X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fnetdev-vport.c;h=d85d528454b46201675ff252ab022482bd914217;hb=f613a0d72c521ca3a4eeb2c29ac523f6fdf72667;hp=e930e1bef7e02ddffc2a3dc9f41f91cc38f0c7de;hpb=c283069c71adc49c182a1ac569a05e2dca949eda;p=sliver-openvswitch.git diff --git a/lib/netdev-vport.c b/lib/netdev-vport.c index e930e1bef..d85d52845 100644 --- a/lib/netdev-vport.c +++ b/lib/netdev-vport.c @@ -26,44 +26,34 @@ #include #include "byte-order.h" +#include "daemon.h" +#include "dirs.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/datapath-protocol.h" #include "openvswitch/tunnel.h" #include "packets.h" -#include "rtnetlink.h" #include "route-table.h" -#include "rtnetlink-link.h" #include "shash.h" #include "socket-util.h" #include "vlog.h" VLOG_DEFINE_THIS_MODULE(netdev_vport); -static struct hmap name_map; -static struct rtnetlink_notifier netdev_vport_link_notifier; - -struct name_node { - struct hmap_node node; /* Node in name_map. */ - uint32_t ifi_index; /* Kernel interface index. */ - - char ifname[IFNAMSIZ]; /* Interface name. */ -}; - -struct netdev_vport_notifier { - struct netdev_notifier notifier; - struct list list_node; - struct shash_node *shash_node; -}; - struct netdev_dev_vport { struct netdev_dev netdev_dev; - uint64_t config[VPORT_CONFIG_SIZE / 8]; + struct ofpbuf *options; + int dp_ifindex; /* -1 if unknown. */ + uint32_t port_no; /* UINT32_MAX if unknown. */ + unsigned int change_seq; }; struct netdev_vport { @@ -71,27 +61,24 @@ struct netdev_vport { }; struct vport_class { - enum odp_vport_type type; + enum ovs_vport_type type; struct netdev_class netdev_class; int (*parse_config)(const char *name, const char *type, - const struct shash *args, void *config); + const struct shash *args, struct ofpbuf *options); int (*unparse_config)(const char *name, const char *type, - const void *config, struct shash *args); + const struct nlattr *options, size_t options_len, + struct shash *args); }; -static struct shash netdev_vport_notifiers = - SHASH_INITIALIZER(&netdev_vport_notifiers); - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); -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 **); + 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 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 bool @@ -122,86 +109,86 @@ netdev_vport_cast(const struct netdev *netdev) return CONTAINER_OF(netdev, struct netdev_vport, netdev); } -/* 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) +/* 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) { const struct netdev_dev *dev = netdev_get_dev(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 (is_vport_class(netdev_dev_get_class(dev)) + ? netdev_dev_vport_cast(dev)->options + : NULL); } -static int -netdev_vport_init(void) +enum ovs_vport_type +netdev_vport_get_vport_type(const struct netdev *netdev) { - netdev_vport_tnl_iface_init(); - route_table_register(); - return 0; + 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 ? OVS_VPORT_TYPE_NETDEV + : OVS_VPORT_TYPE_UNSPEC); } -static int -netdev_vport_create(const struct netdev_class *netdev_class, const char *name, - const struct shash *args, - struct netdev_dev **netdev_devp) +const char * +netdev_vport_get_netdev_type(const struct dpif_linux_vport *vport) { - const struct vport_class *vport_class = vport_class_cast(netdev_class); - uint64_t config[VPORT_CONFIG_SIZE / 8]; - struct shash fetched_args; - int error; + struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]; - memset(config, 0, sizeof config); - shash_init(&fetched_args); + switch (vport->type) { + case OVS_VPORT_TYPE_UNSPEC: + break; - if (!shash_is_empty(args)) { - /* Parse the provided configuration. */ - error = vport_class->parse_config(name, netdev_class->type, - args, config); - } else { - /* Fetch an existing configuration from the kernel. - * - * This case could be ambiguous with initializing a new vport with an - * empty configuration, but none of the existing vport classes accept - * an empty configuration. */ - struct odp_port odp_port; - - memset(&odp_port, 0, sizeof odp_port); - strncpy(odp_port.devname, name, sizeof odp_port.devname); - error = netdev_vport_do_ioctl(ODP_VPORT_QUERY, &odp_port); - if (!error) { - /* XXX verify correct type */ - memcpy(config, odp_port.config, sizeof config); - error = vport_class->unparse_config(name, netdev_class->type, - odp_port.config, - &fetched_args); - if (error) { - VLOG_ERR_RL(&rl, "%s: failed to parse kernel config (%s)", - name, strerror(error)); - } - } else { - VLOG_ERR_RL(&rl, "%s: vport query failed (%s)", - name, strerror(error)); + case OVS_VPORT_TYPE_NETDEV: + return "system"; + + case OVS_VPORT_TYPE_INTERNAL: + return "internal"; + + case OVS_VPORT_TYPE_PATCH: + return "patch"; + + case OVS_VPORT_TYPE_GRE: + if (tnl_port_config_from_nlattr(vport->options, vport->options_len, + a)) { + break; } + return (nl_attr_get_u32(a[OVS_TUNNEL_ATTR_FLAGS]) & TNL_F_IPSEC + ? "ipsec_gre" : "gre"); + + case OVS_VPORT_TYPE_CAPWAP: + return "capwap"; + + case __OVS_VPORT_TYPE_MAX: + break; } - if (!error) { - struct netdev_dev_vport *dev; + VLOG_WARN_RL(&rl, "dp%d: port `%s' has unsupported type %u", + vport->dp_ifindex, vport->name, (unsigned int) vport->type); + return "unknown"; +} - dev = xmalloc(sizeof *dev); - netdev_dev_init(&dev->netdev_dev, name, - shash_is_empty(&fetched_args) ? args : &fetched_args, - netdev_class); - memcpy(dev->config, config, sizeof dev->config); +static int +netdev_vport_create(const struct netdev_class *netdev_class, const char *name, + struct netdev_dev **netdev_devp) +{ + struct netdev_dev_vport *dev; - *netdev_devp = &dev->netdev_dev; - } + dev = xmalloc(sizeof *dev); + netdev_dev_init(&dev->netdev_dev, name, netdev_class); + dev->options = NULL; + dev->dp_ifindex = -1; + dev->port_no = UINT32_MAX; + dev->change_seq = 1; - shash_destroy(&fetched_args); + *netdev_devp = &dev->netdev_dev; + route_table_register(); - return error; + return 0; } static void @@ -214,8 +201,7 @@ netdev_vport_destroy(struct netdev_dev *netdev_dev_) } static int -netdev_vport_open(struct netdev_dev *netdev_dev_, int ethertype OVS_UNUSED, - struct netdev **netdevp) +netdev_vport_open(struct netdev_dev *netdev_dev_, struct netdev **netdevp) { struct netdev_vport *netdev; @@ -233,125 +219,205 @@ netdev_vport_close(struct netdev *netdev_) free(netdev); } +static int +netdev_vport_get_config(struct netdev_dev *dev_, 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_); + 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; + } + + dev->options = ofpbuf_clone_data(reply.options, reply.options_len); + dev->dp_ifindex = reply.dp_ifindex; + dev->port_no = reply.port_no; + ofpbuf_delete(buf); + } + + 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; +} + static int netdev_vport_set_config(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; + const char *name = netdev_dev_get_name(dev_); + struct ofpbuf *options; int error; - memset(&port, 0, sizeof port); - strncpy(port.devname, netdev_dev_get_name(dev_), sizeof port.devname); - port.type = vport_class->type; - error = vport_class->parse_config(netdev_dev_get_name(dev_), - netdev_dev_get_type(dev_), - args, port.config); - if (!error && memcmp(port.config, dev->config, sizeof dev->config)) { - error = netdev_vport_do_ioctl(ODP_VPORT_MOD, &port); + 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()). */ - memcpy(dev->config, port.config, sizeof dev->config); + ofpbuf_delete(dev->options); + dev->options = options; + options = NULL; + error = 0; } } + ofpbuf_delete(options); + return error; } static int -netdev_vport_set_etheraddr(struct netdev *netdev, - const uint8_t mac[ETH_ADDR_LEN]) +netdev_vport_send(struct netdev *netdev, const void *data, size_t size) { - struct odp_vport_ether vport_ether; - int err; - - ovs_strlcpy(vport_ether.devname, netdev_get_name(netdev), - sizeof vport_ether.devname); + struct netdev_dev *dev_ = netdev_get_dev(netdev); + struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_); - memcpy(vport_ether.ether_addr, mac, ETH_ADDR_LEN); + if (dev->dp_ifindex == -1) { + const char *name = netdev_get_name(netdev); + struct dpif_linux_vport reply; + struct ofpbuf *buf; + int error; - err = netdev_vport_do_ioctl(ODP_VPORT_ETHER_SET, &vport_ether); - if (err) { - return err; + error = dpif_linux_vport_get(name, &reply, &buf); + if (error) { + VLOG_ERR_RL(&rl, "%s: failed to query vport for send (%s)", + name, strerror(error)); + return error; + } + dev->dp_ifindex = reply.dp_ifindex; + dev->port_no = reply.port_no; + ofpbuf_delete(buf); } - netdev_vport_poll_notify(netdev); - return 0; + return dpif_linux_vport_send(dev->dp_ifindex, dev->port_no, data, size); } static int -netdev_vport_get_etheraddr(const struct netdev *netdev, - uint8_t mac[ETH_ADDR_LEN]) +netdev_vport_set_etheraddr(struct netdev *netdev, + const uint8_t mac[ETH_ADDR_LEN]) { - struct odp_vport_ether vport_ether; - int err; + struct dpif_linux_vport vport; + int error; - ovs_strlcpy(vport_ether.devname, netdev_get_name(netdev), - sizeof vport_ether.devname); + dpif_linux_vport_init(&vport); + vport.cmd = OVS_VPORT_CMD_SET; + vport.name = netdev_get_name(netdev); + vport.address = mac; - err = netdev_vport_do_ioctl(ODP_VPORT_ETHER_GET, &vport_ether); - if (err) { - return err; + error = dpif_linux_vport_transact(&vport, NULL, NULL); + if (!error) { + netdev_vport_poll_notify(netdev); } - - memcpy(mac, vport_ether.ether_addr, ETH_ADDR_LEN); - return 0; + return error; } static int -netdev_vport_get_mtu(const struct netdev *netdev, int *mtup) +netdev_vport_get_etheraddr(const struct netdev *netdev, + uint8_t mac[ETH_ADDR_LEN]) { - struct odp_vport_mtu vport_mtu; - int err; - - ovs_strlcpy(vport_mtu.devname, netdev_get_name(netdev), - sizeof vport_mtu.devname); + struct dpif_linux_vport reply; + struct ofpbuf *buf; + int error; - err = netdev_vport_do_ioctl(ODP_VPORT_MTU_GET, &vport_mtu); - if (err) { - return err; + error = dpif_linux_vport_get(netdev_get_name(netdev), &reply, &buf); + if (!error) { + if (reply.address) { + memcpy(mac, reply.address, ETH_ADDR_LEN); + } else { + error = EOPNOTSUPP; + } + ofpbuf_delete(buf); } + return error; +} - *mtup = vport_mtu.mtu; - return 0; +#define COPY_OVS_STATS \ + dst->rx_packets = src->rx_packets; \ + dst->tx_packets = src->tx_packets; \ + dst->rx_bytes = src->rx_bytes; \ + dst->tx_bytes = src->tx_bytes; \ + dst->rx_errors = src->rx_errors; \ + dst->tx_errors = src->tx_errors; \ + dst->rx_dropped = src->rx_dropped; \ + dst->tx_dropped = src->tx_dropped; + +/* Copies 'src' into 'dst', performing format conversion in the process. */ +static void +netdev_stats_from_ovs_vport_stats(struct netdev_stats *dst, + const struct ovs_vport_stats *src) +{ + COPY_OVS_STATS + 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; +} + +/* Copies 'src' into 'dst', performing format conversion in the process. */ +static void +netdev_stats_to_ovs_vport_stats(struct ovs_vport_stats *dst, + const struct netdev_stats *src) +{ + COPY_OVS_STATS } int netdev_vport_get_stats(const struct netdev *netdev, struct netdev_stats *stats) { - const char *name = netdev_get_name(netdev); - struct odp_vport_stats_req ovsr; - int err; + 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; + } - ovs_strlcpy(ovsr.devname, name, sizeof ovsr.devname); - err = netdev_vport_do_ioctl(ODP_VPORT_STATS_GET, &ovsr); - if (err) { - return err; - } - - 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; + netdev_stats_from_ovs_vport_stats(stats, reply.stats); + + ofpbuf_delete(buf); return 0; } @@ -359,34 +425,18 @@ netdev_vport_get_stats(const struct netdev *netdev, struct netdev_stats *stats) int netdev_vport_set_stats(struct netdev *netdev, const struct netdev_stats *stats) { - struct odp_vport_stats_req ovsr; + struct ovs_vport_stats rtnl_stats; + struct dpif_linux_vport vport; int err; - 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); + netdev_stats_to_ovs_vport_stats(&rtnl_stats, stats); + + dpif_linux_vport_init(&vport); + vport.cmd = OVS_VPORT_CMD_SET; + vport.name = netdev_get_name(netdev); + vport.stats = &rtnl_stats; + + err = dpif_linux_vport_transact(&vport, NULL, NULL); /* 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 @@ -409,7 +459,7 @@ netdev_vport_get_status(const struct netdev *netdev, struct shash *sh) shash_add(sh, "tunnel_egress_iface", xstrdup(iface)); - if (!netdev_open_default(iface, &egress_netdev)) { + if (!netdev_open(iface, "system", &egress_netdev)) { shash_add(sh, "tunnel_egress_iface_carrier", xstrdup(netdev_get_carrier(egress_netdev) ? "up" : "down")); @@ -433,204 +483,42 @@ netdev_vport_update_flags(struct netdev *netdev OVS_UNUSED, return 0; } -static char * -make_poll_name(const struct netdev *netdev) -{ - return xasprintf("%s:%s", netdev_get_type(netdev), netdev_get_name(netdev)); -} - -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); - } else { - list = shash_node->data; - } - - notifier = xmalloc(sizeof *notifier); - netdev_notifier_init(¬ifier->notifier, netdev, cb, aux); - list_push_back(list, ¬ifier->list_node); - notifier->shash_node = shash_node; - - *notifierp = ¬ifier->notifier; - free(poll_name); - - return 0; -} - -static void -netdev_vport_poll_remove(struct netdev_notifier *notifier_) +static unsigned int +netdev_vport_change_seq(const struct netdev *netdev) { - struct netdev_vport_notifier *notifier = - CONTAINER_OF(notifier_, struct netdev_vport_notifier, notifier); - - struct list *list; - - list = list_remove(¬ifier->list_node); - if (list_is_empty(list)) { - shash_delete(&netdev_vport_notifiers, notifier->shash_node); - free(list); - } - - free(notifier); + return netdev_dev_vport_cast(netdev_get_dev(netdev))->change_seq; } static void netdev_vport_run(void) { - rtnetlink_link_notifier_run(); route_table_run(); } static void netdev_vport_wait(void) { - rtnetlink_link_notifier_wait(); route_table_wait(); } /* get_tnl_iface() implementation. */ - -static struct name_node * -name_node_lookup(int ifi_index) -{ - struct name_node *nn; - - HMAP_FOR_EACH_WITH_HASH(nn, node, hash_int(ifi_index, 0), &name_map) { - if (nn->ifi_index == ifi_index) { - return nn; - } - } - - return NULL; -} - -/* Queries the kernel for fresh data to populate the name map with. */ -static int -netdev_vport_reset_names(void) -{ - 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; - - HMAP_FOR_EACH_SAFE(nn, nn_next, node, &name_map) { - hmap_remove(&name_map, &nn->node); - free(nn); - } - - error = nl_sock_create(NETLINK_ROUTE, &rtnl_sock); - if (error) { - VLOG_WARN_RL(&rl, "Failed to create NETLINK_ROUTE socket"); - return error; - } - - ofpbuf_init(&request, 0); - - nl_msg_put_nlmsghdr(&request, sizeof *rtmsg, RTM_GETLINK, NLM_F_REQUEST); - - rtmsg = ofpbuf_put_zeros(&request, sizeof *rtmsg); - rtmsg->rtgen_family = AF_INET; - - nl_dump_start(&dump, rtnl_sock, &request); - - while (nl_dump_next(&dump, &reply)) { - struct rtnetlink_link_change change; - - if (rtnetlink_link_parse(&reply, &change)) { - netdev_vport_link_change(&change, NULL); - } - } - nl_sock_destroy(rtnl_sock); - - return nl_dump_done(&dump); -} - -static void -netdev_vport_link_change(const struct rtnetlink_link_change *change, - void *aux OVS_UNUSED) -{ - - if (!change) { - netdev_vport_reset_names(); - } else if (change->nlmsg_type == RTM_NEWLINK) { - struct name_node *nn; - - if (name_node_lookup(change->ifi_index)) { - return; - } - - nn = xzalloc(sizeof *nn); - nn->ifi_index = change->ifi_index; - - strncpy(nn->ifname, change->ifname, IFNAMSIZ); - nn->ifname[IFNAMSIZ - 1] = '\0'; - - hmap_insert(&name_map, &nn->node, hash_int(nn->ifi_index, 0)); - } else if (change->nlmsg_type == RTM_DELLINK) { - struct name_node *nn; - - nn = name_node_lookup(change->ifi_index); - - if (nn) { - hmap_remove(&name_map, &nn->node); - free(nn); - } - - } else { - VLOG_WARN_RL(&rl, "Received unexpected rtnetlink message type %d", - change->nlmsg_type); - } -} - -static void -netdev_vport_tnl_iface_init(void) -{ - static bool tnl_iface_is_init = false; - - if (!tnl_iface_is_init) { - hmap_init(&name_map); - - rtnetlink_link_notifier_register(&netdev_vport_link_notifier, - netdev_vport_link_change, NULL); - - netdev_vport_reset_names(); - tnl_iface_is_init = true; - } -} - static const char * netdev_vport_get_tnl_iface(const struct netdev *netdev) { - int ifindex; - uint32_t route; + struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]; + ovs_be32 route; struct netdev_dev_vport *ndv; - struct tnl_port_config *config; + static char name[IFNAMSIZ]; ndv = netdev_dev_vport_cast(netdev_get_dev(netdev)); - config = (struct tnl_port_config *) ndv->config; - route = config->daddr; - - 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; - } - } + if (tnl_port_config_from_nlattr(ndv->options->data, ndv->options->size, + a)) { + return NULL; + } + route = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_DST_IPV4]); + + if (route_table_get_name(route, name)) { + return name; } return NULL; @@ -638,67 +526,61 @@ netdev_vport_get_tnl_iface(const struct netdev *netdev) /* Helper functions. */ -static int -netdev_vport_do_ioctl(int cmd, void *arg) +static void +netdev_vport_poll_notify(const struct netdev *netdev) { - static int ioctl_fd = -1; + struct netdev_dev_vport *ndv; - 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; - } - } + ndv = netdev_dev_vport_cast(netdev_get_dev(netdev)); - return ioctl(ioctl_fd, cmd, arg) ? errno : 0; + ndv->change_seq++; + if (!ndv->change_seq) { + ndv->change_seq++; + } } + +/* Code specific to individual vport types. */ static void -netdev_vport_poll_notify(const struct netdev *netdev) +set_key(const struct shash *args, const char *name, uint16_t type, + struct ofpbuf *options) { - char *poll_name = make_poll_name(netdev); - struct list *list = shash_find_data(&netdev_vport_notifiers, - poll_name); + const char *s; - if (list) { - struct netdev_vport_notifier *notifier; - - LIST_FOR_EACH (notifier, list_node, list) { - struct netdev_notifier *n = ¬ifier->notifier; - n->cb(n); + s = shash_find_data(args, name); + if (!s) { + s = shash_find_data(args, "key"); + if (!s) { + s = "0"; } } - free(poll_name); + if (!strcmp(s, "flow")) { + /* This is the default if no attribute is present. */ + } else { + nl_msg_put_be64(options, type, htonll(strtoull(s, NULL, 0))); + } } - -/* Code specific to individual vport types. */ static int parse_tunnel_config(const char *name, const char *type, - const struct shash *args, void *configp) + const struct shash *args, struct ofpbuf *options) { bool is_gre = false; bool is_ipsec = false; - struct tnl_port_config config; struct shash_node *node; bool ipsec_mech_set = false; + ovs_be32 daddr = htonl(0); + uint32_t flags; - memset(&config, 0, sizeof config); - config.flags |= TNL_F_PMTUD; - config.flags |= TNL_F_HDR_CACHE; - + flags = TNL_F_DF_DEFAULT | TNL_F_PMTUD | TNL_F_HDR_CACHE; if (!strcmp(type, "gre")) { is_gre = true; } else if (!strcmp(type, "ipsec_gre")) { is_gre = true; is_ipsec = true; - - config.flags |= TNL_F_IPSEC; - - /* IPsec doesn't work when header caching is enabled. */ - config.flags &= ~TNL_F_HDR_CACHE; + flags |= TNL_F_IPSEC; + flags &= ~TNL_F_HDR_CACHE; } SHASH_FOR_EACH (node, args) { @@ -707,58 +589,47 @@ parse_tunnel_config(const char *name, const char *type, if (lookup_ip(node->data, &in_addr)) { VLOG_WARN("%s: bad %s 'remote_ip'", name, type); } else { - config.daddr = in_addr.s_addr; + 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)); + nl_msg_put_be32(options, OVS_TUNNEL_ATTR_SRC_IPV4, + in_addr.s_addr); } } else if (!strcmp(node->name, "tos")) { if (!strcmp(node->data, "inherit")) { - config.flags |= TNL_F_TOS_INHERIT; + flags |= TNL_F_TOS_INHERIT; } else { - config.tos = atoi(node->data); + nl_msg_put_u8(options, OVS_TUNNEL_ATTR_TOS, atoi(node->data)); } } else if (!strcmp(node->name, "ttl")) { if (!strcmp(node->data, "inherit")) { - config.flags |= TNL_F_TTL_INHERIT; + flags |= TNL_F_TTL_INHERIT; } else { - config.ttl = atoi(node->data); + nl_msg_put_u8(options, OVS_TUNNEL_ATTR_TTL, atoi(node->data)); } } else if (!strcmp(node->name, "csum") && is_gre) { if (!strcmp(node->data, "true")) { - config.flags |= TNL_F_CSUM; + flags |= TNL_F_CSUM; + } + } else if (!strcmp(node->name, "df_inherit")) { + if (!strcmp(node->data, "true")) { + flags |= TNL_F_DF_INHERIT; + } + } else if (!strcmp(node->name, "df_default")) { + if (!strcmp(node->data, "false")) { + flags &= ~TNL_F_DF_DEFAULT; } } else if (!strcmp(node->name, "pmtud")) { if (!strcmp(node->data, "false")) { - config.flags &= ~TNL_F_PMTUD; + flags &= ~TNL_F_PMTUD; } } else if (!strcmp(node->name, "header_cache")) { if (!strcmp(node->data, "false")) { - config.flags &= ~TNL_F_HDR_CACHE; + flags &= ~TNL_F_HDR_CACHE; } } else if (!strcmp(node->name, "peer_cert") && is_ipsec) { if (shash_find(args, "certificate")) { @@ -774,8 +645,8 @@ parse_tunnel_config(const char *name, const char *type, */ 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); + VLOG_ERR("%s: 'peer_cert' requires 'certificate' argument", + name); return EINVAL; } ipsec_mech_set = true; @@ -787,84 +658,158 @@ parse_tunnel_config(const char *name, const char *type, || !strcmp(node->name, "private_key") || !strcmp(node->name, "use_ssl_cert"))) { /* Ignore options not used by the netdev. */ + } else if (!strcmp(node->name, "key") || + !strcmp(node->name, "in_key") || + !strcmp(node->name, "out_key")) { + /* Handled separately below. */ } else { - VLOG_WARN("%s: unknown %s argument '%s'", - name, type, node->name); + VLOG_WARN("%s: unknown %s argument '%s'", name, type, node->name); } } if (is_ipsec) { + char *file_name = xasprintf("%s/%s", ovs_rundir(), + "ovs-monitor-ipsec.pid"); + pid_t pid = read_pidfile(file_name); + free(file_name); + if (pid < 0) { + VLOG_ERR("%s: IPsec requires the ovs-monitor-ipsec daemon", + name); + return EINVAL; + } + if (shash_find(args, "peer_cert") && shash_find(args, "psk")) { - VLOG_WARN("%s: cannot define both 'peer_cert' and 'psk'", name); + VLOG_ERR("%s: cannot define both 'peer_cert' and 'psk'", name); return EINVAL; } if (!ipsec_mech_set) { - VLOG_WARN("%s: IPsec requires an 'peer_cert' or psk' argument", - name); + VLOG_ERR("%s: IPsec requires an 'peer_cert' or psk' argument", + name); return EINVAL; } } - if (!config.daddr) { - VLOG_WARN("%s: %s type requires valid 'remote_ip' argument", - name, 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) { + 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); + + nl_msg_put_u32(options, OVS_TUNNEL_ATTR_FLAGS, flags); + + 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 }, + [OVS_TUNNEL_ATTR_DST_IPV4] = { .type = NL_A_BE32 }, + [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 }, + }; + struct ofpbuf buf; - BUILD_ASSERT(sizeof config <= VPORT_CONFIG_SIZE); - memcpy(configp, &config, sizeof config); + ofpbuf_use_const(&buf, options, options_len); + if (!nl_policy_parse(&buf, 0, ovs_tunnel_policy, + a, ARRAY_SIZE(ovs_tunnel_policy))) { + return EINVAL; + } return 0; } +static uint64_t +get_be64_or_zero(const struct nlattr *a) +{ + return a ? ntohll(nl_attr_get_be64(a)) : 0; +} + static int unparse_tunnel_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, - const void *config_, struct shash *args) + const struct nlattr *options, size_t options_len, + struct shash *args) { - const struct tnl_port_config *config = config_; + struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]; + ovs_be32 daddr; + uint32_t flags; + int error; + + error = tnl_port_config_from_nlattr(options, options_len, a); + if (error) { + return error; + } - if (!(config->flags & TNL_F_HDR_CACHE) == !(config->flags & TNL_F_IPSEC)) { + flags = nl_attr_get_u32(a[OVS_TUNNEL_ATTR_FLAGS]); + if (!(flags & TNL_F_HDR_CACHE) == !(flags & TNL_F_IPSEC)) { smap_add(args, "header_cache", - config->flags & TNL_F_HDR_CACHE ? "true" : "false"); + flags & TNL_F_HDR_CACHE ? "true" : "false"); } - shash_add(args, "remote_ip", xasprintf(IP_FMT, IP_ARGS(&config->daddr))); - if (config->saddr) { - shash_add(args, "local_ip", - xasprintf(IP_FMT, IP_ARGS(&config->saddr))); + + daddr = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_DST_IPV4]); + shash_add(args, "remote_ip", xasprintf(IP_FMT, IP_ARGS(&daddr))); + + if (a[OVS_TUNNEL_ATTR_SRC_IPV4]) { + ovs_be32 saddr = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_SRC_IPV4]); + shash_add(args, "local_ip", xasprintf(IP_FMT, IP_ARGS(&saddr))); } - if ((config->flags & (TNL_F_IN_KEY_MATCH | TNL_F_OUT_KEY_ACTION)) - == (TNL_F_IN_KEY_MATCH | TNL_F_OUT_KEY_ACTION)) { + if (!a[OVS_TUNNEL_ATTR_IN_KEY] && !a[OVS_TUNNEL_ATTR_OUT_KEY]) { smap_add(args, "key", "flow"); - } else if (config->in_key && config->in_key == config->out_key) { - shash_add(args, "key", - xasprintf("%"PRIu64, ntohll(config->in_key))); } else { - if (config->flags & TNL_F_IN_KEY_MATCH) { - smap_add(args, "in_key", "flow"); - } else if (config->in_key) { - shash_add(args, "in_key", - xasprintf("%"PRIu64, ntohll(config->in_key))); - } + 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) { + shash_add(args, "key", xasprintf("%"PRIu64, in_key)); + } else { + if (!a[OVS_TUNNEL_ATTR_IN_KEY]) { + smap_add(args, "in_key", "flow"); + } else if (in_key) { + shash_add(args, "in_key", xasprintf("%"PRIu64, in_key)); + } - if (config->flags & TNL_F_OUT_KEY_ACTION) { - smap_add(args, "out_key", "flow"); - } else if (config->out_key) { - shash_add(args, "out_key", - xasprintf("%"PRIu64, ntohll(config->out_key))); + if (!a[OVS_TUNNEL_ATTR_OUT_KEY]) { + smap_add(args, "out_key", "flow"); + } else if (out_key) { + shash_add(args, "out_key", xasprintf("%"PRIu64, out_key)); + } } } - if (config->flags & TNL_F_TTL_INHERIT) { + if (flags & TNL_F_TTL_INHERIT) { smap_add(args, "tos", "inherit"); - } else if (config->ttl) { - shash_add(args, "tos", xasprintf("%"PRIu8, config->ttl)); + } else if (a[OVS_TUNNEL_ATTR_TTL]) { + int ttl = nl_attr_get_u8(a[OVS_TUNNEL_ATTR_TTL]); + shash_add(args, "tos", xasprintf("%d", ttl)); } - if (config->flags & TNL_F_CSUM) { + if (flags & TNL_F_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]); + shash_add(args, "tos", xasprintf("%d", tos)); + } + + if (flags & TNL_F_CSUM) { smap_add(args, "csum", "true"); } - if (!(config->flags & TNL_F_PMTUD)) { + if (flags & TNL_F_DF_INHERIT) { + smap_add(args, "df_inherit", "true"); + } + if (!(flags & TNL_F_DF_DEFAULT)) { + smap_add(args, "df_default", "false"); + } + if (!(flags & TNL_F_PMTUD)) { smap_add(args, "pmtud", "false"); } @@ -873,57 +818,68 @@ unparse_tunnel_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, static int parse_patch_config(const char *name, const char *type OVS_UNUSED, - const struct shash *args, void *configp) + const struct shash *args, struct ofpbuf *options) { const char *peer; peer = shash_find_data(args, "peer"); if (!peer) { - VLOG_WARN("%s: patch type requires valid 'peer' argument", name); + VLOG_ERR("%s: patch type requires valid 'peer' argument", name); return EINVAL; } if (shash_count(args) > 1) { - VLOG_WARN("%s: patch type takes only a 'peer' argument", name); + VLOG_ERR("%s: patch type takes only a 'peer' argument", name); return EINVAL; } - if (strlen(peer) >= MIN(IFNAMSIZ, VPORT_CONFIG_SIZE)) { - VLOG_WARN("%s: patch 'peer' arg too long", name); + if (strlen(peer) >= IFNAMSIZ) { + VLOG_ERR("%s: patch 'peer' arg too long", name); return EINVAL; } if (!strcmp(name, peer)) { - VLOG_WARN("%s: patch peer must not be self", name); + VLOG_ERR("%s: patch peer must not be self", name); return EINVAL; } - strncpy(configp, peer, VPORT_CONFIG_SIZE); + nl_msg_put_string(options, OVS_PATCH_ATTR_PEER, peer); return 0; } static int unparse_patch_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, - const void *config_, struct shash *args) + const struct nlattr *options, size_t options_len, + struct shash *args) { - char peer[IFNAMSIZ]; + static const struct nl_policy ovs_patch_policy[] = { + [OVS_PATCH_ATTR_PEER] = { .type = NL_A_STRING, + .max_len = IFNAMSIZ, + .optional = false } + }; + + struct nlattr *a[ARRAY_SIZE(ovs_patch_policy)]; + struct ofpbuf buf; - ovs_strlcpy(peer, config_, MIN(sizeof peer, VPORT_CONFIG_SIZE)); - if (peer[0]) { - smap_add(args, "peer", peer); + ofpbuf_use_const(&buf, options, options_len); + if (!nl_policy_parse(&buf, 0, ovs_patch_policy, + a, ARRAY_SIZE(ovs_patch_policy))) { + return EINVAL; } + smap_add(args, "peer", nl_attr_get_string(a[OVS_PATCH_ATTR_PEER])); return 0; } #define VPORT_FUNCTIONS(GET_STATUS) \ - netdev_vport_init, \ + NULL, \ netdev_vport_run, \ netdev_vport_wait, \ \ netdev_vport_create, \ netdev_vport_destroy, \ + netdev_vport_get_config, \ netdev_vport_set_config, \ \ netdev_vport_open, \ @@ -931,16 +887,18 @@ unparse_patch_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, \ NULL, /* enumerate */ \ \ + NULL, /* listen */ \ NULL, /* recv */ \ NULL, /* recv_wait */ \ NULL, /* drain */ \ \ - NULL, /* send */ \ + netdev_vport_send, /* 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_miimon */ \ @@ -973,26 +931,25 @@ unparse_patch_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, \ netdev_vport_update_flags, \ \ - netdev_vport_poll_add, \ - netdev_vport_poll_remove, + netdev_vport_change_seq void netdev_vport_register(void) { static const struct vport_class vport_classes[] = { - { ODP_VPORT_TYPE_GRE, + { OVS_VPORT_TYPE_GRE, { "gre", VPORT_FUNCTIONS(netdev_vport_get_status) }, parse_tunnel_config, unparse_tunnel_config }, - { ODP_VPORT_TYPE_GRE, + { OVS_VPORT_TYPE_GRE, { "ipsec_gre", VPORT_FUNCTIONS(netdev_vport_get_status) }, parse_tunnel_config, unparse_tunnel_config }, - { ODP_VPORT_TYPE_CAPWAP, + { OVS_VPORT_TYPE_CAPWAP, { "capwap", VPORT_FUNCTIONS(netdev_vport_get_status) }, parse_tunnel_config, unparse_tunnel_config }, - { ODP_VPORT_TYPE_PATCH, + { OVS_VPORT_TYPE_PATCH, { "patch", VPORT_FUNCTIONS(NULL) }, parse_patch_config, unparse_patch_config } };