Simple minded ethernet-over-UDP tunnels
authorGiuseppe Lettieri <g.lettieri@iet.unipi.it>
Sun, 29 Apr 2012 12:59:17 +0000 (14:59 +0200)
committerGiuseppe Lettieri <g.lettieri@iet.unipi.it>
Sun, 29 Apr 2012 12:59:17 +0000 (14:59 +0200)
We provide very simple point-to-point UDP tunnels which openvswitch can
use as network interfaces. The implementation uses a pair of connected
UDP sockets for each tunnel. Each ethernet frame is sent as a single
UDP datagram.

Example usage:
The following commands create a tunnel between hosts 10.0.0.1 and
10.0.0.2 (port and bridge names are arbitrary).

 - On host 10.0.0.1:
ovs-vsctl add-br br0
ovs-vsctl add-port br0 tun0 -- set interface tun0 type=tunnel \
options:remote_ip=10.0.0.2
ovs-appctl netdev-tunnel/get-lport tun0
<(1) the command replies with the local port number>

 - On host 10.0.0.2:
ovs-vsctl add-br br1
ovs-vsctl add-port br0 tun0 -- set interface tun1 type=tunnel \
options:remote_ip=10.0.0.1
ovs-appctl netdev-tunnel/get-lport tun0
<(2) the command replies with the local port number>
ovs-appctl netdev-tunnel/set-rport tun0 <the port obtained in (1)>

 - Back on host 10.0.0.1:
ovs-appctl netdev-tunnel/set-rport tun0 <the port obtained in (2)>

lib/automake.mk
lib/netdev-provider.h
lib/netdev-tunnel.c [new file with mode: 0644]
lib/netdev.c

index 2d2617e..31e07f3 100644 (file)
@@ -84,6 +84,7 @@ lib_libopenvswitch_a_SOURCES = \
        lib/multipath.c \
        lib/multipath.h \
        lib/netdev-dummy.c \
+       lib/netdev-tunnel.c \
        lib/netdev-provider.h \
        lib/netdev.c \
        lib/netdev.h \
index daf46ad..e486ee6 100644 (file)
@@ -594,6 +594,8 @@ extern const struct netdev_class netdev_linux_class;
 extern const struct netdev_class netdev_internal_class;
 extern const struct netdev_class netdev_tap_class;
 
+extern const struct netdev_class netdev_tunnel_class;
+
 #ifdef  __cplusplus
 }
 #endif
diff --git a/lib/netdev-tunnel.c b/lib/netdev-tunnel.c
new file mode 100644 (file)
index 0000000..41b0fd5
--- /dev/null
@@ -0,0 +1,532 @@
+/*
+ * Copyright (c) 2010, 2011, 2012 Nicira Networks.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <config.h>
+
+#include <unistd.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <arpa/inet.h>
+#include <errno.h>
+
+#include "flow.h"
+#include "list.h"
+#include "netdev-provider.h"
+#include "odp-util.h"
+#include "ofp-print.h"
+#include "ofpbuf.h"
+#include "packets.h"
+#include "poll-loop.h"
+#include "shash.h"
+#include "sset.h"
+#include "unixctl.h"
+#include "socket-util.h"
+#include "vlog.h"
+
+VLOG_DEFINE_THIS_MODULE(netdev_tunnel);
+
+struct netdev_dev_tunnel {
+    struct netdev_dev netdev_dev;
+    uint8_t hwaddr[ETH_ADDR_LEN];
+    int mtu;
+    struct netdev_stats stats;
+    enum netdev_flags flags;
+    int sockfd;
+    struct sockaddr_in local_addr;
+    struct sockaddr_in remote_addr;
+    bool got_remote_ip;
+    bool connected;
+    unsigned int change_seq;
+};
+
+struct netdev_tunnel {
+    struct netdev netdev;
+    bool listening;
+};
+
+static struct shash tunnel_netdev_devs = SHASH_INITIALIZER(&tunnel_netdev_devs);
+
+static int netdev_tunnel_create(const struct netdev_class *, const char *,
+                               struct netdev_dev **);
+static void netdev_tunnel_poll_notify(const struct netdev *);
+
+static bool
+is_tunnel_class(const struct netdev_class *class)
+{
+    return class->create == netdev_tunnel_create;
+}
+
+static struct netdev_dev_tunnel *
+netdev_dev_tunnel_cast(const struct netdev_dev *netdev_dev)
+{
+    assert(is_tunnel_class(netdev_dev_get_class(netdev_dev)));
+    return CONTAINER_OF(netdev_dev, struct netdev_dev_tunnel, netdev_dev);
+}
+
+static struct netdev_tunnel *
+netdev_tunnel_cast(const struct netdev *netdev)
+{
+    struct netdev_dev *netdev_dev = netdev_get_dev(netdev);
+    assert(is_tunnel_class(netdev_dev_get_class(netdev_dev)));
+    return CONTAINER_OF(netdev, struct netdev_tunnel, netdev);
+}
+
+static int
+netdev_tunnel_create(const struct netdev_class *class, const char *name,
+                    struct netdev_dev **netdev_devp)
+{
+    static unsigned int n = 0xaa550000;
+    struct netdev_dev_tunnel *netdev_dev;
+    int error;
+
+    netdev_dev = xzalloc(sizeof *netdev_dev);
+    netdev_dev_init(&netdev_dev->netdev_dev, name, class);
+    netdev_dev->hwaddr[0] = 0x55;
+    netdev_dev->hwaddr[1] = 0xaa;
+    netdev_dev->hwaddr[2] = n >> 24;
+    netdev_dev->hwaddr[3] = n >> 16;
+    netdev_dev->hwaddr[4] = n >> 8;
+    netdev_dev->hwaddr[5] = n;
+    netdev_dev->mtu = 1500;
+    netdev_dev->flags = 0;
+    netdev_dev->change_seq = 1;
+    memset(&netdev_dev->remote_addr, 0, sizeof(netdev_dev->remote_addr));
+    netdev_dev->got_remote_ip = false;
+    netdev_dev->connected = false;
+
+
+    netdev_dev->sockfd = inet_open_passive(SOCK_DGRAM, "", 0, &netdev_dev->local_addr, 0);
+    if (netdev_dev->sockfd < 0) {
+       error = netdev_dev->sockfd;
+        goto error;
+    }
+
+
+    shash_add(&tunnel_netdev_devs, name, netdev_dev);
+
+    n++;
+
+    *netdev_devp = &netdev_dev->netdev_dev;
+
+    VLOG_DBG("tunnel_create: name=%s, fd=%d, port=%d", name, netdev_dev->sockfd, netdev_dev->local_addr.sin_port);
+
+    return 0;
+
+error:
+    free(netdev_dev);
+    return error;
+}
+
+static void
+netdev_tunnel_destroy(struct netdev_dev *netdev_dev_)
+{
+    struct netdev_dev_tunnel *netdev_dev = netdev_dev_tunnel_cast(netdev_dev_);
+
+    if (netdev_dev->sockfd != -1)
+       close(netdev_dev->sockfd);
+
+    shash_find_and_delete(&tunnel_netdev_devs,
+                          netdev_dev_get_name(netdev_dev_));
+    free(netdev_dev);
+}
+
+static int
+netdev_tunnel_open(struct netdev_dev *netdev_dev_, struct netdev **netdevp)
+{
+    struct netdev_dev_tunnel *netdev_dev = netdev_dev_tunnel_cast(netdev_dev_);
+    struct netdev_tunnel *netdev;
+
+    netdev = xmalloc(sizeof *netdev);
+    netdev_init(&netdev->netdev, netdev_dev_);
+    netdev->listening = false;
+
+    *netdevp = &netdev->netdev;
+    return 0;
+}
+
+static void
+netdev_tunnel_close(struct netdev *netdev_)
+{
+    struct netdev_tunnel *netdev = netdev_tunnel_cast(netdev_);
+    free(netdev);
+}
+
+static int
+netdev_tunnel_get_config(struct netdev_dev *dev_, struct shash *args)
+{
+    return 0;
+}
+
+static int
+netdev_tunnel_connect(struct netdev_dev_tunnel *dev)
+{
+    if (dev->sockfd < 0)
+        return EBADF;
+    if (!dev->got_remote_ip || !dev->remote_addr.sin_port)
+        return 0;
+    dev->remote_addr.sin_family = AF_INET;
+    if (connect(dev->sockfd, (struct sockaddr*) &dev->remote_addr, sizeof(dev->remote_addr)) < 0) {
+        return errno;
+    }
+    dev->connected = true;
+    VLOG_DBG("%s: connected to (%s, %d)", netdev_dev_get_name(&dev->netdev_dev),
+        inet_ntoa(dev->remote_addr.sin_addr), ntohs(dev->remote_addr.sin_port));
+    return 0;
+}
+
+static int
+netdev_tunnel_set_config(struct netdev_dev *dev_, const struct shash *args)
+{
+    struct netdev_dev_tunnel *netdev_dev = netdev_dev_tunnel_cast(dev_);
+    struct shash_node *node;
+    void *data;
+
+    VLOG_DBG("tunnel_set_config(%s)", netdev_dev_get_name(dev_));
+    SHASH_FOR_EACH(node, args) {
+        VLOG_DBG("arg: %s->%s", node->name, node->data);
+    }
+    data = shash_find_data(args, "remote_ip");
+    if (data) {
+        struct in_addr addr;
+        int error = lookup_ip(data, &addr);
+        if (error)
+            return error;
+        netdev_dev->remote_addr.sin_addr = addr;
+        netdev_dev->got_remote_ip = true;
+        return netdev_tunnel_connect(netdev_dev);        
+    }
+    return 0;
+}
+
+static int
+netdev_tunnel_listen(struct netdev *netdev_)
+{
+    struct netdev_tunnel *netdev = netdev_tunnel_cast(netdev_);
+    VLOG_DBG("tunnel_listen(%s)", netdev_get_name(netdev_));
+    netdev->listening = true;
+    return 0;
+}
+
+static int
+netdev_tunnel_recv(struct netdev *netdev_, void *buffer, size_t size)
+{
+    struct netdev_dev_tunnel *dev = 
+       netdev_dev_tunnel_cast(netdev_get_dev(netdev_));
+    if (!dev->connected)
+        return -EAGAIN;
+    for (;;) {
+        ssize_t retval;
+        retval = recv(dev->sockfd, buffer, size, MSG_TRUNC);
+       VLOG_DBG("%s: recv(%x, %d, MSG_TRUNC) = %d",
+                       netdev_get_name(netdev_), buffer, size, retval);
+        if (retval >= 0) {
+            return retval <= size ? retval : -EMSGSIZE;
+        } else if (errno != EINTR) {
+            if (errno != EAGAIN) {
+                VLOG_WARN("error receiveing Ethernet packet on %s: %s",
+                    strerror(errno), netdev_get_name(netdev_));
+            }
+            return -errno;
+        }
+    }
+}
+
+static void
+netdev_tunnel_recv_wait(struct netdev *netdev_)
+{
+    struct netdev_dev_tunnel *dev = 
+       netdev_dev_tunnel_cast(netdev_get_dev(netdev_));
+    if (dev->sockfd >= 0) {
+        poll_fd_wait(dev->sockfd, POLLIN);
+    }
+}
+
+static int
+netdev_tunnel_send(struct netdev *netdev_, const void *buffer, size_t size)
+{
+    struct netdev_dev_tunnel *dev = 
+       netdev_dev_tunnel_cast(netdev_get_dev(netdev_));
+    if (!dev->connected)
+        return -EAGAIN;
+    for (;;) {
+        ssize_t retval;
+        retval = send(dev->sockfd, buffer, size, 0);
+       VLOG_DBG("%s: send(%x, %d) = %d", netdev_get_name(netdev_), buffer, size, retval);
+        if (retval >= 0) {
+            return retval;
+        } else if (errno != EINTR) {
+            if (errno != EAGAIN) {
+                VLOG_WARN("error sending Ethernet packet on %s: %s",
+                    strerror(errno), netdev_get_name(netdev_));
+            }
+            return -errno;
+        }
+    }
+}
+
+static void
+netdev_tunnel_send_wait(struct netdev *netdev_)
+{
+    struct netdev_dev_tunnel *dev = 
+       netdev_dev_tunnel_cast(netdev_get_dev(netdev_));
+    if (dev->sockfd >= 0) {
+        poll_fd_wait(dev->sockfd, POLLOUT);
+    }
+}
+
+static int
+netdev_tunnel_drain(struct netdev *netdev_)
+{
+    struct netdev_tunnel *netdev = netdev_tunnel_cast(netdev_);
+    return 0;
+}
+
+static int
+netdev_tunnel_set_etheraddr(struct netdev *netdev,
+                           const uint8_t mac[ETH_ADDR_LEN])
+{
+    struct netdev_dev_tunnel *dev =
+        netdev_dev_tunnel_cast(netdev_get_dev(netdev));
+
+    if (!eth_addr_equals(dev->hwaddr, mac)) {
+        memcpy(dev->hwaddr, mac, ETH_ADDR_LEN);
+        netdev_tunnel_poll_notify(netdev);
+    }
+
+    return 0;
+}
+
+static int
+netdev_tunnel_get_etheraddr(const struct netdev *netdev,
+                           uint8_t mac[ETH_ADDR_LEN])
+{
+    const struct netdev_dev_tunnel *dev =
+        netdev_dev_tunnel_cast(netdev_get_dev(netdev));
+
+    memcpy(mac, dev->hwaddr, ETH_ADDR_LEN);
+    return 0;
+}
+
+static int
+netdev_tunnel_get_mtu(const struct netdev *netdev, int *mtup)
+{
+    const struct netdev_dev_tunnel *dev =
+        netdev_dev_tunnel_cast(netdev_get_dev(netdev));
+
+    *mtup = dev->mtu;
+    return 0;
+}
+
+static int
+netdev_tunnel_set_mtu(const struct netdev *netdev, int mtu)
+{
+    struct netdev_dev_tunnel *dev =
+        netdev_dev_tunnel_cast(netdev_get_dev(netdev));
+
+    dev->mtu = mtu;
+    return 0;
+}
+
+static int
+netdev_tunnel_get_carrier(const struct netdev *netdev, bool *carrier)
+{
+    struct netdev_dev_tunnel *dev =
+        netdev_dev_tunnel_cast(netdev_get_dev(netdev));
+    VLOG_DBG("tunnel_get_carrier(%s)", netdev_get_name(netdev));
+    *carrier = dev->connected;
+    return 0;
+}
+
+static int
+netdev_tunnel_get_stats(const struct netdev *netdev, struct netdev_stats *stats)
+{
+    const struct netdev_dev_tunnel *dev =
+        netdev_dev_tunnel_cast(netdev_get_dev(netdev));
+
+    *stats = dev->stats;
+    return 0;
+}
+
+static int
+netdev_tunnel_set_stats(struct netdev *netdev, const struct netdev_stats *stats)
+{
+    struct netdev_dev_tunnel *dev =
+        netdev_dev_tunnel_cast(netdev_get_dev(netdev));
+
+    dev->stats = *stats;
+    return 0;
+}
+
+static int
+netdev_tunnel_update_flags(struct netdev *netdev,
+                          enum netdev_flags off, enum netdev_flags on,
+                          enum netdev_flags *old_flagsp)
+{
+    struct netdev_dev_tunnel *dev =
+        netdev_dev_tunnel_cast(netdev_get_dev(netdev));
+
+    if ((off | on) & ~(NETDEV_UP | NETDEV_PROMISC)) {
+        return EINVAL;
+    }
+
+    *old_flagsp = dev->flags;
+    dev->flags |= on;
+    dev->flags &= ~off;
+    if (*old_flagsp != dev->flags) {
+        netdev_tunnel_poll_notify(netdev);
+    }
+    return 0;
+}
+
+static unsigned int
+netdev_tunnel_change_seq(const struct netdev *netdev)
+{
+    return netdev_dev_tunnel_cast(netdev_get_dev(netdev))->change_seq;
+}
+\f
+/* Helper functions. */
+
+static void
+netdev_tunnel_poll_notify(const struct netdev *netdev)
+{
+    struct netdev_dev_tunnel *dev =
+        netdev_dev_tunnel_cast(netdev_get_dev(netdev));
+
+    dev->change_seq++;
+    if (!dev->change_seq) {
+        dev->change_seq++;
+    }
+}
+
+#define LPORT ((void*)1)
+#define RPORT ((void*)2)
+
+static void
+netdev_tunnel_get_port(struct unixctl_conn *conn,
+                     int argc, const char *argv[], void *aux OVS_UNUSED)
+{
+    struct netdev_dev_tunnel *tunnel_dev;
+    uint16_t port;
+    char buf[6];
+
+    tunnel_dev = shash_find_data(&tunnel_netdev_devs, argv[1]);
+    if (!tunnel_dev) {
+        unixctl_command_reply_error(conn, "no such tunnel netdev");
+        return;
+    }
+
+    port = (aux == LPORT ? tunnel_dev->local_addr.sin_port
+                        : tunnel_dev->remote_addr.sin_port);
+    
+    sprintf(buf, "%d", ntohs(port));
+    unixctl_command_reply(conn, buf);
+}
+
+static void
+netdev_tunnel_set_rport(struct unixctl_conn *conn,
+                     int argc, const char *argv[], void *aux OVS_UNUSED)
+{
+    struct netdev_dev_tunnel *tunnel_dev;
+    int error;
+
+    tunnel_dev = shash_find_data(&tunnel_netdev_devs, argv[1]);
+    if (!tunnel_dev) {
+        unixctl_command_reply_error(conn, "no such tunnel netdev");
+        return;
+    }
+
+    tunnel_dev->remote_addr.sin_port = htons(atoi(argv[2]));
+
+    error = netdev_tunnel_connect(tunnel_dev);
+    if (error) {
+        unixctl_command_reply_error(conn, strerror(errno));
+    } else {
+        unixctl_command_reply(conn, "ok");
+    }
+}
+
+static int
+netdev_tunnel_init(void)
+{
+    unixctl_command_register("netdev-tunnel/get-lport", "NAME",
+                             1, 1, netdev_tunnel_get_port, LPORT);
+    unixctl_command_register("netdev-tunnel/get-rport", "NAME",
+                             1, 1, netdev_tunnel_get_port, RPORT);
+    unixctl_command_register("netdev-tunnel/set-rport", "NAME PORT",
+                             2, 2, netdev_tunnel_set_rport, 0);
+    return 0;
+}
+
+const struct netdev_class netdev_tunnel_class = {
+    "tunnel",
+    netdev_tunnel_init,         /* init */
+    NULL,                       /* run */
+    NULL,                       /* wait */
+
+    netdev_tunnel_create,
+    netdev_tunnel_destroy,
+    netdev_tunnel_get_config,
+    netdev_tunnel_set_config, 
+
+    netdev_tunnel_open,
+    netdev_tunnel_close,
+
+    netdev_tunnel_listen,
+    netdev_tunnel_recv,
+    netdev_tunnel_recv_wait,
+    netdev_tunnel_drain,
+
+    netdev_tunnel_send, 
+    netdev_tunnel_send_wait,  
+
+    netdev_tunnel_set_etheraddr,
+    netdev_tunnel_get_etheraddr,
+    netdev_tunnel_get_mtu,
+    netdev_tunnel_set_mtu,
+    NULL,                       /* get_ifindex */
+    netdev_tunnel_get_carrier,  /* get_carrier */
+    NULL,                       /* get_carrier_resets */
+    NULL,                       /* get_miimon */
+    netdev_tunnel_get_stats,
+    netdev_tunnel_set_stats,
+
+    NULL,                       /* get_features */
+    NULL,                       /* set_advertisements */
+
+    NULL,                       /* set_policing */
+    NULL,                       /* get_qos_types */
+    NULL,                       /* get_qos_capabilities */
+    NULL,                       /* get_qos */
+    NULL,                       /* set_qos */
+    NULL,                       /* get_queue */
+    NULL,                       /* set_queue */
+    NULL,                       /* delete_queue */
+    NULL,                       /* get_queue_stats */
+    NULL,                       /* dump_queues */
+    NULL,                       /* dump_queue_stats */
+
+    NULL,                       /* get_in4 */
+    NULL,                       /* set_in4 */
+    NULL,                       /* get_in6 */
+    NULL,                       /* add_router */
+    NULL,                       /* get_next_hop */
+    NULL,                       /* get_drv_info */
+    NULL,                       /* arp_lookup */
+
+    netdev_tunnel_update_flags,
+
+    netdev_tunnel_change_seq
+};
index 2652a3f..165009a 100644 (file)
@@ -80,6 +80,7 @@ netdev_initialize(void)
         netdev_register_provider(&netdev_tap_class);
         netdev_vport_register();
 #endif
+       netdev_register_provider(&netdev_tunnel_class);
     }
 }