Process RARP packets with ethertype 0x8035 similar to ARP packets.
[sliver-openvswitch.git] / datapath / flow.c
index fb4fc21..44e71e6 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2007-2011 Nicira Networks.
+ * Copyright (c) 2007-2011 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
@@ -184,8 +184,10 @@ void ovs_flow_used(struct sw_flow *flow, struct sk_buff *skb)
 {
        u8 tcp_flags = 0;
 
-       if (flow->key.eth.type == htons(ETH_P_IP) &&
-           flow->key.ip.proto == IPPROTO_TCP) {
+       if ((flow->key.eth.type == htons(ETH_P_IP) ||
+            flow->key.eth.type == htons(ETH_P_IPV6)) &&
+           flow->key.ip.proto == IPPROTO_TCP &&
+           likely(skb->len >= skb_transport_offset(skb) + sizeof(struct tcphdr))) {
                u8 *tcp = (u8 *)tcp_hdr(skb);
                tcp_flags = *(tcp + TCP_FLAGS_OFFSET) & TCP_FLAG_MASK;
        }
@@ -363,6 +365,14 @@ 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)
+{
+       struct hlist_head *head;
+       head = find_bucket(table, flow->hash);
+       hlist_add_head_rcu(&flow->hash_node[table->node_ver], head);
+       table->count++;
+}
+
 static void flow_table_copy_flows(struct flow_table *old, struct flow_table *new)
 {
        int old_ver;
@@ -380,7 +390,7 @@ 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, n, head, hash_node[old_ver])
-                       ovs_flow_tbl_insert(new, flow);
+                       __flow_tbl_insert(new, flow);
        }
        old->keep_flows = true;
 }
@@ -627,7 +637,8 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key,
        memset(key, 0, sizeof(*key));
 
        key->phy.priority = skb->priority;
-       key->phy.tun_id = OVS_CB(skb)->tun_id;
+       if (OVS_CB(skb)->tun_key)
+               memcpy(&key->phy.tun.tun_key, OVS_CB(skb)->tun_key, sizeof(key->phy.tun.tun_key));
        key->phy.in_port = in_port;
 
        skb_reset_mac_header(skb);
@@ -714,7 +725,8 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key,
                        }
                }
 
-       } else if (key->eth.type == htons(ETH_P_ARP) && arphdr_ok(skb)) {
+       } else if ((key->eth.type == htons(ETH_P_ARP) ||
+                  key->eth.type == htons(ETH_P_RARP)) && arphdr_ok(skb)) {
                struct arp_eth_header *arp;
 
                arp = (struct arp_eth_header *)skb_network_header(skb);
@@ -727,15 +739,11 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key,
                        /* We only match on the lower 8 bits of the opcode. */
                        if (ntohs(arp->ar_op) <= 0xff)
                                key->ip.proto = ntohs(arp->ar_op);
-
-                       if (key->ip.proto == ARPOP_REQUEST
-                                       || key->ip.proto == ARPOP_REPLY) {
-                               memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src));
-                               memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst));
-                               memcpy(key->ipv4.arp.sha, arp->ar_sha, ETH_ALEN);
-                               memcpy(key->ipv4.arp.tha, arp->ar_tha, ETH_ALEN);
-                               key_len = SW_FLOW_KEY_OFFSET(ipv4.arp);
-                       }
+                       memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src));
+                       memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst));
+                       memcpy(key->ipv4.arp.sha, arp->ar_sha, ETH_ALEN);
+                       memcpy(key->ipv4.arp.tha, arp->ar_tha, ETH_ALEN);
+                       key_len = SW_FLOW_KEY_OFFSET(ipv4.arp);
                }
        } else if (key->eth.type == htons(ETH_P_IPV6)) {
                int nh_len;             /* IPv6 Header + Extensions */
@@ -784,9 +792,18 @@ out:
        return error;
 }
 
-u32 ovs_flow_hash(const struct sw_flow_key *key, int key_len)
+static u32 ovs_flow_hash(const struct sw_flow_key *key, int key_start, int key_len)
+{
+       return jhash2((u32 *)((u8 *)key + key_start),
+                     DIV_ROUND_UP(key_len - key_start, sizeof(u32)), 0);
+}
+
+static int flow_key_start(struct sw_flow_key *key)
 {
-       return jhash2((u32 *)key, DIV_ROUND_UP(key_len, sizeof(u32)), 0);
+       if (key->phy.tun.tun_key.ipv4_dst)
+               return 0;
+       else
+               return offsetof(struct sw_flow_key, phy.priority);
 }
 
 struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *table,
@@ -795,28 +812,31 @@ struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *table,
        struct sw_flow *flow;
        struct hlist_node *n;
        struct hlist_head *head;
+       u8 *_key;
+       int key_start;
        u32 hash;
 
-       hash = ovs_flow_hash(key, key_len);
+       key_start = flow_key_start(key);
+       hash = ovs_flow_hash(key, key_start, key_len);
 
+       _key = (u8 *) key + key_start;
        head = find_bucket(table, hash);
        hlist_for_each_entry_rcu(flow, n, head, hash_node[table->node_ver]) {
 
                if (flow->hash == hash &&
-                   !memcmp(&flow->key, key, key_len)) {
+                   !memcmp((u8 *)&flow->key + key_start, _key, key_len - key_start)) {
                        return flow;
                }
        }
        return NULL;
 }
 
-void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow)
+void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow,
+                        struct sw_flow_key *key, int key_len)
 {
-       struct hlist_head *head;
-
-       head = find_bucket(table, flow->hash);
-       hlist_add_head_rcu(&flow->hash_node[table->node_ver], head);
-       table->count++;
+       flow->hash = ovs_flow_hash(key, flow_key_start(key), key_len);
+       memcpy(&flow->key, key, sizeof(flow->key));
+       __flow_tbl_insert(table, flow);
 }
 
 void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow)
@@ -845,6 +865,7 @@ const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
 
        /* Not upstream. */
        [OVS_KEY_ATTR_TUN_ID] = sizeof(__be64),
+       [OVS_KEY_ATTR_IPV4_TUNNEL] = sizeof(struct ovs_key_ipv4_tunnel),
 };
 
 static int ipv4_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_len,
@@ -1020,9 +1041,39 @@ int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp,
                swkey->phy.in_port = DP_MAX_PORTS;
        }
 
-       if (attrs & (1ULL << OVS_KEY_ATTR_TUN_ID)) {
-               swkey->phy.tun_id = nla_get_be64(a[OVS_KEY_ATTR_TUN_ID]);
+       if (attrs & (1ULL << OVS_KEY_ATTR_TUN_ID) &&
+           attrs & (1ULL << OVS_KEY_ATTR_IPV4_TUNNEL)) {
+               struct ovs_key_ipv4_tunnel *tun_key;
+               __be64 tun_id;
+
+               tun_key = nla_data(a[OVS_KEY_ATTR_IPV4_TUNNEL]);
+
+               if (!tun_key->ipv4_dst)
+                       return -EINVAL;
+               if (!(tun_key->tun_flags & OVS_FLOW_TNL_F_KEY))
+                       return -EINVAL;
+
+               tun_id = nla_get_be64(a[OVS_KEY_ATTR_TUN_ID]);
+               if (tun_id != tun_key->tun_id)
+                       return -EINVAL;
+
+               memcpy(&swkey->phy.tun.tun_key, tun_key, sizeof(swkey->phy.tun.tun_key));
+               attrs &= ~(1ULL << OVS_KEY_ATTR_TUN_ID);
+               attrs &= ~(1ULL << OVS_KEY_ATTR_IPV4_TUNNEL);
+       } else if (attrs & (1ULL << OVS_KEY_ATTR_TUN_ID)) {
+               swkey->phy.tun.tun_key.tun_id = nla_get_be64(a[OVS_KEY_ATTR_TUN_ID]);
+               swkey->phy.tun.tun_key.tun_flags |= OVS_FLOW_TNL_F_KEY;
+
                attrs &= ~(1ULL << OVS_KEY_ATTR_TUN_ID);
+       } else if (attrs & (1ULL << OVS_KEY_ATTR_IPV4_TUNNEL)) {
+               struct ovs_key_ipv4_tunnel *tun_key;
+               tun_key = nla_data(a[OVS_KEY_ATTR_IPV4_TUNNEL]);
+
+               if (!tun_key->ipv4_dst)
+                       return -EINVAL;
+
+               memcpy(&swkey->phy.tun.tun_key, tun_key, sizeof(swkey->phy.tun.tun_key));
+               attrs &= ~(1ULL << OVS_KEY_ATTR_IPV4_TUNNEL);
        }
 
        /* Data attributes. */
@@ -1123,7 +1174,8 @@ int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp,
                        if (err)
                                return err;
                }
-       } else if (swkey->eth.type == htons(ETH_P_ARP)) {
+       } else if (swkey->eth.type == htons(ETH_P_ARP) ||
+                  swkey->eth.type == htons(ETH_P_RARP)) {
                const struct ovs_key_arp *arp_key;
 
                if (!(attrs & (1 << OVS_KEY_ATTR_ARP)))
@@ -1160,15 +1212,17 @@ int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp,
  * get the metadata, that is, the parts of the flow key that cannot be
  * extracted from the packet itself.
  */
-int ovs_flow_metadata_from_nlattrs(u32 *priority, u16 *in_port, __be64 *tun_id,
-                                  const struct nlattr *attr)
+
+int ovs_flow_metadata_from_nlattrs(struct sw_flow *flow, int key_len, const struct nlattr *attr)
 {
+       struct ovs_key_ipv4_tunnel *tun_key = &flow->key.phy.tun.tun_key;
        const struct nlattr *nla;
        int rem;
+       __be64 tun_id = 0;
 
-       *in_port = DP_MAX_PORTS;
-       *tun_id = 0;
-       *priority = 0;
+       flow->key.phy.in_port = DP_MAX_PORTS;
+       flow->key.phy.priority = 0;
+       memset(tun_key, 0, sizeof(flow->key.phy.tun.tun_key));
 
        nla_for_each_nested(nla, attr, rem) {
                int type = nla_type(nla);
@@ -1179,23 +1233,55 @@ int ovs_flow_metadata_from_nlattrs(u32 *priority, u16 *in_port, __be64 *tun_id,
 
                        switch (type) {
                        case OVS_KEY_ATTR_PRIORITY:
-                               *priority = nla_get_u32(nla);
+                               flow->key.phy.priority = nla_get_u32(nla);
                                break;
 
                        case OVS_KEY_ATTR_TUN_ID:
-                               *tun_id = nla_get_be64(nla);
+                               tun_id = nla_get_be64(nla);
+
+                               if (tun_key->ipv4_dst) {
+                                       if (!(tun_key->tun_flags & OVS_FLOW_TNL_F_KEY))
+                                               return -EINVAL;
+                                       if (tun_key->tun_id != tun_id)
+                                               return -EINVAL;
+                                       break;
+                               }
+                               tun_key->tun_id = tun_id;
+                               tun_key->tun_flags |= OVS_FLOW_TNL_F_KEY;
+
+                               break;
+
+                       case OVS_KEY_ATTR_IPV4_TUNNEL:
+                               if (tun_key->tun_flags & OVS_FLOW_TNL_F_KEY) {
+                                       tun_id = tun_key->tun_id;
+
+                                       memcpy(tun_key, nla_data(nla), sizeof(*tun_key));
+                                       if (!(tun_key->tun_flags & OVS_FLOW_TNL_F_KEY))
+                                               return -EINVAL;
+
+                                       if (tun_key->tun_id != tun_id)
+                                               return -EINVAL;
+                               } else
+                                       memcpy(tun_key, nla_data(nla), sizeof(*tun_key));
+
+                               if (!tun_key->ipv4_dst)
+                                       return -EINVAL;
                                break;
 
                        case OVS_KEY_ATTR_IN_PORT:
                                if (nla_get_u32(nla) >= DP_MAX_PORTS)
                                        return -EINVAL;
-                               *in_port = nla_get_u32(nla);
+                               flow->key.phy.in_port = nla_get_u32(nla);
                                break;
                        }
                }
        }
        if (rem)
                return -EINVAL;
+
+       flow->hash = ovs_flow_hash(&flow->key,
+                                  flow_key_start(&flow->key), key_len);
+
        return 0;
 }
 
@@ -1204,14 +1290,25 @@ int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb)
        struct ovs_key_ethernet *eth_key;
        struct nlattr *nla, *encap;
 
-       if (swkey->phy.priority)
-               NLA_PUT_U32(skb, OVS_KEY_ATTR_PRIORITY, swkey->phy.priority);
+       if (swkey->phy.priority &&
+           nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, swkey->phy.priority))
+               goto nla_put_failure;
 
-       if (swkey->phy.tun_id != cpu_to_be64(0))
-               NLA_PUT_BE64(skb, OVS_KEY_ATTR_TUN_ID, swkey->phy.tun_id);
+       if (swkey->phy.tun.tun_key.ipv4_dst) {
+               struct ovs_key_ipv4_tunnel *tun_key;
+               nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4_TUNNEL, sizeof(*tun_key));
+               if (!nla)
+                       goto nla_put_failure;
+               tun_key = nla_data(nla);
+               memcpy(tun_key, &swkey->phy.tun.tun_key, sizeof(*tun_key));
+       }
+       if ((swkey->phy.tun.tun_key.tun_flags & OVS_FLOW_TNL_F_KEY) &&
+           nla_put_be64(skb, OVS_KEY_ATTR_TUN_ID, swkey->phy.tun.tun_key.tun_id))
+               goto nla_put_failure;
 
-       if (swkey->phy.in_port != DP_MAX_PORTS)
-               NLA_PUT_U32(skb, OVS_KEY_ATTR_IN_PORT, swkey->phy.in_port);
+       if (swkey->phy.in_port != DP_MAX_PORTS &&
+           nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, swkey->phy.in_port))
+               goto nla_put_failure;
 
        nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
        if (!nla)
@@ -1221,8 +1318,9 @@ int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb)
        memcpy(eth_key->eth_dst, swkey->eth.dst, ETH_ALEN);
 
        if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
-               NLA_PUT_BE16(skb, OVS_KEY_ATTR_ETHERTYPE, htons(ETH_P_8021Q));
-               NLA_PUT_BE16(skb, OVS_KEY_ATTR_VLAN, swkey->eth.tci);
+               if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, htons(ETH_P_8021Q)) ||
+                   nla_put_be16(skb, OVS_KEY_ATTR_VLAN, swkey->eth.tci))
+                       goto nla_put_failure;
                encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
                if (!swkey->eth.tci)
                        goto unencap;
@@ -1233,7 +1331,8 @@ int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb)
        if (swkey->eth.type == htons(ETH_P_802_2))
                goto unencap;
 
-       NLA_PUT_BE16(skb, OVS_KEY_ATTR_ETHERTYPE, swkey->eth.type);
+       if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, swkey->eth.type))
+               goto nla_put_failure;
 
        if (swkey->eth.type == htons(ETH_P_IP)) {
                struct ovs_key_ipv4 *ipv4_key;
@@ -1264,7 +1363,8 @@ int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb)
                ipv6_key->ipv6_tclass = swkey->ip.tos;
                ipv6_key->ipv6_hlimit = swkey->ip.ttl;
                ipv6_key->ipv6_frag = swkey->ip.frag;
-       } else if (swkey->eth.type == htons(ETH_P_ARP)) {
+       } else if (swkey->eth.type == htons(ETH_P_ARP) ||
+                  swkey->eth.type == htons(ETH_P_RARP)) {
                struct ovs_key_arp *arp_key;
 
                nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));