X-Git-Url: http://git.cascardo.eti.br/?a=blobdiff_plain;f=lib%2Fnetdev-vport.c;h=a9eb3eb14bc5eb24379ca92e288688c45385b777;hb=35303d715b1f0db46e6a27146815061a60385dc6;hp=bb9e510843b535eb87e9cea5430700f23f962dd5;hpb=cceb11f5b12d09cc8afc87ca4fd03e941234d439;p=cascardo%2Fovs.git diff --git a/lib/netdev-vport.c b/lib/netdev-vport.c index bb9e51084..a3394dd63 100644 --- a/lib/netdev-vport.c +++ b/lib/netdev-vport.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2010, 2011 Nicira Networks. + * Copyright (c) 2010, 2011, 2012, 2013, 2014 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -21,829 +21,1357 @@ #include #include #include -#include #include #include #include "byte-order.h" +#include "csum.h" +#include "daemon.h" +#include "dirs.h" +#include "dpif.h" +#include "dp-packet.h" +#include "dynamic-string.h" +#include "flow.h" #include "hash.h" #include "hmap.h" #include "list.h" #include "netdev-provider.h" -#include "netlink.h" -#include "netlink-socket.h" -#include "ofpbuf.h" -#include "openvswitch/datapath-protocol.h" -#include "openvswitch/tunnel.h" +#include "odp-netlink.h" +#include "dp-packet.h" +#include "ovs-router.h" #include "packets.h" -#include "rtnetlink.h" +#include "poll-loop.h" #include "route-table.h" -#include "rtnetlink-link.h" #include "shash.h" #include "socket-util.h" -#include "vlog.h" +#include "openvswitch/vlog.h" +#include "unaligned.h" +#include "unixctl.h" +#include "util.h" VLOG_DEFINE_THIS_MODULE(netdev_vport); +static struct vlog_rate_limit err_rl = VLOG_RATE_LIMIT_INIT(60, 5); -static struct hmap name_map; -static struct rtnetlink_notifier netdev_vport_link_notifier; +#define GENEVE_DST_PORT 6081 +#define VXLAN_DST_PORT 4789 +#define LISP_DST_PORT 4341 +#define STT_DST_PORT 7471 -struct name_node { - struct hmap_node node; /* Node in name_map. */ - uint32_t ifi_index; /* Kernel interface index. */ +#define VXLAN_HLEN (sizeof(struct eth_header) + \ + sizeof(struct ip_header) + \ + sizeof(struct udp_header) + \ + sizeof(struct vxlanhdr)) - char ifname[IFNAMSIZ]; /* Interface name. */ -}; - -struct netdev_vport_notifier { - struct netdev_notifier notifier; - struct list list_node; - struct shash_node *shash_node; -}; +#define GENEVE_BASE_HLEN (sizeof(struct eth_header) + \ + sizeof(struct ip_header) + \ + sizeof(struct udp_header) + \ + sizeof(struct genevehdr)) -struct netdev_dev_vport { - struct netdev_dev netdev_dev; - uint64_t config[VPORT_CONFIG_SIZE / 8]; -}; +#define DEFAULT_TTL 64 struct netdev_vport { - struct netdev netdev; + struct netdev up; + + /* Protects all members below. */ + struct ovs_mutex mutex; + + uint8_t etheraddr[ETH_ADDR_LEN]; + struct netdev_stats stats; + + /* Tunnels. */ + struct netdev_tunnel_config tnl_cfg; + char egress_iface[IFNAMSIZ]; + bool carrier_status; + + /* Patch Ports. */ + char *peer; }; struct vport_class { + const char *dpif_port; struct netdev_class netdev_class; - int (*parse_config)(const struct netdev_dev *, const struct shash *args, - void *config); }; -static struct shash netdev_vport_notifiers = - SHASH_INITIALIZER(&netdev_vport_notifiers); - -static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); +/* Last read of the route-table's change number. */ +static uint64_t rt_change_seqno; -static int netdev_vport_do_ioctl(int cmd, void *arg); -static int netdev_vport_create(const struct netdev_class *, const char *, - const struct shash *, struct netdev_dev **); -static void netdev_vport_poll_notify(const struct netdev *); +static int netdev_vport_construct(struct netdev *); +static int get_patch_config(const struct netdev *netdev, struct smap *args); +static int get_tunnel_config(const struct netdev *, struct smap *args); +static bool tunnel_check_status_change__(struct netdev_vport *); -static void netdev_vport_tnl_iface_init(void); -static void netdev_vport_link_change(const struct rtnetlink_link_change *, - void *); -static const char *netdev_vport_get_tnl_iface(const struct netdev *netdev); +static uint16_t tnl_udp_port_min = 32768; +static uint16_t tnl_udp_port_max = 61000; static bool is_vport_class(const struct netdev_class *class) { - return class->create == netdev_vport_create; + return class->construct == netdev_vport_construct; +} + +bool +netdev_vport_is_vport_class(const struct netdev_class *class) +{ + return is_vport_class(class); } static const struct vport_class * vport_class_cast(const struct netdev_class *class) { - assert(is_vport_class(class)); + ovs_assert(is_vport_class(class)); return CONTAINER_OF(class, struct vport_class, netdev_class); } -static struct netdev_dev_vport * -netdev_dev_vport_cast(const struct netdev_dev *netdev_dev) +static struct netdev_vport * +netdev_vport_cast(const struct netdev *netdev) { - assert(is_vport_class(netdev_dev_get_class(netdev_dev))); - return CONTAINER_OF(netdev_dev, struct netdev_dev_vport, netdev_dev); + ovs_assert(is_vport_class(netdev_get_class(netdev))); + return CONTAINER_OF(netdev, struct netdev_vport, up); } -static struct netdev_vport * -netdev_vport_cast(const struct netdev *netdev) +static const struct netdev_tunnel_config * +get_netdev_tunnel_config(const struct netdev *netdev) { - struct netdev_dev *netdev_dev = netdev_get_dev(netdev); - assert(is_vport_class(netdev_dev_get_class(netdev_dev))); - return CONTAINER_OF(netdev, struct netdev_vport, netdev); + return &netdev_vport_cast(netdev)->tnl_cfg; } -/* If 'netdev' is a vport netdev, copies its kernel configuration into - * 'config'. Otherwise leaves 'config' untouched. */ -void -netdev_vport_get_config(const struct netdev *netdev, void *config) +bool +netdev_vport_is_patch(const struct netdev *netdev) { - const struct netdev_dev *dev = netdev_get_dev(netdev); + const struct netdev_class *class = netdev_get_class(netdev); - if (is_vport_class(netdev_dev_get_class(dev))) { - const struct netdev_dev_vport *vport = netdev_dev_vport_cast(dev); - memcpy(config, vport->config, VPORT_CONFIG_SIZE); - } + return class->get_config == get_patch_config; } -static int -netdev_vport_init(void) +bool +netdev_vport_is_layer3(const struct netdev *dev) { - netdev_vport_tnl_iface_init(); - route_table_register(); - return 0; + const char *type = netdev_get_type(dev); + + return (!strcmp("lisp", type)); } -static int -netdev_vport_create(const struct netdev_class *netdev_class, const char *name, - const struct shash *args, - struct netdev_dev **netdev_devp) +static bool +netdev_vport_needs_dst_port(const struct netdev *dev) { - const struct vport_class *vport_class = vport_class_cast(netdev_class); - struct netdev_dev_vport *dev; - int error; + const struct netdev_class *class = netdev_get_class(dev); + const char *type = netdev_get_type(dev); - dev = xmalloc(sizeof *dev); - *netdev_devp = &dev->netdev_dev; - netdev_dev_init(&dev->netdev_dev, name, netdev_class); + return (class->get_config == get_tunnel_config && + (!strcmp("geneve", type) || !strcmp("vxlan", type) || + !strcmp("lisp", type) || !strcmp("stt", type)) ); +} - memset(dev->config, 0, sizeof dev->config); - error = vport_class->parse_config(&dev->netdev_dev, args, dev->config); +const char * +netdev_vport_class_get_dpif_port(const struct netdev_class *class) +{ + return is_vport_class(class) ? vport_class_cast(class)->dpif_port : NULL; +} + +const char * +netdev_vport_get_dpif_port(const struct netdev *netdev, + char namebuf[], size_t bufsize) +{ + const struct netdev_class *class = netdev_get_class(netdev); + const char *dpif_port = netdev_vport_class_get_dpif_port(class); - if (error) { - netdev_dev_uninit(&dev->netdev_dev, true); + if (!dpif_port) { + return netdev_get_name(netdev); } - return error; + + if (netdev_vport_needs_dst_port(netdev)) { + const struct netdev_vport *vport = netdev_vport_cast(netdev); + + /* + * Note: IFNAMSIZ is 16 bytes long. Implementations should choose + * a dpif port name that is short enough to fit including any + * port numbers but assert just in case. + */ + BUILD_ASSERT(NETDEV_VPORT_NAME_BUFSIZE >= IFNAMSIZ); + ovs_assert(strlen(dpif_port) + 6 < IFNAMSIZ); + snprintf(namebuf, bufsize, "%s_%d", dpif_port, + ntohs(vport->tnl_cfg.dst_port)); + return namebuf; + } else { + return dpif_port; + } +} + +char * +netdev_vport_get_dpif_port_strdup(const struct netdev *netdev) +{ + char namebuf[NETDEV_VPORT_NAME_BUFSIZE]; + + return xstrdup(netdev_vport_get_dpif_port(netdev, namebuf, + sizeof namebuf)); } +/* Whenever the route-table change number is incremented, + * netdev_vport_route_changed() should be called to update + * the corresponding tunnel interface status. */ static void -netdev_vport_destroy(struct netdev_dev *netdev_dev_) +netdev_vport_route_changed(void) { - struct netdev_dev_vport *netdev_dev = netdev_dev_vport_cast(netdev_dev_); + struct netdev **vports; + size_t i, n_vports; + + vports = netdev_get_vports(&n_vports); + for (i = 0; i < n_vports; i++) { + struct netdev *netdev_ = vports[i]; + struct netdev_vport *netdev = netdev_vport_cast(netdev_); + + ovs_mutex_lock(&netdev->mutex); + /* Finds all tunnel vports. */ + if (netdev->tnl_cfg.ip_dst) { + if (tunnel_check_status_change__(netdev)) { + netdev_change_seq_changed(netdev_); + } + } + ovs_mutex_unlock(&netdev->mutex); + + netdev_close(netdev_); + } - route_table_unregister(); - free(netdev_dev); + free(vports); } -static int -netdev_vport_open(struct netdev_dev *netdev_dev_, int ethertype OVS_UNUSED, - struct netdev **netdevp) +static struct netdev * +netdev_vport_alloc(void) { - struct netdev_vport *netdev; + struct netdev_vport *netdev = xzalloc(sizeof *netdev); + return &netdev->up; +} - netdev = xmalloc(sizeof *netdev); - netdev_init(&netdev->netdev, netdev_dev_); +static int +netdev_vport_construct(struct netdev *netdev_) +{ + struct netdev_vport *dev = netdev_vport_cast(netdev_); + const char *type = netdev_get_type(netdev_); + + ovs_mutex_init(&dev->mutex); + eth_addr_random(dev->etheraddr); + + /* Add a default destination port for tunnel ports if none specified. */ + if (!strcmp(type, "geneve")) { + dev->tnl_cfg.dst_port = htons(GENEVE_DST_PORT); + } else if (!strcmp(type, "vxlan")) { + dev->tnl_cfg.dst_port = htons(VXLAN_DST_PORT); + } else if (!strcmp(type, "lisp")) { + dev->tnl_cfg.dst_port = htons(LISP_DST_PORT); + } else if (!strcmp(type, "stt")) { + dev->tnl_cfg.dst_port = htons(STT_DST_PORT); + } - *netdevp = &netdev->netdev; + dev->tnl_cfg.dont_fragment = true; + dev->tnl_cfg.ttl = DEFAULT_TTL; return 0; } static void -netdev_vport_close(struct netdev *netdev_) +netdev_vport_destruct(struct netdev *netdev_) { struct netdev_vport *netdev = netdev_vport_cast(netdev_); - free(netdev); + + free(netdev->peer); + ovs_mutex_destroy(&netdev->mutex); } -static int -netdev_vport_reconfigure(struct netdev_dev *dev_, - const struct shash *args) -{ - const struct netdev_class *netdev_class = netdev_dev_get_class(dev_); - const struct vport_class *vport_class = vport_class_cast(netdev_class); - struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_); - struct odp_port port; - int error; - - memset(&port, 0, sizeof port); - strncpy(port.devname, netdev_dev_get_name(dev_), sizeof port.devname); - strncpy(port.type, netdev_dev_get_type(dev_), sizeof port.type); - error = vport_class->parse_config(dev_, args, port.config); - if (!error && memcmp(port.config, dev->config, sizeof dev->config)) { - error = netdev_vport_do_ioctl(ODP_VPORT_MOD, &port); - if (!error || error == ENODEV) { - /* Either reconfiguration succeeded or this vport is not installed - * in the kernel (e.g. it hasn't been added to a dpif yet with - * dpif_port_add()). */ - memcpy(dev->config, port.config, sizeof dev->config); - } - } - return error; +static void +netdev_vport_dealloc(struct netdev *netdev_) +{ + struct netdev_vport *netdev = netdev_vport_cast(netdev_); + free(netdev); } static int -netdev_vport_set_etheraddr(struct netdev *netdev, +netdev_vport_set_etheraddr(struct netdev *netdev_, const uint8_t mac[ETH_ADDR_LEN]) { - struct odp_vport_ether vport_ether; - int err; - - ovs_strlcpy(vport_ether.devname, netdev_get_name(netdev), - sizeof vport_ether.devname); + struct netdev_vport *netdev = netdev_vport_cast(netdev_); - memcpy(vport_ether.ether_addr, mac, ETH_ADDR_LEN); + ovs_mutex_lock(&netdev->mutex); + memcpy(netdev->etheraddr, mac, ETH_ADDR_LEN); + ovs_mutex_unlock(&netdev->mutex); + netdev_change_seq_changed(netdev_); - err = netdev_vport_do_ioctl(ODP_VPORT_ETHER_SET, &vport_ether); - if (err) { - return err; - } - - netdev_vport_poll_notify(netdev); return 0; } static int -netdev_vport_get_etheraddr(const struct netdev *netdev, +netdev_vport_get_etheraddr(const struct netdev *netdev_, uint8_t mac[ETH_ADDR_LEN]) { - struct odp_vport_ether vport_ether; - int err; + struct netdev_vport *netdev = netdev_vport_cast(netdev_); - ovs_strlcpy(vport_ether.devname, netdev_get_name(netdev), - sizeof vport_ether.devname); + ovs_mutex_lock(&netdev->mutex); + memcpy(mac, netdev->etheraddr, ETH_ADDR_LEN); + ovs_mutex_unlock(&netdev->mutex); - err = netdev_vport_do_ioctl(ODP_VPORT_ETHER_GET, &vport_ether); - if (err) { - return err; + return 0; +} + +/* Checks if the tunnel status has changed and returns a boolean. + * Updates the tunnel status if it has changed. */ +static bool +tunnel_check_status_change__(struct netdev_vport *netdev) + OVS_REQUIRES(netdev->mutex) +{ + char iface[IFNAMSIZ]; + bool status = false; + ovs_be32 route; + ovs_be32 gw; + + iface[0] = '\0'; + route = netdev->tnl_cfg.ip_dst; + if (ovs_router_lookup(route, iface, &gw)) { + struct netdev *egress_netdev; + + if (!netdev_open(iface, "system", &egress_netdev)) { + status = netdev_get_carrier(egress_netdev); + netdev_close(egress_netdev); + } } - memcpy(mac, vport_ether.ether_addr, ETH_ADDR_LEN); - return 0; + if (strcmp(netdev->egress_iface, iface) + || netdev->carrier_status != status) { + ovs_strlcpy(netdev->egress_iface, iface, IFNAMSIZ); + netdev->carrier_status = status; + + return true; + } + + return false; } static int -netdev_vport_get_mtu(const struct netdev *netdev, int *mtup) +tunnel_get_status(const struct netdev *netdev_, struct smap *smap) { - struct odp_vport_mtu vport_mtu; - int err; + struct netdev_vport *netdev = netdev_vport_cast(netdev_); - ovs_strlcpy(vport_mtu.devname, netdev_get_name(netdev), - sizeof vport_mtu.devname); + if (netdev->egress_iface[0]) { + smap_add(smap, "tunnel_egress_iface", netdev->egress_iface); - err = netdev_vport_do_ioctl(ODP_VPORT_MTU_GET, &vport_mtu); - if (err) { - return err; + smap_add(smap, "tunnel_egress_iface_carrier", + netdev->carrier_status ? "up" : "down"); } - *mtup = vport_mtu.mtu; return 0; } -int -netdev_vport_get_stats(const struct netdev *netdev, struct netdev_stats *stats) +static int +netdev_vport_update_flags(struct netdev *netdev OVS_UNUSED, + enum netdev_flags off, + enum netdev_flags on OVS_UNUSED, + enum netdev_flags *old_flagsp) { - const char *name = netdev_get_name(netdev); - struct odp_vport_stats_req ovsr; - int err; - - ovs_strlcpy(ovsr.devname, name, sizeof ovsr.devname); - err = netdev_vport_do_ioctl(ODP_VPORT_STATS_GET, &ovsr); - if (err) { - return err; + if (off & (NETDEV_UP | NETDEV_PROMISC)) { + return EOPNOTSUPP; } - stats->rx_packets = ovsr.stats.rx_packets; - stats->tx_packets = ovsr.stats.tx_packets; - stats->rx_bytes = ovsr.stats.rx_bytes; - stats->tx_bytes = ovsr.stats.tx_bytes; - stats->rx_errors = ovsr.stats.rx_errors; - stats->tx_errors = ovsr.stats.tx_errors; - stats->rx_dropped = ovsr.stats.rx_dropped; - stats->tx_dropped = ovsr.stats.tx_dropped; - stats->multicast = ovsr.stats.multicast; - stats->collisions = ovsr.stats.collisions; - stats->rx_length_errors = ovsr.stats.rx_length_errors; - stats->rx_over_errors = ovsr.stats.rx_over_errors; - stats->rx_crc_errors = ovsr.stats.rx_crc_errors; - stats->rx_frame_errors = ovsr.stats.rx_frame_errors; - stats->rx_fifo_errors = ovsr.stats.rx_fifo_errors; - stats->rx_missed_errors = ovsr.stats.rx_missed_errors; - stats->tx_aborted_errors = ovsr.stats.tx_aborted_errors; - stats->tx_carrier_errors = ovsr.stats.tx_carrier_errors; - stats->tx_fifo_errors = ovsr.stats.tx_fifo_errors; - stats->tx_heartbeat_errors = ovsr.stats.tx_heartbeat_errors; - stats->tx_window_errors = ovsr.stats.tx_window_errors; - + *old_flagsp = NETDEV_UP | NETDEV_PROMISC; return 0; } -int -netdev_vport_set_stats(struct netdev *netdev, const struct netdev_stats *stats) +static void +netdev_vport_run(void) { - struct odp_vport_stats_req ovsr; - int err; + uint64_t seq; - ovs_strlcpy(ovsr.devname, netdev_get_name(netdev), sizeof ovsr.devname); - - ovsr.stats.rx_packets = stats->rx_packets; - ovsr.stats.tx_packets = stats->tx_packets; - ovsr.stats.rx_bytes = stats->rx_bytes; - ovsr.stats.tx_bytes = stats->tx_bytes; - ovsr.stats.rx_errors = stats->rx_errors; - ovsr.stats.tx_errors = stats->tx_errors; - ovsr.stats.rx_dropped = stats->rx_dropped; - ovsr.stats.tx_dropped = stats->tx_dropped; - ovsr.stats.multicast = stats->multicast; - ovsr.stats.collisions = stats->collisions; - ovsr.stats.rx_length_errors = stats->rx_length_errors; - ovsr.stats.rx_over_errors = stats->rx_over_errors; - ovsr.stats.rx_crc_errors = stats->rx_crc_errors; - ovsr.stats.rx_frame_errors = stats->rx_frame_errors; - ovsr.stats.rx_fifo_errors = stats->rx_fifo_errors; - ovsr.stats.rx_missed_errors = stats->rx_missed_errors; - ovsr.stats.tx_aborted_errors = stats->tx_aborted_errors; - ovsr.stats.tx_carrier_errors = stats->tx_carrier_errors; - ovsr.stats.tx_fifo_errors = stats->tx_fifo_errors; - ovsr.stats.tx_heartbeat_errors = stats->tx_heartbeat_errors; - ovsr.stats.tx_window_errors = stats->tx_window_errors; - - err = netdev_vport_do_ioctl(ODP_VPORT_STATS_SET, &ovsr); - - /* If the vport layer doesn't know about the device, that doesn't mean it - * doesn't exist (after all were able to open it when netdev_open() was - * called), it just means that it isn't attached and we'll be getting - * stats a different way. */ - if (err == ENODEV) { - err = EOPNOTSUPP; - } - - return err; + route_table_run(); + seq = route_table_get_change_seq(); + if (rt_change_seqno != seq) { + rt_change_seqno = seq; + netdev_vport_route_changed(); + } } -static int -netdev_vport_get_status(const struct netdev *netdev, struct shash *sh) +static void +netdev_vport_wait(void) { - const char *iface = netdev_vport_get_tnl_iface(netdev); + uint64_t seq; - if (iface) { - struct netdev *egress_netdev; + route_table_wait(); + seq = route_table_get_change_seq(); + if (rt_change_seqno != seq) { + poll_immediate_wake(); + } +} + +/* Code specific to tunnel types. */ - shash_add(sh, "tunnel_egress_iface", xstrdup(iface)); +static ovs_be64 +parse_key(const struct smap *args, const char *name, + bool *present, bool *flow) +{ + const char *s; - if (!netdev_open_default(iface, &egress_netdev)) { - shash_add(sh, "tunnel_egress_iface_carrier", - xstrdup(netdev_get_carrier(egress_netdev) - ? "up" : "down")); - netdev_close(egress_netdev); + *present = false; + *flow = false; + + s = smap_get(args, name); + if (!s) { + s = smap_get(args, "key"); + if (!s) { + return 0; } } - return 0; + *present = true; + + if (!strcmp(s, "flow")) { + *flow = true; + return 0; + } else { + return htonll(strtoull(s, NULL, 0)); + } } static int -netdev_vport_update_flags(struct netdev *netdev OVS_UNUSED, - enum netdev_flags off, enum netdev_flags on OVS_UNUSED, - enum netdev_flags *old_flagsp) +set_tunnel_config(struct netdev *dev_, const struct smap *args) { - if (off & (NETDEV_UP | NETDEV_PROMISC)) { - return EOPNOTSUPP; + struct netdev_vport *dev = netdev_vport_cast(dev_); + const char *name = netdev_get_name(dev_); + const char *type = netdev_get_type(dev_); + bool ipsec_mech_set, needs_dst_port, has_csum; + struct netdev_tunnel_config tnl_cfg; + struct smap_node *node; + + has_csum = strstr(type, "gre") || strstr(type, "geneve") || + strstr(type, "stt") || strstr(type, "vxlan"); + ipsec_mech_set = false; + memset(&tnl_cfg, 0, sizeof tnl_cfg); + + /* Add a default destination port for tunnel ports if none specified. */ + if (!strcmp(type, "geneve")) { + tnl_cfg.dst_port = htons(GENEVE_DST_PORT); } - *old_flagsp = NETDEV_UP | NETDEV_PROMISC; + if (!strcmp(type, "vxlan")) { + tnl_cfg.dst_port = htons(VXLAN_DST_PORT); + } + + if (!strcmp(type, "lisp")) { + tnl_cfg.dst_port = htons(LISP_DST_PORT); + } + + if (!strcmp(type, "stt")) { + tnl_cfg.dst_port = htons(STT_DST_PORT); + } + + needs_dst_port = netdev_vport_needs_dst_port(dev_); + tnl_cfg.ipsec = strstr(type, "ipsec"); + tnl_cfg.dont_fragment = true; + + SMAP_FOR_EACH (node, args) { + if (!strcmp(node->key, "remote_ip")) { + struct in_addr in_addr; + if (!strcmp(node->value, "flow")) { + tnl_cfg.ip_dst_flow = true; + tnl_cfg.ip_dst = htonl(0); + } else if (lookup_ip(node->value, &in_addr)) { + VLOG_WARN("%s: bad %s 'remote_ip'", name, type); + } else if (ip_is_multicast(in_addr.s_addr)) { + VLOG_WARN("%s: multicast remote_ip="IP_FMT" not allowed", + name, IP_ARGS(in_addr.s_addr)); + return EINVAL; + } else { + tnl_cfg.ip_dst = in_addr.s_addr; + } + } else if (!strcmp(node->key, "local_ip")) { + struct in_addr in_addr; + if (!strcmp(node->value, "flow")) { + tnl_cfg.ip_src_flow = true; + tnl_cfg.ip_src = htonl(0); + } else if (lookup_ip(node->value, &in_addr)) { + VLOG_WARN("%s: bad %s 'local_ip'", name, type); + } else { + tnl_cfg.ip_src = in_addr.s_addr; + } + } else if (!strcmp(node->key, "tos")) { + if (!strcmp(node->value, "inherit")) { + tnl_cfg.tos_inherit = true; + } else { + char *endptr; + int tos; + tos = strtol(node->value, &endptr, 0); + if (*endptr == '\0' && tos == (tos & IP_DSCP_MASK)) { + tnl_cfg.tos = tos; + } else { + VLOG_WARN("%s: invalid TOS %s", name, node->value); + } + } + } else if (!strcmp(node->key, "ttl")) { + if (!strcmp(node->value, "inherit")) { + tnl_cfg.ttl_inherit = true; + } else { + tnl_cfg.ttl = atoi(node->value); + } + } else if (!strcmp(node->key, "dst_port") && needs_dst_port) { + tnl_cfg.dst_port = htons(atoi(node->value)); + } else if (!strcmp(node->key, "csum") && has_csum) { + if (!strcmp(node->value, "true")) { + tnl_cfg.csum = true; + } + } else if (!strcmp(node->key, "df_default")) { + if (!strcmp(node->value, "false")) { + tnl_cfg.dont_fragment = false; + } + } else if (!strcmp(node->key, "peer_cert") && tnl_cfg.ipsec) { + if (smap_get(args, "certificate")) { + ipsec_mech_set = true; + } else { + const char *use_ssl_cert; + + /* If the "use_ssl_cert" is true, then "certificate" and + * "private_key" will be pulled from the SSL table. The + * use of this option is strongly discouraged, since it + * will like be removed when multiple SSL configurations + * are supported by OVS. + */ + use_ssl_cert = smap_get(args, "use_ssl_cert"); + if (!use_ssl_cert || strcmp(use_ssl_cert, "true")) { + VLOG_ERR("%s: 'peer_cert' requires 'certificate' argument", + name); + return EINVAL; + } + ipsec_mech_set = true; + } + } else if (!strcmp(node->key, "psk") && tnl_cfg.ipsec) { + ipsec_mech_set = true; + } else if (tnl_cfg.ipsec + && (!strcmp(node->key, "certificate") + || !strcmp(node->key, "private_key") + || !strcmp(node->key, "use_ssl_cert"))) { + /* Ignore options not used by the netdev. */ + } else if (!strcmp(node->key, "key") || + !strcmp(node->key, "in_key") || + !strcmp(node->key, "out_key")) { + /* Handled separately below. */ + } else if (!strcmp(node->key, "exts")) { + char *str = xstrdup(node->value); + char *ext, *save_ptr = NULL; + + tnl_cfg.exts = 0; + + ext = strtok_r(str, ",", &save_ptr); + while (ext) { + if (!strcmp(type, "vxlan") && !strcmp(ext, "gbp")) { + tnl_cfg.exts |= (1 << OVS_VXLAN_EXT_GBP); + } else { + VLOG_WARN("%s: unknown extension '%s'", name, ext); + } + + ext = strtok_r(NULL, ",", &save_ptr); + } + + free(str); + } else { + VLOG_WARN("%s: unknown %s argument '%s'", name, type, node->key); + } + } + + if (tnl_cfg.ipsec) { + static struct ovs_mutex mutex = OVS_MUTEX_INITIALIZER; + static pid_t pid = 0; + +#ifndef _WIN32 + ovs_mutex_lock(&mutex); + if (pid <= 0) { + char *file_name = xasprintf("%s/%s", ovs_rundir(), + "ovs-monitor-ipsec.pid"); + pid = read_pidfile(file_name); + free(file_name); + } + ovs_mutex_unlock(&mutex); +#endif + + if (pid < 0) { + VLOG_ERR("%s: IPsec requires the ovs-monitor-ipsec daemon", + name); + return EINVAL; + } + + if (smap_get(args, "peer_cert") && smap_get(args, "psk")) { + VLOG_ERR("%s: cannot define both 'peer_cert' and 'psk'", name); + return EINVAL; + } + + if (!ipsec_mech_set) { + VLOG_ERR("%s: IPsec requires an 'peer_cert' or psk' argument", + name); + return EINVAL; + } + } + + if (!tnl_cfg.ip_dst && !tnl_cfg.ip_dst_flow) { + VLOG_ERR("%s: %s type requires valid 'remote_ip' argument", + name, type); + return EINVAL; + } + if (tnl_cfg.ip_src_flow && !tnl_cfg.ip_dst_flow) { + VLOG_ERR("%s: %s type requires 'remote_ip=flow' with 'local_ip=flow'", + name, type); + return EINVAL; + } + if (!tnl_cfg.ttl) { + tnl_cfg.ttl = DEFAULT_TTL; + } + + tnl_cfg.in_key = parse_key(args, "in_key", + &tnl_cfg.in_key_present, + &tnl_cfg.in_key_flow); + + tnl_cfg.out_key = parse_key(args, "out_key", + &tnl_cfg.out_key_present, + &tnl_cfg.out_key_flow); + + ovs_mutex_lock(&dev->mutex); + if (memcmp(&dev->tnl_cfg, &tnl_cfg, sizeof tnl_cfg)) { + dev->tnl_cfg = tnl_cfg; + tunnel_check_status_change__(dev); + netdev_change_seq_changed(dev_); + } + ovs_mutex_unlock(&dev->mutex); + return 0; } -static char * -make_poll_name(const struct netdev *netdev) +static int +get_tunnel_config(const struct netdev *dev, struct smap *args) { - return xasprintf("%s:%s", netdev_get_type(netdev), netdev_get_name(netdev)); -} + struct netdev_vport *netdev = netdev_vport_cast(dev); + struct netdev_tunnel_config tnl_cfg; -static int -netdev_vport_poll_add(struct netdev *netdev, - void (*cb)(struct netdev_notifier *), void *aux, - struct netdev_notifier **notifierp) -{ - char *poll_name = make_poll_name(netdev); - struct netdev_vport_notifier *notifier; - struct list *list; - struct shash_node *shash_node; - - shash_node = shash_find_data(&netdev_vport_notifiers, poll_name); - if (!shash_node) { - list = xmalloc(sizeof *list); - list_init(list); - shash_node = shash_add(&netdev_vport_notifiers, poll_name, list); + ovs_mutex_lock(&netdev->mutex); + tnl_cfg = netdev->tnl_cfg; + ovs_mutex_unlock(&netdev->mutex); + + if (tnl_cfg.ip_dst) { + smap_add_format(args, "remote_ip", IP_FMT, IP_ARGS(tnl_cfg.ip_dst)); + } else if (tnl_cfg.ip_dst_flow) { + smap_add(args, "remote_ip", "flow"); + } + + if (tnl_cfg.ip_src) { + smap_add_format(args, "local_ip", IP_FMT, IP_ARGS(tnl_cfg.ip_src)); + } else if (tnl_cfg.ip_src_flow) { + smap_add(args, "local_ip", "flow"); + } + + if (tnl_cfg.in_key_flow && tnl_cfg.out_key_flow) { + smap_add(args, "key", "flow"); + } else if (tnl_cfg.in_key_present && tnl_cfg.out_key_present + && tnl_cfg.in_key == tnl_cfg.out_key) { + smap_add_format(args, "key", "%"PRIu64, ntohll(tnl_cfg.in_key)); } else { - list = shash_node->data; + if (tnl_cfg.in_key_flow) { + smap_add(args, "in_key", "flow"); + } else if (tnl_cfg.in_key_present) { + smap_add_format(args, "in_key", "%"PRIu64, + ntohll(tnl_cfg.in_key)); + } + + if (tnl_cfg.out_key_flow) { + smap_add(args, "out_key", "flow"); + } else if (tnl_cfg.out_key_present) { + smap_add_format(args, "out_key", "%"PRIu64, + ntohll(tnl_cfg.out_key)); + } + } + + if (tnl_cfg.ttl_inherit) { + smap_add(args, "ttl", "inherit"); + } else if (tnl_cfg.ttl != DEFAULT_TTL) { + smap_add_format(args, "ttl", "%"PRIu8, tnl_cfg.ttl); + } + + if (tnl_cfg.tos_inherit) { + smap_add(args, "tos", "inherit"); + } else if (tnl_cfg.tos) { + smap_add_format(args, "tos", "0x%x", tnl_cfg.tos); } - notifier = xmalloc(sizeof *notifier); - netdev_notifier_init(¬ifier->notifier, netdev, cb, aux); - list_push_back(list, ¬ifier->list_node); - notifier->shash_node = shash_node; + if (tnl_cfg.dst_port) { + uint16_t dst_port = ntohs(tnl_cfg.dst_port); + const char *type = netdev_get_type(dev); - *notifierp = ¬ifier->notifier; - free(poll_name); + if ((!strcmp("geneve", type) && dst_port != GENEVE_DST_PORT) || + (!strcmp("vxlan", type) && dst_port != VXLAN_DST_PORT) || + (!strcmp("lisp", type) && dst_port != LISP_DST_PORT) || + (!strcmp("stt", type) && dst_port != STT_DST_PORT)) { + smap_add_format(args, "dst_port", "%d", dst_port); + } + } + + if (tnl_cfg.csum) { + smap_add(args, "csum", "true"); + } + + if (!tnl_cfg.dont_fragment) { + smap_add(args, "df_default", "false"); + } return 0; } + +/* Code specific to patch ports. */ -static void -netdev_vport_poll_remove(struct netdev_notifier *notifier_) +/* If 'netdev' is a patch port, returns the name of its peer as a malloc()'d + * string that the caller must free. + * + * If 'netdev' is not a patch port, returns NULL. */ +char * +netdev_vport_patch_peer(const struct netdev *netdev_) { - struct netdev_vport_notifier *notifier = - CONTAINER_OF(notifier_, struct netdev_vport_notifier, notifier); + char *peer = NULL; - struct list *list; + if (netdev_vport_is_patch(netdev_)) { + struct netdev_vport *netdev = netdev_vport_cast(netdev_); - list = list_remove(¬ifier->list_node); - if (list_is_empty(list)) { - shash_delete(&netdev_vport_notifiers, notifier->shash_node); - free(list); + ovs_mutex_lock(&netdev->mutex); + if (netdev->peer) { + peer = xstrdup(netdev->peer); + } + ovs_mutex_unlock(&netdev->mutex); } - free(notifier); + return peer; } -static void -netdev_vport_run(void) +void +netdev_vport_inc_rx(const struct netdev *netdev, + const struct dpif_flow_stats *stats) { - rtnetlink_link_notifier_run(); - route_table_run(); + if (is_vport_class(netdev_get_class(netdev))) { + struct netdev_vport *dev = netdev_vport_cast(netdev); + + ovs_mutex_lock(&dev->mutex); + dev->stats.rx_packets += stats->n_packets; + dev->stats.rx_bytes += stats->n_bytes; + ovs_mutex_unlock(&dev->mutex); + } } -static void -netdev_vport_wait(void) +void +netdev_vport_inc_tx(const struct netdev *netdev, + const struct dpif_flow_stats *stats) { - rtnetlink_link_notifier_wait(); - route_table_wait(); + if (is_vport_class(netdev_get_class(netdev))) { + struct netdev_vport *dev = netdev_vport_cast(netdev); + + ovs_mutex_lock(&dev->mutex); + dev->stats.tx_packets += stats->n_packets; + dev->stats.tx_bytes += stats->n_bytes; + ovs_mutex_unlock(&dev->mutex); + } } - -/* get_tnl_iface() implementation. */ -static struct name_node * -name_node_lookup(int ifi_index) +static int +get_patch_config(const struct netdev *dev_, struct smap *args) { - struct name_node *nn; + struct netdev_vport *dev = netdev_vport_cast(dev_); - HMAP_FOR_EACH_WITH_HASH(nn, node, hash_int(ifi_index, 0), &name_map) { - if (nn->ifi_index == ifi_index) { - return nn; - } + ovs_mutex_lock(&dev->mutex); + if (dev->peer) { + smap_add(args, "peer", dev->peer); } + ovs_mutex_unlock(&dev->mutex); - return NULL; + return 0; } -/* Queries the kernel for fresh data to populate the name map with. */ static int -netdev_vport_reset_names(void) +set_patch_config(struct netdev *dev_, const struct smap *args) { - int error; - struct nl_dump dump; - struct rtgenmsg *rtmsg; - struct ofpbuf request, reply; - static struct nl_sock *rtnl_sock; - struct name_node *nn, *nn_next; + struct netdev_vport *dev = netdev_vport_cast(dev_); + const char *name = netdev_get_name(dev_); + const char *peer; - HMAP_FOR_EACH_SAFE(nn, nn_next, node, &name_map) { - hmap_remove(&name_map, &nn->node); - free(nn); + peer = smap_get(args, "peer"); + if (!peer) { + VLOG_ERR("%s: patch type requires valid 'peer' argument", name); + return EINVAL; } - error = nl_sock_create(NETLINK_ROUTE, &rtnl_sock); - if (error) { - VLOG_WARN_RL(&rl, "Failed to create NETLINK_ROUTE socket"); - return error; + if (smap_count(args) > 1) { + VLOG_ERR("%s: patch type takes only a 'peer' argument", name); + return EINVAL; } - ofpbuf_init(&request, 0); + if (!strcmp(name, peer)) { + VLOG_ERR("%s: patch peer must not be self", name); + return EINVAL; + } - nl_msg_put_nlmsghdr(&request, sizeof *rtmsg, RTM_GETLINK, NLM_F_REQUEST); + ovs_mutex_lock(&dev->mutex); + if (!dev->peer || strcmp(dev->peer, peer)) { + free(dev->peer); + dev->peer = xstrdup(peer); + netdev_change_seq_changed(dev_); + } + ovs_mutex_unlock(&dev->mutex); - rtmsg = ofpbuf_put_zeros(&request, sizeof *rtmsg); - rtmsg->rtgen_family = AF_INET; + return 0; +} - nl_dump_start(&dump, rtnl_sock, &request); +static int +get_stats(const struct netdev *netdev, struct netdev_stats *stats) +{ + struct netdev_vport *dev = netdev_vport_cast(netdev); - while (nl_dump_next(&dump, &reply)) { - struct rtnetlink_link_change change; + ovs_mutex_lock(&dev->mutex); + *stats = dev->stats; + ovs_mutex_unlock(&dev->mutex); - if (rtnetlink_link_parse(&reply, &change)) { - netdev_vport_link_change(&change, NULL); - } + return 0; +} + + +/* Tunnel push pop ops. */ + +static struct ip_header * +ip_hdr(void *eth) +{ + return (void *)((char *)eth + sizeof (struct eth_header)); +} + +static struct gre_base_hdr * +gre_hdr(struct ip_header *ip) +{ + return (void *)((char *)ip + sizeof (struct ip_header)); +} + +static void * +ip_extract_tnl_md(struct dp_packet *packet, struct flow_tnl *tnl) +{ + struct ip_header *nh; + void *l4; + + nh = dp_packet_l3(packet); + l4 = dp_packet_l4(packet); + + if (!nh || !l4) { + return NULL; } - error = nl_dump_done(&dump); - nl_sock_destroy(rtnl_sock); + tnl->ip_src = get_16aligned_be32(&nh->ip_src); + tnl->ip_dst = get_16aligned_be32(&nh->ip_dst); + tnl->ip_tos = nh->ip_tos; + tnl->ip_ttl = nh->ip_ttl; - return error; + return l4; } -static void -netdev_vport_link_change(const struct rtnetlink_link_change *change, - void *aux OVS_UNUSED) +/* Pushes the 'size' bytes of 'header' into the headroom of 'packet', + * reallocating the packet if necessary. 'header' should contain an Ethernet + * header, followed by an IPv4 header (without options), and an L4 header. + * + * This function sets the IP header's ip_tot_len field (which should be zeroed + * as part of 'header') and puts its value into '*ip_tot_size' as well. Also + * updates IP header checksum. + * + * Return pointer to the L4 header added to 'packet'. */ +static void * +push_ip_header(struct dp_packet *packet, + const void *header, int size, int *ip_tot_size) { + struct eth_header *eth; + struct ip_header *ip; - if (!change) { - netdev_vport_reset_names(); - } else if (change->nlmsg_type == RTM_NEWLINK) { - struct name_node *nn; + eth = dp_packet_push_uninit(packet, size); + *ip_tot_size = dp_packet_size(packet) - sizeof (struct eth_header); - if (name_node_lookup(change->ifi_index)) { - return; - } + memcpy(eth, header, size); + ip = ip_hdr(eth); + ip->ip_tot_len = htons(*ip_tot_size); - nn = xzalloc(sizeof *nn); - nn->ifi_index = change->ifi_index; - strncpy(nn->ifname, change->ifname, IFNAMSIZ); - nn->ifname[IFNAMSIZ - 1] = '\0'; + ip->ip_csum = recalc_csum16(ip->ip_csum, 0, ip->ip_tot_len); - hmap_insert(&name_map, &nn->node, hash_int(nn->ifi_index, 0)); - } else if (change->nlmsg_type == RTM_DELLINK) { - struct name_node *nn; + return ip + 1; +} - nn = name_node_lookup(change->ifi_index); +static void * +udp_extract_tnl_md(struct dp_packet *packet, struct flow_tnl *tnl) +{ + struct udp_header *udp; - if (nn) { - hmap_remove(&name_map, &nn->node); - free(nn); - } + udp = ip_extract_tnl_md(packet, tnl); + if (!udp) { + return NULL; + } - } else { - VLOG_WARN_RL(&rl, "Received unexpected rtnetlink message type %d", - change->nlmsg_type); + if (udp->udp_csum) { + uint32_t csum = packet_csum_pseudoheader(dp_packet_l3(packet)); + + csum = csum_continue(csum, udp, dp_packet_size(packet) - + ((const unsigned char *)udp - + (const unsigned char *)dp_packet_l2(packet))); + if (csum_finish(csum)) { + return NULL; + } + tnl->flags |= FLOW_TNL_F_CSUM; } + + tnl->tp_src = udp->udp_src; + tnl->tp_dst = udp->udp_dst; + + return udp + 1; +} + +static ovs_be16 +get_src_port(struct dp_packet *packet) +{ + uint32_t hash; + + hash = dp_packet_get_rss_hash(packet); + + return htons((((uint64_t) hash * (tnl_udp_port_max - tnl_udp_port_min)) >> 32) + + tnl_udp_port_min); } static void -netdev_vport_tnl_iface_init(void) +push_udp_header(struct dp_packet *packet, + const struct ovs_action_push_tnl *data) { - static bool tnl_iface_is_init = false; + struct udp_header *udp; + int ip_tot_size; - if (!tnl_iface_is_init) { - hmap_init(&name_map); + udp = push_ip_header(packet, data->header, data->header_len, &ip_tot_size); - rtnetlink_link_notifier_register(&netdev_vport_link_notifier, - netdev_vport_link_change, NULL); + /* set udp src port */ + udp->udp_src = get_src_port(packet); + udp->udp_len = htons(ip_tot_size - sizeof (struct ip_header)); - netdev_vport_reset_names(); - tnl_iface_is_init = true; + if (udp->udp_csum) { + uint32_t csum = packet_csum_pseudoheader(ip_hdr(dp_packet_data(packet))); + + csum = csum_continue(csum, udp, + ip_tot_size - sizeof (struct ip_header)); + udp->udp_csum = csum_finish(csum); + + if (!udp->udp_csum) { + udp->udp_csum = htons(0xffff); + } } } -static const char * -netdev_vport_get_tnl_iface(const struct netdev *netdev) +static void * +udp_build_header(struct netdev_tunnel_config *tnl_cfg, + const struct flow *tnl_flow, + struct ovs_action_push_tnl *data) { - int ifindex; - uint32_t route; - struct netdev_dev_vport *ndv; - struct tnl_port_config *config; + struct ip_header *ip; + struct udp_header *udp; - ndv = netdev_dev_vport_cast(netdev_get_dev(netdev)); - config = (struct tnl_port_config *) ndv->config; - route = config->daddr; + ip = ip_hdr(data->header); + ip->ip_proto = IPPROTO_UDP; - if (route_table_get_ifindex(route, &ifindex)) { - struct name_node *nn; - HMAP_FOR_EACH_WITH_HASH(nn, node, hash_int(ifindex, 0), &name_map) { - if (nn->ifi_index == ifindex) { - return nn->ifname; - } - } + udp = (struct udp_header *) (ip + 1); + udp->udp_dst = tnl_cfg->dst_port; + + if (tnl_flow->tunnel.flags & FLOW_TNL_F_CSUM) { + /* Write a value in now to mark that we should compute the checksum + * later. 0xffff is handy because it is transparent to the + * calculation. */ + udp->udp_csum = htons(0xffff); } - return NULL; + return udp + 1; } - -/* Helper functions. */ static int -netdev_vport_do_ioctl(int cmd, void *arg) +gre_header_len(ovs_be16 flags) { - static int ioctl_fd = -1; + int hlen = sizeof(struct eth_header) + + sizeof(struct ip_header) + 4; - if (ioctl_fd < 0) { - ioctl_fd = open("/dev/net/dp0", O_RDONLY | O_NONBLOCK); - if (ioctl_fd < 0) { - VLOG_ERR_RL(&rl, "failed to open ioctl fd: %s", strerror(errno)); - return errno; - } + if (flags & htons(GRE_CSUM)) { + hlen += 4; } - - return ioctl(ioctl_fd, cmd, arg) ? errno : 0; + if (flags & htons(GRE_KEY)) { + hlen += 4; + } + if (flags & htons(GRE_SEQ)) { + hlen += 4; + } + return hlen; } -static void -netdev_vport_poll_notify(const struct netdev *netdev) +static int +parse_gre_header(struct dp_packet *packet, + struct flow_tnl *tnl) { - char *poll_name = make_poll_name(netdev); - struct list *list = shash_find_data(&netdev_vport_notifiers, - poll_name); + const struct gre_base_hdr *greh; + ovs_16aligned_be32 *options; + int hlen; - if (list) { - struct netdev_vport_notifier *notifier; + greh = ip_extract_tnl_md(packet, tnl); + if (!greh) { + return -EINVAL; + } - LIST_FOR_EACH (notifier, list_node, list) { - struct netdev_notifier *n = ¬ifier->notifier; - n->cb(n); + if (greh->flags & ~(htons(GRE_CSUM | GRE_KEY | GRE_SEQ))) { + return -EINVAL; + } + + if (greh->protocol != htons(ETH_TYPE_TEB)) { + return -EINVAL; + } + + hlen = gre_header_len(greh->flags); + if (hlen > dp_packet_size(packet)) { + return -EINVAL; + } + + options = (ovs_16aligned_be32 *)(greh + 1); + if (greh->flags & htons(GRE_CSUM)) { + ovs_be16 pkt_csum; + + pkt_csum = csum(greh, dp_packet_size(packet) - + ((const unsigned char *)greh - + (const unsigned char *)dp_packet_l2(packet))); + if (pkt_csum) { + return -EINVAL; } + tnl->flags = FLOW_TNL_F_CSUM; + options++; } - free(poll_name); + if (greh->flags & htons(GRE_KEY)) { + tnl->tun_id = (OVS_FORCE ovs_be64) ((OVS_FORCE uint64_t)(get_16aligned_be32(options)) << 32); + tnl->flags |= FLOW_TNL_F_KEY; + options++; + } + + if (greh->flags & htons(GRE_SEQ)) { + options++; + } + + return hlen; +} + +static void +pkt_metadata_init_tnl(struct pkt_metadata *md) +{ + memset(md, 0, offsetof(struct pkt_metadata, tunnel.metadata)); + + /* If 'opt_map' is zero then none of the rest of the tunnel metadata + * will be read, so we can skip clearing it. */ + md->tunnel.metadata.opt_map = 0; } - -/* Code specific to individual vport types. */ static int -parse_tunnel_config(const struct netdev_dev *dev, const struct shash *args, - void *configp) +netdev_gre_pop_header(struct dp_packet *packet) { - const char *name = netdev_dev_get_name(dev); - const char *type = netdev_dev_get_type(dev); - bool is_gre = false; - bool is_ipsec = false; - struct tnl_port_config config; - struct shash_node *node; - bool ipsec_mech_set = false; + struct pkt_metadata *md = &packet->md; + struct flow_tnl *tnl = &md->tunnel; + int hlen = sizeof(struct eth_header) + + sizeof(struct ip_header) + 4; + + pkt_metadata_init_tnl(md); + if (hlen > dp_packet_size(packet)) { + return EINVAL; + } - memset(&config, 0, sizeof config); - config.flags |= TNL_F_PMTUD; - config.flags |= TNL_F_HDR_CACHE; + hlen = parse_gre_header(packet, tnl); + if (hlen < 0) { + return -hlen; + } + + dp_packet_reset_packet(packet, hlen); - if (!strcmp(type, "gre")) { - is_gre = true; - } else if (!strcmp(type, "ipsec_gre")) { - is_gre = true; - is_ipsec = true; + return 0; +} + +static void +netdev_gre_push_header(struct dp_packet *packet, + const struct ovs_action_push_tnl *data) +{ + struct gre_base_hdr *greh; + int ip_tot_size; - config.flags |= TNL_F_IPSEC; + greh = push_ip_header(packet, data->header, data->header_len, &ip_tot_size); - /* IPsec doesn't work when header caching is enabled. */ - config.flags &= ~TNL_F_HDR_CACHE; + if (greh->flags & htons(GRE_CSUM)) { + ovs_be16 *csum_opt = (ovs_be16 *) (greh + 1); + *csum_opt = csum(greh, ip_tot_size - sizeof (struct ip_header)); } +} - 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") && is_gre) { - if (!strcmp(node->data, "flow")) { - config.flags |= TNL_F_IN_KEY_MATCH; - config.flags |= TNL_F_OUT_KEY_ACTION; - } else { - uint64_t key = strtoull(node->data, NULL, 0); - config.out_key = config.in_key = htonll(key); - } - } else if (!strcmp(node->name, "in_key") && is_gre) { - if (!strcmp(node->data, "flow")) { - config.flags |= TNL_F_IN_KEY_MATCH; - } else { - config.in_key = htonll(strtoull(node->data, NULL, 0)); - } - } else if (!strcmp(node->name, "out_key") && is_gre) { - if (!strcmp(node->data, "flow")) { - config.flags |= TNL_F_OUT_KEY_ACTION; - } else { - config.out_key = htonll(strtoull(node->data, NULL, 0)); - } - } 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; - } else { - config.ttl = atoi(node->data); - } - } else if (!strcmp(node->name, "csum") && is_gre) { - if (!strcmp(node->data, "true")) { - config.flags |= TNL_F_CSUM; - } - } else if (!strcmp(node->name, "pmtud")) { - if (!strcmp(node->data, "false")) { - config.flags &= ~TNL_F_PMTUD; - } - } else if (!strcmp(node->name, "header_cache")) { - if (!strcmp(node->data, "false")) { - config.flags &= ~TNL_F_HDR_CACHE; - } - } else if (!strcmp(node->name, "peer_cert") && is_ipsec) { - if (shash_find(args, "certificate")) { - ipsec_mech_set = true; - } else { - const char *use_ssl_cert; +static int +netdev_gre_build_header(const struct netdev *netdev, + struct ovs_action_push_tnl *data, + const struct flow *tnl_flow) +{ + struct netdev_vport *dev = netdev_vport_cast(netdev); + struct netdev_tunnel_config *tnl_cfg; + struct ip_header *ip; + struct gre_base_hdr *greh; + ovs_16aligned_be32 *options; + int hlen; + + /* XXX: RCUfy tnl_cfg. */ + ovs_mutex_lock(&dev->mutex); + tnl_cfg = &dev->tnl_cfg; + + ip = ip_hdr(data->header); + ip->ip_proto = IPPROTO_GRE; + + greh = gre_hdr(ip); + greh->protocol = htons(ETH_TYPE_TEB); + greh->flags = 0; + + options = (ovs_16aligned_be32 *) (greh + 1); + if (tnl_flow->tunnel.flags & FLOW_TNL_F_CSUM) { + greh->flags |= htons(GRE_CSUM); + put_16aligned_be32(options, 0); + options++; + } - /* If the "use_ssl_cert" is true, then "certificate" and - * "private_key" will be pulled from the SSL table. The - * use of this option is strongly discouraged, since it - * will like be removed when multiple SSL configurations - * are supported by OVS. - */ - use_ssl_cert = shash_find_data(args, "use_ssl_cert"); - if (!use_ssl_cert || strcmp(use_ssl_cert, "true")) { - VLOG_WARN("%s: 'peer_cert' requires 'certificate' argument", - name); - return EINVAL; - } - ipsec_mech_set = true; - } - } else if (!strcmp(node->name, "psk") && is_ipsec) { - ipsec_mech_set = true; - } else if (is_ipsec - && (!strcmp(node->name, "certificate") - || !strcmp(node->name, "private_key") - || !strcmp(node->name, "use_ssl_cert"))) { - /* Ignore options not used by the netdev. */ - } else { - VLOG_WARN("%s: unknown %s argument '%s'", - name, type, node->name); - } + if (tnl_cfg->out_key_present) { + greh->flags |= htons(GRE_KEY); + put_16aligned_be32(options, (OVS_FORCE ovs_be32) + ((OVS_FORCE uint64_t) tnl_flow->tunnel.tun_id >> 32)); + options++; } - if (is_ipsec) { - if (shash_find(args, "peer_cert") && shash_find(args, "psk")) { - VLOG_WARN("%s: cannot define both 'peer_cert' and 'psk'", name); - return EINVAL; - } + ovs_mutex_unlock(&dev->mutex); - if (!ipsec_mech_set) { - VLOG_WARN("%s: IPsec requires an 'peer_cert' or psk' argument", - name); - return EINVAL; - } + hlen = (uint8_t *) options - (uint8_t *) greh; + + data->header_len = sizeof(struct eth_header) + + sizeof(struct ip_header) + hlen; + data->tnl_type = OVS_VPORT_TYPE_GRE; + return 0; +} + +static int +netdev_vxlan_pop_header(struct dp_packet *packet) +{ + struct pkt_metadata *md = &packet->md; + struct flow_tnl *tnl = &md->tunnel; + struct vxlanhdr *vxh; + + pkt_metadata_init_tnl(md); + if (VXLAN_HLEN > dp_packet_size(packet)) { + return EINVAL; } - if (!config.daddr) { - VLOG_WARN("%s: %s type requires valid 'remote_ip' argument", - name, type); + vxh = udp_extract_tnl_md(packet, tnl); + if (!vxh) { return EINVAL; } - BUILD_ASSERT(sizeof config <= VPORT_CONFIG_SIZE); - memcpy(configp, &config, sizeof config); + if (get_16aligned_be32(&vxh->vx_flags) != htonl(VXLAN_FLAGS) || + (get_16aligned_be32(&vxh->vx_vni) & htonl(0xff))) { + VLOG_WARN_RL(&err_rl, "invalid vxlan flags=%#x vni=%#x\n", + ntohl(get_16aligned_be32(&vxh->vx_flags)), + ntohl(get_16aligned_be32(&vxh->vx_vni))); + return EINVAL; + } + tnl->tun_id = htonll(ntohl(get_16aligned_be32(&vxh->vx_vni)) >> 8); + tnl->flags |= FLOW_TNL_F_KEY; + + dp_packet_reset_packet(packet, VXLAN_HLEN); + return 0; } static int -parse_patch_config(const struct netdev_dev *dev, const struct shash *args, - void *configp) +netdev_vxlan_build_header(const struct netdev *netdev, + struct ovs_action_push_tnl *data, + const struct flow *tnl_flow) { - const char *name = netdev_dev_get_name(dev); - const char *peer; + struct netdev_vport *dev = netdev_vport_cast(netdev); + struct netdev_tunnel_config *tnl_cfg; + struct vxlanhdr *vxh; - peer = shash_find_data(args, "peer"); - if (!peer) { - VLOG_WARN("%s: patch type requires valid 'peer' argument", name); + /* XXX: RCUfy tnl_cfg. */ + ovs_mutex_lock(&dev->mutex); + tnl_cfg = &dev->tnl_cfg; + + vxh = udp_build_header(tnl_cfg, tnl_flow, data); + + put_16aligned_be32(&vxh->vx_flags, htonl(VXLAN_FLAGS)); + put_16aligned_be32(&vxh->vx_vni, htonl(ntohll(tnl_flow->tunnel.tun_id) << 8)); + + ovs_mutex_unlock(&dev->mutex); + data->header_len = VXLAN_HLEN; + data->tnl_type = OVS_VPORT_TYPE_VXLAN; + return 0; +} + +static int +netdev_geneve_pop_header(struct dp_packet *packet) +{ + struct pkt_metadata *md = &packet->md; + struct flow_tnl *tnl = &md->tunnel; + struct genevehdr *gnh; + unsigned int hlen; + int err; + + pkt_metadata_init_tnl(md); + if (GENEVE_BASE_HLEN > dp_packet_size(packet)) { + VLOG_WARN_RL(&err_rl, "geneve packet too small: min header=%u packet size=%u\n", + (unsigned int)GENEVE_BASE_HLEN, dp_packet_size(packet)); return EINVAL; } - if (shash_count(args) > 1) { - VLOG_WARN("%s: patch type takes only a 'peer' argument", name); + gnh = udp_extract_tnl_md(packet, tnl); + if (!gnh) { return EINVAL; } - if (strlen(peer) >= MIN(IFNAMSIZ, VPORT_CONFIG_SIZE)) { - VLOG_WARN("%s: patch 'peer' arg too long", name); + hlen = GENEVE_BASE_HLEN + gnh->opt_len * 4; + if (hlen > dp_packet_size(packet)) { + VLOG_WARN_RL(&err_rl, "geneve packet too small: header len=%u packet size=%u\n", + hlen, dp_packet_size(packet)); return EINVAL; } - if (!strcmp(name, peer)) { - VLOG_WARN("%s: patch peer must not be self", name); + if (gnh->ver != 0) { + VLOG_WARN_RL(&err_rl, "unknown geneve version: %"PRIu8"\n", gnh->ver); + return EINVAL; + } + + if (gnh->proto_type != htons(ETH_TYPE_TEB)) { + VLOG_WARN_RL(&err_rl, "unknown geneve encapsulated protocol: %#x\n", + ntohs(gnh->proto_type)); return EINVAL; } - strncpy(configp, peer, VPORT_CONFIG_SIZE); + tnl->flags |= gnh->oam ? FLOW_TNL_F_OAM : 0; + tnl->tun_id = htonll(ntohl(get_16aligned_be32(&gnh->vni)) >> 8); + tnl->flags |= FLOW_TNL_F_KEY; + + err = tun_metadata_from_geneve_header(gnh->options, gnh->opt_len * 4, + &tnl->metadata); + if (err) { + VLOG_WARN_RL(&err_rl, "invalid geneve options"); + return err; + } + + dp_packet_reset_packet(packet, hlen); return 0; } + +static int +netdev_geneve_build_header(const struct netdev *netdev, + struct ovs_action_push_tnl *data, + const struct flow *tnl_flow) +{ + struct netdev_vport *dev = netdev_vport_cast(netdev); + struct netdev_tunnel_config *tnl_cfg; + struct genevehdr *gnh; + int opt_len; + bool crit_opt; + + /* XXX: RCUfy tnl_cfg. */ + ovs_mutex_lock(&dev->mutex); + tnl_cfg = &dev->tnl_cfg; + + gnh = udp_build_header(tnl_cfg, tnl_flow, data); + + put_16aligned_be32(&gnh->vni, htonl(ntohll(tnl_flow->tunnel.tun_id) << 8)); + + ovs_mutex_unlock(&dev->mutex); + + opt_len = tun_metadata_to_geneve_header(&tnl_flow->tunnel.metadata, + gnh->options, &crit_opt); + + gnh->opt_len = opt_len / 4; + gnh->oam = !!(tnl_flow->tunnel.flags & FLOW_TNL_F_OAM); + gnh->critical = crit_opt ? 1 : 0; + gnh->proto_type = htons(ETH_TYPE_TEB); + + data->header_len = GENEVE_BASE_HLEN + opt_len; + data->tnl_type = OVS_VPORT_TYPE_GENEVE; + return 0; +} + +static void +netdev_vport_range(struct unixctl_conn *conn, int argc, + const char *argv[], void *aux OVS_UNUSED) +{ + int val1, val2; + + if (argc < 3) { + struct ds ds = DS_EMPTY_INITIALIZER; + + ds_put_format(&ds, "Tunnel UDP source port range: %"PRIu16"-%"PRIu16"\n", + tnl_udp_port_min, tnl_udp_port_max); + + unixctl_command_reply(conn, ds_cstr(&ds)); + ds_destroy(&ds); + return; + } + + if (argc != 3) { + return; + } + + val1 = atoi(argv[1]); + if (val1 <= 0 || val1 > UINT16_MAX) { + unixctl_command_reply(conn, "Invalid min."); + return; + } + val2 = atoi(argv[2]); + if (val2 <= 0 || val2 > UINT16_MAX) { + unixctl_command_reply(conn, "Invalid max."); + return; + } + + if (val1 > val2) { + tnl_udp_port_min = val2; + tnl_udp_port_max = val1; + } else { + tnl_udp_port_min = val1; + tnl_udp_port_max = val2; + } + seq_change(tnl_conf_seq); + + unixctl_command_reply(conn, "OK"); +} + -#define VPORT_FUNCTIONS(GET_STATUS) \ - netdev_vport_init, \ +#define VPORT_FUNCTIONS(GET_CONFIG, SET_CONFIG, \ + GET_TUNNEL_CONFIG, GET_STATUS, \ + BUILD_HEADER, \ + PUSH_HEADER, POP_HEADER) \ + NULL, \ netdev_vport_run, \ netdev_vport_wait, \ \ - netdev_vport_create, \ - netdev_vport_destroy, \ - netdev_vport_reconfigure, \ - \ - netdev_vport_open, \ - netdev_vport_close, \ - \ - NULL, /* enumerate */ \ - \ - NULL, /* recv */ \ - NULL, /* recv_wait */ \ - NULL, /* drain */ \ + netdev_vport_alloc, \ + netdev_vport_construct, \ + netdev_vport_destruct, \ + netdev_vport_dealloc, \ + GET_CONFIG, \ + SET_CONFIG, \ + GET_TUNNEL_CONFIG, \ + BUILD_HEADER, \ + PUSH_HEADER, \ + POP_HEADER, \ + NULL, /* get_numa_id */ \ + NULL, /* set_multiq */ \ \ NULL, /* send */ \ NULL, /* send_wait */ \ \ netdev_vport_set_etheraddr, \ netdev_vport_get_etheraddr, \ - netdev_vport_get_mtu, \ + NULL, /* get_mtu */ \ + NULL, /* set_mtu */ \ NULL, /* get_ifindex */ \ NULL, /* get_carrier */ \ + NULL, /* get_carrier_resets */ \ NULL, /* get_miimon */ \ - netdev_vport_get_stats, \ - netdev_vport_set_stats, \ + get_stats, \ \ NULL, /* get_features */ \ NULL, /* set_advertisements */ \ - NULL, /* get_vlan_vid */ \ \ NULL, /* set_policing */ \ NULL, /* get_qos_types */ \ @@ -854,7 +1382,9 @@ parse_patch_config(const struct netdev_dev *dev, const struct shash *args, NULL, /* set_queue */ \ NULL, /* delete_queue */ \ NULL, /* get_queue_stats */ \ - NULL, /* dump_queues */ \ + NULL, /* queue_dump_start */ \ + NULL, /* queue_dump_next */ \ + NULL, /* queue_dump_done */ \ NULL, /* dump_queue_stats */ \ \ NULL, /* get_in4 */ \ @@ -867,25 +1397,68 @@ parse_patch_config(const struct netdev_dev *dev, const struct shash *args, \ netdev_vport_update_flags, \ \ - netdev_vport_poll_add, \ - netdev_vport_poll_remove, + NULL, /* rx_alloc */ \ + NULL, /* rx_construct */ \ + NULL, /* rx_destruct */ \ + NULL, /* rx_dealloc */ \ + NULL, /* rx_recv */ \ + NULL, /* rx_wait */ \ + NULL, /* rx_drain */ + + +#define TUNNEL_CLASS(NAME, DPIF_PORT, BUILD_HEADER, PUSH_HEADER, POP_HEADER) \ + { DPIF_PORT, \ + { NAME, VPORT_FUNCTIONS(get_tunnel_config, \ + set_tunnel_config, \ + get_netdev_tunnel_config, \ + tunnel_get_status, \ + BUILD_HEADER, PUSH_HEADER, POP_HEADER) }} void -netdev_vport_register(void) +netdev_vport_tunnel_register(void) { + /* The name of the dpif_port should be short enough to accomodate adding + * a port number to the end if one is necessary. */ static const struct vport_class vport_classes[] = { - { { "gre", VPORT_FUNCTIONS(netdev_vport_get_status) }, - parse_tunnel_config }, - { { "ipsec_gre", VPORT_FUNCTIONS(netdev_vport_get_status) }, - parse_tunnel_config }, - { { "capwap", VPORT_FUNCTIONS(netdev_vport_get_status) }, - parse_tunnel_config }, - { { "patch", VPORT_FUNCTIONS(NULL) }, parse_patch_config } + TUNNEL_CLASS("geneve", "genev_sys", netdev_geneve_build_header, + push_udp_header, + netdev_geneve_pop_header), + TUNNEL_CLASS("gre", "gre_sys", netdev_gre_build_header, + netdev_gre_push_header, + netdev_gre_pop_header), + TUNNEL_CLASS("ipsec_gre", "gre_sys", NULL, NULL, NULL), + TUNNEL_CLASS("gre64", "gre64_sys", NULL, NULL, NULL), + TUNNEL_CLASS("ipsec_gre64", "gre64_sys", NULL, NULL, NULL), + TUNNEL_CLASS("vxlan", "vxlan_sys", netdev_vxlan_build_header, + push_udp_header, + netdev_vxlan_pop_header), + TUNNEL_CLASS("lisp", "lisp_sys", NULL, NULL, NULL), + TUNNEL_CLASS("stt", "stt_sys", NULL, NULL, NULL), }; + static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER; - int i; + if (ovsthread_once_start(&once)) { + int i; - for (i = 0; i < ARRAY_SIZE(vport_classes); i++) { - netdev_register_provider(&vport_classes[i].netdev_class); + for (i = 0; i < ARRAY_SIZE(vport_classes); i++) { + netdev_register_provider(&vport_classes[i].netdev_class); + } + + unixctl_command_register("tnl/egress_port_range", "min max", 0, 2, + netdev_vport_range, NULL); + + ovsthread_once_done(&once); } } + +void +netdev_vport_patch_register(void) +{ + static const struct vport_class patch_class = + { NULL, + { "patch", VPORT_FUNCTIONS(get_patch_config, + set_patch_config, + NULL, + NULL, NULL, NULL, NULL) }}; + netdev_register_provider(&patch_class.netdev_class); +}