datapath: correct an invalid BUG_ON
[sliver-openvswitch.git] / datapath / flow.c
index 2acdd05..2aeb704 100644 (file)
@@ -200,20 +200,18 @@ void ovs_flow_used(struct sw_flow *flow, struct sk_buff *skb)
        spin_unlock(&flow->lock);
 }
 
-struct sw_flow_actions *ovs_flow_actions_alloc(const struct nlattr *actions)
+struct sw_flow_actions *ovs_flow_actions_alloc(int size)
 {
-       int actions_len = nla_len(actions);
        struct sw_flow_actions *sfa;
 
-       if (actions_len > MAX_ACTIONS_BUFSIZE)
+       if (size > MAX_ACTIONS_BUFSIZE)
                return ERR_PTR(-EINVAL);
 
-       sfa = kmalloc(sizeof(*sfa) + actions_len, GFP_KERNEL);
+       sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
        if (!sfa)
                return ERR_PTR(-ENOMEM);
 
-       sfa->actions_len = actions_len;
-       memcpy(sfa->actions, nla_data(actions), actions_len);
+       sfa->actions_len = 0;
        return sfa;
 }
 
@@ -301,10 +299,10 @@ void ovs_flow_tbl_destroy(struct flow_table *table)
        for (i = 0; i < table->n_buckets; i++) {
                struct sw_flow *flow;
                struct hlist_head *head = flex_array_get(table->buckets, i);
-               struct hlist_node *node, *n;
+               struct hlist_node *n;
                int ver = table->node_ver;
 
-               hlist_for_each_entry_safe(flow, node, n, head, hash_node[ver]) {
+               hlist_for_each_entry_safe(flow, n, head, hash_node[ver]) {
                        hlist_del_rcu(&flow->hash_node[ver]);
                        ovs_flow_free(flow);
                }
@@ -334,7 +332,6 @@ struct sw_flow *ovs_flow_tbl_next(struct flow_table *table, u32 *bucket, u32 *la
 {
        struct sw_flow *flow;
        struct hlist_head *head;
-       struct hlist_node *n;
        int ver;
        int i;
 
@@ -342,7 +339,7 @@ struct sw_flow *ovs_flow_tbl_next(struct flow_table *table, u32 *bucket, u32 *la
        while (*bucket < table->n_buckets) {
                i = 0;
                head = flex_array_get(table->buckets, *bucket);
-               hlist_for_each_entry_rcu(flow, n, head, hash_node[ver]) {
+               hlist_for_each_entry_rcu(flow, head, hash_node[ver]) {
                        if (i < *last) {
                                i++;
                                continue;
@@ -377,11 +374,10 @@ static void flow_table_copy_flows(struct flow_table *old, struct flow_table *new
        for (i = 0; i < old->n_buckets; i++) {
                struct sw_flow *flow;
                struct hlist_head *head;
-               struct hlist_node *n;
 
                head = flex_array_get(old->buckets, i);
 
-               hlist_for_each_entry(flow, n, head, hash_node[old_ver])
+               hlist_for_each_entry(flow, head, hash_node[old_ver])
                        __flow_tbl_insert(new, flow);
        }
        old->keep_flows = true;
@@ -502,7 +498,11 @@ static __be16 parse_ethertype(struct sk_buff *skb)
                return htons(ETH_P_802_2);
 
        __skb_pull(skb, sizeof(struct llc_snap_hdr));
-       return llc->ethertype;
+
+       if (ntohs(llc->ethertype) >= 1536)
+               return llc->ethertype;
+
+       return htons(ETH_P_802_2);
 }
 
 static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
@@ -622,8 +622,9 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key,
 
        key->phy.priority = skb->priority;
        if (OVS_CB(skb)->tun_key)
-               memcpy(&key->phy.tun.tun_key, OVS_CB(skb)->tun_key, sizeof(key->phy.tun.tun_key));
+               memcpy(&key->tun_key, OVS_CB(skb)->tun_key, sizeof(key->tun_key));
        key->phy.in_port = in_port;
+       key->phy.skb_mark = skb_get_mark(skb);
 
        skb_reset_mac_header(skb);
 
@@ -784,17 +785,16 @@ static u32 ovs_flow_hash(const struct sw_flow_key *key, int key_start, int key_l
 
 static int flow_key_start(struct sw_flow_key *key)
 {
-       if (key->phy.tun.tun_key.ipv4_dst)
+       if (key->tun_key.ipv4_dst)
                return 0;
        else
-               return offsetof(struct sw_flow_key, phy.priority);
+               return offsetof(struct sw_flow_key, phy);
 }
 
 struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *table,
                                struct sw_flow_key *key, int key_len)
 {
        struct sw_flow *flow;
-       struct hlist_node *n;
        struct hlist_head *head;
        u8 *_key;
        int key_start;
@@ -805,7 +805,7 @@ struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *table,
 
        _key = (u8 *) key + key_start;
        head = find_bucket(table, hash);
-       hlist_for_each_entry_rcu(flow, n, head, hash_node[table->node_ver]) {
+       hlist_for_each_entry_rcu(flow, head, hash_node[table->node_ver]) {
 
                if (flow->hash == hash &&
                    !memcmp((u8 *)&flow->key + key_start, _key, key_len - key_start)) {
@@ -825,9 +825,9 @@ void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow,
 
 void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow)
 {
+       BUG_ON(table->count == 0);
        hlist_del_rcu(&flow->hash_node[table->node_ver]);
        table->count--;
-       BUG_ON(table->count < 0);
 }
 
 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute.  */
@@ -835,6 +835,7 @@ const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
        [OVS_KEY_ATTR_ENCAP] = -1,
        [OVS_KEY_ATTR_PRIORITY] = sizeof(u32),
        [OVS_KEY_ATTR_IN_PORT] = sizeof(u32),
+       [OVS_KEY_ATTR_SKB_MARK] = sizeof(u32),
        [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
        [OVS_KEY_ATTR_VLAN] = sizeof(__be16),
        [OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16),
@@ -846,10 +847,7 @@ const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
        [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
        [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
        [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
-       [OVS_KEY_ATTR_IPV4_TUNNEL] = sizeof(struct ovs_key_ipv4_tunnel),
-
-       /* Not upstream. */
-       [OVS_KEY_ATTR_TUN_ID] = sizeof(__be64),
+       [OVS_KEY_ATTR_TUNNEL] = -1,
 };
 
 static int ipv4_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_len,
@@ -987,6 +985,105 @@ static int parse_flow_nlattrs(const struct nlattr *attr,
        return 0;
 }
 
+int ipv4_tun_from_nlattr(const struct nlattr *attr,
+                        struct ovs_key_ipv4_tunnel *tun_key)
+{
+       struct nlattr *a;
+       int rem;
+       bool ttl = false;
+
+       memset(tun_key, 0, sizeof(*tun_key));
+
+       nla_for_each_nested(a, attr, rem) {
+               int type = nla_type(a);
+               static const u32 ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
+                       [OVS_TUNNEL_KEY_ATTR_ID] = sizeof(u64),
+                       [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = sizeof(u32),
+                       [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = sizeof(u32),
+                       [OVS_TUNNEL_KEY_ATTR_TOS] = 1,
+                       [OVS_TUNNEL_KEY_ATTR_TTL] = 1,
+                       [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = 0,
+                       [OVS_TUNNEL_KEY_ATTR_CSUM] = 0,
+               };
+
+               if (type > OVS_TUNNEL_KEY_ATTR_MAX ||
+                       ovs_tunnel_key_lens[type] != nla_len(a))
+                       return -EINVAL;
+
+               switch (type) {
+               case OVS_TUNNEL_KEY_ATTR_ID:
+                       tun_key->tun_id = nla_get_be64(a);
+                       tun_key->tun_flags |= OVS_TNL_F_KEY;
+                       break;
+               case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
+                       tun_key->ipv4_src = nla_get_be32(a);
+                       break;
+               case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
+                       tun_key->ipv4_dst = nla_get_be32(a);
+                       break;
+               case OVS_TUNNEL_KEY_ATTR_TOS:
+                       tun_key->ipv4_tos = nla_get_u8(a);
+                       break;
+               case OVS_TUNNEL_KEY_ATTR_TTL:
+                       tun_key->ipv4_ttl = nla_get_u8(a);
+                       ttl = true;
+                       break;
+               case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
+                       tun_key->tun_flags |= OVS_TNL_F_DONT_FRAGMENT;
+                       break;
+               case OVS_TUNNEL_KEY_ATTR_CSUM:
+                       tun_key->tun_flags |= OVS_TNL_F_CSUM;
+                       break;
+               default:
+                       return -EINVAL;
+
+               }
+       }
+       if (rem > 0)
+               return -EINVAL;
+
+       if (!tun_key->ipv4_dst)
+               return -EINVAL;
+
+       if (!ttl)
+               return -EINVAL;
+
+       return 0;
+}
+
+int ipv4_tun_to_nlattr(struct sk_buff *skb,
+                       const struct ovs_key_ipv4_tunnel *tun_key)
+{
+       struct nlattr *nla;
+
+       nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
+       if (!nla)
+               return -EMSGSIZE;
+
+       if (tun_key->tun_flags & OVS_TNL_F_KEY &&
+           nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, tun_key->tun_id))
+               return -EMSGSIZE;
+       if (tun_key->ipv4_src &&
+           nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC, tun_key->ipv4_src))
+               return -EMSGSIZE;
+       if (nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST, tun_key->ipv4_dst))
+               return -EMSGSIZE;
+       if (tun_key->ipv4_tos &&
+           nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, tun_key->ipv4_tos))
+               return -EMSGSIZE;
+       if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, tun_key->ipv4_ttl))
+               return -EMSGSIZE;
+       if ((tun_key->tun_flags & OVS_TNL_F_DONT_FRAGMENT) &&
+               nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
+               return -EMSGSIZE;
+       if ((tun_key->tun_flags & OVS_TNL_F_CSUM) &&
+               nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
+               return -EMSGSIZE;
+
+       nla_nest_end(skb, nla);
+       return 0;
+}
+
 /**
  * ovs_flow_from_nlattrs - parses Netlink attributes into a flow key.
  * @swkey: receives the extracted flow key.
@@ -1024,40 +1121,22 @@ int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp,
        } else {
                swkey->phy.in_port = DP_MAX_PORTS;
        }
-
-       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))
+       if (attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
+               uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20) && !defined(CONFIG_NETFILTER)
+               if (mark != 0)
                        return -EINVAL;
+#endif
+               swkey->phy.skb_mark = mark;
+               attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
+       }
 
-               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;
+       if (attrs & (1ULL << OVS_KEY_ATTR_TUNNEL)) {
+               err = ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], &swkey->tun_key);
+               if (err)
+                       return err;
 
-               memcpy(&swkey->phy.tun.tun_key, tun_key, sizeof(swkey->phy.tun.tun_key));
-               attrs &= ~(1ULL << OVS_KEY_ATTR_IPV4_TUNNEL);
+               attrs &= ~(1ULL << OVS_KEY_ATTR_TUNNEL);
        }
 
        /* Data attributes. */
@@ -1199,19 +1278,21 @@ int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp,
 
 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;
+       struct ovs_key_ipv4_tunnel *tun_key = &flow->key.tun_key;
        const struct nlattr *nla;
        int rem;
-       __be64 tun_id = 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));
+       flow->key.phy.skb_mark = 0;
+       memset(tun_key, 0, sizeof(flow->key.tun_key));
 
        nla_for_each_nested(nla, attr, rem) {
                int type = nla_type(nla);
 
                if (type <= OVS_KEY_ATTR_MAX && ovs_key_lens[type] > 0) {
+                       int err;
+
                        if (nla_len(nla) != ovs_key_lens[type])
                                return -EINVAL;
 
@@ -1220,36 +1301,10 @@ int ovs_flow_metadata_from_nlattrs(struct sw_flow *flow, int key_len, const stru
                                flow->key.phy.priority = nla_get_u32(nla);
                                break;
 
-                       case OVS_KEY_ATTR_TUN_ID:
-                               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;
+                       case OVS_KEY_ATTR_TUNNEL:
+                               err = ipv4_tun_from_nlattr(nla, tun_key);
+                               if (err)
+                                       return err;
                                break;
 
                        case OVS_KEY_ATTR_IN_PORT:
@@ -1257,6 +1312,14 @@ int ovs_flow_metadata_from_nlattrs(struct sw_flow *flow, int key_len, const stru
                                        return -EINVAL;
                                flow->key.phy.in_port = nla_get_u32(nla);
                                break;
+
+                       case OVS_KEY_ATTR_SKB_MARK:
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20) && !defined(CONFIG_NETFILTER)
+                               if (nla_get_u32(nla) != 0)
+                                       return -EINVAL;
+#endif
+                               flow->key.phy.skb_mark = nla_get_u32(nla);
+                               break;
                        }
                }
        }
@@ -1278,22 +1341,18 @@ int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb)
            nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, swkey->phy.priority))
                goto nla_put_failure;
 
-       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))
+       if (swkey->tun_key.ipv4_dst &&
+           ipv4_tun_to_nlattr(skb, &swkey->tun_key))
                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))
                goto nla_put_failure;
 
+       if (swkey->phy.skb_mark &&
+           nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, swkey->phy.skb_mark))
+               goto nla_put_failure;
+
        nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
        if (!nla)
                goto nla_put_failure;