X-Git-Url: http://git.cascardo.eti.br/?a=blobdiff_plain;f=lib%2Fnetdev-vport.c;h=df6d8cfd9d7f4749801e42fe840b82f4836592be;hb=HEAD;hp=d6c4aa64992e5696b1442ef5c83a0bac19dbb8c1;hpb=db078f85c709300b5e2d7d13921b0fc553205ee5;p=cascardo%2Fovs.git diff --git a/lib/netdev-vport.c b/lib/netdev-vport.c index d6c4aa649..df6d8cfd9 100644 --- a/lib/netdev-vport.c +++ b/lib/netdev-vport.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2010, 2011, 2012, 2013 Nicira, Inc. + * 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. @@ -21,387 +21,334 @@ #include #include #include -#include -#include #include +#include +#include #include #include "byte-order.h" +#include "csum.h" #include "daemon.h" #include "dirs.h" -#include "dpif-linux.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-linux.h" #include "netdev-provider.h" -#include "netlink.h" -#include "netlink-notifier.h" -#include "netlink-socket.h" -#include "ofpbuf.h" -#include "openvswitch/tunnel.h" +#include "odp-netlink.h" +#include "dp-packet.h" +#include "ovs-router.h" #include "packets.h" +#include "poll-loop.h" #include "route-table.h" #include "shash.h" #include "socket-util.h" +#include "openvswitch/vlog.h" #include "unaligned.h" -#include "vlog.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); -/* Default to the OTV port, per the VXLAN IETF draft. */ -#define VXLAN_DST_PORT 8472 +#define GENEVE_DST_PORT 6081 +#define VXLAN_DST_PORT 4789 +#define LISP_DST_PORT 4341 +#define STT_DST_PORT 7471 -struct netdev_dev_vport { - struct netdev_dev netdev_dev; - struct ofpbuf *options; - unsigned int change_seq; - uint8_t etheraddr[ETH_ADDR_LEN]; +#define VXLAN_HLEN (sizeof(struct udp_header) + \ + sizeof(struct vxlanhdr)) + +#define GENEVE_BASE_HLEN (sizeof(struct udp_header) + \ + sizeof(struct genevehdr)) + +#define DEFAULT_TTL 64 + +struct netdev_vport { + struct netdev up; + + /* Protects all members below. */ + struct ovs_mutex mutex; + + struct eth_addr etheraddr; + 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 { - enum ovs_vport_type type; + const char *dpif_port; struct netdev_class netdev_class; - int (*parse_config)(const char *name, const char *type, - const struct smap *args, struct ofpbuf *options); - int (*unparse_config)(const char *name, const char *type, - const struct nlattr *options, size_t options_len, - struct smap *args); }; -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_create(const struct netdev_class *, const char *, - struct netdev_dev **); -static void netdev_vport_poll_notify(const struct netdev *); -static int tnl_port_config_from_nlattr(const struct nlattr *options, - size_t options_len, - struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]); +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 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_dev_vport * -netdev_vport_get_dev(const struct netdev *netdev) +static const struct netdev_tunnel_config * +get_netdev_tunnel_config(const struct netdev *netdev) { - return netdev_dev_vport_cast(netdev_get_dev(netdev)); + return &netdev_vport_cast(netdev)->tnl_cfg; } -/* If 'netdev' is a vport netdev, returns an ofpbuf that contains Netlink - * options to include in OVS_VPORT_ATTR_OPTIONS for configuring that vport. - * Otherwise returns NULL. */ -const struct ofpbuf * -netdev_vport_get_options(const struct netdev *netdev) +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); - return (is_vport_class(netdev_dev_get_class(dev)) - ? netdev_dev_vport_cast(dev)->options - : NULL); + return class->get_config == get_patch_config; } -enum ovs_vport_type -netdev_vport_get_vport_type(const struct netdev *netdev) +bool +netdev_vport_is_layer3(const struct netdev *dev) { - const struct netdev_dev *dev = netdev_get_dev(netdev); - const struct netdev_class *class = netdev_dev_get_class(dev); + const char *type = netdev_get_type(dev); - return (is_vport_class(class) ? vport_class_cast(class)->type - : class == &netdev_internal_class ? OVS_VPORT_TYPE_INTERNAL - : (class == &netdev_linux_class || - class == &netdev_tap_class) ? OVS_VPORT_TYPE_NETDEV - : OVS_VPORT_TYPE_UNSPEC); + return (!strcmp("lisp", type)); } -static uint32_t -get_u32_or_zero(const struct nlattr *a) +static bool +netdev_vport_needs_dst_port(const struct netdev *dev) { - return a ? nl_attr_get_u32(a) : 0; + const struct netdev_class *class = netdev_get_class(dev); + const char *type = netdev_get_type(dev); + + return (class->get_config == get_tunnel_config && + (!strcmp("geneve", type) || !strcmp("vxlan", type) || + !strcmp("lisp", type) || !strcmp("stt", type)) ); } const char * -netdev_vport_get_netdev_type(const struct dpif_linux_vport *vport) +netdev_vport_class_get_dpif_port(const struct netdev_class *class) { - struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]; + return is_vport_class(class) ? vport_class_cast(class)->dpif_port : NULL; +} - switch (vport->type) { - case OVS_VPORT_TYPE_UNSPEC: - break; +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); - case OVS_VPORT_TYPE_NETDEV: - return "system"; + if (!dpif_port) { + return netdev_get_name(netdev); + } - case OVS_VPORT_TYPE_INTERNAL: - return "internal"; + 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; + } +} - case OVS_VPORT_TYPE_PATCH: - return "patch"; +char * +netdev_vport_get_dpif_port_strdup(const struct netdev *netdev) +{ + char namebuf[NETDEV_VPORT_NAME_BUFSIZE]; - case OVS_VPORT_TYPE_GRE: - if (tnl_port_config_from_nlattr(vport->options, vport->options_len, - a)) { - break; - } - return (get_u32_or_zero(a[OVS_TUNNEL_ATTR_FLAGS]) & TNL_F_IPSEC - ? "ipsec_gre" : "gre"); + return xstrdup(netdev_vport_get_dpif_port(netdev, namebuf, + sizeof namebuf)); +} - case OVS_VPORT_TYPE_GRE64: - if (tnl_port_config_from_nlattr(vport->options, vport->options_len, - a)) { - break; +/* 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_route_changed(void) +{ + 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 (ipv6_addr_is_set(&netdev->tnl_cfg.ipv6_dst)) { + if (tunnel_check_status_change__(netdev)) { + netdev_change_seq_changed(netdev_); + } } - return (get_u32_or_zero(a[OVS_TUNNEL_ATTR_FLAGS]) & TNL_F_IPSEC - ? "ipsec_gre64" : "gre64"); - - case OVS_VPORT_TYPE_CAPWAP: - return "capwap"; - - case OVS_VPORT_TYPE_VXLAN: - return "vxlan"; + ovs_mutex_unlock(&netdev->mutex); - case OVS_VPORT_TYPE_FT_GRE: - case __OVS_VPORT_TYPE_MAX: - break; + netdev_close(netdev_); } - VLOG_WARN_RL(&rl, "dp%d: port `%s' has unsupported type %u", - vport->dp_ifindex, vport->name, (unsigned int) vport->type); - return "unknown"; + free(vports); +} + +static struct netdev * +netdev_vport_alloc(void) +{ + struct netdev_vport *netdev = xzalloc(sizeof *netdev); + return &netdev->up; } static int -netdev_vport_create(const struct netdev_class *netdev_class, const char *name, - struct netdev_dev **netdev_devp) +netdev_vport_construct(struct netdev *netdev_) { - struct netdev_dev_vport *dev; + struct netdev_vport *dev = netdev_vport_cast(netdev_); + const char *type = netdev_get_type(netdev_); - dev = xmalloc(sizeof *dev); - netdev_dev_init(&dev->netdev_dev, name, netdev_class); - dev->options = NULL; - dev->change_seq = 1; - eth_addr_random(dev->etheraddr); + ovs_mutex_init(&dev->mutex); + eth_addr_random(&dev->etheraddr); - *netdev_devp = &dev->netdev_dev; - route_table_register(); + /* 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); + } + dev->tnl_cfg.dont_fragment = true; + dev->tnl_cfg.ttl = DEFAULT_TTL; return 0; } static void -netdev_vport_destroy(struct netdev_dev *netdev_dev_) +netdev_vport_destruct(struct netdev *netdev_) { - struct netdev_dev_vport *netdev_dev = netdev_dev_vport_cast(netdev_dev_); - - ofpbuf_delete(netdev_dev->options); - route_table_unregister(); - free(netdev_dev); -} + struct netdev_vport *netdev = netdev_vport_cast(netdev_); -static int -netdev_vport_open(struct netdev_dev *netdev_dev, struct netdev **netdevp) -{ - *netdevp = xmalloc(sizeof **netdevp); - netdev_init(*netdevp, netdev_dev); - return 0; + free(netdev->peer); + ovs_mutex_destroy(&netdev->mutex); } static void -netdev_vport_close(struct netdev *netdev) +netdev_vport_dealloc(struct netdev *netdev_) { + struct netdev_vport *netdev = netdev_vport_cast(netdev_); free(netdev); } static int -netdev_vport_get_config(struct netdev_dev *dev_, struct smap *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_); - const char *name = netdev_dev_get_name(dev_); - int error; - - if (!dev->options) { - struct dpif_linux_vport reply; - struct ofpbuf *buf; - - error = dpif_linux_vport_get(name, &reply, &buf); - if (error) { - VLOG_ERR_RL(&rl, "%s: vport query failed (%s)", - name, strerror(error)); - return error; - } +netdev_vport_set_etheraddr(struct netdev *netdev_, const struct eth_addr mac) +{ + struct netdev_vport *netdev = netdev_vport_cast(netdev_); - dev->options = ofpbuf_clone_data(reply.options, reply.options_len); - ofpbuf_delete(buf); - } + ovs_mutex_lock(&netdev->mutex); + netdev->etheraddr = mac; + ovs_mutex_unlock(&netdev->mutex); + netdev_change_seq_changed(netdev_); - error = vport_class->unparse_config(name, netdev_class->type, - dev->options->data, - dev->options->size, - args); - if (error) { - VLOG_ERR_RL(&rl, "%s: failed to parse kernel config (%s)", - name, strerror(error)); - } - return error; + return 0; } static int -netdev_vport_set_config(struct netdev_dev *dev_, const struct smap *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_); - const char *name = netdev_dev_get_name(dev_); - struct ofpbuf *options; - int error; - - options = ofpbuf_new(64); - error = vport_class->parse_config(name, netdev_dev_get_type(dev_), - args, options); - if (!error - && (!dev->options - || options->size != dev->options->size - || memcmp(options->data, dev->options->data, options->size))) { - struct dpif_linux_vport vport; - - dpif_linux_vport_init(&vport); - vport.cmd = OVS_VPORT_CMD_SET; - vport.name = name; - vport.options = options->data; - vport.options_len = options->size; - error = dpif_linux_vport_transact(&vport, NULL, NULL); - 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()). */ - ofpbuf_delete(dev->options); - dev->options = options; - options = NULL; - error = 0; - } - } - ofpbuf_delete(options); +netdev_vport_get_etheraddr(const struct netdev *netdev_, struct eth_addr *mac) +{ + struct netdev_vport *netdev = netdev_vport_cast(netdev_); - return error; -} + ovs_mutex_lock(&netdev->mutex); + *mac = netdev->etheraddr; + ovs_mutex_unlock(&netdev->mutex); -static int -netdev_vport_set_etheraddr(struct netdev *netdev, - const uint8_t mac[ETH_ADDR_LEN]) -{ - memcpy(netdev_vport_get_dev(netdev)->etheraddr, mac, ETH_ADDR_LEN); - netdev_vport_poll_notify(netdev); return 0; } -static int -netdev_vport_get_etheraddr(const struct netdev *netdev, - uint8_t mac[ETH_ADDR_LEN]) +/* 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) { - memcpy(mac, netdev_vport_get_dev(netdev)->etheraddr, ETH_ADDR_LEN); - return 0; -} + char iface[IFNAMSIZ]; + bool status = false; + struct in6_addr *route; + struct in6_addr gw; + + iface[0] = '\0'; + route = &netdev->tnl_cfg.ipv6_dst; + if (ovs_router_lookup(route, iface, &gw)) { + struct netdev *egress_netdev; -/* Copies 'src' into 'dst', performing format conversion in the process. - * - * 'src' is allowed to be misaligned. */ -static void -netdev_stats_from_ovs_vport_stats(struct netdev_stats *dst, - const struct ovs_vport_stats *src) -{ - dst->rx_packets = get_unaligned_u64(&src->rx_packets); - dst->tx_packets = get_unaligned_u64(&src->tx_packets); - dst->rx_bytes = get_unaligned_u64(&src->rx_bytes); - dst->tx_bytes = get_unaligned_u64(&src->tx_bytes); - dst->rx_errors = get_unaligned_u64(&src->rx_errors); - dst->tx_errors = get_unaligned_u64(&src->tx_errors); - dst->rx_dropped = get_unaligned_u64(&src->rx_dropped); - dst->tx_dropped = get_unaligned_u64(&src->tx_dropped); - dst->multicast = 0; - dst->collisions = 0; - dst->rx_length_errors = 0; - dst->rx_over_errors = 0; - dst->rx_crc_errors = 0; - dst->rx_frame_errors = 0; - dst->rx_fifo_errors = 0; - dst->rx_missed_errors = 0; - dst->tx_aborted_errors = 0; - dst->tx_carrier_errors = 0; - dst->tx_fifo_errors = 0; - dst->tx_heartbeat_errors = 0; - dst->tx_window_errors = 0; -} - -int -netdev_vport_get_stats(const struct netdev *netdev, struct netdev_stats *stats) -{ - struct dpif_linux_vport reply; - struct ofpbuf *buf; - int error; - - error = dpif_linux_vport_get(netdev_get_name(netdev), &reply, &buf); - if (error) { - return error; - } else if (!reply.stats) { - ofpbuf_delete(buf); - return EOPNOTSUPP; + if (!netdev_open(iface, "system", &egress_netdev)) { + status = netdev_get_carrier(egress_netdev); + netdev_close(egress_netdev); + } } - netdev_stats_from_ovs_vport_stats(stats, reply.stats); + if (strcmp(netdev->egress_iface, iface) + || netdev->carrier_status != status) { + ovs_strlcpy(netdev->egress_iface, iface, IFNAMSIZ); + netdev->carrier_status = status; - ofpbuf_delete(buf); + return true; + } - return 0; + return false; } static int -tunnel_get_status(const struct netdev *netdev, struct smap *smap) +tunnel_get_status(const struct netdev *netdev_, struct smap *smap) { - struct netdev_dev_vport *ndv = netdev_vport_get_dev(netdev); - struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]; - static char iface[IFNAMSIZ]; - ovs_be32 route; + struct netdev_vport *netdev = netdev_vport_cast(netdev_); - if (!ndv->options) { - /* Race condition when 'ndv' was created, but did not have it's - * configuration set yet. */ - return 0; - } + if (netdev->egress_iface[0]) { + smap_add(smap, "tunnel_egress_iface", netdev->egress_iface); - if (tnl_port_config_from_nlattr(ndv->options->data, - ndv->options->size, a)) { - return 0; - } - route = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_DST_IPV4]); - - if (route_table_get_name(route, iface)) { - struct netdev *egress_netdev; - - smap_add(smap, "tunnel_egress_iface", iface); - - if (!netdev_open(iface, "system", &egress_netdev)) { - smap_add(smap, "tunnel_egress_iface_carrier", - netdev_get_carrier(egress_netdev) ? "up" : "down"); - netdev_close(egress_netdev); - } + smap_add(smap, "tunnel_egress_iface_carrier", + netdev->carrier_status ? "up" : "down"); } return 0; @@ -409,8 +356,9 @@ tunnel_get_status(const struct netdev *netdev, struct smap *smap) 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) + enum netdev_flags off, + enum netdev_flags on OVS_UNUSED, + enum netdev_flags *old_flagsp) { if (off & (NETDEV_UP | NETDEV_PROMISC)) { return EOPNOTSUPP; @@ -420,140 +368,189 @@ netdev_vport_update_flags(struct netdev *netdev OVS_UNUSED, return 0; } -static unsigned int -netdev_vport_change_seq(const struct netdev *netdev) -{ - return netdev_vport_get_dev(netdev)->change_seq; -} - static void netdev_vport_run(void) { + uint64_t seq; + route_table_run(); + seq = route_table_get_change_seq(); + if (rt_change_seqno != seq) { + rt_change_seqno = seq; + netdev_vport_route_changed(); + } } static void netdev_vport_wait(void) { - route_table_wait(); -} - -/* Helper functions. */ - -static void -netdev_vport_poll_notify(const struct netdev *netdev) -{ - struct netdev_dev_vport *ndv = netdev_vport_get_dev(netdev); + uint64_t seq; - ndv->change_seq++; - if (!ndv->change_seq) { - ndv->change_seq++; + route_table_wait(); + seq = route_table_get_change_seq(); + if (rt_change_seqno != seq) { + poll_immediate_wake(); } } -/* Code specific to individual vport types. */ +/* Code specific to tunnel types. */ -static void -set_key(const struct smap *args, const char *name, uint16_t type, - struct ofpbuf *options) +static ovs_be64 +parse_key(const struct smap *args, const char *name, + bool *present, bool *flow) { const char *s; + *present = false; + *flow = false; + s = smap_get(args, name); if (!s) { s = smap_get(args, "key"); if (!s) { - s = "0"; + return 0; } } + *present = true; + if (!strcmp(s, "flow")) { - /* This is the default if no attribute is present. */ + *flow = true; + return 0; } else { - nl_msg_put_be64(options, type, htonll(strtoull(s, NULL, 0))); + return htonll(strtoull(s, NULL, 0)); } } static int -parse_tunnel_config(const char *name, const char *type, - const struct smap *args, struct ofpbuf *options) +parse_tunnel_ip(const char *value, bool accept_mcast, bool *flow, + struct in6_addr *ipv6, uint16_t *protocol) { - bool is_gre = false; - bool is_ipsec = false; - bool needs_dst_port = false; - bool found_dst_port = false; + if (!strcmp(value, "flow")) { + *flow = true; + *protocol = 0; + return 0; + } + if (addr_is_ipv6(value)) { + if (lookup_ipv6(value, ipv6)) { + return ENOENT; + } + if (!accept_mcast && ipv6_addr_is_multicast(ipv6)) { + return EINVAL; + } + *protocol = ETH_TYPE_IPV6; + } else { + struct in_addr ip; + if (lookup_ip(value, &ip)) { + return ENOENT; + } + if (!accept_mcast && ip_is_multicast(ip.s_addr)) { + return EINVAL; + } + in6_addr_set_mapped_ipv4(ipv6, ip.s_addr); + *protocol = ETH_TYPE_IP; + } + return 0; +} + +static int +set_tunnel_config(struct netdev *dev_, const struct smap *args) +{ + 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; + uint16_t dst_proto = 0, src_proto = 0; + struct netdev_tunnel_config tnl_cfg; struct smap_node *node; - bool ipsec_mech_set = false; - ovs_be32 daddr = htonl(0); - ovs_be32 saddr = htonl(0); - uint32_t flags; - - if (!strcmp(type, "capwap")) { - VLOG_WARN_ONCE("CAPWAP tunnel support is deprecated."); - } - - flags = TNL_F_DF_DEFAULT; - if (!strcmp(type, "gre") || !strcmp(type, "gre64")) { - is_gre = true; - } else if (!strcmp(type, "ipsec_gre") || !strcmp(type, "ipsec_gre64")) { - is_gre = true; - is_ipsec = true; - flags |= TNL_F_IPSEC; - } else if (!strcmp(type, "vxlan")) { - needs_dst_port = true; + + 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); + } + + 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 (lookup_ip(node->value, &in_addr)) { + int err; + err = parse_tunnel_ip(node->value, false, &tnl_cfg.ip_dst_flow, + &tnl_cfg.ipv6_dst, &dst_proto); + switch (err) { + case ENOENT: VLOG_WARN("%s: bad %s 'remote_ip'", name, type); - } else { - daddr = in_addr.s_addr; + break; + case EINVAL: + VLOG_WARN("%s: multicast remote_ip=%s not allowed", + name, node->value); + return EINVAL; + } + if (dst_proto == ETH_TYPE_IPV6) { + VLOG_WARN("%s: IPv6 'remote_ip' is not supported", name); + return EOPNOTSUPP; } } else if (!strcmp(node->key, "local_ip")) { - struct in_addr in_addr; - if (lookup_ip(node->value, &in_addr)) { + int err; + err = parse_tunnel_ip(node->value, true, &tnl_cfg.ip_src_flow, + &tnl_cfg.ipv6_src, &src_proto); + switch (err) { + case ENOENT: VLOG_WARN("%s: bad %s 'local_ip'", name, type); - } else { - saddr = in_addr.s_addr; + break; + } + if (src_proto == ETH_TYPE_IPV6) { + VLOG_WARN("%s: IPv6 'local_ip' is not supported", name); + return EOPNOTSUPP; } } else if (!strcmp(node->key, "tos")) { if (!strcmp(node->value, "inherit")) { - flags |= TNL_F_TOS_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)) { - nl_msg_put_u8(options, OVS_TUNNEL_ATTR_TOS, tos); + 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")) { - flags |= TNL_F_TTL_INHERIT; + tnl_cfg.ttl_inherit = true; } else { - nl_msg_put_u8(options, OVS_TUNNEL_ATTR_TTL, atoi(node->value)); + tnl_cfg.ttl = atoi(node->value); } } else if (!strcmp(node->key, "dst_port") && needs_dst_port) { - nl_msg_put_u16(options, OVS_TUNNEL_ATTR_DST_PORT, - atoi(node->value)); - found_dst_port = true; - } else if (!strcmp(node->key, "csum") && is_gre) { - if (!strcmp(node->value, "true")) { - flags |= TNL_F_CSUM; - } - } else if (!strcmp(node->key, "df_inherit")) { + tnl_cfg.dst_port = htons(atoi(node->value)); + } else if (!strcmp(node->key, "csum") && has_csum) { if (!strcmp(node->value, "true")) { - flags |= TNL_F_DF_INHERIT; + tnl_cfg.csum = true; } } else if (!strcmp(node->key, "df_default")) { if (!strcmp(node->value, "false")) { - flags &= ~TNL_F_DF_DEFAULT; + tnl_cfg.dont_fragment = false; } - } else if (!strcmp(node->key, "peer_cert") && is_ipsec) { + } else if (!strcmp(node->key, "peer_cert") && tnl_cfg.ipsec) { if (smap_get(args, "certificate")) { ipsec_mech_set = true; } else { @@ -573,9 +570,9 @@ parse_tunnel_config(const char *name, const char *type, } ipsec_mech_set = true; } - } else if (!strcmp(node->key, "psk") && is_ipsec) { + } else if (!strcmp(node->key, "psk") && tnl_cfg.ipsec) { ipsec_mech_set = true; - } else if (is_ipsec + } else if (tnl_cfg.ipsec && (!strcmp(node->key, "certificate") || !strcmp(node->key, "private_key") || !strcmp(node->key, "use_ssl_cert"))) { @@ -584,24 +581,43 @@ parse_tunnel_config(const char *name, const char *type, !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); } } - /* Add a default destination port for VXLAN if none specified. */ - if (needs_dst_port && !found_dst_port) { - nl_msg_put_u16(options, OVS_TUNNEL_ATTR_DST_PORT, VXLAN_DST_PORT); - } - - if (is_ipsec) { + 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", @@ -621,146 +637,193 @@ parse_tunnel_config(const char *name, const char *type, } } - set_key(args, "in_key", OVS_TUNNEL_ATTR_IN_KEY, options); - set_key(args, "out_key", OVS_TUNNEL_ATTR_OUT_KEY, options); - - if (!daddr) { + if (!ipv6_addr_is_set(&tnl_cfg.ipv6_dst) && !tnl_cfg.ip_dst_flow) { VLOG_ERR("%s: %s type requires valid 'remote_ip' argument", name, type); return EINVAL; } - 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); - } + 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 (src_proto && dst_proto && src_proto != dst_proto) { + VLOG_ERR("%s: 'remote_ip' and 'local_ip' has to be of the same address family", + name); + return EINVAL; + } + if (!tnl_cfg.ttl) { + tnl_cfg.ttl = DEFAULT_TTL; } - nl_msg_put_u32(options, OVS_TUNNEL_ATTR_FLAGS, flags); + tnl_cfg.in_key = parse_key(args, "in_key", + &tnl_cfg.in_key_present, + &tnl_cfg.in_key_flow); - return 0; -} - -static int -tnl_port_config_from_nlattr(const struct nlattr *options, size_t options_len, - struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]) -{ - static const struct nl_policy ovs_tunnel_policy[] = { - [OVS_TUNNEL_ATTR_FLAGS] = { .type = NL_A_U32, .optional = true }, - [OVS_TUNNEL_ATTR_DST_IPV4] = { .type = NL_A_BE32, .optional = true }, - [OVS_TUNNEL_ATTR_SRC_IPV4] = { .type = NL_A_BE32, .optional = true }, - [OVS_TUNNEL_ATTR_IN_KEY] = { .type = NL_A_BE64, .optional = true }, - [OVS_TUNNEL_ATTR_OUT_KEY] = { .type = NL_A_BE64, .optional = true }, - [OVS_TUNNEL_ATTR_TOS] = { .type = NL_A_U8, .optional = true }, - [OVS_TUNNEL_ATTR_TTL] = { .type = NL_A_U8, .optional = true }, - [OVS_TUNNEL_ATTR_DST_PORT] = { .type = NL_A_U16, .optional = true }, - }; - struct ofpbuf buf; + tnl_cfg.out_key = parse_key(args, "out_key", + &tnl_cfg.out_key_present, + &tnl_cfg.out_key_flow); - ofpbuf_use_const(&buf, options, options_len); - if (!nl_policy_parse(&buf, 0, ovs_tunnel_policy, - a, ARRAY_SIZE(ovs_tunnel_policy))) { - return EINVAL; + 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_); } - return 0; -} + ovs_mutex_unlock(&dev->mutex); -static uint64_t -get_be64_or_zero(const struct nlattr *a) -{ - return a ? ntohll(nl_attr_get_be64(a)) : 0; + return 0; } static int -unparse_tunnel_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, - const struct nlattr *options, size_t options_len, - struct smap *args) +get_tunnel_config(const struct netdev *dev, struct smap *args) { - struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]; - uint32_t flags; - int error; + struct netdev_vport *netdev = netdev_vport_cast(dev); + struct netdev_tunnel_config tnl_cfg; - error = tnl_port_config_from_nlattr(options, options_len, a); - if (error) { - return error; - } + ovs_mutex_lock(&netdev->mutex); + tnl_cfg = netdev->tnl_cfg; + ovs_mutex_unlock(&netdev->mutex); - if (a[OVS_TUNNEL_ATTR_DST_IPV4]) { - ovs_be32 daddr = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_DST_IPV4]); - smap_add_format(args, "remote_ip", IP_FMT, IP_ARGS(daddr)); + if (ipv6_addr_is_set(&tnl_cfg.ipv6_dst)) { + smap_add_ipv6(args, "remote_ip", &tnl_cfg.ipv6_dst); + } else if (tnl_cfg.ip_dst_flow) { + smap_add(args, "remote_ip", "flow"); } - if (a[OVS_TUNNEL_ATTR_SRC_IPV4]) { - ovs_be32 saddr = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_SRC_IPV4]); - smap_add_format(args, "local_ip", IP_FMT, IP_ARGS(saddr)); + if (ipv6_addr_is_set(&tnl_cfg.ipv6_src)) { + smap_add_ipv6(args, "local_ip", &tnl_cfg.ipv6_src); + } else if (tnl_cfg.ip_src_flow) { + smap_add(args, "local_ip", "flow"); } - if (!a[OVS_TUNNEL_ATTR_IN_KEY] && !a[OVS_TUNNEL_ATTR_OUT_KEY]) { + 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 { - uint64_t in_key = get_be64_or_zero(a[OVS_TUNNEL_ATTR_IN_KEY]); - uint64_t out_key = get_be64_or_zero(a[OVS_TUNNEL_ATTR_OUT_KEY]); - - if (in_key && in_key == out_key) { - smap_add_format(args, "key", "%"PRIu64, in_key); - } else { - if (!a[OVS_TUNNEL_ATTR_IN_KEY]) { - smap_add(args, "in_key", "flow"); - } else if (in_key) { - smap_add_format(args, "in_key", "%"PRIu64, in_key); - } + 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 (!a[OVS_TUNNEL_ATTR_OUT_KEY]) { - smap_add(args, "out_key", "flow"); - } else if (out_key) { - smap_add_format(args, "out_key", "%"PRIu64, out_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)); } } - flags = get_u32_or_zero(a[OVS_TUNNEL_ATTR_FLAGS]); - - if (flags & TNL_F_TTL_INHERIT) { + if (tnl_cfg.ttl_inherit) { smap_add(args, "ttl", "inherit"); - } else if (a[OVS_TUNNEL_ATTR_TTL]) { - int ttl = nl_attr_get_u8(a[OVS_TUNNEL_ATTR_TTL]); - smap_add_format(args, "ttl", "%d", ttl); + } else if (tnl_cfg.ttl != DEFAULT_TTL) { + smap_add_format(args, "ttl", "%"PRIu8, tnl_cfg.ttl); } - if (flags & TNL_F_TOS_INHERIT) { + if (tnl_cfg.tos_inherit) { smap_add(args, "tos", "inherit"); - } else if (a[OVS_TUNNEL_ATTR_TOS]) { - int tos = nl_attr_get_u8(a[OVS_TUNNEL_ATTR_TOS]); - smap_add_format(args, "tos", "0x%x", tos); + } else if (tnl_cfg.tos) { + smap_add_format(args, "tos", "0x%x", tnl_cfg.tos); } - if (a[OVS_TUNNEL_ATTR_DST_PORT]) { - uint16_t dst_port = nl_attr_get_u16(a[OVS_TUNNEL_ATTR_DST_PORT]); - if (dst_port != VXLAN_DST_PORT) { + if (tnl_cfg.dst_port) { + uint16_t dst_port = ntohs(tnl_cfg.dst_port); + const char *type = netdev_get_type(dev); + + 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 (flags & TNL_F_CSUM) { + if (tnl_cfg.csum) { smap_add(args, "csum", "true"); } - if (flags & TNL_F_DF_INHERIT) { - smap_add(args, "df_inherit", "true"); - } - if (!(flags & TNL_F_DF_DEFAULT)) { + + if (!tnl_cfg.dont_fragment) { smap_add(args, "df_default", "false"); } return 0; } + +/* Code specific to patch ports. */ + +/* 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_) +{ + char *peer = NULL; + + if (netdev_vport_is_patch(netdev_)) { + struct netdev_vport *netdev = netdev_vport_cast(netdev_); + + ovs_mutex_lock(&netdev->mutex); + if (netdev->peer) { + peer = xstrdup(netdev->peer); + } + ovs_mutex_unlock(&netdev->mutex); + } + + return peer; +} + +void +netdev_vport_inc_rx(const struct netdev *netdev, + const struct dpif_flow_stats *stats) +{ + 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); + } +} + +void +netdev_vport_inc_tx(const struct netdev *netdev, + const struct dpif_flow_stats *stats) +{ + 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); + } +} + +static int +get_patch_config(const struct netdev *dev_, struct smap *args) +{ + struct netdev_vport *dev = netdev_vport_cast(dev_); + + ovs_mutex_lock(&dev->mutex); + if (dev->peer) { + smap_add(args, "peer", dev->peer); + } + ovs_mutex_unlock(&dev->mutex); + + return 0; +} static int -parse_patch_config(const char *name, const char *type OVS_UNUSED, - const struct smap *args, struct ofpbuf *options) +set_patch_config(struct netdev *dev_, const struct smap *args) { + struct netdev_vport *dev = netdev_vport_cast(dev_); + const char *name = netdev_get_name(dev_); const char *peer; peer = smap_get(args, "peer"); @@ -774,62 +837,666 @@ parse_patch_config(const char *name, const char *type OVS_UNUSED, return EINVAL; } - if (strlen(peer) >= IFNAMSIZ) { - VLOG_ERR("%s: patch 'peer' arg too long", name); + if (!strcmp(name, peer)) { + VLOG_ERR("%s: patch peer must not be self", name); return EINVAL; } - if (!strcmp(name, peer)) { - VLOG_ERR("%s: patch peer must not be self", name); + 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); + + return 0; +} + +static int +get_stats(const struct netdev *netdev, struct netdev_stats *stats) +{ + struct netdev_vport *dev = netdev_vport_cast(netdev); + + ovs_mutex_lock(&dev->mutex); + *stats = dev->stats; + ovs_mutex_unlock(&dev->mutex); + + return 0; +} + + +/* Tunnel push pop ops. */ + +static struct ip_header * +ip_hdr(void *eth) +{ + return (void *)((char *)eth + sizeof (struct eth_header)); +} + +static struct ovs_16aligned_ip6_hdr * +ipv6_hdr(void *eth) +{ + return (void *)((char *)eth + sizeof (struct eth_header)); +} + +static void * +ip_extract_tnl_md(struct dp_packet *packet, struct flow_tnl *tnl, + unsigned int *hlen) +{ + void *nh; + struct ip_header *ip; + struct ovs_16aligned_ip6_hdr *ip6; + void *l4; + int l3_size; + + nh = dp_packet_l3(packet); + ip = nh; + ip6 = nh; + l4 = dp_packet_l4(packet); + + if (!nh || !l4) { + return NULL; + } + + *hlen = sizeof(struct eth_header); + + l3_size = dp_packet_size(packet) - + ((char *)nh - (char *)dp_packet_data(packet)); + + if (IP_VER(ip->ip_ihl_ver) == 4) { + + ovs_be32 ip_src, ip_dst; + + if (csum(ip, IP_IHL(ip->ip_ihl_ver) * 4)) { + VLOG_WARN_RL(&err_rl, "ip packet has invalid checksum"); + return NULL; + } + + if (ntohs(ip->ip_tot_len) > l3_size) { + VLOG_WARN_RL(&err_rl, "ip packet is truncated (IP length %d, actual %d)", + ntohs(ip->ip_tot_len), l3_size); + return NULL; + } + if (IP_IHL(ip->ip_ihl_ver) * 4 > sizeof(struct ip_header)) { + VLOG_WARN_RL(&err_rl, "ip options not supported on tunnel packets " + "(%d bytes)", IP_IHL(ip->ip_ihl_ver) * 4); + return NULL; + } + + ip_src = get_16aligned_be32(&ip->ip_src); + ip_dst = get_16aligned_be32(&ip->ip_dst); + + tnl->ip_src = ip_src; + tnl->ip_dst = ip_dst; + tnl->ip_tos = ip->ip_tos; + tnl->ip_ttl = ip->ip_ttl; + + *hlen += IP_HEADER_LEN; + + } else if (IP_VER(ip->ip_ihl_ver) == 6) { + + memcpy(tnl->ipv6_src.s6_addr, ip6->ip6_src.be16, sizeof ip6->ip6_src); + memcpy(tnl->ipv6_dst.s6_addr, ip6->ip6_dst.be16, sizeof ip6->ip6_dst); + tnl->ip_tos = 0; + tnl->ip_ttl = ip6->ip6_hlim; + + *hlen += IPV6_HEADER_LEN; + + } else { + VLOG_WARN_RL(&err_rl, "ipv4 packet has invalid version (%d)", + IP_VER(ip->ip_ihl_ver)); + return NULL; + } + + return l4; +} + +static bool +is_header_ipv6(const void *header) +{ + const struct eth_header *eth; + eth = header; + return eth->eth_type == htons(ETH_TYPE_IPV6); +} + +/* 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; + struct ovs_16aligned_ip6_hdr *ip6; + + eth = dp_packet_push_uninit(packet, size); + *ip_tot_size = dp_packet_size(packet) - sizeof (struct eth_header); + + memcpy(eth, header, size); + + if (is_header_ipv6(header)) { + ip6 = ipv6_hdr(eth); + *ip_tot_size -= IPV6_HEADER_LEN; + ip6->ip6_plen = htons(*ip_tot_size); + return ip6 + 1; + } else { + ip = ip_hdr(eth); + ip->ip_tot_len = htons(*ip_tot_size); + ip->ip_csum = recalc_csum16(ip->ip_csum, 0, ip->ip_tot_len); + *ip_tot_size -= IP_HEADER_LEN; + return ip + 1; + } +} + +static void * +udp_extract_tnl_md(struct dp_packet *packet, struct flow_tnl *tnl, + unsigned int *hlen) +{ + struct udp_header *udp; + + udp = ip_extract_tnl_md(packet, tnl, hlen); + if (!udp) { + return NULL; + } + + if (udp->udp_csum) { + uint32_t csum; + if (is_header_ipv6(dp_packet_data(packet))) { + csum = packet_csum_pseudoheader6(dp_packet_l3(packet)); + } else { + 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 +push_udp_header(struct dp_packet *packet, + const struct ovs_action_push_tnl *data) +{ + struct udp_header *udp; + int ip_tot_size; + + udp = push_ip_header(packet, data->header, data->header_len, &ip_tot_size); + + /* set udp src port */ + udp->udp_src = get_src_port(packet); + udp->udp_len = htons(ip_tot_size); + + if (udp->udp_csum) { + uint32_t csum; + if (is_header_ipv6(dp_packet_data(packet))) { + csum = packet_csum_pseudoheader6(ipv6_hdr(dp_packet_data(packet))); + } else { + csum = packet_csum_pseudoheader(ip_hdr(dp_packet_data(packet))); + } + + csum = csum_continue(csum, udp, ip_tot_size); + udp->udp_csum = csum_finish(csum); + + if (!udp->udp_csum) { + udp->udp_csum = htons(0xffff); + } + } +} + +static void * +udp_build_header(struct netdev_tunnel_config *tnl_cfg, + const struct flow *tnl_flow, + struct ovs_action_push_tnl *data, + unsigned int *hlen) +{ + struct ip_header *ip; + struct ovs_16aligned_ip6_hdr *ip6; + struct udp_header *udp; + bool is_ipv6; + + *hlen = sizeof(struct eth_header); + + is_ipv6 = is_header_ipv6(data->header); + + if (is_ipv6) { + ip6 = ipv6_hdr(data->header); + ip6->ip6_nxt = IPPROTO_UDP; + udp = (struct udp_header *) (ip6 + 1); + *hlen += IPV6_HEADER_LEN; + } else { + ip = ip_hdr(data->header); + ip->ip_proto = IPPROTO_UDP; + udp = (struct udp_header *) (ip + 1); + *hlen += IP_HEADER_LEN; + } + + udp->udp_dst = tnl_cfg->dst_port; + + if (is_ipv6 || 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 udp + 1; +} + +static int +gre_header_len(ovs_be16 flags) +{ + int hlen = 4; + + if (flags & htons(GRE_CSUM)) { + hlen += 4; + } + if (flags & htons(GRE_KEY)) { + hlen += 4; + } + if (flags & htons(GRE_SEQ)) { + hlen += 4; + } + return hlen; +} + +static int +parse_gre_header(struct dp_packet *packet, + struct flow_tnl *tnl) +{ + const struct gre_base_hdr *greh; + ovs_16aligned_be32 *options; + int hlen; + unsigned int ulen; + + greh = ip_extract_tnl_md(packet, tnl, &ulen); + if (!greh) { + return -EINVAL; + } + + if (greh->flags & ~(htons(GRE_CSUM | GRE_KEY | GRE_SEQ))) { + return -EINVAL; + } + + if (greh->protocol != htons(ETH_TYPE_TEB)) { + return -EINVAL; + } + + hlen = ulen + 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++; + } + + 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) +{ + /* Zero up through the tunnel metadata options. The length and table + * are before this and as long as they are empty, the options won't + * be looked at. */ + memset(md, 0, offsetof(struct pkt_metadata, tunnel.metadata.opts)); +} + +static int +netdev_gre_pop_header(struct dp_packet *packet) +{ + struct pkt_metadata *md = &packet->md; + struct flow_tnl *tnl = &md->tunnel; + int hlen = sizeof(struct eth_header) + 4; + + hlen += is_header_ipv6(dp_packet_data(packet)) ? + IPV6_HEADER_LEN : IP_HEADER_LEN; + + pkt_metadata_init_tnl(md); + if (hlen > dp_packet_size(packet)) { return EINVAL; } - nl_msg_put_string(options, OVS_PATCH_ATTR_PEER, peer); + hlen = parse_gre_header(packet, tnl); + if (hlen < 0) { + return -hlen; + } + + dp_packet_reset_packet(packet, hlen); 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; + + greh = push_ip_header(packet, data->header, data->header_len, &ip_tot_size); + + if (greh->flags & htons(GRE_CSUM)) { + ovs_be16 *csum_opt = (ovs_be16 *) (greh + 1); + *csum_opt = csum(greh, ip_tot_size); + } +} + static int -unparse_patch_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, - const struct nlattr *options, size_t options_len, - struct smap *args) -{ - static const struct nl_policy ovs_patch_policy[] = { - [OVS_PATCH_ATTR_PEER] = { .type = NL_A_STRING, - .max_len = IFNAMSIZ, - .optional = false } - }; +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 ovs_16aligned_ip6_hdr *ip6; + struct gre_base_hdr *greh; + ovs_16aligned_be32 *options; + int hlen; + bool is_ipv6; + + is_ipv6 = is_header_ipv6(data->header); + + /* XXX: RCUfy tnl_cfg. */ + ovs_mutex_lock(&dev->mutex); + tnl_cfg = &dev->tnl_cfg; + + if (is_ipv6) { + ip6 = ipv6_hdr(data->header); + ip6->ip6_nxt = IPPROTO_GRE; + greh = (struct gre_base_hdr *) (ip6 + 1); + } else { + ip = ip_hdr(data->header); + ip->ip_proto = IPPROTO_GRE; + greh = (struct gre_base_hdr *) (ip + 1); + } + + 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 (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++; + } + + ovs_mutex_unlock(&dev->mutex); + + hlen = (uint8_t *) options - (uint8_t *) greh; + + data->header_len = sizeof(struct eth_header) + hlen + + (is_ipv6 ? IPV6_HEADER_LEN : IP_HEADER_LEN); + 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; + unsigned int hlen; + + pkt_metadata_init_tnl(md); + if (VXLAN_HLEN > dp_packet_l4_size(packet)) { + return EINVAL; + } + + vxh = udp_extract_tnl_md(packet, tnl, &hlen); + if (!vxh) { + return EINVAL; + } + + 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, hlen + VXLAN_HLEN); + + return 0; +} + +static int +netdev_vxlan_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 vxlanhdr *vxh; + unsigned int hlen; + + /* XXX: RCUfy tnl_cfg. */ + ovs_mutex_lock(&dev->mutex); + tnl_cfg = &dev->tnl_cfg; + + vxh = udp_build_header(tnl_cfg, tnl_flow, data, &hlen); + + 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 = hlen + 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, opts_len, ulen; + + pkt_metadata_init_tnl(md); + if (GENEVE_BASE_HLEN > dp_packet_l4_size(packet)) { + VLOG_WARN_RL(&err_rl, "geneve packet too small: min header=%u packet size=%"PRIuSIZE"\n", + (unsigned int)GENEVE_BASE_HLEN, dp_packet_l4_size(packet)); + return EINVAL; + } + + gnh = udp_extract_tnl_md(packet, tnl, &ulen); + if (!gnh) { + return EINVAL; + } + + opts_len = gnh->opt_len * 4; + hlen = ulen + GENEVE_BASE_HLEN + opts_len; + 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; + } - struct nlattr *a[ARRAY_SIZE(ovs_patch_policy)]; - struct ofpbuf buf; + if (gnh->ver != 0) { + VLOG_WARN_RL(&err_rl, "unknown geneve version: %"PRIu8"\n", gnh->ver); + return EINVAL; + } - ofpbuf_use_const(&buf, options, options_len); - if (!nl_policy_parse(&buf, 0, ovs_patch_policy, - a, ARRAY_SIZE(ovs_patch_policy))) { + 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; } - smap_add(args, "peer", nl_attr_get_string(a[OVS_PATCH_ATTR_PEER])); + 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; + + memcpy(tnl->metadata.opts.gnv, gnh->options, opts_len); + tnl->metadata.present.len = opts_len; + tnl->flags |= FLOW_TNL_F_UDPIF; + + 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; + unsigned int hlen; + + /* XXX: RCUfy tnl_cfg. */ + ovs_mutex_lock(&dev->mutex); + tnl_cfg = &dev->tnl_cfg; + + gnh = udp_build_header(tnl_cfg, tnl_flow, data, &hlen); + + 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, + 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 = hlen + 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) \ +#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_get_config, \ - netdev_vport_set_config, \ - \ - netdev_vport_open, \ - netdev_vport_close, \ - \ - NULL, /* listen */ \ - 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 */ \ @@ -842,8 +1509,7 @@ unparse_patch_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, NULL, /* get_carrier */ \ NULL, /* get_carrier_resets */ \ NULL, /* get_miimon */ \ - netdev_vport_get_stats, \ - NULL, /* set_stats */ \ + get_stats, \ \ NULL, /* get_features */ \ NULL, /* set_advertisements */ \ @@ -857,7 +1523,9 @@ unparse_patch_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, 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 */ \ @@ -870,32 +1538,66 @@ unparse_patch_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, \ netdev_vport_update_flags, \ \ - netdev_vport_change_seq - -#define TUNNEL_CLASS(NAME, VPORT_TYPE) \ - { VPORT_TYPE, \ - { NAME, VPORT_FUNCTIONS(tunnel_get_status) }, \ - parse_tunnel_config, unparse_tunnel_config } + 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[] = { - TUNNEL_CLASS("gre", OVS_VPORT_TYPE_GRE), - TUNNEL_CLASS("ipsec_gre", OVS_VPORT_TYPE_GRE), - TUNNEL_CLASS("gre64", OVS_VPORT_TYPE_GRE64), - TUNNEL_CLASS("ipsec_gre64", OVS_VPORT_TYPE_GRE64), - TUNNEL_CLASS("capwap", OVS_VPORT_TYPE_CAPWAP), - TUNNEL_CLASS("vxlan", OVS_VPORT_TYPE_VXLAN), - - { OVS_VPORT_TYPE_PATCH, - { "patch", VPORT_FUNCTIONS(NULL) }, - parse_patch_config, unparse_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("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); +}