From b37e6334fd80a6460aabc95f05eb8ab68e5a82ed Mon Sep 17 00:00:00 2001
From: Ben Pfaff
Date: Mon, 24 Oct 2011 12:27:36 -0700
Subject: [PATCH] datapath: Add multicast tunnel support.
Something like this, on two separate vswitches, works to try it out:
route add -net 224.0.0.0 netmask 240.0.0.0 dev eth0
ovs-vsctl \
-- add-port br0 gre0 \
-- set interface gre0 type=gre options:remote_ip=224.0.0.1
Runtime tested on Linux 3.0, build tested on Linux 2.6.18, both i386.
Signed-off-by: Ben Pfaff
Acked-by: Jesse Gross
---
datapath/tunnel.c | 117 +++++++++++++++++++++++++++++++++++--------
datapath/tunnel.h | 3 ++
datapath/vport-gre.c | 2 +
lib/netdev-vport.c | 12 ++++-
lib/packets.h | 5 ++
vswitchd/vswitch.xml | 16 ++++--
6 files changed, 128 insertions(+), 27 deletions(-)
diff --git a/datapath/tunnel.c b/datapath/tunnel.c
index f9138493..372d90ed 100644
--- a/datapath/tunnel.c
+++ b/datapath/tunnel.c
@@ -10,8 +10,10 @@
#include
#include
#include
+#include
#include
#include
+#include
#include
#include
#include
@@ -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);
diff --git a/datapath/tunnel.h b/datapath/tunnel.h
index 9211740a..f80df99b 100644
--- a/datapath/tunnel.h
+++ b/datapath/tunnel.h
@@ -89,6 +89,9 @@ struct tnl_mutable_config {
u32 flags;
u8 tos;
u8 ttl;
+
+ /* Multicast configuration. */
+ int mlink;
};
struct tnl_ops {
diff --git a/datapath/vport-gre.c b/datapath/vport-gre.c
index cc64d129..95ac4bb2 100644
--- a/datapath/vport-gre.c
+++ b/datapath/vport-gre.c
@@ -188,6 +188,8 @@ static void gre_err(struct sk_buff *skb, u32 info)
return;
iph = (struct iphdr *)skb->data;
+ if (ipv4_is_multicast(iph->daddr))
+ return;
tunnel_hdr_len = parse_header(iph, &flags, &key);
if (tunnel_hdr_len < 0)
diff --git a/lib/netdev-vport.c b/lib/netdev-vport.c
index 0577bd3b..f6dbd03e 100644
--- a/lib/netdev-vport.c
+++ b/lib/netdev-vport.c
@@ -578,6 +578,7 @@ parse_tunnel_config(const char *name, const char *type,
struct shash_node *node;
bool ipsec_mech_set = false;
ovs_be32 daddr = htonl(0);
+ ovs_be32 saddr = htonl(0);
uint32_t flags;
flags = TNL_F_DF_DEFAULT | TNL_F_PMTUD | TNL_F_HDR_CACHE;
@@ -603,8 +604,7 @@ parse_tunnel_config(const char *name, const char *type,
if (lookup_ip(node->data, &in_addr)) {
VLOG_WARN("%s: bad %s 'local_ip'", name, type);
} else {
- nl_msg_put_be32(options, OVS_TUNNEL_ATTR_SRC_IPV4,
- in_addr.s_addr);
+ saddr = in_addr.s_addr;
}
} else if (!strcmp(node->name, "tos")) {
if (!strcmp(node->data, "inherit")) {
@@ -707,6 +707,14 @@ parse_tunnel_config(const char *name, const char *type,
}
nl_msg_put_be32(options, OVS_TUNNEL_ATTR_DST_IPV4, daddr);
+ if (saddr) {
+ if (ip_is_multicast(daddr)) {
+ VLOG_WARN("%s: remote_ip is multicast, ignoring local_ip", name);
+ } else {
+ nl_msg_put_be32(options, OVS_TUNNEL_ATTR_SRC_IPV4, saddr);
+ }
+ }
+
nl_msg_put_u32(options, OVS_TUNNEL_ATTR_FLAGS, flags);
return 0;
diff --git a/lib/packets.h b/lib/packets.h
index f5f473ce..84877541 100644
--- a/lib/packets.h
+++ b/lib/packets.h
@@ -291,6 +291,11 @@ ip_is_cidr(ovs_be32 netmask)
uint32_t x = ~ntohl(netmask);
return !(x & (x + 1));
}
+static inline bool
+ip_is_multicast(ovs_be32 ip)
+{
+ return (ip & htonl(0xf0000000)) == htonl(0xe0000000);
+}
int ip_count_cidr_bits(ovs_be32 netmask);
void ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *);
diff --git a/vswitchd/vswitch.xml b/vswitchd/vswitch.xml
index 239a9e88..bcb6b6fd 100644
--- a/vswitchd/vswitch.xml
+++ b/vswitchd/vswitch.xml
@@ -1118,12 +1118,22 @@
- Required. The tunnel endpoint.
+
+ Required. The tunnel endpoint. Unicast and multicast endpoints are
+ both supported.
+
+
+
+ When a multicast endpoint is specified, a routing table lookup occurs
+ only when the tunnel is created. Following a routing change, delete
+ and then re-create the tunnel to force a new routing table lookup.
+
- Optional. The destination IP that received packets must
- match. Default is to match all addresses.
+ Optional. The destination IP that received packets must match.
+ Default is to match all addresses. Must be omitted when is a multicast address.
--
2.30.2