X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=datapath%2Fdatapath.c;h=4d40ac3bde253c1d6d34c2778c96fa4e7e0a82f7;hb=0f4d9dce8150fced85070149e0820707d55ee252;hp=454f96cc20bcc85e725441fd5d38d7804cacbdfa;hpb=d9065a90b6b955aa38586c952e8804ca7a22547e;p=openvswitch

diff --git a/datapath/datapath.c b/datapath/datapath.c
index 454f96cc..4d40ac3b 100644
--- a/datapath/datapath.c
+++ b/datapath/datapath.c
@@ -49,8 +49,8 @@
 #include "datapath.h"
 #include "actions.h"
 #include "flow.h"
-#include "table.h"
 #include "vlan.h"
+#include "tunnel.h"
 #include "vport-internal_dev.h"
 
 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,18) || \
@@ -84,7 +84,7 @@ EXPORT_SYMBOL(dp_ioctl_hook);
 static LIST_HEAD(dps);
 
 static struct vport *new_vport(const struct vport_parms *);
-static int queue_userspace_packets(struct datapath *, struct sk_buff *,
+static int queue_userspace_packets(struct datapath *, u32 pid, struct sk_buff *,
 				 const struct dp_upcall_info *);
 
 /* Must be called with rcu_read_lock, genl_mutex, or RTNL lock. */
@@ -107,7 +107,7 @@ struct datapath *get_dp(int dp_ifindex)
 EXPORT_SYMBOL_GPL(get_dp);
 
 /* Must be called with genl_mutex. */
-static struct tbl *get_table_protected(struct datapath *dp)
+static struct flow_table *get_table_protected(struct datapath *dp)
 {
 	return rcu_dereference_protected(dp->table, lockdep_genl_is_held());
 }
@@ -124,6 +124,24 @@ const char *dp_name(const struct datapath *dp)
 	return vport_get_name(rcu_dereference_rtnl(dp->ports[OVSP_LOCAL]));
 }
 
+static int get_dpifindex(struct datapath *dp)
+{
+	struct vport *local;
+	int ifindex;
+
+	rcu_read_lock();
+
+	local = get_vport_protected(dp, OVSP_LOCAL);
+	if (local)
+		ifindex = vport_get_ifindex(local);
+	else
+		ifindex = 0;
+
+	rcu_read_unlock();
+
+	return ifindex;
+}
+
 static inline size_t br_nlmsg_size(void)
 {
 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
@@ -160,8 +178,7 @@ static int dp_fill_ifinfo(struct sk_buff *skb,
 	hdr->ifi_change = 0;
 
 	NLA_PUT_STRING(skb, IFLA_IFNAME, vport_get_name(port));
-	NLA_PUT_U32(skb, IFLA_MASTER,
-		vport_get_ifindex(get_vport_protected(dp, OVSP_LOCAL)));
+	NLA_PUT_U32(skb, IFLA_MASTER, get_dpifindex(dp));
 	NLA_PUT_U32(skb, IFLA_MTU, vport_get_mtu(port));
 #ifdef IFLA_OPERSTATE
 	NLA_PUT_U8(skb, IFLA_OPERSTATE,
@@ -217,7 +234,7 @@ static void destroy_dp_rcu(struct rcu_head *rcu)
 {
 	struct datapath *dp = container_of(rcu, struct datapath, rcu);
 
-	tbl_destroy((struct tbl __force *)dp->table, flow_free_tbl);
+	flow_tbl_destroy(dp->table);
 	free_percpu(dp->stats_percpu);
 	kobject_put(&dp->ifobj);
 }
@@ -241,7 +258,7 @@ static struct vport *new_vport(const struct vport_parms *parms)
 }
 
 /* Called with RTNL lock. */
-int dp_detach_port(struct vport *p)
+void dp_detach_port(struct vport *p)
 {
 	ASSERT_RTNL();
 
@@ -254,13 +271,14 @@ int dp_detach_port(struct vport *p)
 	rcu_assign_pointer(p->dp->ports[p->port_no], NULL);
 
 	/* Then destroy it. */
-	return vport_del(p);
+	vport_del(p);
 }
 
 /* Must be called with rcu_read_lock. */
 void dp_process_received_packet(struct vport *p, struct sk_buff *skb)
 {
 	struct datapath *dp = p->dp;
+	struct sw_flow *flow;
 	struct dp_stats_percpu *stats;
 	int stats_counter_off;
 	int error;
@@ -269,7 +287,6 @@ void dp_process_received_packet(struct vport *p, struct sk_buff *skb)
 
 	if (!OVS_CB(skb)->flow) {
 		struct sw_flow_key key;
-		struct tbl_node *flow_node;
 		int key_len;
 		bool is_frag;
 
@@ -287,9 +304,8 @@ void dp_process_received_packet(struct vport *p, struct sk_buff *skb)
 		}
 
 		/* Look up flow. */
-		flow_node = tbl_lookup(rcu_dereference(dp->table), &key, key_len,
-				       flow_hash(&key, key_len), flow_cmp);
-		if (unlikely(!flow_node)) {
+		flow = flow_tbl_lookup(rcu_dereference(dp->table), &key, key_len);
+		if (unlikely(!flow)) {
 			struct dp_upcall_info upcall;
 
 			upcall.cmd = OVS_PACKET_CMD_MISS;
@@ -303,7 +319,7 @@ void dp_process_received_packet(struct vport *p, struct sk_buff *skb)
 			goto out;
 		}
 
-		OVS_CB(skb)->flow = flow_cast(flow_node);
+		OVS_CB(skb)->flow = flow;
 	}
 
 	stats_counter_off = offsetof(struct dp_stats_percpu, n_hit);
@@ -345,51 +361,22 @@ static struct genl_family dp_packet_genl_family = {
 	.maxattr = OVS_PACKET_ATTR_MAX
 };
 
-/* Generic Netlink multicast groups for upcalls.
- *
- * We really want three unique multicast groups per datapath, but we can't even
- * get one, because genl_register_mc_group() takes genl_lock, which is also
- * held during Generic Netlink message processing, so trying to acquire
- * multicast groups during OVS_DP_NEW processing deadlocks.  Instead, we
- * preallocate a few groups and use them round-robin for datapaths.  Collision
- * isn't fatal--multicast listeners should check that the family is the one
- * that they want and discard others--but it wastes time and memory to receive
- * unwanted messages.
- */
-#define PACKET_N_MC_GROUPS 16
-static struct genl_multicast_group packet_mc_groups[PACKET_N_MC_GROUPS];
-
-static u32 packet_mc_group(struct datapath *dp, u8 cmd)
-{
-	u32 idx;
-	BUILD_BUG_ON_NOT_POWER_OF_2(PACKET_N_MC_GROUPS);
-
-	idx = jhash_2words(dp->dp_ifindex, cmd, 0) & (PACKET_N_MC_GROUPS - 1);
-	return packet_mc_groups[idx].id;
-}
-
-static int packet_register_mc_groups(void)
-{
-	int i;
-
-	for (i = 0; i < PACKET_N_MC_GROUPS; i++) {
-		struct genl_multicast_group *group = &packet_mc_groups[i];
-		int error;
-
-		sprintf(group->name, "packet%d", i);
-		error = genl_register_mc_group(&dp_packet_genl_family, group);
-		if (error)
-			return error;
-	}
-	return 0;
-}
-
 int dp_upcall(struct datapath *dp, struct sk_buff *skb, const struct dp_upcall_info *upcall_info)
 {
 	struct dp_stats_percpu *stats;
+	u32 pid;
 	int err;
 
-	WARN_ON_ONCE(skb_shared(skb));
+	if (OVS_CB(skb)->flow)
+		pid = OVS_CB(skb)->flow->upcall_pid;
+	else
+		pid = OVS_CB(skb)->vport->upcall_pid;
+
+	if (pid == 0) {
+		err = -ENOTCONN;
+		kfree_skb(skb);
+		goto err;
+	}
 
 	forward_ip_summed(skb, true);
 
@@ -407,7 +394,7 @@ int dp_upcall(struct datapath *dp, struct sk_buff *skb, const struct dp_upcall_i
 		skb = nskb;
 	}
 
-	err = queue_userspace_packets(dp, skb, upcall_info);
+	err = queue_userspace_packets(dp, pid, skb, upcall_info);
 	if (err)
 		goto err;
 
@@ -430,13 +417,21 @@ err:
  * 'upcall_info'.  There will be only one packet unless we broke up a GSO
  * packet.
  */
-static int queue_userspace_packets(struct datapath *dp, struct sk_buff *skb,
-				 const struct dp_upcall_info *upcall_info)
+static int queue_userspace_packets(struct datapath *dp, u32 pid,
+				   struct sk_buff *skb,
+				   const struct dp_upcall_info *upcall_info)
 {
-	u32 group = packet_mc_group(dp, upcall_info->cmd);
+	int dp_ifindex;
 	struct sk_buff *nskb;
 	int err;
 
+	dp_ifindex = get_dpifindex(dp);
+	if (!dp_ifindex) {
+		err = -ENODEV;
+		nskb = skb->next;
+		goto err_kfree_skbs;
+	}
+
 	do {
 		struct ovs_header *upcall;
 		struct sk_buff *user_skb; /* to be queued to userspace */
@@ -450,8 +445,10 @@ static int queue_userspace_packets(struct datapath *dp, struct sk_buff *skb,
 		if (unlikely(err))
 			goto err_kfree_skbs;
 
-		if (nla_attr_size(skb->len) > USHRT_MAX)
+		if (nla_attr_size(skb->len) > USHRT_MAX) {
+			err = -EFBIG;
 			goto err_kfree_skbs;
+		}
 
 		len = sizeof(struct ovs_header);
 		len += nla_total_size(skb->len);
@@ -465,12 +462,12 @@ static int queue_userspace_packets(struct datapath *dp, struct sk_buff *skb,
 
 		user_skb = genlmsg_new(len, GFP_ATOMIC);
 		if (!user_skb) {
-			netlink_set_err(INIT_NET_GENL_SOCK, 0, group, -ENOBUFS);
+			err = -ENOMEM;
 			goto err_kfree_skbs;
 		}
 
 		upcall = genlmsg_put(user_skb, 0, 0, &dp_packet_genl_family, 0, upcall_info->cmd);
-		upcall->dp_ifindex = dp->dp_ifindex;
+		upcall->dp_ifindex = dp_ifindex;
 
 		nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_KEY);
 		flow_to_nlattrs(upcall_info->key, user_skb);
@@ -495,7 +492,7 @@ static int queue_userspace_packets(struct datapath *dp, struct sk_buff *skb,
 		else
 			skb_copy_bits(skb, 0, nla_data(nla), skb->len);
 
-		err = genlmsg_multicast(user_skb, 0, group, GFP_ATOMIC);
+		err = genlmsg_unicast(&init_net, user_skb, pid);
 		if (err)
 			goto err_kfree_skbs;
 
@@ -516,8 +513,8 @@ err_kfree_skbs:
 /* Called with genl_mutex. */
 static int flush_flows(int dp_ifindex)
 {
-	struct tbl *old_table;
-	struct tbl *new_table;
+	struct flow_table *old_table;
+	struct flow_table *new_table;
 	struct datapath *dp;
 
 	dp = get_dp(dp_ifindex);
@@ -525,14 +522,13 @@ static int flush_flows(int dp_ifindex)
 		return -ENODEV;
 
 	old_table = get_table_protected(dp);
-	new_table = tbl_create(TBL_MIN_BUCKETS);
+	new_table = flow_tbl_alloc(TBL_MIN_BUCKETS);
 	if (!new_table)
 		return -ENOMEM;
 
 	rcu_assign_pointer(dp->table, new_table);
 
-	tbl_deferred_destroy(old_table, flow_free_tbl);
-
+	flow_tbl_deferred_destroy(old_table);
 	return 0;
 }
 
@@ -614,24 +610,6 @@ static void clear_stats(struct sw_flow *flow)
 	flow->byte_count = 0;
 }
 
-/* Called with genl_mutex. */
-static int expand_table(struct datapath *dp)
-{
-	struct tbl *old_table = get_table_protected(dp);
-	struct tbl *new_table;
-
-	new_table = tbl_expand(old_table);
-	if (IS_ERR(new_table)) {
-		if (PTR_ERR(new_table) != -ENOSPC)
-			return PTR_ERR(new_table);
-	} else {
-		rcu_assign_pointer(dp->table, new_table);
-		tbl_deferred_destroy(old_table, NULL);
-	}
-
- 	return 0;
-}
-
 static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
 {
 	struct ovs_header *ovs_header = info->userhdr;
@@ -692,7 +670,12 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
 	if (err)
 		goto err_flow_put;
 
-	flow->tbl_node.hash = flow_hash(&flow->key, key_len);
+	flow->hash = flow_hash(&flow->key, key_len);
+
+	if (a[OVS_PACKET_ATTR_UPCALL_PID])
+		flow->upcall_pid = nla_get_u32(a[OVS_PACKET_ATTR_UPCALL_PID]);
+	else
+		flow->upcall_pid = NETLINK_CB(skb).pid;
 
 	acts = flow_actions_alloc(a[OVS_PACKET_ATTR_ACTIONS]);
 	err = PTR_ERR(acts);
@@ -707,6 +690,11 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
 	err = -ENODEV;
 	if (!dp)
 		goto err_unlock;
+
+	if (flow->key.eth.in_port < DP_MAX_PORTS)
+		OVS_CB(packet)->vport = get_vport_protected(dp,
+							flow->key.eth.in_port);
+
 	err = execute_actions(dp, packet);
 	rcu_read_unlock();
 
@@ -727,6 +715,7 @@ static const struct nla_policy packet_policy[OVS_PACKET_ATTR_MAX + 1] = {
 	[OVS_PACKET_ATTR_PACKET] = { .type = NLA_UNSPEC },
 	[OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED },
 	[OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED },
+	[OVS_PACKET_ATTR_UPCALL_PID] = { .type = NLA_U32 },
 };
 
 static struct genl_ops dp_packet_genl_ops[] = {
@@ -740,9 +729,9 @@ static struct genl_ops dp_packet_genl_ops[] = {
 static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats)
 {
 	int i;
-	struct tbl *table = get_table_protected(dp);
+	struct flow_table *table = get_table_protected(dp);
 
-	stats->n_flows = tbl_count(table);
+	stats->n_flows = flow_tbl_count(table);
 
 	stats->n_frags = stats->n_hit = stats->n_missed = stats->n_lost = 0;
 	for_each_possible_cpu(i) {
@@ -764,54 +753,9 @@ static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats)
 	}
 }
 
-/* MTU of the dp pseudo-device: ETH_DATA_LEN or the minimum of the ports.
- * Called with RTNL lock.
- */
-int dp_min_mtu(const struct datapath *dp)
-{
-	struct vport *p;
-	int mtu = 0;
-
-	ASSERT_RTNL();
-
-	list_for_each_entry (p, &dp->port_list, node) {
-		int dev_mtu;
-
-		/* Skip any internal ports, since that's what we're trying to
-		 * set. */
-		if (is_internal_vport(p))
-			continue;
-
-		dev_mtu = vport_get_mtu(p);
-		if (!dev_mtu)
-			continue;
-		if (!mtu || dev_mtu < mtu)
-			mtu = dev_mtu;
-	}
-
-	return mtu ? mtu : ETH_DATA_LEN;
-}
-
-/* Sets the MTU of all datapath devices to the minimum of the ports
- * Called with RTNL lock.
- */
-void set_internal_devs_mtu(const struct datapath *dp)
-{
-	struct vport *p;
-	int mtu;
-
-	ASSERT_RTNL();
-
-	mtu = dp_min_mtu(dp);
-
-	list_for_each_entry (p, &dp->port_list, node) {
-		if (is_internal_vport(p))
-			vport_set_mtu(p, mtu);
-	}
-}
-
 static const struct nla_policy flow_policy[OVS_FLOW_ATTR_MAX + 1] = {
 	[OVS_FLOW_ATTR_KEY] = { .type = NLA_NESTED },
+	[OVS_FLOW_ATTR_UPCALL_PID] = { .type = NLA_U32 },
 	[OVS_FLOW_ATTR_ACTIONS] = { .type = NLA_NESTED },
 	[OVS_FLOW_ATTR_CLEAR] = { .type = NLA_FLAG },
 };
@@ -848,7 +792,7 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
 	if (!ovs_header)
 		return -EMSGSIZE;
 
-	ovs_header->dp_ifindex = dp->dp_ifindex;
+	ovs_header->dp_ifindex = get_dpifindex(dp);
 
 	nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY);
 	if (!nla)
@@ -858,6 +802,8 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
 		goto error;
 	nla_nest_end(skb, nla);
 
+	NLA_PUT_U32(skb, OVS_FLOW_ATTR_UPCALL_PID, flow->upcall_pid);
+
 	spin_lock_bh(&flow->lock);
 	used = flow->used;
 	stats.n_packets = flow->packet_count;
@@ -933,13 +879,11 @@ 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 tbl_node *flow_node;
 	struct sw_flow_key key;
 	struct sw_flow *flow;
 	struct sk_buff *reply;
 	struct datapath *dp;
-	struct tbl *table;
-	u32 hash;
+	struct flow_table *table;
 	int error;
 	int key_len;
 
@@ -966,10 +910,9 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 	if (!dp)
 		goto error;
 
-	hash = flow_hash(&key, key_len);
 	table = get_table_protected(dp);
-	flow_node = tbl_lookup(table, &key, key_len, hash, flow_cmp);
-	if (!flow_node) {
+	flow = flow_tbl_lookup(table, &key, key_len);
+	if (!flow) {
 		struct sw_flow_actions *acts;
 
 		/* Bail out if we're not allowed to create a new flow. */
@@ -978,11 +921,15 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 			goto error;
 
 		/* Expand table, if necessary, to make room. */
-		if (tbl_count(table) >= tbl_n_buckets(table)) {
-			error = expand_table(dp);
-			if (error)
-				goto error;
-			table = get_table_protected(dp);
+		if (flow_tbl_need_to_expand(table)) {
+			struct flow_table *new_table;
+
+			new_table = flow_tbl_expand(table);
+			if (!IS_ERR(new_table)) {
+				rcu_assign_pointer(dp->table, new_table);
+				flow_tbl_deferred_destroy(table);
+				table = get_table_protected(dp);
+			}
 		}
 
 		/* Allocate flow. */
@@ -994,6 +941,11 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 		flow->key = key;
 		clear_stats(flow);
 
+		if (a[OVS_FLOW_ATTR_UPCALL_PID])
+			flow->upcall_pid = nla_get_u32(a[OVS_FLOW_ATTR_UPCALL_PID]);
+		else
+			flow->upcall_pid = NETLINK_CB(skb).pid;
+
 		/* Obtain actions. */
 		acts = flow_actions_alloc(a[OVS_FLOW_ATTR_ACTIONS]);
 		error = PTR_ERR(acts);
@@ -1002,9 +954,8 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 		rcu_assign_pointer(flow->sf_acts, acts);
 
 		/* Put flow in bucket. */
-		error = tbl_insert(table, &flow->tbl_node, hash);
-		if (error)
-			goto error_free_flow;
+		flow->hash = flow_hash(&key, key_len);
+		flow_tbl_insert(table, flow);
 
 		reply = ovs_flow_cmd_build_info(flow, dp, info->snd_pid,
 						info->snd_seq, OVS_FLOW_CMD_NEW);
@@ -1024,7 +975,6 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 			goto error;
 
 		/* Update actions. */
-		flow = flow_cast(flow_node);
 		old_acts = rcu_dereference_protected(flow->sf_acts,
 						     lockdep_genl_is_held());
 		if (a[OVS_FLOW_ATTR_ACTIONS] &&
@@ -1045,6 +995,9 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 		reply = ovs_flow_cmd_build_info(flow, dp, info->snd_pid,
 						info->snd_seq, OVS_FLOW_CMD_NEW);
 
+		if (a[OVS_FLOW_ATTR_UPCALL_PID])
+			flow->upcall_pid = nla_get_u32(a[OVS_FLOW_ATTR_UPCALL_PID]);
+
 		/* Clear stats. */
 		if (a[OVS_FLOW_ATTR_CLEAR]) {
 			spin_lock_bh(&flow->lock);
@@ -1072,11 +1025,10 @@ static int ovs_flow_cmd_get(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 tbl_node *flow_node;
 	struct sk_buff *reply;
 	struct sw_flow *flow;
 	struct datapath *dp;
-	struct tbl *table;
+	struct flow_table *table;
 	int err;
 	int key_len;
 
@@ -1091,12 +1043,10 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
 		return -ENODEV;
 
 	table = get_table_protected(dp);
-	flow_node = tbl_lookup(table, &key, key_len, flow_hash(&key, key_len),
-			       flow_cmp);
-	if (!flow_node)
+	flow = flow_tbl_lookup(table, &key, key_len);
+	if (!flow)
 		return -ENOENT;
 
-	flow = flow_cast(flow_node);
 	reply = ovs_flow_cmd_build_info(flow, dp, info->snd_pid, info->snd_seq, OVS_FLOW_CMD_NEW);
 	if (IS_ERR(reply))
 		return PTR_ERR(reply);
@@ -1109,11 +1059,10 @@ static int ovs_flow_cmd_del(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 tbl_node *flow_node;
 	struct sk_buff *reply;
 	struct sw_flow *flow;
 	struct datapath *dp;
-	struct tbl *table;
+	struct flow_table *table;
 	int err;
 	int key_len;
 
@@ -1128,21 +1077,15 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
  		return -ENODEV;
 
 	table = get_table_protected(dp);
-	flow_node = tbl_lookup(table, &key, key_len, flow_hash(&key, key_len),
-			       flow_cmp);
-	if (!flow_node)
+	flow = flow_tbl_lookup(table, &key, key_len);
+	if (!flow)
 		return -ENOENT;
-	flow = flow_cast(flow_node);
 
 	reply = ovs_flow_cmd_alloc_info(flow);
 	if (!reply)
 		return -ENOMEM;
 
-	err = tbl_remove(table, flow_node);
-	if (err) {
-		kfree_skb(reply);
-		return err;
-	}
+	flow_tbl_remove(table, flow);
 
 	err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_pid,
 				     info->snd_seq, 0, OVS_FLOW_CMD_DEL);
@@ -1165,17 +1108,15 @@ static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
 		return -ENODEV;
 
 	for (;;) {
-		struct tbl_node *flow_node;
 		struct sw_flow *flow;
 		u32 bucket, obj;
 
 		bucket = cb->args[0];
 		obj = cb->args[1];
-		flow_node = tbl_next(get_table_protected(dp), &bucket, &obj);
-		if (!flow_node)
+		flow = flow_tbl_next(get_table_protected(dp), &bucket, &obj);
+		if (!flow)
 			break;
 
-		flow = flow_cast(flow_node);
 		if (ovs_flow_cmd_fill_info(flow, dp, skb, NETLINK_CB(cb->skb).pid,
 					   cb->nlh->nlmsg_seq, NLM_F_MULTI,
 					   OVS_FLOW_CMD_NEW) < 0)
@@ -1215,6 +1156,7 @@ static const struct nla_policy datapath_policy[OVS_DP_ATTR_MAX + 1] = {
 #ifdef HAVE_NLA_NUL_STRING
 	[OVS_DP_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
 #endif
+	[OVS_DP_ATTR_UPCALL_PID] = { .type = NLA_U32 },
 	[OVS_DP_ATTR_IPV4_FRAGS] = { .type = NLA_U32 },
 	[OVS_DP_ATTR_SAMPLING] = { .type = NLA_U32 },
 };
@@ -1243,7 +1185,7 @@ static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
 	if (!ovs_header)
 		goto error;
 
-	ovs_header->dp_ifindex = dp->dp_ifindex;
+	ovs_header->dp_ifindex = get_dpifindex(dp);
 
 	rcu_read_lock();
 	err = nla_put_string(skb, OVS_DP_ATTR_NAME, dp_name(dp));
@@ -1262,14 +1204,6 @@ static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
 	if (dp->sflow_probability)
 		NLA_PUT_U32(skb, OVS_DP_ATTR_SAMPLING, dp->sflow_probability);
 
-	nla = nla_nest_start(skb, OVS_DP_ATTR_MCGROUPS);
-	if (!nla)
-		goto nla_put_failure;
-	NLA_PUT_U32(skb, OVS_PACKET_CMD_MISS, packet_mc_group(dp, OVS_PACKET_CMD_MISS));
-	NLA_PUT_U32(skb, OVS_PACKET_CMD_ACTION, packet_mc_group(dp, OVS_PACKET_CMD_ACTION));
-	NLA_PUT_U32(skb, OVS_PACKET_CMD_SAMPLE, packet_mc_group(dp, OVS_PACKET_CMD_SAMPLE));
-	nla_nest_end(skb, nla);
-
 	return genlmsg_end(skb, ovs_header);
 
 nla_put_failure:
@@ -1370,35 +1304,39 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
 
 	/* Allocate table. */
 	err = -ENOMEM;
-	rcu_assign_pointer(dp->table, tbl_create(TBL_MIN_BUCKETS));
+	rcu_assign_pointer(dp->table, flow_tbl_alloc(TBL_MIN_BUCKETS));
 	if (!dp->table)
 		goto err_free_dp;
 
+	dp->drop_frags = 0;
+	dp->stats_percpu = alloc_percpu(struct dp_stats_percpu);
+	if (!dp->stats_percpu) {
+		err = -ENOMEM;
+		goto err_destroy_table;
+	}
+
+	change_datapath(dp, a);
+
 	/* Set up our datapath device. */
 	parms.name = nla_data(a[OVS_DP_ATTR_NAME]);
 	parms.type = OVS_VPORT_TYPE_INTERNAL;
 	parms.options = NULL;
 	parms.dp = dp;
 	parms.port_no = OVSP_LOCAL;
+	if (a[OVS_DP_ATTR_UPCALL_PID])
+		parms.upcall_pid = nla_get_u32(a[OVS_DP_ATTR_UPCALL_PID]);
+	else
+		parms.upcall_pid = NETLINK_CB(skb).pid;
+
 	vport = new_vport(&parms);
 	if (IS_ERR(vport)) {
 		err = PTR_ERR(vport);
 		if (err == -EBUSY)
 			err = -EEXIST;
 
-		goto err_destroy_table;
-	}
-	dp->dp_ifindex = vport_get_ifindex(vport);
-
-	dp->drop_frags = 0;
-	dp->stats_percpu = alloc_percpu(struct dp_stats_percpu);
-	if (!dp->stats_percpu) {
-		err = -ENOMEM;
-		goto err_destroy_local_port;
+		goto err_destroy_percpu;
 	}
 
-	change_datapath(dp, a);
-
 	reply = ovs_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, OVS_DP_CMD_NEW);
 	err = PTR_ERR(reply);
 	if (IS_ERR(reply))
@@ -1415,8 +1353,10 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
 
 err_destroy_local_port:
 	dp_detach_port(get_vport_protected(dp, OVSP_LOCAL));
+err_destroy_percpu:
+	free_percpu(dp->stats_percpu);
 err_destroy_table:
-	tbl_destroy(get_table_protected(dp), NULL);
+	flow_tbl_destroy(get_table_protected(dp));
 err_free_dp:
 	kfree(dp);
 err_put_module:
@@ -1576,15 +1516,15 @@ static struct genl_ops dp_datapath_genl_ops[] = {
 static const struct nla_policy vport_policy[OVS_VPORT_ATTR_MAX + 1] = {
 #ifdef HAVE_NLA_NUL_STRING
 	[OVS_VPORT_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
-	[OVS_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 },
-	[OVS_VPORT_ATTR_TYPE] = { .type = NLA_U32 },
-	[OVS_VPORT_ATTR_STATS] = { .len = sizeof(struct rtnl_link_stats64) },
+	[OVS_VPORT_ATTR_STATS] = { .len = sizeof(struct ovs_vport_stats) },
 	[OVS_VPORT_ATTR_ADDRESS] = { .len = ETH_ALEN },
 #else
-	[OVS_VPORT_ATTR_STATS] = { .minlen = sizeof(struct rtnl_link_stats64) },
+	[OVS_VPORT_ATTR_STATS] = { .minlen = sizeof(struct ovs_vport_stats) },
 	[OVS_VPORT_ATTR_ADDRESS] = { .minlen = ETH_ALEN },
 #endif
-	[OVS_VPORT_ATTR_MTU] = { .type = NLA_U32 },
+	[OVS_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 },
+	[OVS_VPORT_ATTR_TYPE] = { .type = NLA_U32 },
+	[OVS_VPORT_ATTR_UPCALL_PID] = { .type = NLA_U32 },
 	[OVS_VPORT_ATTR_OPTIONS] = { .type = NLA_NESTED },
 };
 
@@ -1607,7 +1547,6 @@ static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb,
 	struct ovs_header *ovs_header;
 	struct nlattr *nla;
 	int ifindex;
-	int mtu;
 	int err;
 
 	ovs_header = genlmsg_put(skb, pid, seq, &dp_vport_genl_family,
@@ -1615,23 +1554,20 @@ static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb,
 	if (!ovs_header)
 		return -EMSGSIZE;
 
-	ovs_header->dp_ifindex = vport->dp->dp_ifindex;
+	ovs_header->dp_ifindex = get_dpifindex(vport->dp);
 
 	NLA_PUT_U32(skb, OVS_VPORT_ATTR_PORT_NO, vport->port_no);
 	NLA_PUT_U32(skb, OVS_VPORT_ATTR_TYPE, vport_get_type(vport));
 	NLA_PUT_STRING(skb, OVS_VPORT_ATTR_NAME, vport_get_name(vport));
+	NLA_PUT_U32(skb, OVS_VPORT_ATTR_UPCALL_PID, vport->upcall_pid);
 
-	nla = nla_reserve(skb, OVS_VPORT_ATTR_STATS, sizeof(struct rtnl_link_stats64));
+	nla = nla_reserve(skb, OVS_VPORT_ATTR_STATS, sizeof(struct ovs_vport_stats));
 	if (!nla)
 		goto nla_put_failure;
-	if (vport_get_stats(vport, nla_data(nla)))
-		__skb_trim(skb, skb->len - nla->nla_len);
 
-	NLA_PUT(skb, OVS_VPORT_ATTR_ADDRESS, ETH_ALEN, vport_get_addr(vport));
+	vport_get_stats(vport, nla_data(nla));
 
-	mtu = vport_get_mtu(vport);
-	if (mtu)
-		NLA_PUT_U32(skb, OVS_VPORT_ATTR_MTU, mtu);
+	NLA_PUT(skb, OVS_VPORT_ATTR_ADDRESS, ETH_ALEN, vport_get_addr(vport));
 
 	err = vport_get_options(vport, skb);
 	if (err == -EMSGSIZE)
@@ -1708,12 +1644,13 @@ static struct vport *lookup_vport(struct ovs_header *ovs_header,
 static int change_vport(struct vport *vport, struct nlattr *a[OVS_VPORT_ATTR_MAX + 1])
 {
 	int err = 0;
+
 	if (a[OVS_VPORT_ATTR_STATS])
-		err = vport_set_stats(vport, nla_data(a[OVS_VPORT_ATTR_STATS]));
-	if (!err && a[OVS_VPORT_ATTR_ADDRESS])
+		vport_set_stats(vport, nla_data(a[OVS_VPORT_ATTR_STATS]));
+
+	if (a[OVS_VPORT_ATTR_ADDRESS])
 		err = vport_set_addr(vport, nla_data(a[OVS_VPORT_ATTR_ADDRESS]));
-	if (!err && a[OVS_VPORT_ATTR_MTU])
-		err = vport_set_mtu(vport, nla_get_u32(a[OVS_VPORT_ATTR_MTU]));
+
 	return err;
 }
 
@@ -1770,13 +1707,16 @@ static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info)
 	parms.options = a[OVS_VPORT_ATTR_OPTIONS];
 	parms.dp = dp;
 	parms.port_no = port_no;
+	if (a[OVS_VPORT_ATTR_UPCALL_PID])
+		parms.upcall_pid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
+	else
+		parms.upcall_pid = NETLINK_CB(skb).pid;
 
 	vport = new_vport(&parms);
 	err = PTR_ERR(vport);
 	if (IS_ERR(vport))
 		goto exit_unlock;
 
- 	set_internal_devs_mtu(dp);
  	dp_sysfs_add_if(vport);
 
 	err = change_vport(vport, a);
@@ -1822,6 +1762,8 @@ static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info)
 		err = vport_set_options(vport, a[OVS_VPORT_ATTR_OPTIONS]);
 	if (!err)
 		err = change_vport(vport, a);
+	if (!err && a[OVS_VPORT_ATTR_UPCALL_PID])
+		vport->upcall_pid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
 
 	reply = ovs_vport_cmd_build_info(vport, info->snd_pid, info->snd_seq,
 					 OVS_VPORT_CMD_NEW);
@@ -1869,7 +1811,7 @@ static int ovs_vport_cmd_del(struct sk_buff *skb, struct genl_info *info)
 	if (IS_ERR(reply))
 		goto exit_unlock;
 
-	err = dp_detach_port(vport);
+	dp_detach_port(vport);
 
 	genl_notify(reply, genl_info_net(info), info->snd_pid,
 		    dp_vport_multicast_group.id, info->nlhdr, GFP_KERNEL);
@@ -2023,9 +1965,6 @@ static int dp_register_genl(void)
 		}
 	}
 
-	err = packet_register_mc_groups();
-	if (err)
-		goto error;
 	return 0;
 
 error:
@@ -2042,10 +1981,14 @@ static int __init dp_init(void)
 
 	printk("Open vSwitch %s, built "__DATE__" "__TIME__"\n", VERSION BUILDNR);
 
-	err = flow_init();
+	err = tnl_init();
 	if (err)
 		goto error;
 
+	err = flow_init();
+	if (err)
+		goto error_tnl_exit;
+
 	err = vport_init();
 	if (err)
 		goto error_flow_exit;
@@ -2066,6 +2009,8 @@ error_vport_exit:
 	vport_exit();
 error_flow_exit:
 	flow_exit();
+error_tnl_exit:
+	tnl_exit();
 error:
 	return err;
 }
@@ -2077,6 +2022,7 @@ static void dp_cleanup(void)
 	unregister_netdevice_notifier(&dp_device_notifier);
 	vport_exit();
 	flow_exit();
+	tnl_exit();
 }
 
 module_init(dp_init);