meta-flow: Correctly set destination MAC in mf_set_flow_value().
[sliver-openvswitch.git] / datapath / tunnel.c
index 9bb9cef..33d2fe9 100644 (file)
@@ -1,9 +1,19 @@
 /*
- * Copyright (c) 2010, 2011 Nicira Networks.
- * Distributed under the terms of the GNU GPL version 2.
+ * Copyright (c) 2007-2011 Nicira Networks.
  *
- * Significant portions of this file may be copied from parts of the Linux
- * kernel, by Linus Torvalds and others.
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of version 2 of the GNU General Public
+ * License as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA
  */
 
 #include <linux/if_arp.h>
@@ -32,7 +42,6 @@
 #include <net/route.h>
 #include <net/xfrm.h>
 
-#include "actions.h"
 #include "checksum.h"
 #include "datapath.h"
 #include "tunnel.h"
@@ -86,8 +95,10 @@ static DECLARE_DELAYED_WORK(cache_cleaner_wq, cache_cleaner);
  */
 static unsigned int key_local_remote_ports __read_mostly;
 static unsigned int key_remote_ports __read_mostly;
+static unsigned int key_multicast_ports __read_mostly;
 static unsigned int local_remote_ports __read_mostly;
 static unsigned int remote_ports __read_mostly;
+static unsigned int multicast_ports __read_mostly;
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
 #define rt_dst(rt) (rt->dst)
@@ -132,7 +143,7 @@ static void free_cache(struct tnl_cache *cache)
        if (!cache)
                return;
 
-       flow_put(cache->flow);
+       ovs_flow_put(cache->flow);
        ip_rt_put(cache->rt);
        kfree(cache);
 }
@@ -191,14 +202,20 @@ static void assign_cache_rcu(struct vport *vport, struct tnl_cache *new_cache)
 
 static unsigned int *find_port_pool(const struct tnl_mutable_config *mutable)
 {
+       bool is_multicast = ipv4_is_multicast(mutable->key.daddr);
+
        if (mutable->flags & TNL_F_IN_KEY_MATCH) {
                if (mutable->key.saddr)
                        return &local_remote_ports;
+               else if (is_multicast)
+                       return &multicast_ports;
                else
                        return &remote_ports;
        } else {
                if (mutable->key.saddr)
                        return &key_local_remote_ports;
+               else if (is_multicast)
+                       return &key_multicast_ports;
                else
                        return &key_remote_ports;
        }
@@ -282,32 +299,13 @@ static struct vport *port_table_lookup(struct port_lookup_key *key,
        return NULL;
 }
 
-struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
-                           int tunnel_type,
-                           const struct tnl_mutable_config **mutable)
+struct vport *ovs_tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
+                               int tunnel_type,
+                               const struct tnl_mutable_config **mutable)
 {
        struct port_lookup_key lookup;
        struct vport *vport;
-
-       if (ipv4_is_multicast(saddr)) {
-               lookup.saddr = 0;
-               lookup.daddr = saddr;
-               if (key_remote_ports) {
-                       lookup.tunnel_type = tunnel_type | TNL_T_KEY_EXACT;
-                       lookup.in_key = key;
-                       vport = port_table_lookup(&lookup, mutable);
-                       if (vport)
-                               return vport;
-               }
-               if (remote_ports) {
-                       lookup.tunnel_type = tunnel_type | TNL_T_KEY_MATCH;
-                       lookup.in_key = 0;
-                       vport = port_table_lookup(&lookup, mutable);
-                       if (vport)
-                               return vport;
-               }
-               return NULL;
-       }
+       bool is_multicast = ipv4_is_multicast(saddr);
 
        lookup.saddr = saddr;
        lookup.daddr = daddr;
@@ -315,7 +313,7 @@ struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
        /* First try for exact match on in_key. */
        lookup.in_key = key;
        lookup.tunnel_type = tunnel_type | TNL_T_KEY_EXACT;
-       if (key_local_remote_ports) {
+       if (!is_multicast && key_local_remote_ports) {
                vport = port_table_lookup(&lookup, mutable);
                if (vport)
                        return vport;
@@ -332,7 +330,7 @@ struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
        /* Then try matches that wildcard in_key. */
        lookup.in_key = 0;
        lookup.tunnel_type = tunnel_type | TNL_T_KEY_MATCH;
-       if (local_remote_ports) {
+       if (!is_multicast && local_remote_ports) {
                vport = port_table_lookup(&lookup, mutable);
                if (vport)
                        return vport;
@@ -344,6 +342,25 @@ struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
                        return vport;
        }
 
+       if (is_multicast) {
+               lookup.saddr = 0;
+               lookup.daddr = saddr;
+               if (key_multicast_ports) {
+                       lookup.tunnel_type = tunnel_type | TNL_T_KEY_EXACT;
+                       lookup.in_key = key;
+                       vport = port_table_lookup(&lookup, mutable);
+                       if (vport)
+                               return vport;
+               }
+               if (multicast_ports) {
+                       lookup.tunnel_type = tunnel_type | TNL_T_KEY_MATCH;
+                       lookup.in_key = 0;
+                       vport = port_table_lookup(&lookup, mutable);
+                       if (vport)
+                               return vport;
+               }
+       }
+
        return NULL;
 }
 
@@ -382,7 +399,7 @@ static void ecn_decapsulate(struct sk_buff *skb, u8 tos)
 }
 
 /**
- *     tnl_rcv - ingress point for generic tunnel code
+ *     ovs_tnl_rcv - ingress point for generic tunnel code
  *
  * @vport: port this packet was received on
  * @skb: received packet
@@ -396,7 +413,7 @@ static void ecn_decapsulate(struct sk_buff *skb, u8 tos)
  * - skb->csum does not include the inner Ethernet header.
  * - The layer pointers are undefined.
  */
-void tnl_rcv(struct vport *vport, struct sk_buff *skb, u8 tos)
+void ovs_tnl_rcv(struct vport *vport, struct sk_buff *skb, u8 tos)
 {
        struct ethhdr *eh;
 
@@ -421,7 +438,7 @@ void tnl_rcv(struct vport *vport, struct sk_buff *skb, u8 tos)
                return;
        }
 
-       vport_receive(vport, skb);
+       ovs_vport_receive(vport, skb);
 }
 
 static bool check_ipv4_address(__be32 addr)
@@ -521,6 +538,7 @@ static bool ipv6_should_icmp(struct sk_buff *skb)
        int addr_type;
        int payload_off = (u8 *)(old_ipv6h + 1) - skb->data;
        u8 nexthdr = ipv6_hdr(skb)->nexthdr;
+       __be16 frag_off;
 
        /* Check source address is valid. */
        addr_type = ipv6_addr_type(&old_ipv6h->saddr);
@@ -532,7 +550,7 @@ static bool ipv6_should_icmp(struct sk_buff *skb)
                return false;
 
        /* Don't respond to ICMP error messages. */
-       payload_off = ipv6_skip_exthdr(skb, payload_off, &nexthdr);
+       payload_off = ipv6_skip_exthdr(skb, payload_off, &nexthdr, &frag_off);
        if (payload_off < 0)
                return false;
 
@@ -570,8 +588,8 @@ static void ipv6_build_icmp(struct sk_buff *skb, struct sk_buff *nskb,
                                              + payload_length);
        ipv6h->nexthdr          =       NEXTHDR_ICMP;
        ipv6h->hop_limit        =       IPV6_DEFAULT_HOPLIMIT;
-       ipv6_addr_copy(&ipv6h->daddr, &old_ipv6h->saddr);
-       ipv6_addr_copy(&ipv6h->saddr, &old_ipv6h->daddr);
+       ipv6h->daddr            =       old_ipv6h->saddr;
+       ipv6h->saddr            =       old_ipv6h->daddr;
 
        /* ICMPv6 */
        icmp6h->icmp6_type      =       ICMPV6_PKT_TOOBIG;
@@ -590,9 +608,9 @@ static void ipv6_build_icmp(struct sk_buff *skb, struct sk_buff *nskb,
 }
 #endif /* IPv6 */
 
-bool tnl_frag_needed(struct vport *vport,
-                    const struct tnl_mutable_config *mutable,
-                    struct sk_buff *skb, unsigned int mtu, __be64 flow_key)
+bool ovs_tnl_frag_needed(struct vport *vport,
+                        const struct tnl_mutable_config *mutable,
+                        struct sk_buff *skb, unsigned int mtu, __be64 flow_key)
 {
        unsigned int eth_hdr_len = ETH_HLEN;
        unsigned int total_length = 0, header_length = 0, payload_length;
@@ -692,7 +710,7 @@ bool tnl_frag_needed(struct vport *vport,
                return false;
        }
 
-       vport_receive(vport, nskb);
+       ovs_vport_receive(vport, nskb);
 
        return true;
 }
@@ -739,8 +757,8 @@ static bool check_mtu(struct sk_buff *skb,
                        mtu = max(mtu, IP_MIN_MTU);
 
                        if (packet_length > mtu &&
-                           tnl_frag_needed(vport, mutable, skb, mtu,
-                                           OVS_CB(skb)->tun_id))
+                           ovs_tnl_frag_needed(vport, mutable, skb, mtu,
+                                               OVS_CB(skb)->tun_id))
                                return false;
                }
        }
@@ -756,8 +774,8 @@ static bool check_mtu(struct sk_buff *skb,
                        mtu = max(mtu, IPV6_MIN_MTU);
 
                        if (packet_length > mtu &&
-                           tnl_frag_needed(vport, mutable, skb, mtu,
-                                           OVS_CB(skb)->tun_id))
+                           ovs_tnl_frag_needed(vport, mutable, skb, mtu,
+                                               OVS_CB(skb)->tun_id))
                                return false;
                }
        }
@@ -813,7 +831,7 @@ static bool check_cache_valid(const struct tnl_cache *cache,
                hh->hh_lock.sequence == cache->hh_seq &&
 #endif
                mutable->seq == cache->mutable_seq &&
-               (!is_internal_dev(rt_dst(cache->rt).dev) ||
+               (!ovs_is_internal_dev(rt_dst(cache->rt).dev) ||
                (cache->flow && !cache->flow->dead));
 }
 
@@ -928,7 +946,7 @@ static struct tnl_cache *build_cache(struct vport *vport,
        cache->expiration = jiffies + tnl_vport->cache_exp_interval;
 #endif
 
-       if (is_internal_dev(rt_dst(rt).dev)) {
+       if (ovs_is_internal_dev(rt_dst(rt).dev)) {
                struct sw_flow_key flow_key;
                struct vport *dst_vport;
                struct sk_buff *skb;
@@ -936,7 +954,7 @@ static struct tnl_cache *build_cache(struct vport *vport,
                int flow_key_len;
                struct sw_flow *flow;
 
-               dst_vport = internal_dev_get_vport(rt_dst(rt).dev);
+               dst_vport = ovs_internal_dev_get_vport(rt_dst(rt).dev);
                if (!dst_vport)
                        goto done;
 
@@ -947,18 +965,18 @@ static struct tnl_cache *build_cache(struct vport *vport,
                __skb_put(skb, cache->len);
                memcpy(skb->data, get_cached_header(cache), cache->len);
 
-               err = flow_extract(skb, dst_vport->port_no, &flow_key,
-                                  &flow_key_len);
+               err = ovs_flow_extract(skb, dst_vport->port_no, &flow_key,
+                                      &flow_key_len);
 
                consume_skb(skb);
                if (err)
                        goto done;
 
-               flow = flow_tbl_lookup(rcu_dereference(dst_vport->dp->table),
-                                        &flow_key, flow_key_len);
+               flow = ovs_flow_tbl_lookup(rcu_dereference(dst_vport->dp->table),
+                                          &flow_key, flow_key_len);
                if (flow) {
                        cache->flow = flow;
-                       flow_hold(flow);
+                       ovs_flow_hold(flow);
                }
        }
 
@@ -1037,7 +1055,7 @@ static bool need_linearize(const struct sk_buff *skb)
         * change them from underneath us and we can skip the linearization.
         */
        for (i = 0; i < skb_shinfo(skb)->nr_frags; i++)
-               if (unlikely(page_count(skb_shinfo(skb)->frags[i].page) > 1))
+               if (unlikely(page_count(skb_frag_page(&skb_shinfo(skb)->frags[i])) > 1))
                        return true;
 
        return false;
@@ -1134,11 +1152,11 @@ free_frags:
         * dropped so just free the rest.  This may help improve the congestion
         * that caused the first packet to be dropped.
         */
-       tnl_free_linked_skbs(skb);
+       ovs_tnl_free_linked_skbs(skb);
        return sent_len;
 }
 
-int tnl_send(struct vport *vport, struct sk_buff *skb)
+int ovs_tnl_send(struct vport *vport, struct sk_buff *skb)
 {
        struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
        const struct tnl_mutable_config *mutable = rcu_dereference(tnl_vport->mutable);
@@ -1285,7 +1303,7 @@ int tnl_send(struct vport *vport, struct sk_buff *skb)
                        int orig_len = skb->len - cache->len;
                        struct vport *cache_vport;
 
-                       cache_vport = internal_dev_get_vport(rt_dst(rt).dev);
+                       cache_vport = ovs_internal_dev_get_vport(rt_dst(rt).dev);
                        skb->protocol = htons(ETH_P_IP);
                        iph = ip_hdr(skb);
                        iph->tot_len = htons(skb->len - skb_network_offset(skb));
@@ -1298,7 +1316,7 @@ int tnl_send(struct vport *vport, struct sk_buff *skb)
                                }
 
                                OVS_CB(skb)->flow = cache->flow;
-                               vport_receive(cache_vport, skb);
+                               ovs_vport_receive(cache_vport, skb);
                                sent_len += orig_len;
                        } else {
                                int xmit_err;
@@ -1317,14 +1335,14 @@ next:
        }
 
        if (unlikely(sent_len == 0))
-               vport_record_error(vport, VPORT_E_TX_DROPPED);
+               ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
 
        goto out;
 
 error_free:
-       tnl_free_linked_skbs(skb);
+       ovs_tnl_free_linked_skbs(skb);
 error:
-       vport_record_error(vport, err);
+       ovs_vport_record_error(vport, err);
 out:
        dst_release(unattached_dst);
        return sent_len;
@@ -1422,9 +1440,9 @@ static int tnl_set_config(struct nlattr *options, const struct tnl_ops *tnl_ops,
        return 0;
 }
 
-struct vport *tnl_create(const struct vport_parms *parms,
-                        const struct vport_ops *vport_ops,
-                        const struct tnl_ops *tnl_ops)
+struct vport *ovs_tnl_create(const struct vport_parms *parms,
+                            const struct vport_ops *vport_ops,
+                            const struct tnl_ops *tnl_ops)
 {
        struct vport *vport;
        struct tnl_vport *tnl_vport;
@@ -1432,7 +1450,7 @@ struct vport *tnl_create(const struct vport_parms *parms,
        int initial_frag_id;
        int err;
 
-       vport = vport_alloc(sizeof(struct tnl_vport), vport_ops, parms);
+       vport = ovs_vport_alloc(sizeof(struct tnl_vport), vport_ops, parms);
        if (IS_ERR(vport)) {
                err = PTR_ERR(vport);
                goto error;
@@ -1449,7 +1467,7 @@ struct vport *tnl_create(const struct vport_parms *parms,
                goto error_free_vport;
        }
 
-       vport_gen_rand_ether_addr(mutable->eth_addr);
+       random_ether_addr(mutable->eth_addr);
 
        get_random_bytes(&initial_frag_id, sizeof(int));
        atomic_set(&tnl_vport->frag_id, initial_frag_id);
@@ -1474,12 +1492,12 @@ error_free_mutable:
        free_mutable_rtnl(mutable);
        kfree(mutable);
 error_free_vport:
-       vport_free(vport);
+       ovs_vport_free(vport);
 error:
        return ERR_PTR(err);
 }
 
-int tnl_set_options(struct vport *vport, struct nlattr *options)
+int ovs_tnl_set_options(struct vport *vport, struct nlattr *options)
 {
        struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
        const struct tnl_mutable_config *old_mutable;
@@ -1516,7 +1534,7 @@ error:
        return err;
 }
 
-int tnl_get_options(const struct vport *vport, struct sk_buff *skb)
+int ovs_tnl_get_options(const struct vport *vport, struct sk_buff *skb)
 {
        const struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
        const struct tnl_mutable_config *mutable = rcu_dereference_rtnl(tnl_vport->mutable);
@@ -1548,10 +1566,10 @@ static void free_port_rcu(struct rcu_head *rcu)
 
        free_cache((struct tnl_cache __force *)tnl_vport->cache);
        kfree((struct tnl_mutable __force *)tnl_vport->mutable);
-       vport_free(tnl_vport_to_vport(tnl_vport));
+       ovs_vport_free(tnl_vport_to_vport(tnl_vport));
 }
 
-void tnl_destroy(struct vport *vport)
+void ovs_tnl_destroy(struct vport *vport)
 {
        struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
        struct tnl_mutable_config *mutable;
@@ -1562,7 +1580,7 @@ void tnl_destroy(struct vport *vport)
        call_rcu(&tnl_vport->rcu, free_port_rcu);
 }
 
-int tnl_set_addr(struct vport *vport, const unsigned char *addr)
+int ovs_tnl_set_addr(struct vport *vport, const unsigned char *addr)
 {
        struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
        struct tnl_mutable_config *old_mutable, *mutable;
@@ -1580,19 +1598,19 @@ int tnl_set_addr(struct vport *vport, const unsigned char *addr)
        return 0;
 }
 
-const char *tnl_get_name(const struct vport *vport)
+const char *ovs_tnl_get_name(const struct vport *vport)
 {
        const struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
        return tnl_vport->name;
 }
 
-const unsigned char *tnl_get_addr(const struct vport *vport)
+const unsigned char *ovs_tnl_get_addr(const struct vport *vport)
 {
        const struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
        return rcu_dereference_rtnl(tnl_vport->mutable)->eth_addr;
 }
 
-void tnl_free_linked_skbs(struct sk_buff *skb)
+void ovs_tnl_free_linked_skbs(struct sk_buff *skb)
 {
        while (skb) {
                struct sk_buff *next = skb->next;
@@ -1601,7 +1619,7 @@ void tnl_free_linked_skbs(struct sk_buff *skb)
        }
 }
 
-int tnl_init(void)
+int ovs_tnl_init(void)
 {
        int i;
 
@@ -1616,7 +1634,7 @@ int tnl_init(void)
        return 0;
 }
 
-void tnl_exit(void)
+void ovs_tnl_exit(void)
 {
        int i;