openvswitch: Mega flow implementation

Add wildcarded flow support in kernel datapath.

Wildcarded flow can improve OVS flow set up performance by avoid sending
matching new flows to the user space program. The exact performance boost
will largely dependent on wildcarded flow hit rate.

In case all new flows hits wildcard flows, the flow set up rate is
within 5% of that of linux bridge module.

Pravin has made significant contributions to this patch. Including API
clean ups and bug fixes.

Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>

7 files changed, 1171 insertions, 513 deletions

diff --git a/Documentation/networking/openvswitch.txt b/Documentation/networking/openvswitch.txt
index 8fa2dd1e792..37c20ee2455 100644
--- a/Documentation/networking/openvswitch.txt
+++ b/Documentation/networking/openvswitch.txt
@@ -91,6 +91,46 @@ Often we ellipsize arguments not important to the discussion, e.g.:
     in_port(1), eth(...), eth_type(0x0800), ipv4(...), tcp(...)
 
 
+Wildcarded flow key format
+--------------------------
+
+A wildcarded flow is described with two sequences of Netlink attributes
+passed over the Netlink socket. A flow key, exactly as described above, and an
+optional corresponding flow mask.
+
+A wildcarded flow can represent a group of exact match flows. Each '1' bit
+in the mask specifies a exact match with the corresponding bit in the flow key.
+A '0' bit specifies a don't care bit, which will match either a '1' or '0' bit
+of a incoming packet. Using wildcarded flow can improve the flow set up rate
+by reduce the number of new flows need to be processed by the user space program.
+
+Support for the mask Netlink attribute is optional for both the kernel and user
+space program. The kernel can ignore the mask attribute, installing an exact
+match flow, or reduce the number of don't care bits in the kernel to less than
+what was specified by the user space program. In this case, variations in bits
+that the kernel does not implement will simply result in additional flow setups.
+The kernel module will also work with user space programs that neither support
+nor supply flow mask attributes.
+
+Since the kernel may ignore or modify wildcard bits, it can be difficult for
+the userspace program to know exactly what matches are installed. There are
+two possible approaches: reactively install flows as they miss the kernel
+flow table (and therefore not attempt to determine wildcard changes at all)
+or use the kernel's response messages to determine the installed wildcards.
+
+When interacting with userspace, the kernel should maintain the match portion
+of the key exactly as originally installed. This will provides a handle to
+identify the flow for all future operations. However, when reporting the
+mask of an installed flow, the mask should include any restrictions imposed
+by the kernel.
+
+The behavior when using overlapping wildcarded flows is undefined. It is the
+responsibility of the user space program to ensure that any incoming packet
+can match at most one flow, wildcarded or not. The current implementation
+performs best-effort detection of overlapping wildcarded flows and may reject
+some but not all of them. However, this behavior may change in future versions.
+
+
 Basic rule for evolving flow keys
 ---------------------------------
 
diff --git a/include/uapi/linux/openvswitch.h b/include/uapi/linux/openvswitch.h
index 52490b0e62b..de1fa5d3780 100644
--- a/include/uapi/linux/openvswitch.h
+++ b/include/uapi/linux/openvswitch.h
@@ -1,6 +1,6 @@
 
 /*
- * Copyright (c) 2007-2011 Nicira Networks.
+ * Copyright (c) 2007-2013 Nicira, Inc.
  *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of version 2 of the GNU General Public
@@ -379,6 +379,12 @@ struct ovs_key_nd {
  * @OVS_FLOW_ATTR_CLEAR: If present in a %OVS_FLOW_CMD_SET request, clears the
  * last-used time, accumulated TCP flags, and statistics for this flow.
  * Otherwise ignored in requests.  Never present in notifications.
+ * @OVS_FLOW_ATTR_MASK: Nested %OVS_KEY_ATTR_* attributes specifying the
+ * mask bits for wildcarded flow match. Mask bit value '1' specifies exact
+ * match with corresponding flow key bit, while mask bit value '0' specifies
+ * a wildcarded match. Omitting attribute is treated as wildcarding all
+ * corresponding fields. Optional for all requests. If not present,
+ * all flow key bits are exact match bits.
  *
  * These attributes follow the &struct ovs_header within the Generic Netlink
  * payload for %OVS_FLOW_* commands.
@@ -391,6 +397,7 @@ enum ovs_flow_attr {
 	OVS_FLOW_ATTR_TCP_FLAGS, /* 8-bit OR'd TCP flags. */
 	OVS_FLOW_ATTR_USED,      /* u64 msecs last used in monotonic time. */
 	OVS_FLOW_ATTR_CLEAR,     /* Flag to clear stats, tcp_flags, used. */
+	OVS_FLOW_ATTR_MASK,      /* Sequence of OVS_KEY_ATTR_* attributes. */
 	__OVS_FLOW_ATTR_MAX
 };
 
diff --git a/net/openvswitch/actions.c b/net/openvswitch/actions.c
index ab101f71544..1f680222f4f 100644
--- a/net/openvswitch/actions.c
+++ b/net/openvswitch/actions.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2007-2012 Nicira, Inc.
+ * Copyright (c) 2007-2013 Nicira, Inc.
  *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of version 2 of the GNU General Public
@@ -376,8 +376,10 @@ static int output_userspace(struct datapath *dp, struct sk_buff *skb,
 	const struct nlattr *a;
 	int rem;
 
+	BUG_ON(!OVS_CB(skb)->pkt_key);
+
 	upcall.cmd = OVS_PACKET_CMD_ACTION;
-	upcall.key = &OVS_CB(skb)->flow->key;
+	upcall.key = OVS_CB(skb)->pkt_key;
 	upcall.userdata = NULL;
 	upcall.portid = 0;
 
diff --git a/net/openvswitch/datapath.c b/net/openvswitch/datapath.c
index 9d97ef3c983..d29cd9aa4a6 100644
--- a/net/openvswitch/datapath.c
+++ b/net/openvswitch/datapath.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2007-2012 Nicira, Inc.
+ * Copyright (c) 2007-2013 Nicira, Inc.
  *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of version 2 of the GNU General Public
@@ -165,7 +165,7 @@ static void destroy_dp_rcu(struct rcu_head *rcu)
 {
 	struct datapath *dp = container_of(rcu, struct datapath, rcu);
 
-	ovs_flow_tbl_destroy((__force struct flow_table *)dp->table);
+	ovs_flow_tbl_destroy((__force struct flow_table *)dp->table, false);
 	free_percpu(dp->stats_percpu);
 	release_net(ovs_dp_get_net(dp));
 	kfree(dp->ports);
@@ -226,19 +226,18 @@ void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
 	struct sw_flow_key key;
 	u64 *stats_counter;
 	int error;
-	int key_len;
 
 	stats = this_cpu_ptr(dp->stats_percpu);
 
 	/* Extract flow from 'skb' into 'key'. */
-	error = ovs_flow_extract(skb, p->port_no, &key, &key_len);
+	error = ovs_flow_extract(skb, p->port_no, &key);
 	if (unlikely(error)) {
 		kfree_skb(skb);
 		return;
 	}
 
 	/* Look up flow. */
-	flow = ovs_flow_tbl_lookup(rcu_dereference(dp->table), &key, key_len);
+	flow = ovs_flow_lookup(rcu_dereference(dp->table), &key);
 	if (unlikely(!flow)) {
 		struct dp_upcall_info upcall;
 
@@ -253,6 +252,7 @@ void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
 	}
 
 	OVS_CB(skb)->flow = flow;
+	OVS_CB(skb)->pkt_key = &key;
 
 	stats_counter = &stats->n_hit;
 	ovs_flow_used(OVS_CB(skb)->flow, skb);
@@ -435,7 +435,7 @@ static int queue_userspace_packet(struct net *net, int dp_ifindex,
 	upcall->dp_ifindex = dp_ifindex;
 
 	nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_KEY);
-	ovs_flow_to_nlattrs(upcall_info->key, user_skb);
+	ovs_flow_to_nlattrs(upcall_info->key, upcall_info->key, user_skb);
 	nla_nest_end(user_skb, nla);
 
 	if (upcall_info->userdata)
@@ -468,7 +468,7 @@ static int flush_flows(struct datapath *dp)
 
 	rcu_assign_pointer(dp->table, new_table);
 
-	ovs_flow_tbl_deferred_destroy(old_table);
+	ovs_flow_tbl_destroy(old_table, true);
 	return 0;
 }
 
@@ -611,10 +611,12 @@ static int validate_tp_port(const struct sw_flow_key *flow_key)
 static int validate_and_copy_set_tun(const struct nlattr *attr,
 				     struct sw_flow_actions **sfa)
 {
-	struct ovs_key_ipv4_tunnel tun_key;
+	struct sw_flow_match match;
+	struct sw_flow_key key;
 	int err, start;
 
-	err = ovs_ipv4_tun_from_nlattr(nla_data(attr), &tun_key);
+	ovs_match_init(&match, &key, NULL);
+	err = ovs_ipv4_tun_from_nlattr(nla_data(attr), &match, false);
 	if (err)
 		return err;
 
@@ -622,7 +624,8 @@ static int validate_and_copy_set_tun(const struct nlattr *attr,
 	if (start < 0)
 		return start;
 
-	err = add_action(sfa, OVS_KEY_ATTR_IPV4_TUNNEL, &tun_key, sizeof(tun_key));
+	err = add_action(sfa, OVS_KEY_ATTR_IPV4_TUNNEL, &match.key->tun_key,
+			sizeof(match.key->tun_key));
 	add_nested_action_end(*sfa, start);
 
 	return err;
@@ -857,7 +860,6 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
 	struct ethhdr *eth;
 	int len;
 	int err;
-	int key_len;
 
 	err = -EINVAL;
 	if (!a[OVS_PACKET_ATTR_PACKET] || !a[OVS_PACKET_ATTR_KEY] ||
@@ -890,11 +892,11 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
 	if (IS_ERR(flow))
 		goto err_kfree_skb;
 
-	err = ovs_flow_extract(packet, -1, &flow->key, &key_len);
+	err = ovs_flow_extract(packet, -1, &flow->key);
 	if (err)
 		goto err_flow_free;
 
-	err = ovs_flow_metadata_from_nlattrs(flow, key_len, a[OVS_PACKET_ATTR_KEY]);
+	err = ovs_flow_metadata_from_nlattrs(flow, a[OVS_PACKET_ATTR_KEY]);
 	if (err)
 		goto err_flow_free;
 	acts = ovs_flow_actions_alloc(nla_len(a[OVS_PACKET_ATTR_ACTIONS]));
@@ -908,6 +910,7 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
 		goto err_flow_free;
 
 	OVS_CB(packet)->flow = flow;
+	OVS_CB(packet)->pkt_key = &flow->key;
 	packet->priority = flow->key.phy.priority;
 	packet->mark = flow->key.phy.skb_mark;
 
@@ -922,13 +925,13 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
 	local_bh_enable();
 	rcu_read_unlock();
 
-	ovs_flow_free(flow);
+	ovs_flow_free(flow, false);
 	return err;
 
 err_unlock:
 	rcu_read_unlock();
 err_flow_free:
-	ovs_flow_free(flow);
+	ovs_flow_free(flow, false);
 err_kfree_skb:
 	kfree_skb(packet);
 err:
@@ -1045,7 +1048,8 @@ static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
 		if (!start)
 			return -EMSGSIZE;
 
-		err = ovs_ipv4_tun_to_nlattr(skb, nla_data(ovs_key));
+		err = ovs_ipv4_tun_to_nlattr(skb, nla_data(ovs_key),
+					     nla_data(ovs_key));
 		if (err)
 			return err;
 		nla_nest_end(skb, start);
@@ -1093,6 +1097,7 @@ static size_t ovs_flow_cmd_msg_size(const struct sw_flow_actions *acts)
 {
 	return NLMSG_ALIGN(sizeof(struct ovs_header))
 		+ nla_total_size(key_attr_size()) /* OVS_FLOW_ATTR_KEY */
+		+ nla_total_size(key_attr_size()) /* OVS_FLOW_ATTR_MASK */
 		+ nla_total_size(sizeof(struct ovs_flow_stats)) /* OVS_FLOW_ATTR_STATS */
 		+ nla_total_size(1) /* OVS_FLOW_ATTR_TCP_FLAGS */
 		+ nla_total_size(8) /* OVS_FLOW_ATTR_USED */
@@ -1119,12 +1124,25 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
 
 	ovs_header->dp_ifindex = get_dpifindex(dp);
 
+	/* Fill flow key. */
 	nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY);
 	if (!nla)
 		goto nla_put_failure;
-	err = ovs_flow_to_nlattrs(&flow->key, skb);
+
+	err = ovs_flow_to_nlattrs(&flow->unmasked_key,
+			&flow->unmasked_key, skb);
+	if (err)
+		goto error;
+	nla_nest_end(skb, nla);
+
+	nla = nla_nest_start(skb, OVS_FLOW_ATTR_MASK);
+	if (!nla)
+		goto nla_put_failure;
+
+	err = ovs_flow_to_nlattrs(&flow->key, &flow->mask->key, skb);
 	if (err)
 		goto error;
+
 	nla_nest_end(skb, nla);
 
 	spin_lock_bh(&flow->lock);
@@ -1214,20 +1232,24 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 {
 	struct nlattr **a = info->attrs;
 	struct ovs_header *ovs_header = info->userhdr;
-	struct sw_flow_key key;
-	struct sw_flow *flow;
+	struct sw_flow_key key, masked_key;
+	struct sw_flow *flow = NULL;
+	struct sw_flow_mask mask;
 	struct sk_buff *reply;
 	struct datapath *dp;
 	struct flow_table *table;
 	struct sw_flow_actions *acts = NULL;
+	struct sw_flow_match match;
 	int error;
-	int key_len;
 
 	/* Extract key. */
 	error = -EINVAL;
 	if (!a[OVS_FLOW_ATTR_KEY])
 		goto error;
-	error = ovs_flow_from_nlattrs(&key, &key_len, a[OVS_FLOW_ATTR_KEY]);
+
+	ovs_match_init(&match, &key, &mask);
+	error = ovs_match_from_nlattrs(&match,
+			a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]);
 	if (error)
 		goto error;
 
@@ -1238,9 +1260,13 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 		if (IS_ERR(acts))
 			goto error;
 
-		error = validate_and_copy_actions(a[OVS_FLOW_ATTR_ACTIONS], &key,  0, &acts);
-		if (error)
+		ovs_flow_key_mask(&masked_key, &key, &mask);
+		error = validate_and_copy_actions(a[OVS_FLOW_ATTR_ACTIONS],
+						  &masked_key, 0, &acts);
+		if (error) {
+			OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
 			goto err_kfree;
+		}
 	} else if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW) {
 		error = -EINVAL;
 		goto error;
@@ -1253,8 +1279,11 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 		goto err_unlock_ovs;
 
 	table = ovsl_dereference(dp->table);
-	flow = ovs_flow_tbl_lookup(table, &key, key_len);
+
+	/* Check if this is a duplicate flow */
+	flow = ovs_flow_lookup(table, &key);
 	if (!flow) {
+		struct sw_flow_mask *mask_p;
 		/* Bail out if we're not allowed to create a new flow. */
 		error = -ENOENT;
 		if (info->genlhdr->cmd == OVS_FLOW_CMD_SET)
@@ -1267,7 +1296,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 			new_table = ovs_flow_tbl_expand(table);
 			if (!IS_ERR(new_table)) {
 				rcu_assign_pointer(dp->table, new_table);
-				ovs_flow_tbl_deferred_destroy(table);
+				ovs_flow_tbl_destroy(table, true);
 				table = ovsl_dereference(dp->table);
 			}
 		}
@@ -1280,14 +1309,30 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 		}
 		clear_stats(flow);
 
+		flow->key = masked_key;
+		flow->unmasked_key = key;
+
+		/* Make sure mask is unique in the system */
+		mask_p = ovs_sw_flow_mask_find(table, &mask);
+		if (!mask_p) {
+			/* Allocate a new mask if none exsits. */
+			mask_p = ovs_sw_flow_mask_alloc();
+			if (!mask_p)
+				goto err_flow_free;
+			mask_p->key = mask.key;
+			mask_p->range = mask.range;
+			ovs_sw_flow_mask_insert(table, mask_p);
+		}
+
+		ovs_sw_flow_mask_add_ref(mask_p);
+		flow->mask = mask_p;
 		rcu_assign_pointer(flow->sf_acts, acts);
 
 		/* Put flow in bucket. */
-		ovs_flow_tbl_insert(table, flow, &key, key_len);
+		ovs_flow_insert(table, flow);
 
 		reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
-						info->snd_seq,
-						OVS_FLOW_CMD_NEW);
+						info->snd_seq, OVS_FLOW_CMD_NEW);
 	} else {
 		/* We found a matching flow. */
 		struct sw_flow_actions *old_acts;
@@ -1303,6 +1348,13 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 		    info->nlhdr->nlmsg_flags & (NLM_F_CREATE | NLM_F_EXCL))
 			goto err_unlock_ovs;
 
+		/* The unmasked key has to be the same for flow updates. */
+		error = -EINVAL;
+		if (!ovs_flow_cmp_unmasked_key(flow, &key, match.range.end)) {
+			OVS_NLERR("Flow modification message rejected, unmasked key does not match.\n");
+			goto err_unlock_ovs;
+		}
+
 		/* Update actions. */
 		old_acts = ovsl_dereference(flow->sf_acts);
 		rcu_assign_pointer(flow->sf_acts, acts);
@@ -1327,6 +1379,8 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 				ovs_dp_flow_multicast_group.id, PTR_ERR(reply));
 	return 0;
 
+err_flow_free:
+	ovs_flow_free(flow, false);
 err_unlock_ovs:
 	ovs_unlock();
 err_kfree:
@@ -1344,12 +1398,16 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
 	struct sw_flow *flow;
 	struct datapath *dp;
 	struct flow_table *table;
+	struct sw_flow_match match;
 	int err;
-	int key_len;
 
-	if (!a[OVS_FLOW_ATTR_KEY])
+	if (!a[OVS_FLOW_ATTR_KEY]) {
+		OVS_NLERR("Flow get message rejected, Key attribute missing.\n");
 		return -EINVAL;
-	err = ovs_flow_from_nlattrs(&key, &key_len, a[OVS_FLOW_ATTR_KEY]);
+	}
+
+	ovs_match_init(&match, &key, NULL);
+	err = ovs_match_from_nlattrs(&match, a[OVS_FLOW_ATTR_KEY], NULL);
 	if (err)
 		return err;
 
@@ -1361,7 +1419,7 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
 	}
 
 	table = ovsl_dereference(dp->table);
-	flow = ovs_flow_tbl_lookup(table, &key, key_len);
+	flow = ovs_flow_lookup_unmasked_key(table, &match);
 	if (!flow) {
 		err = -ENOENT;
 		goto unlock;
@@ -1390,8 +1448,8 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
 	struct sw_flow *flow;
 	struct datapath *dp;
 	struct flow_table *table;
+	struct sw_flow_match match;
 	int err;
-	int key_len;
 
 	ovs_lock();
 	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
@@ -1404,12 +1462,14 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
 		err = flush_flows(dp);
 		goto unlock;
 	}
-	err = ovs_flow_from_nlattrs(&key, &key_len, a[OVS_FLOW_ATTR_KEY]);
+
+	ovs_match_init(&match, &key, NULL);
+	err = ovs_match_from_nlattrs(&match, a[OVS_FLOW_ATTR_KEY], NULL);
 	if (err)
 		goto unlock;
 
 	table = ovsl_dereference(dp->table);
-	flow = ovs_flow_tbl_lookup(table, &key, key_len);
+	flow = ovs_flow_lookup_unmasked_key(table, &match);
 	if (!flow) {
 		err = -ENOENT;
 		goto unlock;
@@ -1421,13 +1481,13 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
 		goto unlock;
 	}
 
-	ovs_flow_tbl_remove(table, flow);
+	ovs_flow_remove(table, flow);
 
 	err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_portid,
 				     info->snd_seq, 0, OVS_FLOW_CMD_DEL);
 	BUG_ON(err < 0);
 
-	ovs_flow_deferred_free(flow);
+	ovs_flow_free(flow, true);
 	ovs_unlock();
 
 	ovs_notify(reply, info, &ovs_dp_flow_multicast_group);
@@ -1457,7 +1517,7 @@ static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
 
 		bucket = cb->args[0];
 		obj = cb->args[1];
-		flow = ovs_flow_tbl_next(table, &bucket, &obj);
+		flow = ovs_flow_dump_next(table, &bucket, &obj);
 		if (!flow)
 			break;
 
@@ -1680,7 +1740,7 @@ err_destroy_ports_array:
 err_destroy_percpu:
 	free_percpu(dp->stats_percpu);
 err_destroy_table:
-	ovs_flow_tbl_destroy(ovsl_dereference(dp->table));
+	ovs_flow_tbl_destroy(ovsl_dereference(dp->table), false);
 err_free_dp:
 	release_net(ovs_dp_get_net(dp));
 	kfree(dp);
@@ -2287,7 +2347,7 @@ static void rehash_flow_table(struct work_struct *work)
 			new_table = ovs_flow_tbl_rehash(old_table);
 			if (!IS_ERR(new_table)) {
 				rcu_assign_pointer(dp->table, new_table);
-				ovs_flow_tbl_deferred_destroy(old_table);
+				ovs_flow_tbl_destroy(old_table, true);
 			}
 		}
 	}
diff --git a/net/openvswitch/datapath.h b/net/openvswitch/datapath.h
index a9148648491..4d109c176ef 100644
--- a/net/openvswitch/datapath.h
+++ b/net/openvswitch/datapath.h
@@ -88,11 +88,13 @@ struct datapath {
 /**
  * struct ovs_skb_cb - OVS data in skb CB
  * @flow: The flow associated with this packet.  May be %NULL if no flow.
+ * @pkt_key: The flow information extracted from the packet.  Must be nonnull.
  * @tun_key: Key for the tunnel that encapsulated this packet. NULL if the
  * packet is not being tunneled.
  */
 struct ovs_skb_cb {
 	struct sw_flow		*flow;
+	struct sw_flow_key	*pkt_key;
 	struct ovs_key_ipv4_tunnel  *tun_key;
 };
 #define OVS_CB(skb) ((struct ovs_skb_cb *)(skb)->cb)
@@ -183,4 +185,8 @@ struct sk_buff *ovs_vport_cmd_build_info(struct vport *, u32 pid, u32 seq,
 
 int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb);
 void ovs_dp_notify_wq(struct work_struct *work);
+
+#define OVS_NLERR(fmt, ...) \
+	pr_info_once("netlink: " fmt, ##__VA_ARGS__)
+
 #endif /* datapath.h */
diff --git a/net/openvswitch/flow.c b/net/openvswitch/flow.c
index fca282520ce..1fceb965359 100644
--- a/net/openvswitch/flow.c
+++ b/net/openvswitch/flow.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2007-2011 Nicira, Inc.
+ * Copyright (c) 2007-2013 Nicira, Inc.
  *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of version 2 of the GNU General Public
@@ -46,6 +46,184 @@
 
 static struct kmem_cache *flow_cache;
 
+static void ovs_sw_flow_mask_set(struct sw_flow_mask *mask,
+		struct sw_flow_key_range *range, u8 val);
+
+static void update_range__(struct sw_flow_match *match,
+			  size_t offset, size_t size, bool is_mask)
+{
+	struct sw_flow_key_range *range = NULL;
+	size_t start = offset;
+	size_t end = offset + size;
+
+	if (!is_mask)
+		range = &match->range;
+	else if (match->mask)
+		range = &match->mask->range;
+
+	if (!range)
+		return;
+
+	if (range->start == range->end) {
+		range->start = start;
+		range->end = end;
+		return;
+	}
+
+	if (range->start > start)
+		range->start = start;
+
+	if (range->end < end)
+		range->end = end;
+}
+
+#define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
+	do { \
+		update_range__(match, offsetof(struct sw_flow_key, field),  \
+				     sizeof((match)->key->field), is_mask); \
+		if (is_mask) {						    \
+			if ((match)->mask)				    \
+				(match)->mask->key.field = value;	    \
+		} else {                                                    \
+			(match)->key->field = value;		            \
+		}                                                           \
+	} while (0)
+
+#define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
+	do { \
+		update_range__(match, offsetof(struct sw_flow_key, field),  \
+				len, is_mask);                              \
+		if (is_mask) {						    \
+			if ((match)->mask)				    \
+				memcpy(&(match)->mask->key.field, value_p, len);\
+		} else {                                                    \
+			memcpy(&(match)->key->field, value_p, len);         \
+		}                                                           \
+	} while (0)
+
+void ovs_match_init(struct sw_flow_match *match,
+		    struct sw_flow_key *key,
+		    struct sw_flow_mask *mask)
+{
+	memset(match, 0, sizeof(*match));
+	match->key = key;
+	match->mask = mask;
+
+	memset(key, 0, sizeof(*key));
+
+	if (mask) {
+		memset(&mask->key, 0, sizeof(mask->key));
+		mask->range.start = mask->range.end = 0;
+	}
+}
+
+static bool ovs_match_validate(const struct sw_flow_match *match,
+		u64 key_attrs, u64 mask_attrs)
+{
+	u64 key_expected = 1 << OVS_KEY_ATTR_ETHERNET;
+	u64 mask_allowed = key_attrs;  /* At most allow all key attributes */
+
+	/* The following mask attributes allowed only if they
+	 * pass the validation tests. */
+	mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
+			| (1 << OVS_KEY_ATTR_IPV6)
+			| (1 << OVS_KEY_ATTR_TCP)
+			| (1 << OVS_KEY_ATTR_UDP)
+			| (1 << OVS_KEY_ATTR_ICMP)
+			| (1 << OVS_KEY_ATTR_ICMPV6)
+			| (1 << OVS_KEY_ATTR_ARP)
+			| (1 << OVS_KEY_ATTR_ND));
+
+	/* Always allowed mask fields. */
+	mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
+		       | (1 << OVS_KEY_ATTR_IN_PORT)
+		       | (1 << OVS_KEY_ATTR_ETHERTYPE));
+
+	/* Check key attributes. */
+	if (match->key->eth.type == htons(ETH_P_ARP)
+			|| match->key->eth.type == htons(ETH_P_RARP)) {
+		key_expected |= 1 << OVS_KEY_ATTR_ARP;
+		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
+			mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
+	}
+
+	if (match->key->eth.type == htons(ETH_P_IP)) {
+		key_expected |= 1 << OVS_KEY_ATTR_IPV4;
+		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
+			mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
+
+		if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
+			if (match->key->ip.proto == IPPROTO_UDP) {
+				key_expected |= 1 << OVS_KEY_ATTR_UDP;
+				if (match->mask && (match->mask->key.ip.proto == 0xff))
+					mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
+			}
+
+			if (match->key->ip.proto == IPPROTO_TCP) {
+				key_expected |= 1 << OVS_KEY_ATTR_TCP;
+				if (match->mask && (match->mask->key.ip.proto == 0xff))
+					mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
+			}
+
+			if (match->key->ip.proto == IPPROTO_ICMP) {
+				key_expected |= 1 << OVS_KEY_ATTR_ICMP;
+				if (match->mask && (match->mask->key.ip.proto == 0xff))
+					mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
+			}
+		}
+	}
+
+	if (match->key->eth.type == htons(ETH_P_IPV6)) {
+		key_expected |= 1 << OVS_KEY_ATTR_IPV6;
+		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
+			mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
+
+		if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
+			if (match->key->ip.proto == IPPROTO_UDP) {
+				key_expected |= 1 << OVS_KEY_ATTR_UDP;
+				if (match->mask && (match->mask->key.ip.proto == 0xff))
+					mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
+			}
+
+			if (match->key->ip.proto == IPPROTO_TCP) {
+				key_expected |= 1 << OVS_KEY_ATTR_TCP;
+				if (match->mask && (match->mask->key.ip.proto == 0xff))
+					mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
+			}
+
+			if (match->key->ip.proto == IPPROTO_ICMPV6) {
+				key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
+				if (match->mask && (match->mask->key.ip.proto == 0xff))
+					mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
+
+				if (match->key->ipv6.tp.src ==
+						htons(NDISC_NEIGHBOUR_SOLICITATION) ||
+				    match->key->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
+					key_expected |= 1 << OVS_KEY_ATTR_ND;
+					if (match->mask && (match->mask->key.ipv6.tp.src == htons(0xffff)))
+						mask_allowed |= 1 << OVS_KEY_ATTR_ND;
+				}
+			}
+		}
+	}
+
+	if ((key_attrs & key_expected) != key_expected) {
+		/* Key attributes check failed. */
+		OVS_NLERR("Missing expected key attributes (key_attrs=%llx, expected=%llx).\n",
+				key_attrs, key_expected);
+		return false;
+	}
+
+	if ((mask_attrs & mask_allowed) != mask_attrs) {
+		/* Mask attributes check failed. */
+		OVS_NLERR("Contain more than allowed mask fields (mask_attrs=%llx, mask_allowed=%llx).\n",
+				mask_attrs, mask_allowed);
+		return false;
+	}
+
+	return true;
+}
+
 static int check_header(struct sk_buff *skb, int len)
 {
 	if (unlikely(skb->len < len))
@@ -121,12 +299,7 @@ u64 ovs_flow_used_time(unsigned long flow_jiffies)
 	return cur_ms - idle_ms;
 }
 
-#define SW_FLOW_KEY_OFFSET(field)		\
-	(offsetof(struct sw_flow_key, field) +	\
-	 FIELD_SIZEOF(struct sw_flow_key, field))
-
-static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key,
-			 int *key_lenp)
+static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key)
 {
 	unsigned int nh_ofs = skb_network_offset(skb);
 	unsigned int nh_len;
@@ -136,8 +309,6 @@ static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key,
 	__be16 frag_off;
 	int err;
 
-	*key_lenp = SW_FLOW_KEY_OFFSET(ipv6.label);
-
 	err = check_header(skb, nh_ofs + sizeof(*nh));
 	if (unlikely(err))
 		return err;
@@ -176,6 +347,21 @@ static bool icmp6hdr_ok(struct sk_buff *skb)
 				  sizeof(struct icmp6hdr));
 }
 
+void ovs_flow_key_mask(struct sw_flow_key *dst, const struct sw_flow_key *src,
+		       const struct sw_flow_mask *mask)
+{
+	u8 *m = (u8 *)&mask->key + mask->range.start;
+	u8 *s = (u8 *)src + mask->range.start;
+	u8 *d = (u8 *)dst + mask->range.start;
+	int i;
+
+	memset(dst, 0, sizeof(*dst));
+	for (i = 0; i < ovs_sw_flow_mask_size_roundup(mask); i++) {
+		*d = *s & *m;
+		d++, s++, m++;
+	}
+}
+
 #define TCP_FLAGS_OFFSET 13
 #define TCP_FLAG_MASK 0x3f
 
@@ -224,6 +410,7 @@ struct sw_flow *ovs_flow_alloc(void)
 
 	spin_lock_init(&flow->lock);
 	flow->sf_acts = NULL;
+	flow->mask = NULL;
 
 	return flow;
 }
@@ -263,7 +450,7 @@ static void free_buckets(struct flex_array *buckets)
 	flex_array_free(buckets);
 }
 
-struct flow_table *ovs_flow_tbl_alloc(int new_size)
+static struct flow_table *__flow_tbl_alloc(int new_size)
 {
 	struct flow_table *table = kmalloc(sizeof(*table), GFP_KERNEL);
 
@@ -281,17 +468,15 @@ struct flow_table *ovs_flow_tbl_alloc(int new_size)
 	table->node_ver = 0;
 	table->keep_flows = false;
 	get_random_bytes(&table->hash_seed, sizeof(u32));
+	table->mask_list = NULL;
 
 	return table;
 }
 
-void ovs_flow_tbl_destroy(struct flow_table *table)
+static void __flow_tbl_destroy(struct flow_table *table)
 {
 	int i;
 
-	if (!table)
-		return;
-
 	if (table->keep_flows)
 		goto skip_flows;
 
@@ -303,31 +488,55 @@ void ovs_flow_tbl_destroy(struct flow_table *table)
 
 		hlist_for_each_entry_safe(flow, n, head, hash_node[ver]) {
 			hlist_del(&flow->hash_node[ver]);
-			ovs_flow_free(flow);
+			ovs_flow_free(flow, false);
 		}
 	}
 
+	BUG_ON(!list_empty(table->mask_list));
+	kfree(table->mask_list);
+
 skip_flows:
 	free_buckets(table->buckets);
 	kfree(table);
 }
 
+struct flow_table *ovs_flow_tbl_alloc(int new_size)
+{
+	struct flow_table *table = __flow_tbl_alloc(new_size);
+
+	if (!table)
+		return NULL;
+
+	table->mask_list = kmalloc(sizeof(struct list_head), GFP_KERNEL);
+	if (!table->mask_list) {
+		table->keep_flows = true;
+		__flow_tbl_destroy(table);
+		return NULL;
+	}
+	INIT_LIST_HEAD(table->mask_list);
+
+	return table;
+}
+
 static void flow_tbl_destroy_rcu_cb(struct rcu_head *rcu)
 {
 	struct flow_table *table = container_of(rcu, struct flow_table, rcu);
 
-	ovs_flow_tbl_destroy(table);
+	__flow_tbl_destroy(table);
 }
 
-void ovs_flow_tbl_deferred_destroy(struct flow_table *table)
+void ovs_flow_tbl_destroy(struct flow_table *table, bool deferred)
 {
 	if (!table)
 		return;
 
-	call_rcu(&table->rcu, flow_tbl_destroy_rcu_cb);
+	if (deferred)
+		call_rcu(&table->rcu, flow_tbl_destroy_rcu_cb);
+	else
+		__flow_tbl_destroy(table);
 }
 
-struct sw_flow *ovs_flow_tbl_next(struct flow_table *table, u32 *bucket, u32 *last)
+struct sw_flow *ovs_flow_dump_next(struct flow_table *table, u32 *bucket, u32 *last)
 {
 	struct sw_flow *flow;
 	struct hlist_head *head;
@@ -353,11 +562,13 @@ struct sw_flow *ovs_flow_tbl_next(struct flow_table *table, u32 *bucket, u32 *la
 	return NULL;
 }
 
-static void __flow_tbl_insert(struct flow_table *table, struct sw_flow *flow)
+static void __tbl_insert(struct flow_table *table, struct sw_flow *flow)
 {
 	struct hlist_head *head;
+
 	head = find_bucket(table, flow->hash);
 	hlist_add_head_rcu(&flow->hash_node[table->node_ver], head);
+
 	table->count++;
 }
 
@@ -377,8 +588,10 @@ static void flow_table_copy_flows(struct flow_table *old, struct flow_table *new
 		head = flex_array_get(old->buckets, i);
 
 		hlist_for_each_entry(flow, head, hash_node[old_ver])
-			__flow_tbl_insert(new, flow);
+			__tbl_insert(new, flow);
 	}
+
+	new->mask_list = old->mask_list;
 	old->keep_flows = true;
 }
 
@@ -386,7 +599,7 @@ static struct flow_table *__flow_tbl_rehash(struct flow_table *table, int n_buck
 {
 	struct flow_table *new_table;
 
-	new_table = ovs_flow_tbl_alloc(n_buckets);
+	new_table = __flow_tbl_alloc(n_buckets);
 	if (!new_table)
 		return ERR_PTR(-ENOMEM);
 
@@ -405,28 +618,30 @@ struct flow_table *ovs_flow_tbl_expand(struct flow_table *table)
 	return __flow_tbl_rehash(table, table->n_buckets * 2);
 }
 
-void ovs_flow_free(struct sw_flow *flow)
+static void __flow_free(struct sw_flow *flow)
 {
-	if (unlikely(!flow))
-		return;
-
 	kfree((struct sf_flow_acts __force *)flow->sf_acts);
 	kmem_cache_free(flow_cache, flow);
 }
 
-/* RCU callback used by ovs_flow_deferred_free. */
 static void rcu_free_flow_callback(struct rcu_head *rcu)
 {
 	struct sw_flow *flow = container_of(rcu, struct sw_flow, rcu);
 
-	ovs_flow_free(flow);
+	__flow_free(flow);
 }
 
-/* Schedules 'flow' to be freed after the next RCU grace period.
- * The caller must hold rcu_read_lock for this to be sensible. */
-void ovs_flow_deferred_free(struct sw_flow *flow)
+void ovs_flow_free(struct sw_flow *flow, bool deferred)
 {
-	call_rcu(&flow->rcu, rcu_free_flow_callback);
+	if (!flow)
+		return;
+
+	ovs_sw_flow_mask_del_ref(flow->mask, deferred);
+
+	if (deferred)
+		call_rcu(&flow->rcu, rcu_free_flow_callback);
+	else
+		__flow_free(flow);
 }
 
 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
@@ -497,18 +712,15 @@ static __be16 parse_ethertype(struct sk_buff *skb)
 }
 
 static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
-			int *key_lenp, int nh_len)
+			int nh_len)
 {
 	struct icmp6hdr *icmp = icmp6_hdr(skb);
-	int error = 0;
-	int key_len;
 
 	/* The ICMPv6 type and code fields use the 16-bit transport port
 	 * fields, so we need to store them in 16-bit network byte order.
 	 */
 	key->ipv6.tp.src = htons(icmp->icmp6_type);
 	key->ipv6.tp.dst = htons(icmp->icmp6_code);
-	key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
 
 	if (icmp->icmp6_code == 0 &&
 	    (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION ||
@@ -517,21 +729,17 @@ static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
 		struct nd_msg *nd;
 		int offset;
 
-		key_len = SW_FLOW_KEY_OFFSET(ipv6.nd);
-
 		/* In order to process neighbor discovery options, we need the
 		 * entire packet.
 		 */
 		if (unlikely(icmp_len < sizeof(*nd)))
-			goto out;
-		if (unlikely(skb_linearize(skb))) {
-			error = -ENOMEM;
-			goto out;
-		}
+			return 0;
+
+		if (unlikely(skb_linearize(skb)))
+			return -ENOMEM;
 
 		nd = (struct nd_msg *)skb_transport_header(skb);
 		key->ipv6.nd.target = nd->target;
-		key_len = SW_FLOW_KEY_OFFSET(ipv6.nd);
 
 		icmp_len -= sizeof(*nd);
 		offset = 0;
@@ -541,7 +749,7 @@ static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
 			int opt_len = nd_opt->nd_opt_len * 8;