev.h"
@@ -185,7 +186,7 @@ struct port_lookup_key {
  * Modifies 'target' to store the rcu_dereferenced pointer that was used to do
  * the comparision.
  */
-static int port_cmp(const struct tbl_node *node, void *target)
+static int port_cmp(const struct tbl_node *node, void *target, int unused)
 {
 	const struct tnl_vport *tnl_vport = tnl_vport_table_cast(node);
 	struct port_lookup_key *lookup = target;
@@ -248,11 +249,13 @@ static int add_port(struct vport *vport)
 		struct tbl *new_table;
 
 		new_table = tbl_expand(cur_table);
-		if (IS_ERR(new_table))
-			return PTR_ERR(new_table);
-
-		rcu_assign_pointer(port_table, new_table);
-		tbl_deferred_destroy(cur_table, NULL);
+		if (IS_ERR(new_table)) {
+			if (PTR_ERR(new_table) != -ENOSPC)
+				return PTR_ERR(new_table);
+		} else {
+			rcu_assign_pointer(port_table, new_table);
+			tbl_deferred_destroy(cur_table, NULL);
+		}
 	}
 
 	err = tbl_insert(rtnl_dereference(port_table), &tnl_vport->tbl_node,
@@ -336,7 +339,8 @@ struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
 		lookup.tunnel_type = tunnel_type & ~TNL_T_KEY_MATCH;
 
 		if (key_local_remote_ports) {
-			tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp);
+			tbl_node = tbl_lookup(table, &lookup, sizeof(lookup),
+					      port_hash(&lookup), port_cmp);
 			if (tbl_node)
 				goto found;
 		}
@@ -344,7 +348,8 @@ struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
 		if (key_remote_ports) {
 			lookup.saddr = 0;
 
-			tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp);
+			tbl_node = tbl_lookup(table, &lookup, sizeof(lookup),
+					      port_hash(&lookup), port_cmp);
 			if (tbl_node)
 				goto found;
 
@@ -357,7 +362,8 @@ struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
 		lookup.tunnel_type = tunnel_type & ~TNL_T_KEY_EXACT;
 
 		if (local_remote_ports) {
-			tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp);
+			tbl_node = tbl_lookup(table, &lookup, sizeof(lookup),
+					      port_hash(&lookup), port_cmp);
 			if (tbl_node)
 				goto found;
 		}
@@ -365,7 +371,8 @@ struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
 		if (remote_ports) {
 			lookup.saddr = 0;
 
-			tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp);
+			tbl_node = tbl_lookup(table, &lookup, sizeof(lookup),
+					      port_hash(&lookup), port_cmp);
 			if (tbl_node)
 				goto found;
 		}
@@ -378,18 +385,14 @@ found:
 	return tnl_vport_to_vport(tnl_vport_table_cast(tbl_node));
 }
 
-static inline void ecn_decapsulate(struct sk_buff *skb)
+static void ecn_decapsulate(struct sk_buff *skb, u8 tos)
 {
-	/* This is accessing the outer IP header of the tunnel, which we've
-	 * already validated to be OK.  skb->data is currently set to the start
-	 * of the inner Ethernet header, and we've validated ETH_HLEN.
-	 */
-	if (unlikely(INET_ECN_is_ce(ip_hdr(skb)->tos))) {
+	if (unlikely(INET_ECN_is_ce(tos))) {
 		__be16 protocol = skb->protocol;
 
 		skb_set_network_header(skb, ETH_HLEN);
 
-		if (skb->protocol == htons(ETH_P_8021Q)) {
+		if (protocol == htons(ETH_P_8021Q)) {
 			if (unlikely(!pskb_may_pull(skb, VLAN_ETH_HLEN)))
 				return;
 
@@ -416,17 +419,27 @@ static inline void ecn_decapsulate(struct sk_buff *skb)
 	}
 }
 
-/* Called with rcu_read_lock. */
-void tnl_rcv(struct vport *vport, struct sk_buff *skb)
+/**
+ *	tnl_rcv - ingress point for generic tunnel code
+ *
+ * @vport: port this packet was received on
+ * @skb: received packet
+ * @tos: ToS from encapsulating IP packet, used to copy ECN bits
+ *
+ * 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.
+ * - skb->csum does not include the inner Ethernet header.
+ * - The layer pointers are undefined.
+ */
+void tnl_rcv(struct vport *vport, struct sk_buff *skb, u8 tos)
 {
-	/* 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.
-	 * - skb->csum does not include the inner Ethernet header.
-	 * - The layer pointers point at the outer headers.
-	 */
+	struct ethhdr *eh;
 
-	struct ethhdr *eh = (struct ethhdr *)skb->data;
+	skb_reset_mac_header(skb);
+	eh = eth_hdr(skb);
 
 	if (likely(ntohs(eh->h_proto) >= 1536))
 		skb->protocol = eh->h_proto;
@@ -435,10 +448,16 @@ void tnl_rcv(struct vport *vport, struct sk_buff *skb)
 
 	skb_dst_drop(skb);
 	nf_reset(skb);
+	skb_clear_rxhash(skb);
 	secpath_reset(skb);
 
-	ecn_decapsulate(skb);
-	compute_ip_summed(skb, false);
+	ecn_decapsulate(skb, tos);
+	vlan_set_tci(skb, 0);
+
+	if (unlikely(compute_ip_summed(skb, false))) {
+		kfree_skb(skb);
+		return;
+	}
 
 	vport_receive(vport, skb);
 }
@@ -682,7 +701,8 @@ bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutab
 
 		vh->h_vlan_TCI = vlan_eth_hdr(skb)->h_vlan_TCI;
 		vh->h_vlan_encapsulated_proto = skb->protocol;
-	}
+	} else
+		vlan_set_tci(nskb, vlan_get_tci(skb));
 	skb_reset_mac_header(nskb);
 
 	/* Protocol */
@@ -704,7 +724,11 @@ bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutab
 	    (TNL_F_IN_KEY_MATCH | TNL_F_OUT_KEY_ACTION))
 		OVS_CB(nskb)->tun_id = flow_key;
 
-	compute_ip_summed(nskb, false);
+	if (unlikely(compute_ip_summed(nskb, false))) {
+		kfree_skb(nskb);
+		return false;
+	}
+
 	vport_receive(vport, nskb);
 
 	return true;
@@ -715,29 +739,43 @@ static bool check_mtu(struct sk_buff *skb,
 		      const struct tnl_mutable_config *mutable,
 		      const struct rtable *rt, __be16 *frag_offp)
 {
+	bool df_inherit = mutable->flags & TNL_F_DF_INHERIT;
 	bool pmtud = mutable->flags & TNL_F_PMTUD;
-	__be16 frag_off = 0;
-	int mtu;
+	__be16 frag_off = mutable->flags & TNL_F_DF_DEFAULT ? htons(IP_DF) : 0;
+	int mtu = 0;
+	unsigned int packet_length = skb->len - ETH_HLEN;
+
+	/* Allow for one level of tagging in the packet length. */
+	if (!vlan_tx_tag_present(skb) &&
+	    eth_hdr(skb)->h_proto == htons(ETH_P_8021Q))
+		packet_length -= VLAN_HLEN;
 
 	if (pmtud) {
-		frag_off = htons(IP_DF);
+		int vlan_header = 0;
+
+		/* The tag needs to go in packet regardless of where it
+		 * currently is, so subtract it from the MTU.
+		 */
+		if (vlan_tx_tag_present(skb) ||
+		    eth_hdr(skb)->h_proto == htons(ETH_P_8021Q))
+			vlan_header = VLAN_HLEN;
 
 		mtu = dst_mtu(&rt_dst(rt))
 			- ETH_HLEN
 			- mutable->tunnel_hlen
-			- (eth_hdr(skb)->h_proto == htons(ETH_P_8021Q) ?
-				VLAN_HLEN : 0);
+			- vlan_header;
 	}
 
 	if (skb->protocol == htons(ETH_P_IP)) {
 		struct iphdr *iph = ip_hdr(skb);
 
-		frag_off |= iph->frag_off & htons(IP_DF);
+		if (df_inherit)
+			frag_off = iph->frag_off & htons(IP_DF);
 
 		if (pmtud && iph->frag_off & htons(IP_DF)) {
 			mtu = max(mtu, IP_MIN_MTU);
 
-			if (ntohs(iph->tot_len) > mtu &&
+			if (packet_length > mtu &&
 			    tnl_frag_needed(vport, mutable, skb, mtu,
 					    OVS_CB(skb)->tun_id))
 				return false;
@@ -745,12 +783,10 @@ static bool check_mtu(struct sk_buff *skb,
 	}
 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 	else if (skb->protocol == htons(ETH_P_IPV6)) {
-		unsigned int packet_length = skb->len - ETH_HLEN
-			- (eth_hdr(skb)->h_proto == htons(ETH_P_8021Q) ?
-				VLAN_HLEN : 0);
-
-		/* IPv6 requires PMTUD if the packet is above the minimum MTU. */
-		if (packet_length > IPV6_MIN_MTU)
+		/* IPv6 requires end hosts to do fragmentation
+		 * if the packet is above the minimum MTU.
+		 */
+		if (df_inherit && packet_length > IPV6_MIN_MTU)
 			frag_off = htons(IP_DF);
 
 		if (pmtud) {
@@ -784,7 +820,7 @@ static void create_tunnel_header(const struct vport *vport,
 	iph->saddr	= rt->rt_src;
 	iph->ttl	= mutable->ttl;
 	if (!iph->ttl)
-		iph->ttl = dst_metric(&rt_dst(rt), RTAX_HOPLIMIT);
+		iph->ttl = ip4_dst_hoplimit(&rt_dst(rt));
 
 	tnl_vport->tnl_ops->build_header(vport, mutable, iph + 1);
 }
@@ -918,6 +954,7 @@ static struct tnl_cache *build_cache(struct vport *vport,
 		struct sk_buff *skb;
 		bool is_frag;
 		int err;
+		int flow_key_len;
 
 		dst_vport = internal_dev_get_vport(rt_dst(rt).dev);
 		if (!dst_vport)
@@ -930,14 +967,16 @@ static struct tnl_cache *build_cache(struct vport *vport,
 		__skb_put(skb, cache->len);
 		memcpy(skb->data, get_cached_header(cache), cache->len);
 
-		err = flow_extract(skb, dst_vport->port_no, &flow_key, &is_frag);
+		err = flow_extract(skb, dst_vport->port_no, &flow_key,
+				   &flow_key_len, &is_frag);
 
-		kfree_skb(skb);
+		consume_skb(skb);
 		if (err || is_frag)
 			goto done;
 
 		flow_node = tbl_lookup(rcu_dereference(dst_vport->dp->table),
-				       &flow_key, flow_hash(&flow_key),
+				       &flow_key, flow_key_len,
+				       flow_hash(&flow_key, flow_key_len),
 				       flow_cmp);
 		if (flow_node) {
 			struct sw_flow *flow = flow_cast(flow_node);
@@ -971,6 +1010,7 @@ static struct rtable *find_route(struct vport *vport,
 		return cur_cache->rt;
 	} else {
 		struct rtable *rt;
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39)
 		struct flowi fl = { .nl_u = { .ip4_u =
 					      { .daddr = mutable->daddr,
 						.saddr = mutable->saddr,
@@ -979,6 +1019,16 @@ static struct rtable *find_route(struct vport *vport,
 
 		if (unlikely(ip_route_output_key(&init_net, &rt, &fl)))
 			return NULL;
+#else
+		struct flowi4 fl = { .daddr = mutable->daddr,
+				     .saddr = mutable->saddr,
+				     .flowi4_tos = tos,
+				     .flowi4_proto = tnl_vport->tnl_ops->ipproto };
+
+		rt = ip_route_output_key(&init_net, &fl);
+		if (IS_ERR(rt))
+			return NULL;
+#endif
 
 		if (likely(tos == mutable->tos))
 			*cache = build_cache(vport, mutable, rt);
@@ -987,27 +1037,6 @@ static struct rtable *find_route(struct vport *vport,
 	}
 }
 
-static struct sk_buff *check_headroom(struct sk_buff *skb, int headroom)
-{
-	if (skb_headroom(skb) < headroom || skb_header_cloned(skb)) {
-		struct sk_buff *nskb = skb_realloc_headroom(skb, headroom + 16);
-		if (unlikely(!nskb)) {
-			kfree_skb(skb);
-			return ERR_PTR(-ENOMEM);
-		}
-
-		set_skb_csum_bits(skb, nskb);
-
-		if (skb->sk)
-			skb_set_owner_w(nskb, skb->sk);
-
-		kfree_skb(skb);
-		return nskb;
-	}
-
-	return skb;
-}
-
 static inline bool need_linearize(const struct sk_buff *skb)
 {
 	int i;
@@ -1021,7 +1050,7 @@ static inline bool need_linearize(const struct sk_buff *skb)
 	 * 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_shinfo(skb)->frags[0].page) > 1))
+		if (unlikely(page_count(skb_shinfo(skb)->frags[i].page) > 1))
 			return true;
 
 	return false;
@@ -1034,70 +1063,53 @@ static struct sk_buff *handle_offloads(struct sk_buff *skb,
 	int min_headroom;
 	int err;
 
-	forward_ip_summed(skb);
-
-	err = vswitch_skb_checksum_setup(skb);
-	if (unlikely(err))
-		goto error_free;
-
 	min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
-			+ mutable->tunnel_hlen;
+			+ mutable->tunnel_hlen
+			+ (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
+
+	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 error_free;
+	}
+
+	forward_ip_summed(skb, true);
 
 	if (skb_is_gso(skb)) {
 		struct sk_buff *nskb;
 
-		/*
-		 * If we are doing GSO on a pskb it is better to make sure that
-		 * the headroom is correct now.  We will only have to copy the
-		 * portion in the linear data area and GSO will preserve
-		 * headroom when it creates the segments.  This is particularly
-		 * beneficial on Xen where we get a lot of GSO pskbs.
-		 * Conversely, we avoid copying if it is just to get our own
-		 * writable clone because GSO will do the copy for us.
-		 */
-		if (skb_headroom(skb) < min_headroom) {
-			skb = check_headroom(skb, min_headroom);
-			if (IS_ERR(skb)) {
-				err = PTR_ERR(skb);
-				goto error;
-			}
-		}
-
 		nskb = skb_gso_segment(skb, 0);
-		kfree_skb(skb);
 		if (IS_ERR(nskb)) {
+			kfree_skb(skb);
 			err = PTR_ERR(nskb);
 			goto error;
 		}
 
+		consume_skb(skb);
 		skb = nskb;
-	} else {
-		skb = check_headroom(skb, min_headroom);
-		if (IS_ERR(skb)) {
-			err = PTR_ERR(skb);
-			goto error;
-		}
-
-		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_free;
-			}
-
-			err = skb_checksum_help(skb);
+	} else if (get_ip_summed(skb) == OVS_CSUM_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_free;
-		} else if (skb->ip_summed == CHECKSUM_COMPLETE)
-			skb->ip_summed = CHECKSUM_NONE;
+		}
+
+		err = skb_checksum_help(skb);
+		if (unlikely(err))
+			goto error_free;
 	}
 
+	set_ip_summed(skb, OVS_CSUM_NONE);
+
 	return skb;
 
 error_free:
@@ -1110,25 +1122,21 @@ static int send_frags(struct sk_buff *skb,
 		      const struct tnl_mutable_config *mutable)
 {
 	int sent_len;
-	int err;
 
 	sent_len = 0;
 	while (skb) {
 		struct sk_buff *next = skb->next;
 		int frag_len = skb->len - mutable->tunnel_hlen;
+		int err;
 
 		skb->next = NULL;
 		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
 
 		err = ip_local_out(skb);
-		if (likely(net_xmit_eval(err) == 0))
-			sent_len += frag_len;
-		else {
-			skb = next;
-			goto free_frags;
-		}
-
 		skb = next;
+		if (unlikely(net_xmit_eval(err)))
+			goto free_frags;
+		sent_len += frag_len;
 	}
 
 	return sent_len;
@@ -1159,7 +1167,8 @@ int tnl_send(struct vport *vport, struct sk_buff *skb)
 	u8 tos;
 
 	/* Validate the protocol headers before we try to use them. */
-	if (skb->protocol == htons(ETH_P_8021Q)) {
+	if (skb->protocol == htons(ETH_P_8021Q) &&
+	    !vlan_tx_tag_present(skb)) {
 		if (unlikely(!pskb_may_pull(skb, VLAN_ETH_HLEN)))
 			goto error_free;
 
@@ -1208,6 +1217,7 @@ int tnl_send(struct vport *vport, struct sk_buff *skb)
 	nf_reset(skb);
 	secpath_reset(skb);
 	skb_dst_drop(skb);
+	skb_clear_rxhash(skb);
 
 	/* Offloading */
 	skb = handle_offloads(skb, mutable, rt);
@@ -1234,7 +1244,7 @@ int tnl_send(struct vport *vport, struct sk_buff *skb)
 	/* TTL */
 	ttl = mutable->ttl;
 	if (!ttl)
-		ttl = dst_metric(&rt_dst(rt), RTAX_HOPLIMIT);
+		ttl = ip4_dst_hoplimit(&rt_dst(rt));
 
 	if (mutable->flags & TNL_F_TTL_INHERIT) {
 		if (skb->protocol == htons(ETH_P_IP))
@@ -1250,6 +1260,9 @@ int tnl_send(struct vport *vport, struct sk_buff *skb)
 		struct sk_buff *next_skb = skb->next;
 		skb->next = NULL;
 
+		if (unlikely(vlan_deaccel_tag(skb)))
+			goto next;
+
 		if (likely(cache)) {
 			skb_push(skb, cache->len);
 			memcpy(skb->data, get_cached_header(cache), cache->len);
@@ -1290,8 +1303,12 @@ int tnl_send(struct vport *vport, struct sk_buff *skb)
 			ip_send_check(iph);
 
 			if (cache_vport) {
+				if (unlikely(compute_ip_summed(skb, true))) {
+					kfree_skb(skb);
+					goto next;
+				}
+
 				OVS_CB(skb)->flow = cache->flow;
-				compute_ip_summed(skb, true);
 				vport_receive(cache_vport, skb);
 				sent_len += orig_len;
 			} else {
@@ -1318,9 +1335,9 @@ next:
 error_free:
 	tnl_free_linked_skbs(skb);
 error:
-	dst_release(unattached_dst);
 	vport_record_error(vport, err);
 out:
+	dst_release(unattached_dst);
 	return sent_len;
 }
 
@@ -1369,12 +1386,6 @@ static int tnl_set_config(struct nlattr *options, const struct tnl_ops *tnl_ops,
 	if (a[ODP_TUNNEL_ATTR_TTL])
 		mutable->ttl = nla_get_u8(a[ODP_TUNNEL_ATTR_TTL]);
 
-	mutable->tunnel_hlen = tnl_ops->hdr_len(mutable);
-	if (mutable->tunnel_hlen < 0)
-		return mutable->tunnel_hlen;
-
-	mutable->tunnel_hlen += sizeof(struct iphdr);
-
 	mutable->tunnel_type = tnl_ops->tunnel_type;
 	if (!a[ODP_TUNNEL_ATTR_IN_KEY]) {
 		mutable->tunnel_type |= TNL_T_KEY_MATCH;
@@ -1389,6 +1400,12 @@ static int tnl_set_config(struct nlattr *options, const struct tnl_ops *tnl_ops,
 	else
 		mutable->out_key = nla_get_be64(a[ODP_TUNNEL_ATTR_OUT_KEY]);
 
+	mutable->tunnel_hlen = tnl_ops->hdr_len(mutable);
+	if (mutable->tunnel_hlen < 0)
+		return mutable->tunnel_hlen;
+
+	mutable->tunnel_hlen += sizeof(struct iphdr);
+
 	old_vport = tnl_find_port(mutable->saddr, mutable->daddr,
 				  mutable->in_key, mutable->tunnel_type,
 				  &old_mutable);
@@ -1575,9 +1592,6 @@ const unsigned char *tnl_get_addr(const struct vport *vport)
 
 void tnl_free_linked_skbs(struct sk_buff *skb)
 {
-	if (unlikely(!skb))
-		return;
-
 	while (skb) {
 		struct sk_buff *next = skb->next;
 		kfree_skb(skb);