X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fnetdev-vport.c;h=4214b383d2414006d93dc369e8277356e390d90c;hb=5453ae2067671c0d40a5b3ac3cb3d4027bed6abb;hp=72175b77ff7fb4bc6dadaf74110014db3752ac6a;hpb=b9ad7294a5b4faf886a5022c5ec53b500f93cf3f;p=sliver-openvswitch.git diff --git a/lib/netdev-vport.c b/lib/netdev-vport.c index 72175b77f..4214b383d 100644 --- a/lib/netdev-vport.c +++ b/lib/netdev-vport.c @@ -21,8 +21,6 @@ #include #include #include -#include -#include #include #include @@ -30,33 +28,26 @@ #include "daemon.h" #include "dirs.h" #include "dpif.h" -#include "dpif-linux.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 "packets.h" #include "route-table.h" #include "shash.h" #include "socket-util.h" -#include "unaligned.h" #include "vlog.h" VLOG_DEFINE_THIS_MODULE(netdev_vport); -/* Default to the OTV port, per the VXLAN IETF draft. */ -#define VXLAN_DST_PORT 8472 +#define VXLAN_DST_PORT 4789 +#define LISP_DST_PORT 4341 #define DEFAULT_TTL 64 -struct netdev_dev_vport { - struct netdev_dev netdev_dev; +struct netdev_vport { + struct netdev up; unsigned int change_seq; uint8_t etheraddr[ETH_ADDR_LEN]; struct netdev_stats stats; @@ -69,15 +60,15 @@ struct netdev_dev_vport { }; struct vport_class { - enum ovs_vport_type type; const char *dpif_port; struct netdev_class netdev_class; }; static int netdev_vport_create(const struct netdev_class *, const char *, - struct netdev_dev **); -static int get_patch_config(struct netdev_dev *, struct smap *args); -static void netdev_vport_poll_notify(struct netdev_dev_vport *); + struct netdev **); +static int get_patch_config(const struct netdev *, struct smap *args); +static int get_tunnel_config(const struct netdev *, struct smap *args); +static void netdev_vport_poll_notify(struct netdev_vport *); static bool is_vport_class(const struct netdev_class *class) @@ -92,108 +83,112 @@ vport_class_cast(const struct netdev_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) { - ovs_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; } -static const struct netdev_tunnel_config * -get_netdev_tunnel_config(const struct netdev_dev *netdev_dev) +bool +netdev_vport_is_patch(const struct netdev *netdev) { - return &netdev_dev_vport_cast(netdev_dev)->tnl_cfg; + const struct netdev_class *class = netdev_get_class(netdev); + + return class->get_config == get_patch_config; } -enum ovs_vport_type -netdev_vport_get_vport_type(const struct netdev *netdev) +static bool +netdev_vport_needs_dst_port(const struct netdev *dev) { - const struct netdev_dev *dev = netdev_get_dev(netdev); - const struct netdev_class *class = netdev_dev_get_class(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); + const struct netdev_class *class = netdev_get_class(dev); + const char *type = netdev_get_type(dev); + + return (class->get_config == get_tunnel_config && + (!strcmp("vxlan", type) || !strcmp("lisp", type))); } -bool -netdev_vport_is_patch(const struct netdev *netdev) +const char * +netdev_vport_class_get_dpif_port(const struct netdev_class *class) { - const struct netdev_dev *dev = netdev_get_dev(netdev); - const struct netdev_class *class = netdev_dev_get_class(dev); - - return class->get_config == get_patch_config; + return is_vport_class(class) ? vport_class_cast(class)->dpif_port : NULL; } const char * -netdev_vport_get_dpif_port(const struct netdev *netdev) +netdev_vport_get_dpif_port(const struct netdev *netdev, + char namebuf[], size_t bufsize) +{ + if (netdev_vport_needs_dst_port(netdev)) { + const struct netdev_vport *vport = netdev_vport_cast(netdev); + const char *type = netdev_get_type(netdev); + + /* + * Note: IFNAMSIZ is 16 bytes long. The maximum length of a VXLAN + * or LISP port name below is 15 or 14 bytes respectively. Still, + * assert here on the size of strlen(type) in case that changes + * in the future. + */ + BUILD_ASSERT(NETDEV_VPORT_NAME_BUFSIZE >= IFNAMSIZ); + ovs_assert(strlen(type) + 10 < IFNAMSIZ); + snprintf(namebuf, bufsize, "%s_sys_%d", type, + ntohs(vport->tnl_cfg.dst_port)); + return namebuf; + } else { + const struct netdev_class *class = netdev_get_class(netdev); + const char *dpif_port = netdev_vport_class_get_dpif_port(class); + return dpif_port ? dpif_port : netdev_get_name(netdev); + } +} + +char * +netdev_vport_get_dpif_port_strdup(const struct netdev *netdev) { - const struct netdev_dev *dev = netdev_get_dev(netdev); - const struct netdev_class *class = netdev_dev_get_class(dev); - const char *dpif_port; + char namebuf[NETDEV_VPORT_NAME_BUFSIZE]; - dpif_port = (is_vport_class(class) - ? vport_class_cast(class)->dpif_port - : NULL); - return dpif_port ? dpif_port : netdev_get_name(netdev); + return xstrdup(netdev_vport_get_dpif_port(netdev, namebuf, + sizeof namebuf)); } static int netdev_vport_create(const struct netdev_class *netdev_class, const char *name, - struct netdev_dev **netdev_devp) + struct netdev **netdevp) { - struct netdev_dev_vport *dev; + struct netdev_vport *dev; dev = xzalloc(sizeof *dev); - netdev_dev_init(&dev->netdev_dev, name, netdev_class); + netdev_init(&dev->up, name, netdev_class); dev->change_seq = 1; eth_addr_random(dev->etheraddr); - *netdev_devp = &dev->netdev_dev; + *netdevp = &dev->up; route_table_register(); return 0; } static void -netdev_vport_destroy(struct netdev_dev *netdev_dev_) +netdev_vport_destroy(struct netdev *netdev_) { - struct netdev_dev_vport *netdev_dev = netdev_dev_vport_cast(netdev_dev_); + struct netdev_vport *netdev = netdev_vport_cast(netdev_); route_table_unregister(); - free(netdev_dev->peer); - free(netdev_dev); -} - -static int -netdev_vport_open(struct netdev_dev *netdev_dev, struct netdev **netdevp) -{ - *netdevp = xmalloc(sizeof **netdevp); - netdev_init(*netdevp, netdev_dev); - return 0; -} - -static void -netdev_vport_close(struct netdev *netdev) -{ + free(netdev->peer); 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 netdev_dev_vport *dev = netdev_vport_get_dev(netdev); - memcpy(dev->etheraddr, mac, ETH_ADDR_LEN); - netdev_vport_poll_notify(dev); + struct netdev_vport *netdev = netdev_vport_cast(netdev_); + memcpy(netdev->etheraddr, mac, ETH_ADDR_LEN); + netdev_vport_poll_notify(netdev); return 0; } @@ -201,69 +196,17 @@ static int netdev_vport_get_etheraddr(const struct netdev *netdev, uint8_t mac[ETH_ADDR_LEN]) { - memcpy(mac, netdev_vport_get_dev(netdev)->etheraddr, ETH_ADDR_LEN); - return 0; -} - -/* 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; - } - - netdev_stats_from_ovs_vport_stats(stats, reply.stats); - - ofpbuf_delete(buf); - + memcpy(mac, netdev_vport_cast(netdev)->etheraddr, ETH_ADDR_LEN); return 0; } static int tunnel_get_status(const struct netdev *netdev, struct smap *smap) { - static char iface[IFNAMSIZ]; + char iface[IFNAMSIZ]; ovs_be32 route; - route = netdev_vport_get_dev(netdev)->tnl_cfg.ip_dst; + route = netdev_vport_cast(netdev)->tnl_cfg.ip_dst; if (route_table_get_name(route, iface)) { struct netdev *egress_netdev; @@ -281,8 +224,9 @@ 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; @@ -295,7 +239,7 @@ netdev_vport_update_flags(struct netdev *netdev OVS_UNUSED, static unsigned int netdev_vport_change_seq(const struct netdev *netdev) { - return netdev_vport_get_dev(netdev)->change_seq; + return netdev_vport_cast(netdev)->change_seq; } static void @@ -313,7 +257,7 @@ netdev_vport_wait(void) /* Helper functions. */ static void -netdev_vport_poll_notify(struct netdev_dev_vport *ndv) +netdev_vport_poll_notify(struct netdev_vport *ndv) { ndv->change_seq++; if (!ndv->change_seq) { @@ -351,11 +295,11 @@ parse_key(const struct smap *args, const char *name, } static int -set_tunnel_config(struct netdev_dev *dev_, const struct smap *args) +set_tunnel_config(struct netdev *dev_, const struct smap *args) { - struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_); - const char *name = netdev_dev_get_name(dev_); - const char *type = netdev_dev_get_type(dev_); + 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; @@ -364,25 +308,31 @@ set_tunnel_config(struct netdev_dev *dev_, const struct smap *args) ipsec_mech_set = false; memset(&tnl_cfg, 0, sizeof tnl_cfg); - if (!strcmp(type, "capwap")) { - VLOG_WARN_ONCE("CAPWAP tunnel support is deprecated."); - } - - needs_dst_port = !strcmp(type, "vxlan"); + 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)) { + 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 (lookup_ip(node->value, &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; @@ -453,18 +403,27 @@ set_tunnel_config(struct netdev_dev *dev_, const struct smap *args) } /* Add a default destination port for VXLAN if none specified. */ - if (needs_dst_port && !tnl_cfg.dst_port) { + if (!strcmp(type, "vxlan") && !tnl_cfg.dst_port) { tnl_cfg.dst_port = htons(VXLAN_DST_PORT); } + /* Add a default destination port for LISP if none specified. */ + if (!strcmp(type, "lisp") && !tnl_cfg.dst_port) { + tnl_cfg.dst_port = htons(LISP_DST_PORT); + } + if (tnl_cfg.ipsec) { + static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER; static pid_t pid = 0; + + pthread_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); } + pthread_mutex_unlock(&mutex); if (pid < 0) { VLOG_ERR("%s: IPsec requires the ovs-monitor-ipsec daemon", @@ -484,19 +443,16 @@ set_tunnel_config(struct netdev_dev *dev_, const struct smap *args) } } - if (!tnl_cfg.ip_dst) { + 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) { - if (ip_is_multicast(tnl_cfg.ip_dst)) { - VLOG_WARN("%s: remote_ip is multicast, ignoring local_ip", name); - tnl_cfg.ip_src = 0; - } + 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; } @@ -516,17 +472,21 @@ set_tunnel_config(struct netdev_dev *dev_, const struct smap *args) } static int -get_tunnel_config(struct netdev_dev *dev, struct smap *args) +get_tunnel_config(const struct netdev *dev, struct smap *args) { const struct netdev_tunnel_config *tnl_cfg = - &netdev_dev_vport_cast(dev)->tnl_cfg; + &netdev_vport_cast(dev)->tnl_cfg; 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) { @@ -564,7 +524,10 @@ get_tunnel_config(struct netdev_dev *dev, struct smap *args) if (tnl_cfg->dst_port) { uint16_t dst_port = ntohs(tnl_cfg->dst_port); - if (dst_port != VXLAN_DST_PORT) { + const char *type = netdev_get_type(dev); + + if ((!strcmp("vxlan", type) && dst_port != VXLAN_DST_PORT) || + (!strcmp("lisp", type) && dst_port != LISP_DST_PORT)) { smap_add_format(args, "dst_port", "%d", dst_port); } } @@ -585,17 +548,17 @@ get_tunnel_config(struct netdev_dev *dev, struct smap *args) const char * netdev_vport_patch_peer(const struct netdev *netdev) { - return netdev_vport_is_patch(netdev) - ? netdev_vport_get_dev(netdev)->peer - : NULL; + return (netdev_vport_is_patch(netdev) + ? netdev_vport_cast(netdev)->peer + : NULL); } void netdev_vport_inc_rx(const struct netdev *netdev, const struct dpif_flow_stats *stats) { - if (is_vport_class(netdev_dev_get_class(netdev_get_dev(netdev)))) { - struct netdev_dev_vport *dev = netdev_vport_get_dev(netdev); + if (is_vport_class(netdev_get_class(netdev))) { + struct netdev_vport *dev = netdev_vport_cast(netdev); dev->stats.rx_packets += stats->n_packets; dev->stats.rx_bytes += stats->n_bytes; } @@ -605,17 +568,17 @@ void netdev_vport_inc_tx(const struct netdev *netdev, const struct dpif_flow_stats *stats) { - if (is_vport_class(netdev_dev_get_class(netdev_get_dev(netdev)))) { - struct netdev_dev_vport *dev = netdev_vport_get_dev(netdev); + if (is_vport_class(netdev_get_class(netdev))) { + struct netdev_vport *dev = netdev_vport_cast(netdev); dev->stats.tx_packets += stats->n_packets; dev->stats.tx_bytes += stats->n_bytes; } } static int -get_patch_config(struct netdev_dev *dev_, struct smap *args) +get_patch_config(const struct netdev *dev_, struct smap *args) { - struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_); + struct netdev_vport *dev = netdev_vport_cast(dev_); if (dev->peer) { smap_add(args, "peer", dev->peer); @@ -624,10 +587,10 @@ get_patch_config(struct netdev_dev *dev_, struct smap *args) } static int -set_patch_config(struct netdev_dev *dev_, const struct smap *args) +set_patch_config(struct netdev *dev_, const struct smap *args) { - struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_); - const char *name = netdev_dev_get_name(dev_); + struct netdev_vport *dev = netdev_vport_cast(dev_); + const char *name = netdev_get_name(dev_); const char *peer; peer = smap_get(args, "peer"); @@ -648,14 +611,14 @@ set_patch_config(struct netdev_dev *dev_, const struct smap *args) free(dev->peer); dev->peer = xstrdup(peer); - + netdev_vport_poll_notify(dev); return 0; } static int get_stats(const struct netdev *netdev, struct netdev_stats *stats) { - struct netdev_dev_vport *dev = netdev_vport_get_dev(netdev); + struct netdev_vport *dev = netdev_vport_cast(netdev); memcpy(stats, &dev->stats, sizeof *stats); return 0; } @@ -672,13 +635,7 @@ get_stats(const struct netdev *netdev, struct netdev_stats *stats) SET_CONFIG, \ GET_TUNNEL_CONFIG, \ \ - netdev_vport_open, \ - netdev_vport_close, \ - \ - NULL, /* listen */ \ - NULL, /* recv */ \ - NULL, /* recv_wait */ \ - NULL, /* drain */ \ + NULL, /* rx_open */ \ \ NULL, /* send */ \ NULL, /* send_wait */ \ @@ -721,34 +678,44 @@ get_stats(const struct netdev *netdev, struct netdev_stats *stats) \ netdev_vport_change_seq -#define TUNNEL_CLASS(NAME, VPORT_TYPE, DPIF_PORT) \ - { VPORT_TYPE, DPIF_PORT, \ +#define TUNNEL_CLASS(NAME, DPIF_PORT) \ + { DPIF_PORT, \ { NAME, VPORT_FUNCTIONS(get_tunnel_config, \ set_tunnel_config, \ get_netdev_tunnel_config, \ tunnel_get_status) }} void -netdev_vport_register(void) +netdev_vport_tunnel_register(void) { static const struct vport_class vport_classes[] = { - TUNNEL_CLASS("gre", OVS_VPORT_TYPE_GRE, "gre_system"), - TUNNEL_CLASS("ipsec_gre", OVS_VPORT_TYPE_GRE, "gre_system"), - TUNNEL_CLASS("gre64", OVS_VPORT_TYPE_GRE64, "gre64_system"), - TUNNEL_CLASS("ipsec_gre64", OVS_VPORT_TYPE_GRE64, "gre64_system"), - TUNNEL_CLASS("capwap", OVS_VPORT_TYPE_CAPWAP, "capwap_system"), - TUNNEL_CLASS("vxlan", OVS_VPORT_TYPE_VXLAN, "vxlan_system"), - - { OVS_VPORT_TYPE_UNSPEC, NULL, - { "patch", VPORT_FUNCTIONS(get_patch_config, - set_patch_config, - NULL, - NULL) }}, + TUNNEL_CLASS("gre", "gre_system"), + TUNNEL_CLASS("ipsec_gre", "gre_system"), + TUNNEL_CLASS("gre64", "gre64_system"), + TUNNEL_CLASS("ipsec_gre64", "gre64_system"), + TUNNEL_CLASS("vxlan", "vxlan_system"), + TUNNEL_CLASS("lisp", "lisp_system") }; + static bool inited; int i; - for (i = 0; i < ARRAY_SIZE(vport_classes); i++) { - netdev_register_provider(&vport_classes[i].netdev_class); + if (!inited) { + inited = true; + for (i = 0; i < ARRAY_SIZE(vport_classes); i++) { + netdev_register_provider(&vport_classes[i].netdev_class); + } } } + +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) }}; + netdev_register_provider(&patch_class.netdev_class); +}