X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=378a08cb5674173def36ad77f99dea14d7bb0632;hb=6161d3fd928edf7016abae60f549a135a2f83f09;hp=6caf086a83262cec3210bde6b5cbe6018efb971e;hpb=892815f564700501888139823ad282355a691e10;p=openvswitch diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index 6caf086a..378a08cb 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -15,60 +15,36 @@ #include #include "bridge.h" -#include "byte-order.h" #include #include -#include -#include #include -#include -#include -#include -#include #include -#include -#include -#include -#include -#include #include "bitmap.h" #include "bond.h" #include "cfm.h" -#include "classifier.h" #include "coverage.h" #include "daemon.h" #include "dirs.h" -#include "dpif.h" #include "dynamic-string.h" -#include "flow.h" #include "hash.h" #include "hmap.h" #include "jsonrpc.h" #include "lacp.h" #include "list.h" -#include "mac-learning.h" #include "netdev.h" -#include "netlink.h" -#include "odp-util.h" #include "ofp-print.h" #include "ofpbuf.h" -#include "ofproto/netflow.h" #include "ofproto/ofproto.h" -#include "ovsdb-data.h" -#include "packets.h" #include "poll-loop.h" -#include "process.h" #include "sha1.h" #include "shash.h" #include "socket-util.h" #include "stream-ssl.h" #include "sset.h" -#include "svec.h" #include "system-stats.h" #include "timeval.h" #include "util.h" #include "unixctl.h" -#include "vconn.h" #include "vswitchd/vswitch-idl.h" #include "xenserver.h" #include "vlog.h" @@ -158,8 +134,8 @@ static long long int stats_timer = LLONG_MIN; static long long int db_limiter = LLONG_MIN; static void add_del_bridges(const struct ovsrec_open_vswitch *); -static void bridge_del_dps(void); -static bool bridge_add_dp(struct bridge *); +static void bridge_del_ofprotos(void); +static bool bridge_add_ofprotos(struct bridge *); static void bridge_create(const struct ovsrec_bridge *); static void bridge_destroy(struct bridge *); static struct bridge *bridge_lookup(const char *name); @@ -172,8 +148,11 @@ static void bridge_add_ofproto_ports(struct bridge *); static void bridge_del_ofproto_ports(struct bridge *); static void bridge_refresh_ofp_port(struct bridge *); static void bridge_configure_datapath_id(struct bridge *); +static void bridge_configure_flow_eviction_threshold(struct bridge *); static void bridge_configure_netflow(struct bridge *); +static void bridge_configure_forward_bpdu(struct bridge *); static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number); +static void bridge_configure_stp(struct bridge *); static void bridge_configure_remotes(struct bridge *, const struct sockaddr_in *managers, size_t n_managers); @@ -183,12 +162,16 @@ static void bridge_pick_local_hw_addr(struct bridge *, static uint64_t bridge_pick_datapath_id(struct bridge *, const uint8_t bridge_ea[ETH_ADDR_LEN], struct iface *hw_addr_iface); +static const char *bridge_get_other_config(const struct ovsrec_bridge *, + const char *key); +static const char *get_port_other_config(const struct ovsrec_port *, + const char *key, + const char *default_value); static uint64_t dpid_from_hash(const void *, size_t nbytes); static bool bridge_has_bond_fake_iface(const struct bridge *, const char *name); static bool port_is_bond_fake_iface(const struct port *); -static unixctl_cb_func cfm_unixctl_show; static unixctl_cb_func qos_unixctl_show; static struct port *port_create(struct bridge *, const struct ovsrec_port *); @@ -199,7 +182,9 @@ static struct port *port_lookup(const struct bridge *, const char *name); static void port_configure(struct port *); static struct lacp_settings *port_configure_lacp(struct port *, struct lacp_settings *); -static void port_configure_bond(struct port *, struct bond_settings *); +static void port_configure_bond(struct port *, struct bond_settings *, + uint32_t *bond_stable_ids); +static bool port_is_synthetic(const struct port *); static void bridge_configure_mirrors(struct bridge *); static struct mirror *mirror_create(struct bridge *, @@ -217,10 +202,12 @@ static struct iface *iface_from_ofp_port(const struct bridge *, uint16_t ofp_port); static void iface_set_mac(struct iface *); static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport); +static void iface_clear_db_record(const struct ovsrec_interface *if_cfg); static void iface_configure_qos(struct iface *, const struct ovsrec_qos *); static void iface_configure_cfm(struct iface *); -static bool iface_refresh_cfm_stats(struct iface *iface); -static bool iface_get_carrier(const struct iface *); +static void iface_refresh_cfm_stats(struct iface *); +static void iface_refresh_stats(struct iface *); +static void iface_refresh_status(struct iface *); static bool iface_is_synthetic(const struct iface *); static void shash_from_ovs_idl_map(char **keys, char **values, size_t n, @@ -238,6 +225,7 @@ bridge_init(const char *remote) { /* Create connection to database. */ idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true); + ovsdb_idl_set_lock(idl, "ovs_vswitchd"); ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg); ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics); @@ -248,8 +236,11 @@ bridge_init(const char *remote) ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version); ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id); + ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status); ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids); + ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status); + ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics); ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids); ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge); @@ -257,10 +248,14 @@ bridge_init(const char *remote) ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state); + ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status); + ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault); + ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids); + ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current); ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids); ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected); @@ -268,10 +263,6 @@ bridge_init(const char *remote) ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status); ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids); - ovsdb_idl_omit_alert(idl, &ovsrec_maintenance_point_col_fault); - - ovsdb_idl_omit_alert(idl, &ovsrec_monitor_col_fault); - ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids); ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids); @@ -291,14 +282,14 @@ bridge_init(const char *remote) ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids); /* Register unixctl commands. */ - unixctl_command_register("cfm/show", cfm_unixctl_show, NULL); - unixctl_command_register("qos/show", qos_unixctl_show, NULL); - unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows, - NULL); - unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect, - NULL); + unixctl_command_register("qos/show", "interface", qos_unixctl_show, NULL); + unixctl_command_register("bridge/dump-flows", "bridge", + bridge_unixctl_dump_flows, NULL); + unixctl_command_register("bridge/reconnect", "[bridge]", + bridge_unixctl_reconnect, NULL); lacp_init(); bond_init(); + cfm_init(); } void @@ -395,7 +386,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) * that port already belongs to a different datapath, so we must do all * port deletions before any port additions. A datapath always has a * "local port" so we must delete not-configured datapaths too. */ - bridge_del_dps(); + bridge_del_ofprotos(); HMAP_FOR_EACH (br, node, &all_bridges) { if (br->ofproto) { bridge_del_ofproto_ports(br); @@ -409,7 +400,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) * has at least one iface, every "struct iface" has a valid ofp_port and * netdev. */ HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) { - if (!br->ofproto && !bridge_add_dp(br)) { + if (!br->ofproto && !bridge_add_ofprotos(br)) { bridge_destroy(br); } } @@ -429,7 +420,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) port_configure(port); - HMAP_FOR_EACH (iface, ofp_port_node, &br->ifaces) { + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { iface_configure_cfm(iface); iface_configure_qos(iface, port->cfg->qos); iface_set_mac(iface); @@ -437,9 +428,12 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) } bridge_configure_mirrors(br); bridge_configure_datapath_id(br); + bridge_configure_flow_eviction_threshold(br); + bridge_configure_forward_bpdu(br); bridge_configure_remotes(br, managers, n_managers); bridge_configure_netflow(br); bridge_configure_sflow(br, &sflow_bridge_number); + bridge_configure_stp(br); } free(managers); @@ -448,40 +442,35 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) daemonize_complete(); } -/* Iterate over all system dpifs and delete any of them that do not have a +/* Iterate over all ofprotos and delete any of them that do not have a * configured bridge or that are the wrong type. */ static void -bridge_del_dps(void) +bridge_del_ofprotos(void) { - struct sset dpif_names; - struct sset dpif_types; + struct sset names; + struct sset types; const char *type; - sset_init(&dpif_names); - sset_init(&dpif_types); - dp_enumerate_types(&dpif_types); - SSET_FOR_EACH (type, &dpif_types) { + sset_init(&names); + sset_init(&types); + ofproto_enumerate_types(&types); + SSET_FOR_EACH (type, &types) { const char *name; - dp_enumerate_names(type, &dpif_names); - SSET_FOR_EACH (name, &dpif_names) { + ofproto_enumerate_names(type, &names); + SSET_FOR_EACH (name, &names) { struct bridge *br = bridge_lookup(name); if (!br || strcmp(type, br->type)) { - struct dpif *dpif; - - if (!dpif_open(name, type, &dpif)) { - dpif_delete(dpif); - dpif_close(dpif); - } + ofproto_delete(name, type); } } } - sset_destroy(&dpif_names); - sset_destroy(&dpif_types); + sset_destroy(&names); + sset_destroy(&types); } static bool -bridge_add_dp(struct bridge *br) +bridge_add_ofprotos(struct bridge *br) { int error = ofproto_create(br->name, br->type, &br->ofproto); if (error) { @@ -516,7 +505,6 @@ port_configure(struct port *port) if (list_is_short(&port->ifaces)) { if (*cfg->tag >= 0 && *cfg->tag <= 4095) { s.vlan = *cfg->tag; - VLOG_DBG("port %s: assigning VLAN tag %d", port->name, s.vlan); } } else { /* It's possible that bonded, VLAN-tagged ports make sense. Maybe @@ -528,13 +516,39 @@ port_configure(struct port *port) /* Get VLAN trunks. */ s.trunks = NULL; - if (s.vlan < 0 && cfg->n_trunks) { + if (cfg->n_trunks) { s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks); - } else if (s.vlan >= 0 && cfg->n_trunks) { - VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan", - port->name); } + /* Get VLAN mode. */ + if (cfg->vlan_mode) { + if (!strcmp(cfg->vlan_mode, "access")) { + s.vlan_mode = PORT_VLAN_ACCESS; + } else if (!strcmp(cfg->vlan_mode, "trunk")) { + s.vlan_mode = PORT_VLAN_TRUNK; + } else if (!strcmp(cfg->vlan_mode, "native-tagged")) { + s.vlan_mode = PORT_VLAN_NATIVE_TAGGED; + } else if (!strcmp(cfg->vlan_mode, "native-untagged")) { + s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED; + } else { + /* This "can't happen" because ovsdb-server should prevent it. */ + VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode); + s.vlan_mode = PORT_VLAN_TRUNK; + } + } else { + if (s.vlan >= 0) { + s.vlan_mode = PORT_VLAN_ACCESS; + if (cfg->n_trunks) { + VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan", + port->name); + } + } else { + s.vlan_mode = PORT_VLAN_TRUNK; + } + } + s.use_priority_tags = !strcmp("true", get_port_other_config( + cfg, "priority-tags", "")); + /* Get LACP settings. */ s.lacp = port_configure_lacp(port, &lacp_settings); if (s.lacp) { @@ -550,18 +564,26 @@ port_configure(struct port *port) /* Get bond settings. */ if (s.n_slaves > 1) { - port_configure_bond(port, &bond_settings); s.bond = &bond_settings; + s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids); + port_configure_bond(port, &bond_settings, s.bond_stable_ids); } else { s.bond = NULL; + s.bond_stable_ids = NULL; + + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + netdev_set_miimon_interval(iface->netdev, 0); + } } /* Register. */ ofproto_bundle_register(port->bridge->ofproto, port, &s); /* Clean up. */ + free(s.slaves); free(s.trunks); free(s.lacp_slaves); + free(s.bond_stable_ids); } /* Pick local port hardware address and datapath ID for 'br'. */ @@ -707,6 +729,196 @@ bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number) sset_destroy(&oso.targets); } +static void +port_configure_stp(const struct ofproto *ofproto, struct port *port, + struct ofproto_port_stp_settings *port_s, + int *port_num_counter, unsigned long *port_num_bitmap) +{ + const char *config_str; + struct iface *iface; + + config_str = get_port_other_config(port->cfg, "stp-enable", NULL); + if (config_str && !strcmp(config_str, "false")) { + port_s->enable = false; + return; + } else { + port_s->enable = true; + } + + /* STP over bonds is not supported. */ + if (!list_is_singleton(&port->ifaces)) { + VLOG_ERR("port %s: cannot enable STP on bonds, disabling", + port->name); + port_s->enable = false; + return; + } + + iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem); + + /* Internal ports shouldn't participate in spanning tree, so + * skip them. */ + if (!strcmp(iface->type, "internal")) { + VLOG_DBG("port %s: disable STP on internal ports", port->name); + port_s->enable = false; + return; + } + + /* STP on mirror output ports is not supported. */ + if (ofproto_is_mirror_output_bundle(ofproto, port)) { + VLOG_DBG("port %s: disable STP on mirror ports", port->name); + port_s->enable = false; + return; + } + + config_str = get_port_other_config(port->cfg, "stp-port-num", NULL); + if (config_str) { + unsigned long int port_num = strtoul(config_str, NULL, 0); + int port_idx = port_num - 1; + + if (port_num < 1 || port_num > STP_MAX_PORTS) { + VLOG_ERR("port %s: invalid stp-port-num", port->name); + port_s->enable = false; + return; + } + + if (bitmap_is_set(port_num_bitmap, port_idx)) { + VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling", + port->name, port_num); + port_s->enable = false; + return; + } + bitmap_set1(port_num_bitmap, port_idx); + port_s->port_num = port_idx; + } else { + if (*port_num_counter > STP_MAX_PORTS) { + VLOG_ERR("port %s: too many STP ports, disabling", port->name); + port_s->enable = false; + return; + } + + port_s->port_num = (*port_num_counter)++; + } + + config_str = get_port_other_config(port->cfg, "stp-path-cost", NULL); + if (config_str) { + port_s->path_cost = strtoul(config_str, NULL, 10); + } else { + uint32_t current; + + if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) { + /* Couldn't get speed, so assume 100Mb/s. */ + port_s->path_cost = 19; + } else { + unsigned int mbps; + + mbps = netdev_features_to_bps(current) / 1000000; + port_s->path_cost = stp_convert_speed_to_cost(mbps); + } + } + + config_str = get_port_other_config(port->cfg, "stp-port-priority", NULL); + if (config_str) { + port_s->priority = strtoul(config_str, NULL, 0); + } else { + port_s->priority = STP_DEFAULT_PORT_PRIORITY; + } +} + +/* Set spanning tree configuration on 'br'. */ +static void +bridge_configure_stp(struct bridge *br) +{ + if (!br->cfg->stp_enable) { + ofproto_set_stp(br->ofproto, NULL); + } else { + struct ofproto_stp_settings br_s; + const char *config_str; + struct port *port; + int port_num_counter; + unsigned long *port_num_bitmap; + + config_str = bridge_get_other_config(br->cfg, "stp-system-id"); + if (config_str) { + uint8_t ea[ETH_ADDR_LEN]; + + if (eth_addr_from_string(config_str, ea)) { + br_s.system_id = eth_addr_to_uint64(ea); + } else { + br_s.system_id = eth_addr_to_uint64(br->ea); + VLOG_ERR("bridge %s: invalid stp-system-id, defaulting " + "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea)); + } + } else { + br_s.system_id = eth_addr_to_uint64(br->ea); + } + + config_str = bridge_get_other_config(br->cfg, "stp-priority"); + if (config_str) { + br_s.priority = strtoul(config_str, NULL, 0); + } else { + br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY; + } + + config_str = bridge_get_other_config(br->cfg, "stp-hello-time"); + if (config_str) { + br_s.hello_time = strtoul(config_str, NULL, 10) * 1000; + } else { + br_s.hello_time = STP_DEFAULT_HELLO_TIME; + } + + config_str = bridge_get_other_config(br->cfg, "stp-max-age"); + if (config_str) { + br_s.max_age = strtoul(config_str, NULL, 10) * 1000; + } else { + br_s.max_age = STP_DEFAULT_MAX_AGE; + } + + config_str = bridge_get_other_config(br->cfg, "stp-forward-delay"); + if (config_str) { + br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000; + } else { + br_s.fwd_delay = STP_DEFAULT_FWD_DELAY; + } + + /* Configure STP on the bridge. */ + if (ofproto_set_stp(br->ofproto, &br_s)) { + VLOG_ERR("bridge %s: could not enable STP", br->name); + return; + } + + /* Users must either set the port number with the "stp-port-num" + * configuration on all ports or none. If manual configuration + * is not done, then we allocate them sequentially. */ + port_num_counter = 0; + port_num_bitmap = bitmap_allocate(STP_MAX_PORTS); + HMAP_FOR_EACH (port, hmap_node, &br->ports) { + struct ofproto_port_stp_settings port_s; + struct iface *iface; + + port_configure_stp(br->ofproto, port, &port_s, + &port_num_counter, port_num_bitmap); + + /* As bonds are not supported, just apply configuration to + * all interfaces. */ + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + if (ofproto_port_set_stp(br->ofproto, iface->ofp_port, + &port_s)) { + VLOG_ERR("port %s: could not enable STP", port->name); + continue; + } + } + } + + if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS + && port_num_counter) { + VLOG_ERR("bridge %s: must manually configure all STP port " + "IDs or none, disabling", br->name); + ofproto_set_stp(br->ofproto, NULL); + } + bitmap_free(port_num_bitmap); + } +} + static bool bridge_has_bond_fake_iface(const struct bridge *br, const char *name) { @@ -730,9 +942,16 @@ add_del_bridges(const struct ovsrec_open_vswitch *cfg) /* Collect new bridges' names and types. */ shash_init(&new_br); for (i = 0; i < cfg->n_bridges; i++) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); const struct ovsrec_bridge *br_cfg = cfg->bridges[i]; - if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) { - VLOG_WARN("bridge %s specified twice", br_cfg->name); + + if (strchr(br_cfg->name, '/')) { + /* Prevent remote ovsdb-server users from accessing arbitrary + * directories, e.g. consider a bridge named "../../../etc/". */ + VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"", + br_cfg->name); + } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) { + VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name); } } @@ -740,8 +959,8 @@ add_del_bridges(const struct ovsrec_open_vswitch *cfg) * Update 'cfg' of bridges that still exist. */ HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) { br->cfg = shash_find_data(&new_br, br->name); - if (!br->cfg || strcmp(br->type, - dpif_normalize_type(br->cfg->datapath_type))) { + if (!br->cfg || strcmp(br->type, ofproto_normalize_type( + br->cfg->datapath_type))) { bridge_destroy(br); } } @@ -802,7 +1021,6 @@ bridge_del_ofproto_ports(struct bridge *br) br->name, name, strerror(error)); } if (iface) { - ofproto_port_unregister(br->ofproto, ofproto_port.ofp_port); netdev_close(iface->netdev); iface->netdev = NULL; } @@ -817,7 +1035,7 @@ iface_set_ofp_port(struct iface *iface, int ofp_port) assert(iface->ofp_port < 0 && ofp_port >= 0); iface->ofp_port = ofp_port; hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0)); - + iface_set_ofport(iface->cfg, ofp_port); } static void @@ -854,7 +1072,7 @@ bridge_refresh_ofp_port(struct bridge *br) } } -/* Add a dpif port for any "struct iface" that doesn't have one. +/* Add an ofproto port for any "struct iface" that doesn't have one. * Delete any "struct iface" for which this fails. * Delete any "struct port" that thereby ends up with no ifaces. */ static void @@ -867,29 +1085,36 @@ bridge_add_ofproto_ports(struct bridge *br) struct ofproto_port ofproto_port; LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) { - struct shash args; int error; - /* Open the netdev or reconfigure it. */ - shash_init(&args); - shash_from_ovs_idl_map(iface->cfg->key_options, - iface->cfg->value_options, - iface->cfg->n_options, &args); + /* Open the netdev. */ if (!iface->netdev) { - struct netdev_options options; - options.name = iface->name; - options.type = iface->type; - options.args = &args; - options.ethertype = NETDEV_ETH_TYPE_NONE; - error = netdev_open(&options, &iface->netdev); + error = netdev_open(iface->name, iface->type, &iface->netdev); + if (error) { + VLOG_WARN("could not open network device %s (%s)", + iface->name, strerror(error)); + } } else { - error = netdev_set_config(iface->netdev, &args); + error = 0; } - shash_destroy(&args); - if (error) { - VLOG_WARN("could not %s network device %s (%s)", - iface->netdev ? "reconfigure" : "open", - iface->name, strerror(error)); + + /* Configure the netdev. */ + if (iface->netdev) { + struct shash args; + + shash_init(&args); + shash_from_ovs_idl_map(iface->cfg->key_options, + iface->cfg->value_options, + iface->cfg->n_options, &args); + error = netdev_set_config(iface->netdev, &args); + shash_destroy(&args); + + if (error) { + VLOG_WARN("could not configure network device %s (%s)", + iface->name, strerror(error)); + netdev_close(iface->netdev); + iface->netdev = NULL; + } } /* Add the port, if necessary. */ @@ -907,7 +1132,13 @@ bridge_add_ofproto_ports(struct bridge *br) } } - /* Delete the iface if */ + /* Populate stats columns in new Interface rows. */ + if (iface->netdev && !iface->cfg->mtu) { + iface_refresh_stats(iface); + iface_refresh_status(iface); + } + + /* Delete the iface if we failed. */ if (iface->netdev && iface->ofp_port >= 0) { VLOG_DBG("bridge %s: interface %s is on port %d", br->name, iface->name, iface->ofp_port); @@ -919,7 +1150,7 @@ bridge_add_ofproto_ports(struct bridge *br) /* We already reported a related error, don't bother * duplicating it. */ } - iface_set_ofport(iface->cfg, -1); + iface_clear_db_record(iface->cfg); iface_destroy(iface); } } @@ -933,15 +1164,10 @@ bridge_add_ofproto_ports(struct bridge *br) if (port_is_bond_fake_iface(port)) { if (ofproto_port_query_by_name(br->ofproto, port->name, &ofproto_port)) { - struct netdev_options options; struct netdev *netdev; int error; - options.name = port->name; - options.type = "internal"; - options.args = NULL; - options.ethertype = NETDEV_ETH_TYPE_NONE; - error = netdev_open(&options, &netdev); + error = netdev_open(port->name, "internal", &netdev); if (!error) { ofproto_port_add(br->ofproto, netdev, NULL); netdev_close(netdev); @@ -953,7 +1179,6 @@ bridge_add_ofproto_ports(struct bridge *br) /* Already exists, nothing to do. */ ofproto_port_destroy(&ofproto_port); } - ofproto_port_destroy(&ofproto_port); } } } @@ -980,12 +1205,43 @@ bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key) &ovsrec_bridge_col_other_config, key); } +/* Set Flow eviction threshold */ +static void +bridge_configure_flow_eviction_threshold(struct bridge *br) +{ + const char *threshold_str; + unsigned threshold; + + threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold"); + if (threshold_str) { + threshold = strtoul(threshold_str, NULL, 10); + } else { + threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT; + } + ofproto_set_flow_eviction_threshold(br->ofproto, threshold); +} + +/* Set forward BPDU option. */ +static void +bridge_configure_forward_bpdu(struct bridge *br) +{ + const char *forward_bpdu_str; + bool forward_bpdu = false; + + forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu"); + if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) { + forward_bpdu = true; + } + ofproto_set_forward_bpdu(br->ofproto, forward_bpdu); +} + static void bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], struct iface **hw_addr_iface) { const char *hwaddr; struct port *port; + bool found_addr = false; int error; *hw_addr_iface = NULL; @@ -1005,7 +1261,6 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], /* Otherwise choose the minimum non-local MAC address among all of the * interfaces. */ - memset(ea, 0xff, ETH_ADDR_LEN); HMAP_FOR_EACH (port, hmap_node, &br->ports) { uint8_t iface_ea[ETH_ADDR_LEN]; struct iface *candidate; @@ -1050,9 +1305,6 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], /* Grab MAC. */ error = netdev_get_etheraddr(iface->netdev, iface_ea); if (error) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); - VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s", - iface->name, strerror(error)); continue; } } @@ -1062,20 +1314,21 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], !eth_addr_is_local(iface_ea) && !eth_addr_is_reserved(iface_ea) && !eth_addr_is_zero(iface_ea) && - eth_addr_compare_3way(iface_ea, ea) < 0) + (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0)) { memcpy(ea, iface_ea, ETH_ADDR_LEN); *hw_addr_iface = iface; + found_addr = true; } } - if (eth_addr_is_multicast(ea)) { + if (found_addr) { + VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT, + br->name, ETH_ADDR_ARGS(ea)); + } else { memcpy(ea, br->default_ea, ETH_ADDR_LEN); *hw_addr_iface = NULL; VLOG_WARN("bridge %s: using default bridge Ethernet " "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea)); - } else { - VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT, - br->name, ETH_ADDR_ARGS(ea)); } } @@ -1225,11 +1478,8 @@ iface_refresh_status(struct iface *iface) ovsrec_interface_set_link_speed(iface->cfg, NULL, 0); } - ovsrec_interface_set_link_state(iface->cfg, - iface_get_carrier(iface) ? "up" : "down"); - error = netdev_get_mtu(iface->netdev, &mtu); - if (!error && mtu != INT_MAX) { + if (!error) { mtu_64 = mtu; ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1); } @@ -1238,107 +1488,190 @@ iface_refresh_status(struct iface *iface) } } -/* Writes 'iface''s CFM statistics to the database. Returns true if anything - * changed, false otherwise. */ -static bool +/* Writes 'iface''s CFM statistics to the database. */ +static void iface_refresh_cfm_stats(struct iface *iface) { - const struct ovsrec_monitor *mon; - const struct cfm *cfm; - bool changed = false; - size_t i; + const struct ovsrec_interface *cfg = iface->cfg; + int fault, error; + const uint64_t *rmps; + size_t n_rmps; - mon = iface->cfg->monitor; - cfm = ofproto_port_get_cfm(iface->port->bridge->ofproto, iface->ofp_port); + if (iface_is_synthetic(iface)) { + return; + } - if (!cfm || !mon) { - return false; + fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto, + iface->ofp_port); + if (fault >= 0) { + bool fault_bool = fault; + ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1); + } else { + ovsrec_interface_set_cfm_fault(cfg, NULL, 0); + } + + error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto, + iface->ofp_port, &rmps, &n_rmps); + if (error >= 0) { + ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps, + n_rmps); + } else { + ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0); } +} - for (i = 0; i < mon->n_remote_mps; i++) { - const struct ovsrec_maintenance_point *mp; - const struct remote_mp *rmp; +static void +iface_refresh_stats(struct iface *iface) +{ +#define IFACE_STATS \ + IFACE_STAT(rx_packets, "rx_packets") \ + IFACE_STAT(tx_packets, "tx_packets") \ + IFACE_STAT(rx_bytes, "rx_bytes") \ + IFACE_STAT(tx_bytes, "tx_bytes") \ + IFACE_STAT(rx_dropped, "rx_dropped") \ + IFACE_STAT(tx_dropped, "tx_dropped") \ + IFACE_STAT(rx_errors, "rx_errors") \ + IFACE_STAT(tx_errors, "tx_errors") \ + IFACE_STAT(rx_frame_errors, "rx_frame_err") \ + IFACE_STAT(rx_over_errors, "rx_over_err") \ + IFACE_STAT(rx_crc_errors, "rx_crc_err") \ + IFACE_STAT(collisions, "collisions") + +#define IFACE_STAT(MEMBER, NAME) NAME, + static char *keys[] = { IFACE_STATS }; +#undef IFACE_STAT + int64_t values[ARRAY_SIZE(keys)]; + int i; - mp = mon->remote_mps[i]; - rmp = cfm_get_remote_mp(cfm, mp->mpid); + struct netdev_stats stats; - if (mp->n_fault != 1 || mp->fault[0] != rmp->fault) { - ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1); - changed = true; - } + if (iface_is_synthetic(iface)) { + return; } - if (mon->n_fault != 1 || mon->fault[0] != cfm->fault) { - ovsrec_monitor_set_fault(mon, &cfm->fault, 1); - changed = true; - } + /* Intentionally ignore return value, since errors will set 'stats' to + * all-1s, and we will deal with that correctly below. */ + netdev_get_stats(iface->netdev, &stats); + + /* Copy statistics into values[] array. */ + i = 0; +#define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER; + IFACE_STATS; +#undef IFACE_STAT + assert(i == ARRAY_SIZE(keys)); - return changed; + ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys)); +#undef IFACE_STATS } -static bool -iface_refresh_lacp_stats(struct iface *iface) +static void +br_refresh_stp_status(struct bridge *br) { - struct ofproto *ofproto = iface->port->bridge->ofproto; - int old = iface->cfg->lacp_current ? *iface->cfg->lacp_current : -1; - int new = ofproto_port_is_lacp_current(ofproto, iface->ofp_port); + struct ofproto *ofproto = br->ofproto; + struct ofproto_stp_status status; + char *keys[3], *values[3]; + size_t i; + + if (ofproto_get_stp_status(ofproto, &status)) { + return; + } - if (old != new) { - bool current = new; - ovsrec_interface_set_lacp_current(iface->cfg, ¤t, new >= 0); + if (!status.enabled) { + ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0); + return; + } + + keys[0] = "stp_bridge_id", + values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id)); + keys[1] = "stp_designated_root", + values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root)); + keys[2] = "stp_root_path_cost", + values[2] = xasprintf("%d", status.root_path_cost); + + ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values)); + + for (i = 0; i < ARRAY_SIZE(values); i++) { + free(values[i]); } - return old != new; } static void -iface_refresh_stats(struct iface *iface) +port_refresh_stp_status(struct port *port) { - struct iface_stat { - char *name; - int offset; - }; - static const struct iface_stat iface_stats[] = { - { "rx_packets", offsetof(struct netdev_stats, rx_packets) }, - { "tx_packets", offsetof(struct netdev_stats, tx_packets) }, - { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) }, - { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) }, - { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) }, - { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) }, - { "rx_errors", offsetof(struct netdev_stats, rx_errors) }, - { "tx_errors", offsetof(struct netdev_stats, tx_errors) }, - { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) }, - { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) }, - { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) }, - { "collisions", offsetof(struct netdev_stats, collisions) }, - }; - enum { N_STATS = ARRAY_SIZE(iface_stats) }; - const struct iface_stat *s; - - char *keys[N_STATS]; - int64_t values[N_STATS]; - int n; + struct ofproto *ofproto = port->bridge->ofproto; + struct iface *iface; + struct ofproto_port_stp_status status; + char *keys[4]; + char *str_values[4]; + int64_t int_values[3]; + size_t i; - struct netdev_stats stats; + if (port_is_synthetic(port)) { + return; + } - if (iface_is_synthetic(iface)) { + /* STP doesn't currently support bonds. */ + if (!list_is_singleton(&port->ifaces)) { + ovsrec_port_set_status(port->cfg, NULL, NULL, 0); return; } - /* Intentionally ignore return value, since errors will set 'stats' to - * all-1s, and we will deal with that correctly below. */ - netdev_get_stats(iface->netdev, &stats); + iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem); - n = 0; - for (s = iface_stats; s < &iface_stats[N_STATS]; s++) { - uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset); - if (value != UINT64_MAX) { - keys[n] = s->name; - values[n] = value; - n++; - } + if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) { + return; } - ovsrec_interface_set_statistics(iface->cfg, keys, values, n); + if (!status.enabled) { + ovsrec_port_set_status(port->cfg, NULL, NULL, 0); + ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0); + return; + } + + /* Set Status column. */ + keys[0] = "stp_port_id"; + str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id); + keys[1] = "stp_state"; + str_values[1] = xstrdup(stp_state_name(status.state)); + keys[2] = "stp_sec_in_state"; + str_values[2] = xasprintf("%u", status.sec_in_state); + keys[3] = "stp_role"; + str_values[3] = xstrdup(stp_role_name(status.role)); + + ovsrec_port_set_status(port->cfg, keys, str_values, + ARRAY_SIZE(str_values)); + + for (i = 0; i < ARRAY_SIZE(str_values); i++) { + free(str_values[i]); + } + + /* Set Statistics column. */ + keys[0] = "stp_tx_count"; + int_values[0] = status.tx_count; + keys[1] = "stp_rx_count"; + int_values[1] = status.rx_count; + keys[2] = "stp_error_count"; + int_values[2] = status.error_count; + + ovsrec_port_set_statistics(port->cfg, keys, int_values, + ARRAY_SIZE(int_values)); +} + +static bool +enable_system_stats(const struct ovsrec_open_vswitch *cfg) +{ + const char *enable; + + /* Use other-config:enable-system-stats by preference. */ + enable = get_ovsrec_key_value(&cfg->header_, + &ovsrec_open_vswitch_col_other_config, + "enable-statistics"); + if (enable) { + return !strcmp(enable, "true"); + } + + /* Disable by default. */ + return false; } static void @@ -1348,7 +1681,9 @@ refresh_system_stats(const struct ovsrec_open_vswitch *cfg) struct shash stats; shash_init(&stats); - get_system_stats(&stats); + if (enable_system_stats(cfg)) { + get_system_stats(&stats); + } ovsdb_datum_from_shash(&datum, &stats); ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics, @@ -1371,13 +1706,20 @@ nx_role_to_str(enum nx_role role) } static void -bridge_refresh_controller_status(const struct bridge *br) +refresh_controller_status(void) { + struct bridge *br; struct shash info; const struct ovsrec_controller *cfg; - ofproto_get_ofproto_controller_info(br->ofproto, &info); + shash_init(&info); + /* Accumulate status for controllers on all bridges. */ + HMAP_FOR_EACH (br, node, &all_bridges) { + ofproto_get_ofproto_controller_info(br->ofproto, &info); + } + + /* Update each controller in the database with current status. */ OVSREC_CONTROLLER_FOR_EACH(cfg, idl) { struct ofproto_controller_info *cinfo = shash_find_data(&info, cfg->target); @@ -1398,6 +1740,31 @@ bridge_refresh_controller_status(const struct bridge *br) ofproto_free_ofproto_controller_info(&info); } +static void +refresh_cfm_stats(void) +{ + static struct ovsdb_idl_txn *txn = NULL; + + if (!txn) { + struct bridge *br; + + txn = ovsdb_idl_txn_create(idl); + + HMAP_FOR_EACH (br, node, &all_bridges) { + struct iface *iface; + + HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) { + iface_refresh_cfm_stats(iface); + } + } + } + + if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) { + ovsdb_idl_txn_destroy(txn); + txn = NULL; + } +} + void bridge_run(void) { @@ -1407,6 +1774,24 @@ bridge_run(void) bool database_changed; struct bridge *br; + /* (Re)configure if necessary. */ + database_changed = ovsdb_idl_run(idl); + if (ovsdb_idl_is_lock_contended(idl)) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); + struct bridge *br, *next_br; + + VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, " + "disabling this process until it goes away"); + + HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) { + bridge_destroy(br); + } + return; + } else if (!ovsdb_idl_has_lock(idl)) { + return; + } + cfg = ovsrec_open_vswitch_first(idl); + /* Let each bridge do the work that it needs to do. */ datapath_destroyed = false; HMAP_FOR_EACH (br, node, &all_bridges) { @@ -1419,10 +1804,6 @@ bridge_run(void) } } - /* (Re)configure if necessary. */ - database_changed = ovsdb_idl_run(idl); - cfg = ovsrec_open_vswitch_first(idl); -#ifdef HAVE_OPENSSL /* Re-configure SSL. We do this on every trip through the main loop, * instead of just when the database changes, because the contents of the * key and certificate files can change without the database changing. @@ -1435,7 +1816,7 @@ bridge_run(void) stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate); stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert); } -#endif + if (database_changed || datapath_destroyed) { if (cfg) { struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl); @@ -1471,9 +1852,9 @@ bridge_run(void) iface_refresh_status(iface); } } - bridge_refresh_controller_status(br); } refresh_system_stats(cfg); + refresh_controller_status(); ovsdb_idl_txn_commit(txn); ovsdb_idl_txn_destroy(txn); /* XXX */ } @@ -1483,72 +1864,69 @@ bridge_run(void) if (time_msec() >= db_limiter) { struct ovsdb_idl_txn *txn; - bool changed = false; txn = ovsdb_idl_txn_create(idl); HMAP_FOR_EACH (br, node, &all_bridges) { + struct iface *iface; struct port *port; + br_refresh_stp_status(br); + HMAP_FOR_EACH (port, hmap_node, &br->ports) { - struct iface *iface; + port_refresh_stp_status(port); + } + + HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) { + const char *link_state; + int64_t link_resets; + int current; + + if (iface_is_synthetic(iface)) { + continue; + } - LIST_FOR_EACH (iface, port_elem, &port->ifaces) { - changed = iface_refresh_cfm_stats(iface) || changed; - changed = iface_refresh_lacp_stats(iface) || changed; + current = ofproto_port_is_lacp_current(br->ofproto, + iface->ofp_port); + if (current >= 0) { + bool bl = current; + ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1); + } else { + ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0); } + + link_state = netdev_get_carrier(iface->netdev) ? "up" : "down"; + ovsrec_interface_set_link_state(iface->cfg, link_state); + + link_resets = netdev_get_carrier_resets(iface->netdev); + ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1); } } - if (changed) { + if (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) { db_limiter = time_msec() + DB_LIMIT_INTERVAL; } - - ovsdb_idl_txn_commit(txn); ovsdb_idl_txn_destroy(txn); } + + refresh_cfm_stats(); } void bridge_wait(void) { - struct bridge *br; - - HMAP_FOR_EACH (br, node, &all_bridges) { - ofproto_wait(br->ofproto); - } ovsdb_idl_wait(idl); - poll_timer_wait_until(stats_timer); - - if (db_limiter > time_msec()) { - poll_timer_wait_until(db_limiter); - } -} - -/* CFM unixctl user interface functions. */ -static void -cfm_unixctl_show(struct unixctl_conn *conn, - const char *args, void *aux OVS_UNUSED) -{ - struct ds ds = DS_EMPTY_INITIALIZER; - struct iface *iface; - const struct cfm *cfm; - - iface = iface_find(args); - if (!iface) { - unixctl_command_reply(conn, 501, "no such interface"); - return; - } + if (!hmap_is_empty(&all_bridges)) { + struct bridge *br; - cfm = ofproto_port_get_cfm(iface->port->bridge->ofproto, iface->ofp_port); + HMAP_FOR_EACH (br, node, &all_bridges) { + ofproto_wait(br->ofproto); + } + poll_timer_wait_until(stats_timer); - if (!cfm) { - unixctl_command_reply(conn, 501, "CFM not enabled"); - return; + if (db_limiter > time_msec()) { + poll_timer_wait_until(db_limiter); + } } - - cfm_dump_ds(cfm, &ds); - unixctl_command_reply(conn, 200, ds_cstr(&ds)); - ds_destroy(&ds); } /* QoS unixctl user interface functions. */ @@ -1654,9 +2032,13 @@ bridge_create(const struct ovsrec_bridge *br_cfg) br = xzalloc(sizeof *br); br->name = xstrdup(br_cfg->name); - br->type = xstrdup(dpif_normalize_type(br_cfg->datapath_type)); + br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type)); br->cfg = br_cfg; - eth_addr_nicira_random(br->default_ea); + + /* Derive the default Ethernet address from the bridge's UUID. This should + * be unique and it will be stable between ovs-vswitchd runs. */ + memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN); + eth_addr_mark_random(br->default_ea); hmap_init(&br->ports); hmap_init(&br->ifaces); @@ -1911,6 +2293,26 @@ bridge_configure_local_iface_netdev(struct bridge *br, } } +/* Returns true if 'a' and 'b' are the same except that any number of slashes + * in either string are treated as equal to any number of slashes in the other, + * e.g. "x///y" is equal to "x/y". */ +static bool +equal_pathnames(const char *a, const char *b) +{ + while (*a == *b) { + if (*a == '/') { + a += strspn(a, "/"); + b += strspn(b, "/"); + } else if (*a == '\0') { + return true; + } else { + a++; + b++; + } + } + return false; +} + static void bridge_configure_remotes(struct bridge *br, const struct sockaddr_in *managers, size_t n_managers) @@ -1957,13 +2359,26 @@ bridge_configure_remotes(struct bridge *br, if (!strncmp(c->target, "punix:", 6) || !strncmp(c->target, "unix:", 5)) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + char *whitelist; + + whitelist = xasprintf("unix:%s/%s.controller", + ovs_rundir(), br->name); + if (!equal_pathnames(c->target, whitelist)) { + /* Prevent remote ovsdb-server users from accessing arbitrary + * Unix domain sockets and overwriting arbitrary local + * files. */ + VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket " + "controller \"%s\" due to possibility for remote " + "exploit. Instead, specify whitelisted \"%s\" or " + "connect to \"unix:%s/%s.mgmt\" (which is always " + "available without special configuration).", + br->name, c->target, whitelist, + ovs_rundir(), br->name); + free(whitelist); + continue; + } - /* Prevent remote ovsdb-server users from accessing arbitrary Unix - * domain sockets and overwriting arbitrary local files. */ - VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket " - "controller \"%s\" due to possibility for remote " - "exploit", br->name, c->target); - continue; + free(whitelist); } bridge_configure_local_iface_netdev(br, c); @@ -2051,7 +2466,10 @@ port_del_ifaces(struct port *port) sset_init(&new_ifaces); for (i = 0; i < port->cfg->n_interfaces; i++) { const char *name = port->cfg->interfaces[i]->name; - sset_add(&new_ifaces, name); + const char *type = port->cfg->interfaces[i]->name; + if (strcmp(type, "null")) { + sset_add(&new_ifaces, name); + } } /* Get rid of deleted interfaces. */ @@ -2077,10 +2495,11 @@ port_add_ifaces(struct port *port) shash_init(&new_ifaces); for (i = 0; i < port->cfg->n_interfaces; i++) { const struct ovsrec_interface *cfg = port->cfg->interfaces[i]; - if (!shash_add_once(&new_ifaces, cfg->name, cfg)) { + if (strcmp(cfg->type, "null") + && !shash_add_once(&new_ifaces, cfg->name, cfg)) { VLOG_WARN("port %s: %s specified twice as port interface", port->name, cfg->name); - iface_set_ofport(cfg, -1); + iface_clear_db_record(cfg); } } @@ -2189,9 +2608,10 @@ port_configure_lacp(struct port *port, struct lacp_settings *s) ? priority : UINT16_MAX - !list_is_short(&port->ifaces)); - s->strict = !strcmp(get_port_other_config(port->cfg, "lacp-strict", - "false"), - "true"); + s->heartbeat = !strcmp(get_port_other_config(port->cfg, + "lacp-heartbeat", + "false"), "true"); + lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow"); custom_time = atoi(lacp_time); @@ -2212,11 +2632,13 @@ port_configure_lacp(struct port *port, struct lacp_settings *s) static void iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s) { - int priority, portid; + int priority, portid, key; portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0")); priority = atoi(get_interface_other_config(iface->cfg, "lacp-port-priority", "0")); + key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key", + "0")); if (portid <= 0 || portid > UINT16_MAX) { portid = iface->ofp_port; @@ -2226,15 +2648,24 @@ iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s) priority = UINT16_MAX; } + if (key < 0 || key > UINT16_MAX) { + key = 0; + } + s->name = iface->name; s->id = portid; s->priority = priority; + s->key = key; } static void -port_configure_bond(struct port *port, struct bond_settings *s) +port_configure_bond(struct port *port, struct bond_settings *s, + uint32_t *bond_stable_ids) { const char *detect_s; + struct iface *iface; + int miimon_interval; + size_t i; s->name = port->name; s->balance = BM_SLB; @@ -2244,23 +2675,30 @@ port_configure_bond(struct port *port, struct bond_settings *s) port->name, port->cfg->bond_mode, bond_mode_to_string(s->balance)); } + if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) { + VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, " + "please use another bond type or disable flood_vlans", + port->name); + } - s->detect = BLSM_CARRIER; - detect_s = get_port_other_config(port->cfg, "bond-detect-mode", NULL); - if (detect_s && !bond_detect_mode_from_string(&s->detect, detect_s)) { - VLOG_WARN("port %s: unsupported bond-detect-mode %s, " - "defaulting to %s", - port->name, detect_s, bond_detect_mode_to_string(s->detect)); + miimon_interval = atoi(get_port_other_config(port->cfg, + "bond-miimon-interval", "0")); + if (miimon_interval <= 0) { + miimon_interval = 200; } - s->miimon_interval = atoi( - get_port_other_config(port->cfg, "bond-miimon-interval", "200")); - if (s->miimon_interval < 100) { - s->miimon_interval = 100; + detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier"); + if (!strcmp(detect_s, "carrier")) { + miimon_interval = 0; + } else if (strcmp(detect_s, "miimon")) { + VLOG_WARN("port %s: unsupported bond-detect-mode %s, " + "defaulting to carrier", port->name, detect_s); + miimon_interval = 0; } s->up_delay = MAX(0, port->cfg->bond_updelay); s->down_delay = MAX(0, port->cfg->bond_downdelay); + s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0")); s->rebalance_interval = atoi( get_port_other_config(port->cfg, "bond-rebalance-interval", "10000")); if (s->rebalance_interval < 1000) { @@ -2268,6 +2706,28 @@ port_configure_bond(struct port *port, struct bond_settings *s) } s->fake_iface = port->cfg->bond_fake_iface; + + i = 0; + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + long long stable_id; + + stable_id = atoll(get_interface_other_config(iface->cfg, + "bond-stable-id", "0")); + if (stable_id <= 0 || stable_id >= UINT32_MAX) { + stable_id = iface->ofp_port; + } + bond_stable_ids[i++] = stable_id; + + netdev_set_miimon_interval(iface->netdev, miimon_interval); + } +} + +/* Returns true if 'port' is synthetic, that is, if we constructed it locally + * instead of obtaining it from the database. */ +static bool +port_is_synthetic(const struct port *port) +{ + return ovsdb_idl_row_is_synthetic(&port->cfg->header_); } /* Interface functions. */ @@ -2400,6 +2860,29 @@ iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport) } } +/* Clears all of the fields in 'if_cfg' that indicate interface status, and + * sets the "ofport" field to -1. + * + * This is appropriate when 'if_cfg''s interface cannot be created or is + * otherwise invalid. */ +static void +iface_clear_db_record(const struct ovsrec_interface *if_cfg) +{ + if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) { + iface_set_ofport(if_cfg, -1); + ovsrec_interface_set_status(if_cfg, NULL, NULL, 0); + ovsrec_interface_set_admin_state(if_cfg, NULL); + ovsrec_interface_set_duplex(if_cfg, NULL); + ovsrec_interface_set_link_speed(if_cfg, NULL, 0); + ovsrec_interface_set_link_state(if_cfg, NULL); + ovsrec_interface_set_mtu(if_cfg, NULL, 0); + ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0); + ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0); + ovsrec_interface_set_lacp_current(if_cfg, NULL, 0); + ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0); + } +} + /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'. * * The value strings in '*shash' are taken directly from values[], not copied, @@ -2476,11 +2959,12 @@ iface_delete_queues(unsigned int queue_id, static void iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) { - if (!qos || qos->type[0] == '\0') { + if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) { netdev_set_qos(iface->netdev, NULL, NULL); } else { struct iface_delete_queues_cbdata cbdata; struct shash details; + bool queue_zero; size_t i; /* Configure top-level Qos for 'iface'. */ @@ -2496,16 +2980,28 @@ iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata); /* Configure queues for 'iface'. */ + queue_zero = false; for (i = 0; i < qos->n_queues; i++) { const struct ovsrec_queue *queue = qos->value_queues[i]; unsigned int queue_id = qos->key_queues[i]; + if (queue_id == 0) { + queue_zero = true; + } + shash_from_ovs_idl_map(queue->key_other_config, queue->value_other_config, queue->n_other_config, &details); netdev_set_queue(iface->netdev, queue_id, &details); shash_destroy(&details); } + if (!queue_zero) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); + VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default " + "queue (queue 0). Packets not directed to a " + "correctly configured queue may be dropped.", + iface->name); + } } netdev_set_policing(iface->netdev, @@ -2516,48 +3012,32 @@ iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) static void iface_configure_cfm(struct iface *iface) { - size_t i; - struct cfm cfm; - uint16_t *remote_mps; - struct ovsrec_monitor *mon; - uint8_t maid[CCM_MAID_LEN]; + const struct ovsrec_interface *cfg = iface->cfg; + const char *extended_str, *opstate_str; + struct cfm_settings s; - mon = iface->cfg->monitor; - - if (!mon) { + if (!cfg->n_cfm_mpid) { ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port); return; } - if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) { - VLOG_WARN("interface %s: Failed to generate MAID.", iface->name); - return; + s.mpid = *cfg->cfm_mpid; + s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval", + "0")); + s.ccm_vlan = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_vlan", + "0")); + if (s.interval <= 0) { + s.interval = 1000; } - cfm.mpid = mon->mpid; - cfm.interval = mon->interval ? *mon->interval : 1000; + extended_str = get_interface_other_config(iface->cfg, "cfm_extended", + "false"); + s.extended = !strcasecmp("true", extended_str); - memcpy(cfm.maid, maid, sizeof cfm.maid); + opstate_str = get_interface_other_config(iface->cfg, "cfm_opstate", "up"); + s.opup = !strcasecmp("up", opstate_str); - remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps); - for(i = 0; i < mon->n_remote_mps; i++) { - remote_mps[i] = mon->remote_mps[i]->mpid; - } - - ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, - &cfm, remote_mps, mon->n_remote_mps); - free(remote_mps); -} - -/* Read carrier or miimon status directly from 'iface''s netdev, according to - * how 'iface''s port is configured. - * - * Returns true if 'iface' is up, false otherwise. */ -static bool -iface_get_carrier(const struct iface *iface) -{ - /* XXX */ - return netdev_get_carrier(iface->netdev); + ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s); } /* Returns true if 'iface' is synthetic, that is, if we constructed it locally @@ -2678,8 +3158,6 @@ static bool mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg) { struct ofproto_mirror_settings s; - struct port *out_port; - struct port *port; /* Set name. */ if (strcmp(cfg->name, m->name)) { @@ -2691,7 +3169,7 @@ mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg) /* Get output port or VLAN. */ if (cfg->output_port) { s.out_bundle = port_lookup(m->bridge, cfg->output_port->name); - if (!out_port) { + if (!s.out_bundle) { VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge", m->bridge->name, m->name); return false; @@ -2717,6 +3195,7 @@ mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg) if (cfg->select_all) { size_t n_ports = hmap_count(&m->bridge->ports); void **ports = xmalloc(n_ports * sizeof *ports); + struct port *port; size_t i; i = 0; @@ -2736,7 +3215,6 @@ mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg) &s.srcs, &s.n_srcs); mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port, &s.dsts, &s.n_dsts); - } /* Get VLAN selection. */