X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fnetdev-tunnel.c;h=f8bc74bb18fcc15d20dbd85f2a0a8cba8cbc7ed5;hb=0f795b759cfc75d9effbe245c8714dbea6d20bb1;hp=0497cbbc266e3ad396f713f04bce897135d50033;hpb=17dcd764543b3f31cfa795ddce4bbf56f0038cb0;p=sliver-openvswitch.git diff --git a/lib/netdev-tunnel.c b/lib/netdev-tunnel.c index 0497cbbc2..f8bc74bb1 100644 --- a/lib/netdev-tunnel.c +++ b/lib/netdev-tunnel.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2010 Nicira Networks. + * 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. @@ -15,249 +15,503 @@ */ #include + +#include +#include +#include +#include #include -#include -#include -#include +#include "flow.h" +#include "list.h" #include "netdev-provider.h" -#include "netdev-vport.h" -#include "openflow/openflow.h" -#include "openvswitch/datapath-protocol.h" -#include "openvswitch/tunnel.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) +VLOG_DEFINE_THIS_MODULE(netdev_tunnel); -struct netdev_dev_tunnel { - struct netdev_dev netdev_dev; +struct netdev_tunnel { + struct netdev up; + uint8_t hwaddr[ETH_ADDR_LEN]; + struct netdev_stats stats; + enum netdev_flags flags; + int sockfd; + struct sockaddr_in local_addr; + struct sockaddr_in remote_addr; + bool valid_remote_ip; + bool valid_remote_port; + bool connected; + unsigned int change_seq; }; -struct netdev_tunnel { - struct netdev netdev; +struct netdev_rx_tunnel { + struct netdev_rx up; + int fd; }; -static int netdev_tunnel_create(const char *name, const char *type, - const struct shash *args, struct netdev_dev **); +static const struct netdev_rx_class netdev_rx_tunnel_class; + +static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); + +static struct shash tunnel_netdevs = SHASH_INITIALIZER(&tunnel_netdevs); -static struct netdev_dev_tunnel * -netdev_dev_tunnel_cast(const struct netdev_dev *netdev_dev) +static int netdev_tunnel_create(const struct netdev_class *, const char *, + struct netdev **); +static void netdev_tunnel_update_seq(struct netdev_tunnel *); + +static bool +is_netdev_tunnel_class(const struct netdev_class *class) { - assert(netdev_dev_get_class(netdev_dev)->create == netdev_tunnel_create); - return CONTAINER_OF(netdev_dev, struct netdev_dev_tunnel, netdev_dev); + return class->create == netdev_tunnel_create; } static struct netdev_tunnel * netdev_tunnel_cast(const struct netdev *netdev) { - struct netdev_dev *netdev_dev = netdev_get_dev(netdev); - assert(netdev_dev_get_class(netdev_dev)->create == netdev_tunnel_create); - return CONTAINER_OF(netdev, struct netdev_tunnel, netdev); + ovs_assert(is_netdev_tunnel_class(netdev_get_class(netdev))); + return CONTAINER_OF(netdev, struct netdev_tunnel, up); +} + +static struct netdev_rx_tunnel * +netdev_rx_tunnel_cast(const struct netdev_rx *rx) +{ + netdev_rx_assert_class(rx, &netdev_rx_tunnel_class); + return CONTAINER_OF(rx, struct netdev_rx_tunnel, up); } static int -parse_config(const char *name, const char *type, const struct shash *args, - struct tnl_port_config *config) +netdev_tunnel_create(const struct netdev_class *class, const char *name, + struct netdev **netdevp) { + static unsigned int n = 0; + struct netdev_tunnel *netdev; + int error; + + netdev = xzalloc(sizeof *netdev); + netdev_init(&netdev->up, name, class); + netdev->hwaddr[0] = 0xfe; + netdev->hwaddr[1] = 0xff; + netdev->hwaddr[2] = 0xff; + netdev->hwaddr[3] = n >> 16; + netdev->hwaddr[4] = n >> 8; + netdev->hwaddr[5] = n; + netdev->flags = 0; + netdev->change_seq = 1; + memset(&netdev->remote_addr, 0, sizeof(netdev->remote_addr)); + netdev->valid_remote_ip = false; + netdev->valid_remote_port = false; + netdev->connected = false; + + + netdev->sockfd = inet_open_passive(SOCK_DGRAM, "", 0, &netdev->local_addr, 0); + if (netdev->sockfd < 0) { + error = netdev->sockfd; + goto error; + } + + + shash_add(&tunnel_netdevs, name, netdev); + + n++; + + *netdevp = &netdev->up; + + VLOG_DBG("tunnel_create: name=%s, fd=%d, port=%d", name, netdev->sockfd, netdev->local_addr.sin_port); + + return 0; + +error: + free(netdev); + return error; +} + +static void +netdev_tunnel_destroy(struct netdev *netdev_) +{ + struct netdev_tunnel *netdev = netdev_tunnel_cast(netdev_); + + if (netdev->sockfd != -1) + close(netdev->sockfd); + + shash_find_and_delete(&tunnel_netdevs, + netdev_get_name(netdev_)); + free(netdev); +} + +static int +netdev_tunnel_get_config(const struct netdev *dev_, struct smap *args) +{ + struct netdev_tunnel *netdev = netdev_tunnel_cast(dev_); + + if (netdev->valid_remote_ip) + smap_add_format(args, "remote_ip", IP_FMT, + IP_ARGS(netdev->remote_addr.sin_addr.s_addr)); + if (netdev->valid_remote_port) + smap_add_format(args, "remote_port", "%"PRIu16, + ntohs(netdev->remote_addr.sin_port)); + return 0; +} + +static int +netdev_tunnel_connect(struct netdev_tunnel *dev) +{ + char buf[1024]; + if (dev->sockfd < 0) + return EBADF; + if (!dev->valid_remote_ip || !dev->valid_remote_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; + netdev_tunnel_update_seq(dev); + VLOG_DBG("%s: connected to (%s, %d)", netdev_get_name(&dev->up), + inet_ntop(AF_INET, &dev->remote_addr.sin_addr, buf, 1024), ntohs(dev->remote_addr.sin_port)); + return 0; +} + +static int +netdev_tunnel_set_config(struct netdev *dev_, const struct smap *args) +{ + struct netdev_tunnel *netdev = netdev_tunnel_cast(dev_); struct shash_node *node; - memset(config, 0, sizeof *config); + VLOG_DBG("tunnel_set_config(%s)", netdev_get_name(dev_)); + SMAP_FOR_EACH(node, args) { + VLOG_DBG("arg: %s->%s", node->name, (char*)node->data); + if (!strcmp(node->name, "remote_ip")) { + struct in_addr addr; + if (lookup_ip(node->data, &addr)) { + VLOG_WARN("%s: bad 'remote_ip'", node->name); + } else { + netdev->remote_addr.sin_addr = addr; + netdev->valid_remote_ip = true; + } + } else if (!strcmp(node->name, "remote_port")) { + netdev->remote_addr.sin_port = htons(atoi(node->data)); + netdev->valid_remote_port = true; + } else { + VLOG_WARN("%s: unknown argument '%s'", + netdev_get_name(dev_), node->name); + } + } + return netdev_tunnel_connect(netdev); +} - config->flags |= TNL_F_PMTUD; +static int +netdev_tunnel_rx_open(struct netdev *netdev_, struct netdev_rx **rxp) +{ + struct netdev_tunnel *netdev = + netdev_tunnel_cast(netdev_); + struct netdev_rx_tunnel *rx; + rx = xmalloc(sizeof *rx); + netdev_rx_init(&rx->up, netdev_, &netdev_rx_tunnel_class); + rx->fd = netdev->sockfd; + *rxp = &rx->up; + return 0; +} - SHASH_FOR_EACH (node, args) { - if (!strcmp(node->name, "remote_ip")) { - struct in_addr in_addr; - if (lookup_ip(node->data, &in_addr)) { - VLOG_WARN("%s: bad %s 'remote_ip'", name, type); - } else { - config->daddr = in_addr.s_addr; - } - } else if (!strcmp(node->name, "local_ip")) { - struct in_addr in_addr; - if (lookup_ip(node->data, &in_addr)) { - VLOG_WARN("%s: bad %s 'local_ip'", name, type); - } else { - config->saddr = in_addr.s_addr; - } - } else if (!strcmp(node->name, "key")) { - if (!strcmp(node->data, "flow")) { - config->flags |= TNL_F_IN_KEY_MATCH; - config->flags |= TNL_F_OUT_KEY_ACTION; - } else { - config->out_key = config->in_key = htonl(atoi(node->data)); - } - } else if (!strcmp(node->name, "in_key")) { - if (!strcmp(node->data, "flow")) { - config->flags |= TNL_F_IN_KEY_MATCH; - } else { - config->in_key = htonl(atoi(node->data)); - } - } else if (!strcmp(node->name, "out_key")) { - if (!strcmp(node->data, "flow")) { - config->flags |= TNL_F_OUT_KEY_ACTION; - } else { - config->out_key = htonl(atoi(node->data)); - } - } else if (!strcmp(node->name, "tos")) { - if (!strcmp(node->data, "inherit")) { - config->flags |= TNL_F_TOS_INHERIT; - } else { - config->tos = atoi(node->data); - } - } else if (!strcmp(node->name, "ttl")) { - if (!strcmp(node->data, "inherit")) { - config->flags |= TNL_F_TTL_INHERIT; +static void +netdev_rx_tunnel_destroy(struct netdev_rx *rx_) +{ + struct netdev_rx_tunnel *rx = netdev_rx_tunnel_cast(rx_); + free(rx); +} + +static int +netdev_rx_tunnel_recv(struct netdev_rx *rx_, void *buffer, size_t size) +{ + struct netdev_rx_tunnel *rx = netdev_rx_tunnel_cast(rx_); + struct netdev_tunnel *netdev = + netdev_tunnel_cast(rx_->netdev); + if (!netdev->connected) + return -EAGAIN; + for (;;) { + ssize_t retval; + retval = recv(rx->fd, buffer, size, MSG_TRUNC); + VLOG_DBG("%s: recv(%"PRIxPTR", %zu, MSG_TRUNC) = %zd", + netdev_rx_get_name(rx_), (uintptr_t)buffer, size, retval); + if (retval >= 0) { + netdev->stats.rx_packets++; + netdev->stats.rx_bytes += retval; + if (retval <= size) { + return retval; } else { - config->ttl = atoi(node->data); + netdev->stats.rx_errors++; + netdev->stats.rx_length_errors++; + return -EMSGSIZE; } - } else if (!strcmp(node->name, "csum")) { - if (!strcmp(node->data, "true")) { - config->flags |= TNL_F_CSUM; + } else if (errno != EINTR) { + if (errno != EAGAIN) { + VLOG_WARN_RL(&rl, "error receiveing Ethernet packet on %s: %s", + netdev_rx_get_name(rx_), ovs_strerror(errno)); + netdev->stats.rx_errors++; } - } else if (!strcmp(node->name, "pmtud")) { - if (!strcmp(node->data, "false")) { - config->flags &= ~TNL_F_PMTUD; + return -errno; + } + } +} + +static void +netdev_rx_tunnel_wait(struct netdev_rx *rx_) +{ + struct netdev_rx_tunnel *rx = + netdev_rx_tunnel_cast(rx_); + if (rx->fd >= 0) { + poll_fd_wait(rx->fd, POLLIN); + } +} + +static int +netdev_tunnel_send(struct netdev *netdev_, const void *buffer, size_t size) +{ + struct netdev_tunnel *dev = + netdev_tunnel_cast(netdev_); + if (!dev->connected) + return EAGAIN; + for (;;) { + ssize_t retval; + retval = send(dev->sockfd, buffer, size, 0); + VLOG_DBG("%s: send(%"PRIxPTR", %zu) = %zd", + netdev_get_name(netdev_), (uintptr_t)buffer, size, retval); + if (retval >= 0) { + dev->stats.tx_packets++; + dev->stats.tx_bytes += retval; + if (retval != size) { + VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%zd bytes of " + "%zu) on %s", retval, size, netdev_get_name(netdev_)); + dev->stats.tx_errors++; + } + return 0; + } else if (errno != EINTR) { + if (errno != EAGAIN) { + VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: %s", + netdev_get_name(netdev_), ovs_strerror(errno)); + dev->stats.tx_errors++; } - } else { - VLOG_WARN("%s: unknown %s argument '%s'", name, type, node->name); + return errno; } } +} - if (!config->daddr) { - VLOG_WARN("%s: %s type requires valid 'remote_ip' argument", name, type); - return EINVAL; +static void +netdev_tunnel_send_wait(struct netdev *netdev_) +{ + struct netdev_tunnel *dev = netdev_tunnel_cast(netdev_); + if (dev->sockfd >= 0) { + poll_fd_wait(dev->sockfd, POLLOUT); } +} +static int +netdev_rx_tunnel_drain(struct netdev_rx *rx_) +{ + struct netdev_tunnel *netdev = + netdev_tunnel_cast(rx_->netdev); + struct netdev_rx_tunnel *rx = + netdev_rx_tunnel_cast(rx_); + char buffer[128]; + int error; + + if (!netdev->connected) + return 0; + for (;;) { + error = recv(rx->fd, buffer, 128, MSG_TRUNC); + if (error) { + if (error == -EAGAIN) + break; + else if (error != -EMSGSIZE) + return error; + } + } return 0; } static int -netdev_tunnel_create(const char *name, const char *type, - const struct shash *args, struct netdev_dev **netdev_devp) +netdev_tunnel_set_etheraddr(struct netdev *netdev, + const uint8_t mac[ETH_ADDR_LEN]) { - int err; - struct odp_vport_add ova; - struct tnl_port_config port_config; - struct netdev_dev_tunnel *netdev_dev; - - ovs_strlcpy(ova.port_type, type, sizeof ova.port_type); - ovs_strlcpy(ova.devname, name, sizeof ova.devname); - ova.config = &port_config; - - err = parse_config(name, type, args, &port_config); - if (err) { - return err; + struct netdev_tunnel *dev = netdev_tunnel_cast(netdev); + + if (!eth_addr_equals(dev->hwaddr, mac)) { + memcpy(dev->hwaddr, mac, ETH_ADDR_LEN); + netdev_tunnel_update_seq(dev); } - err = netdev_vport_do_ioctl(ODP_VPORT_ADD, &ova); - if (err == EBUSY) { - VLOG_WARN("%s: destroying existing device", name); + return 0; +} - err = netdev_vport_do_ioctl(ODP_VPORT_DEL, ova.devname); - if (err) { - return err; - } +static int +netdev_tunnel_get_etheraddr(const struct netdev *netdev, + uint8_t mac[ETH_ADDR_LEN]) +{ + const struct netdev_tunnel *dev = netdev_tunnel_cast(netdev); - err = netdev_vport_do_ioctl(ODP_VPORT_ADD, &ova); - } + memcpy(mac, dev->hwaddr, ETH_ADDR_LEN); + return 0; +} - if (err) { - return err; - } - netdev_dev = xmalloc(sizeof *netdev_dev); - netdev_dev_init(&netdev_dev->netdev_dev, name, &netdev_gre_class); +static int +netdev_tunnel_get_stats(const struct netdev *netdev, struct netdev_stats *stats) +{ + const struct netdev_tunnel *dev = netdev_tunnel_cast(netdev); + + *stats = dev->stats; + return 0; +} - *netdev_devp = &netdev_dev->netdev_dev; +static int +netdev_tunnel_set_stats(struct netdev *netdev, const struct netdev_stats *stats) +{ + struct netdev_tunnel *dev = netdev_tunnel_cast(netdev); + + dev->stats = *stats; return 0; } static int -netdev_tunnel_reconfigure(struct netdev_dev *netdev_dev_, const struct shash *args) +netdev_tunnel_update_flags(struct netdev *dev_, + enum netdev_flags off, enum netdev_flags on, + enum netdev_flags *old_flagsp) { - const char *name = netdev_dev_get_name(netdev_dev_); - struct odp_vport_mod ovm; - struct tnl_port_config port_config; - int err; - - ovs_strlcpy(ovm.devname, name, sizeof ovm.devname); - ovm.config = &port_config; - - err = parse_config(name, netdev_dev_get_class(netdev_dev_)->type, args, - &port_config); - if (err) { - return err; + struct netdev_tunnel *netdev = + netdev_tunnel_cast(dev_); + + if ((off | on) & ~(NETDEV_UP | NETDEV_PROMISC)) { + return EINVAL; } - return netdev_vport_do_ioctl(ODP_VPORT_MOD, &ovm); + // XXX should we actually do something with these flags? + *old_flagsp = netdev->flags; + netdev->flags |= on; + netdev->flags &= ~off; + if (*old_flagsp != netdev->flags) { + netdev_tunnel_update_seq(netdev); + } + return 0; } +static unsigned int +netdev_tunnel_change_seq(const struct netdev *netdev) +{ + return netdev_tunnel_cast(netdev)->change_seq; +} + +/* Helper functions. */ + static void -netdev_tunnel_destroy(struct netdev_dev *netdev_dev_) +netdev_tunnel_update_seq(struct netdev_tunnel *dev) { - struct netdev_dev_tunnel *netdev_dev = netdev_dev_tunnel_cast(netdev_dev_); + dev->change_seq++; + if (!dev->change_seq) { + dev->change_seq++; + } +} - netdev_vport_do_ioctl(ODP_VPORT_DEL, (char *)netdev_dev_get_name(netdev_dev_)); - free(netdev_dev); +static void +netdev_tunnel_get_port(struct unixctl_conn *conn, + int argc OVS_UNUSED, const char *argv[], void *aux OVS_UNUSED) +{ + struct netdev_tunnel *tunnel_dev; + char buf[6]; + + tunnel_dev = shash_find_data(&tunnel_netdevs, argv[1]); + if (!tunnel_dev) { + unixctl_command_reply_error(conn, "no such tunnel netdev"); + return; + } + + sprintf(buf, "%d", ntohs(tunnel_dev->local_addr.sin_port)); + unixctl_command_reply(conn, buf); } -static int -netdev_tunnel_open(struct netdev_dev *netdev_dev_, int ethertype OVS_UNUSED, - struct netdev **netdevp) +static void +netdev_tunnel_get_tx_bytes(struct unixctl_conn *conn, + int argc OVS_UNUSED, const char *argv[], void *aux OVS_UNUSED) { - struct netdev_tunnel *netdev; + struct netdev_tunnel *tunnel_dev; + char buf[128]; - netdev = xmalloc(sizeof *netdev); - netdev_init(&netdev->netdev, netdev_dev_); + tunnel_dev = shash_find_data(&tunnel_netdevs, argv[1]); + if (!tunnel_dev) { + unixctl_command_reply_error(conn, "no such tunnel netdev"); + return; + } - *netdevp = &netdev->netdev; - return 0; + sprintf(buf, "%"PRIu64, tunnel_dev->stats.tx_bytes); + unixctl_command_reply(conn, buf); } static void -netdev_tunnel_close(struct netdev *netdev_) +netdev_tunnel_get_rx_bytes(struct unixctl_conn *conn, + int argc OVS_UNUSED, const char *argv[], void *aux OVS_UNUSED) { - struct netdev_tunnel *netdev = netdev_tunnel_cast(netdev_); - free(netdev); + struct netdev_tunnel *tunnel_dev; + char buf[128]; + + tunnel_dev = shash_find_data(&tunnel_netdevs, argv[1]); + if (!tunnel_dev) { + unixctl_command_reply_error(conn, "no such tunnel netdev"); + return; + } + + sprintf(buf, "%"PRIu64, tunnel_dev->stats.rx_bytes); + unixctl_command_reply(conn, buf); } -const struct netdev_class netdev_gre_class = { - "gre", - NULL, /* init */ +static int +netdev_tunnel_init(void) +{ + unixctl_command_register("netdev-tunnel/get-port", "NAME", + 1, 1, netdev_tunnel_get_port, NULL); + unixctl_command_register("netdev-tunnel/get-tx-bytes", "NAME", + 1, 1, netdev_tunnel_get_tx_bytes, NULL); + unixctl_command_register("netdev-tunnel/get-rx-bytes", "NAME", + 1, 1, netdev_tunnel_get_rx_bytes, NULL); + 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_reconfigure, - - netdev_tunnel_open, - netdev_tunnel_close, - - NULL, /* enumerate */ + netdev_tunnel_get_config, + netdev_tunnel_set_config, + NULL, /* get_tunnel_config */ - NULL, /* recv */ - NULL, /* recv_wait */ - NULL, /* drain */ + netdev_tunnel_rx_open, - NULL, /* send */ - NULL, /* send_wait */ + netdev_tunnel_send, + netdev_tunnel_send_wait, - netdev_vport_set_etheraddr, - netdev_vport_get_etheraddr, - netdev_vport_get_mtu, + netdev_tunnel_set_etheraddr, + netdev_tunnel_get_etheraddr, + NULL, /* get_mtu */ + NULL, /* set_mtu */ NULL, /* get_ifindex */ - netdev_vport_get_carrier, - netdev_vport_get_stats, - netdev_vport_set_stats, + NULL, /* get_carrier */ + NULL, /* get_carrier_resets */ + NULL, /* get_miimon */ + netdev_tunnel_get_stats, + netdev_tunnel_set_stats, NULL, /* get_features */ NULL, /* set_advertisements */ - NULL, /* get_vlan_vid */ NULL, /* set_policing */ NULL, /* get_qos_types */ @@ -276,10 +530,19 @@ const struct netdev_class netdev_gre_class = { NULL, /* get_in6 */ NULL, /* add_router */ NULL, /* get_next_hop */ + NULL, /* get_drv_info */ NULL, /* arp_lookup */ - netdev_vport_update_flags, + netdev_tunnel_update_flags, - netdev_vport_poll_add, - netdev_vport_poll_remove, + netdev_tunnel_change_seq }; + + +static const struct netdev_rx_class netdev_rx_tunnel_class = { + netdev_rx_tunnel_destroy, + netdev_rx_tunnel_recv, + netdev_rx_tunnel_wait, + netdev_rx_tunnel_drain, +}; +