datapath: Add multicast tunnel support.
[sliver-openvswitch.git] / datapath / tunnel.c
index f913849..372d90e 100644 (file)
 #include <linux/if_ether.h>
 #include <linux/ip.h>
 #include <linux/if_vlan.h>
+#include <linux/igmp.h>
 #include <linux/in.h>
 #include <linux/in_route.h>
+#include <linux/inetdevice.h>
 #include <linux/jhash.h>
 #include <linux/list.h>
 #include <linux/kernel.h>
@@ -147,6 +149,21 @@ static void free_cache_rcu(struct rcu_head *rcu)
        free_cache(c);
 }
 
+/* Frees the portion of 'mutable' that requires RTNL and thus can't happen
+ * within an RCU callback.  Fortunately this part doesn't require waiting for
+ * an RCU grace period.
+ */
+static void free_mutable_rtnl(struct tnl_mutable_config *mutable)
+{
+       ASSERT_RTNL();
+       if (ipv4_is_multicast(mutable->key.daddr) && mutable->mlink) {
+               struct in_device *in_dev;
+               in_dev = inetdev_by_index(&init_net, mutable->mlink);
+               if (in_dev)
+                       ip_mc_dec_group(in_dev, mutable->key.daddr);
+       }
+}
+
 static void assign_config_rcu(struct vport *vport,
                              struct tnl_mutable_config *new_config)
 {
@@ -155,6 +172,8 @@ static void assign_config_rcu(struct vport *vport,
 
        old_config = rtnl_dereference(tnl_vport->mutable);
        rcu_assign_pointer(tnl_vport->mutable, new_config);
+
+       free_mutable_rtnl(old_config);
        call_rcu(&old_config->rcu, free_config_rcu);
 }
 
@@ -270,6 +289,26 @@ struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
        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;
+       }
+
        lookup.saddr = saddr;
        lookup.daddr = daddr;
 
@@ -932,6 +971,31 @@ unlock:
        return cache;
 }
 
+static struct rtable *__find_route(const struct tnl_mutable_config *mutable,
+                                  u8 ipproto, u8 tos)
+{
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39)
+       struct flowi fl = { .nl_u = { .ip4_u =
+                                     { .daddr = mutable->key.daddr,
+                                       .saddr = mutable->key.saddr,
+                                       .tos = tos } },
+                           .proto = ipproto };
+       struct rtable *rt;
+
+       if (unlikely(ip_route_output_key(&init_net, &rt, &fl)))
+               return ERR_PTR(-EADDRNOTAVAIL);
+
+       return rt;
+#else
+       struct flowi4 fl = { .daddr = mutable->key.daddr,
+                            .saddr = mutable->key.saddr,
+                            .flowi4_tos = tos,
+                            .flowi4_proto = ipproto };
+
+       return ip_route_output_key(&init_net, &fl);
+#endif
+}
+
 static struct rtable *find_route(struct vport *vport,
                                 const struct tnl_mutable_config *mutable,
                                 u8 tos, struct tnl_cache **cache)
@@ -947,25 +1011,10 @@ static struct rtable *find_route(struct vport *vport,
                return cur_cache->rt;
        } else {
                struct rtable *rt;
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39)
-               struct flowi fl = { .nl_u = { .ip4_u =
-                                             { .daddr = mutable->key.daddr,
-                                               .saddr = mutable->key.saddr,
-                                               .tos = tos } },
-                                   .proto = tnl_vport->tnl_ops->ipproto };
-
-               if (unlikely(ip_route_output_key(&init_net, &rt, &fl)))
-                       return NULL;
-#else
-               struct flowi4 fl = { .daddr = mutable->key.daddr,
-                                    .saddr = mutable->key.saddr,
-                                    .flowi4_tos = tos,
-                                    .flowi4_proto = tnl_vport->tnl_ops->ipproto };
 
-               rt = ip_route_output_key(&init_net, &fl);
+               rt = __find_route(mutable, tnl_vport->tnl_ops->ipproto, tos);
                if (IS_ERR(rt))
                        return NULL;
-#endif
 
                if (likely(tos == mutable->tos))
                        *cache = build_cache(vport, mutable, rt);
@@ -1310,9 +1359,12 @@ static int tnl_set_config(struct nlattr *options, const struct tnl_ops *tnl_ops,
 
        mutable->flags = nla_get_u32(a[OVS_TUNNEL_ATTR_FLAGS]) & TNL_F_PUBLIC;
 
-       if (a[OVS_TUNNEL_ATTR_SRC_IPV4])
-               mutable->key.saddr = nla_get_be32(a[OVS_TUNNEL_ATTR_SRC_IPV4]);
        mutable->key.daddr = nla_get_be32(a[OVS_TUNNEL_ATTR_DST_IPV4]);
+       if (a[OVS_TUNNEL_ATTR_SRC_IPV4]) {
+               if (ipv4_is_multicast(mutable->key.daddr))
+                       return -EINVAL;
+               mutable->key.saddr = nla_get_be32(a[OVS_TUNNEL_ATTR_SRC_IPV4]);
+       }
 
        if (a[OVS_TUNNEL_ATTR_TOS]) {
                mutable->tos = nla_get_u8(a[OVS_TUNNEL_ATTR_TOS]);
@@ -1347,6 +1399,22 @@ static int tnl_set_config(struct nlattr *options, const struct tnl_ops *tnl_ops,
        if (old_vport && old_vport != cur_vport)
                return -EEXIST;
 
+       mutable->mlink = 0;
+       if (ipv4_is_multicast(mutable->key.daddr)) {
+               struct net_device *dev;
+               struct rtable *rt;
+
+               rt = __find_route(mutable, tnl_ops->ipproto, mutable->tos);
+               if (IS_ERR(rt))
+                       return -EADDRNOTAVAIL;
+               dev = rt_dst(rt).dev;
+               ip_rt_put(rt);
+               if (__in_dev_get_rtnl(dev) == NULL)
+                       return -EADDRNOTAVAIL;
+               mutable->mlink = dev->ifindex;
+               ip_mc_inc_group(__in_dev_get_rtnl(dev), mutable->key.daddr);
+       }
+
        return 0;
 }
 
@@ -1399,6 +1467,7 @@ struct vport *tnl_create(const struct vport_parms *parms,
        return vport;
 
 error_free_mutable:
+       free_mutable_rtnl(mutable);
        kfree(mutable);
 error_free_vport:
        vport_free(vport);
@@ -1437,6 +1506,7 @@ int tnl_set_options(struct vport *vport, struct nlattr *options)
        return 0;
 
 error_free:
+       free_mutable_rtnl(mutable);
        kfree(mutable);
 error:
        return err;
@@ -1480,23 +1550,26 @@ static void free_port_rcu(struct rcu_head *rcu)
 void tnl_destroy(struct vport *vport)
 {
        struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
-       const struct tnl_mutable_config *mutable;
+       struct tnl_mutable_config *mutable;
 
        mutable = rtnl_dereference(tnl_vport->mutable);
        port_table_remove_port(vport);
+       free_mutable_rtnl(mutable);
        call_rcu(&tnl_vport->rcu, free_port_rcu);
 }
 
 int tnl_set_addr(struct vport *vport, const unsigned char *addr)
 {
        struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
-       struct tnl_mutable_config *mutable;
+       struct tnl_mutable_config *old_mutable, *mutable;
 
-       mutable = kmemdup(rtnl_dereference(tnl_vport->mutable),
-                         sizeof(struct tnl_mutable_config), GFP_KERNEL);
+       old_mutable = rtnl_dereference(tnl_vport->mutable);
+       mutable = kmemdup(old_mutable, sizeof(struct tnl_mutable_config), GFP_KERNEL);
        if (!mutable)
                return -ENOMEM;
 
+       old_mutable->mlink = 0;
+
        memcpy(mutable->eth_addr, addr, ETH_ALEN);
        assign_config_rcu(vport, mutable);