X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=2dc5529666437c7ce4718ce04473f22a3badca51;hb=70fa1ecbf4f013566734fc506f50d90ee570a3e0;hp=fed22d3afee9a08260016d8c3a032a1e5894101e;hpb=8d6db33e350d38f0a4b692117c986fb4158fb05f;p=sliver-openvswitch.git diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index fed22d3af..2dc552966 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -18,6 +18,8 @@ #include #include #include +#include "async-append.h" +#include "bfd.h" #include "bitmap.h" #include "bond.h" #include "cfm.h" @@ -66,13 +68,13 @@ struct if_cfg { struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */ const struct ovsrec_interface *cfg; /* Interface record. */ const struct ovsrec_port *parent; /* Parent port record. */ - int64_t ofport; /* Requested OpenFlow port number. */ + ofp_port_t ofport; /* Requested OpenFlow port number. */ }; /* OpenFlow port slated for removal from ofproto. */ struct ofpp_garbage { struct list list_node; /* Node in bridge's ofpp_garbage. */ - uint16_t ofp_port; /* Port to be deleted. */ + ofp_port_t ofp_port; /* Port to be deleted. */ }; struct iface { @@ -85,7 +87,8 @@ struct iface { /* These members are valid only after bridge_reconfigure() causes them to * be initialized. */ struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */ - int ofp_port; /* OpenFlow port number, -1 if unknown. */ + ofp_port_t ofp_port; /* OpenFlow port number, */ + /* OFPP_NONE if unknown. */ struct netdev *netdev; /* Network device. */ const char *type; /* Usually same as cfg->type. */ const struct ovsrec_interface *cfg; @@ -146,6 +149,23 @@ static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges); /* OVSDB IDL used to obtain configuration. */ static struct ovsdb_idl *idl; +/* We want to complete daemonization, fully detaching from our parent process, + * only after we have completed our initial configuration, committed our state + * to the database, and received confirmation back from the database server + * that it applied the commit. This allows our parent process to know that, + * post-detach, ephemeral fields such as datapath-id and ofport are very likely + * to have already been filled in. (It is only "very likely" rather than + * certain because there is always a slim possibility that the transaction will + * fail or that some other client has added new bridges, ports, etc. while + * ovs-vswitchd was configuring using an old configuration.) + * + * We only need to do this once for our initial configuration at startup, so + * 'initial_config_done' tracks whether we've already done it. While we are + * waiting for a response to our commit, 'daemonize_txn' tracks the transaction + * itself and is otherwise NULL. */ +static bool initial_config_done; +static struct ovsdb_idl_txn *daemonize_txn; + /* Most recently processed IDL sequence number. */ static unsigned int idl_seqno; @@ -177,12 +197,13 @@ static size_t bridge_get_controllers(const struct bridge *br, static void bridge_add_del_ports(struct bridge *, const unsigned long int *splinter_vlans); static void bridge_refresh_ofp_port(struct bridge *); +static void bridge_configure_flow_miss_model(const char *opt); 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_mac_table(struct bridge *); static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number); +static void bridge_configure_ipfix(struct bridge *); static void bridge_configure_stp(struct bridge *); static void bridge_configure_tables(struct bridge *); static void bridge_configure_dp_desc(struct bridge *); @@ -226,7 +247,8 @@ static bool mirror_configure(struct mirror *); static void mirror_refresh_stats(struct mirror *); static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *); -static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port); +static bool iface_create(struct bridge *, struct if_cfg *, + ofp_port_t ofp_port); static bool iface_is_internal(const struct ovsrec_interface *iface, const struct ovsrec_bridge *br); static const char *iface_get_type(const struct ovsrec_interface *, @@ -236,9 +258,9 @@ static struct iface *iface_lookup(const struct bridge *, const char *name); static struct iface *iface_find(const char *name); static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name); static struct iface *iface_from_ofp_port(const struct bridge *, - uint16_t ofp_port); + ofp_port_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_set_ofport(const struct ovsrec_interface *, ofp_port_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 *); @@ -246,7 +268,7 @@ 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 int64_t iface_pick_ofport(const struct ovsrec_interface *); +static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *); /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.) * @@ -318,7 +340,7 @@ void bridge_init(const char *remote) { /* Create connection to database. */ - idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true); + idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true); idl_seqno = ovsdb_idl_get_seqno(idl); ovsdb_idl_set_lock(idl, "ovs_vswitchd"); ovsdb_idl_verify_write_only(idl); @@ -345,6 +367,8 @@ bridge_init(const char *remote) 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_mac_in_use); + ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex); 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); @@ -354,6 +378,7 @@ bridge_init(const char *remote) ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate); + ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current); ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids); @@ -370,8 +395,9 @@ bridge_init(const char *remote) ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics); ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids); - ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids); + ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids); + ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids); ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids); ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe); @@ -472,6 +498,16 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) ovs_assert(!reconfiguring); reconfiguring = true; + ofproto_set_flow_eviction_threshold( + smap_get_int(&ovs_cfg->other_config, "flow-eviction-threshold", + OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT)); + + ofproto_set_n_handler_threads( + smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0)); + + bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config, + "force-miss-model")); + /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal * configuration otherwise. @@ -512,7 +548,6 @@ bridge_reconfigure_ofp(void) long long int deadline; struct bridge *br; - time_refresh(); deadline = time_msec() + OFP_PORT_ACTION_WINDOW; /* The kernel will reject any attempt to add a given port to a datapath if @@ -531,7 +566,6 @@ bridge_reconfigure_ofp(void) list_remove(&garbage->list_node); free(garbage); - time_refresh(); if (time_msec() >= deadline) { return false; } @@ -543,8 +577,7 @@ bridge_reconfigure_ofp(void) struct if_cfg *if_cfg, *next; HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) { - iface_create(br, if_cfg, -1); - time_refresh(); + iface_create(br, if_cfg, OFPP_NONE); if (time_msec() >= deadline) { return false; } @@ -585,29 +618,29 @@ bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg) iface_configure_cfm(iface); iface_configure_qos(iface, port->cfg->qos); iface_set_mac(iface); + ofproto_port_set_bfd(br->ofproto, iface->ofp_port, + &iface->cfg->bfd); } } bridge_configure_mirrors(br); - bridge_configure_flow_eviction_threshold(br); bridge_configure_forward_bpdu(br); bridge_configure_mac_table(br); bridge_configure_remotes(br, managers, n_managers); bridge_configure_netflow(br); bridge_configure_sflow(br, &sflow_bridge_number); + bridge_configure_ipfix(br); bridge_configure_stp(br); bridge_configure_tables(br); bridge_configure_dp_desc(br); - } - free(managers); - - if (done) { - /* ovs-vswitchd has completed initialization, so allow the process that - * forked us to exit successfully. */ - daemonize_complete(); - reconfiguring = false; - VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION); + if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) { + /* XXX: Remove this warning message eventually. */ + VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been" + " moved to the Open_vSwitch table. Ignoring its" + " setting in the bridge table."); + } } + free(managers); return done; } @@ -664,7 +697,7 @@ bridge_update_ofprotos(void) error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port); if (error) { VLOG_ERR("failed to delete port %s: %s", ofproto_port.name, - strerror(error)); + ovs_strerror(error)); } ofproto_port_destroy(&ofproto_port); } @@ -673,7 +706,7 @@ bridge_update_ofprotos(void) error = ofproto_create(br->name, br->type, &br->ofproto); if (error) { VLOG_ERR("failed to create bridge %s: %s", br->name, - strerror(error)); + ovs_strerror(error)); bridge_destroy(br); } } @@ -777,6 +810,22 @@ port_configure(struct port *port) free(s.lacp_slaves); } +static void +bridge_configure_flow_miss_model(const char *opt) +{ + enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO; + + if (opt) { + if (strcmp(opt, "with-facets")) { + model = OFPROTO_HANDLE_MISS_WITH_FACETS; + } else if (strcmp(opt, "without-facets")) { + model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS; + } + } + + ofproto_set_flow_miss_model(model); +} + /* Pick local port hardware address and datapath ID for 'br'. */ static void bridge_configure_datapath_id(struct bridge *br) @@ -795,7 +844,7 @@ bridge_configure_datapath_id(struct bridge *br) static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge " "Ethernet address: %s", - br->name, strerror(error)); + br->name, ovs_strerror(error)); } } memcpy(br->ea, ea, ETH_ADDR_LEN); @@ -935,6 +984,89 @@ bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number) sset_destroy(&oso.targets); } +/* Set IPFIX configuration on 'br'. */ +static void +bridge_configure_ipfix(struct bridge *br) +{ + const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix; + const struct ovsrec_flow_sample_collector_set *fe_cfg; + struct ofproto_ipfix_bridge_exporter_options be_opts; + struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL; + size_t n_fe_opts = 0; + + OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) { + if (fe_cfg->bridge == br->cfg) { + n_fe_opts++; + } + } + + if (!be_cfg && n_fe_opts == 0) { + ofproto_set_ipfix(br->ofproto, NULL, NULL, 0); + return; + } + + if (be_cfg) { + memset(&be_opts, 0, sizeof be_opts); + + sset_init(&be_opts.targets); + sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets); + + if (be_cfg->sampling) { + be_opts.sampling_rate = *be_cfg->sampling; + } else { + be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE; + } + if (be_cfg->obs_domain_id) { + be_opts.obs_domain_id = *be_cfg->obs_domain_id; + } + if (be_cfg->obs_point_id) { + be_opts.obs_point_id = *be_cfg->obs_point_id; + } + if (be_cfg->cache_active_timeout) { + be_opts.cache_active_timeout = *be_cfg->cache_active_timeout; + } + if (be_cfg->cache_max_flows) { + be_opts.cache_max_flows = *be_cfg->cache_max_flows; + } + } + + if (n_fe_opts > 0) { + struct ofproto_ipfix_flow_exporter_options *opts; + fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts); + opts = fe_opts; + OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) { + if (fe_cfg->bridge == br->cfg) { + opts->collector_set_id = fe_cfg->id; + sset_init(&opts->targets); + sset_add_array(&opts->targets, fe_cfg->ipfix->targets, + fe_cfg->ipfix->n_targets); + opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout + ? *fe_cfg->ipfix->cache_active_timeout : 0; + opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows + ? *fe_cfg->ipfix->cache_max_flows : 0; + opts++; + } + } + } + + ofproto_set_ipfix(br->ofproto, be_cfg ? &be_opts : NULL, fe_opts, + n_fe_opts); + + if (be_cfg) { + sset_destroy(&be_opts.targets); + } + + if (n_fe_opts > 0) { + struct ofproto_ipfix_flow_exporter_options *opts = fe_opts; + size_t i; + for (i = 0; i < n_fe_opts; i++) { + sset_destroy(&opts->targets); + opts++; + } + free(fe_opts); + } +} + static void port_configure_stp(const struct ofproto *ofproto, struct port *port, struct ofproto_port_stp_settings *port_s, @@ -1178,13 +1310,14 @@ add_del_bridges(const struct ovsrec_open_vswitch *cfg) } static void -iface_set_ofp_port(struct iface *iface, int ofp_port) +iface_set_ofp_port(struct iface *iface, ofp_port_t ofp_port) { struct bridge *br = iface->port->bridge; - ovs_assert(iface->ofp_port < 0 && ofp_port >= 0); + ovs_assert(iface->ofp_port == OFPP_NONE && ofp_port != OFPP_NONE); iface->ofp_port = ofp_port; - hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0)); + hmap_insert(&br->ifaces, &iface->ofp_port_node, + hash_ofp_port(ofp_port)); iface_set_ofport(iface->cfg, ofp_port); } @@ -1199,7 +1332,7 @@ iface_set_netdev_config(const struct ovsrec_interface *iface_cfg, error = netdev_set_config(netdev, &iface_cfg->options); if (error) { VLOG_WARN("could not configure network device %s (%s)", - iface_cfg->name, strerror(error)); + iface_cfg->name, ovs_strerror(error)); } return error; } @@ -1219,12 +1352,12 @@ bridge_refresh_one_ofp_port(struct bridge *br, { const char *name = ofproto_port->name; const char *type = ofproto_port->type; - uint16_t ofp_port = ofproto_port->ofp_port; + ofp_port_t ofp_port = ofproto_port->ofp_port; struct iface *iface = iface_lookup(br, name); if (iface) { /* Check that the name-to-number mapping is one-to-one. */ - if (iface->ofp_port >= 0) { + if (iface->ofp_port != OFPP_NONE) { VLOG_WARN("bridge %s: interface %s reported twice", br->name, name); return false; @@ -1278,7 +1411,7 @@ bridge_refresh_ofp_port(struct bridge *br) struct iface *iface; LIST_FOR_EACH (iface, port_elem, &port->ifaces) { - iface->ofp_port = -1; + iface->ofp_port = OFPP_NONE; } } @@ -1300,7 +1433,7 @@ bridge_refresh_ofp_port(struct bridge *br) struct iface *iface, *iface_next; LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) { - if (iface->ofp_port < 0) { + if (iface->ofp_port == OFPP_NONE) { bridge_queue_if_cfg(br, iface->cfg, port->cfg); iface_destroy(iface); } @@ -1313,7 +1446,7 @@ bridge_refresh_ofp_port(struct bridge *br) } /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp' - * is negative, adds the network device to br->ofproto and stores the OpenFlow + * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp' * untouched. * @@ -1322,18 +1455,25 @@ bridge_refresh_ofp_port(struct bridge *br) static int iface_do_create(const struct bridge *br, const struct if_cfg *if_cfg, - int *ofp_portp, struct netdev **netdevp) + ofp_port_t *ofp_portp, struct netdev **netdevp) { const struct ovsrec_interface *iface_cfg = if_cfg->cfg; const struct ovsrec_port *port_cfg = if_cfg->parent; - struct netdev *netdev; + struct netdev *netdev = NULL; int error; + if (netdev_is_reserved_name(iface_cfg->name)) { + VLOG_WARN("could not create interface %s, name is reserved", + iface_cfg->name); + error = EINVAL; + goto error; + } + error = netdev_open(iface_cfg->name, iface_get_type(iface_cfg, br->cfg), &netdev); if (error) { VLOG_WARN("could not open network device %s (%s)", - iface_cfg->name, strerror(error)); + iface_cfg->name, ovs_strerror(error)); goto error; } @@ -1342,8 +1482,8 @@ iface_do_create(const struct bridge *br, goto error; } - if (*ofp_portp < 0) { - uint16_t ofp_port = if_cfg->ofport; + if (*ofp_portp == OFPP_NONE) { + ofp_port_t ofp_port = if_cfg->ofport; error = ofproto_port_add(br->ofproto, netdev, &ofp_port); if (error) { @@ -1360,7 +1500,7 @@ iface_do_create(const struct bridge *br, if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) || iface_is_internal(iface_cfg, br->cfg)) { - netdev_turn_flags_on(netdev, NETDEV_UP, true); + netdev_turn_flags_on(netdev, NETDEV_UP, NULL); } *netdevp = netdev; @@ -1373,13 +1513,13 @@ error: } /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow - * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is + * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is * automatically allocated for the iface. Takes ownership of and * deallocates 'if_cfg'. * * Return true if an iface is successfully created, false otherwise. */ static bool -iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port) +iface_create(struct bridge *br, struct if_cfg *if_cfg, ofp_port_t ofp_port) { const struct ovsrec_interface *iface_cfg = if_cfg->cfg; const struct ovsrec_port *port_cfg = if_cfg->parent; @@ -1400,7 +1540,7 @@ iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port) error = iface_do_create(br, if_cfg, &ofp_port, &netdev); bridge_run_fast(); if (error) { - iface_set_ofport(iface_cfg, -1); + iface_set_ofport(iface_cfg, OFPP_NONE); iface_clear_db_record(iface_cfg); ok = false; goto done; @@ -1419,7 +1559,7 @@ iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port) hash_string(iface_cfg->name, 0)); iface->port = port; iface->name = xstrdup(iface_cfg->name); - iface->ofp_port = -1; + iface->ofp_port = OFPP_NONE; iface->netdev = netdev; iface->type = iface_get_type(iface_cfg, br->cfg); iface->cfg = iface_cfg; @@ -1441,13 +1581,13 @@ iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port) error = netdev_open(port->name, "internal", &netdev); if (!error) { - uint16_t fake_ofp_port = if_cfg->ofport; + ofp_port_t fake_ofp_port = if_cfg->ofport; ofproto_port_add(br->ofproto, netdev, &fake_ofp_port); netdev_close(netdev); } else { VLOG_WARN("could not open network device %s (%s)", - port->name, strerror(error)); + port->name, ovs_strerror(error)); } } else { /* Already exists, nothing to do. */ @@ -1462,23 +1602,6 @@ done: return ok; } -/* Set Flow eviction threshold */ -static void -bridge_configure_flow_eviction_threshold(struct bridge *br) -{ - const char *threshold_str; - unsigned threshold; - - threshold_str = smap_get(&br->cfg->other_config, - "flow-eviction-threshold"); - if (threshold_str) { - threshold = strtoul(threshold_str, NULL, 10); - } else { - threshold = OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT; - } - ofproto_set_flow_eviction_threshold(br->ofproto, threshold); -} - /* Set forward BPDU option. */ static void bridge_configure_forward_bpdu(struct bridge *br) @@ -1698,6 +1821,8 @@ iface_refresh_status(struct iface *iface) int64_t bps; int mtu; int64_t mtu_64; + uint8_t mac[ETH_ADDR_LEN]; + int64_t ifindex64; int error; if (iface_is_synthetic(iface)) { @@ -1733,6 +1858,24 @@ iface_refresh_status(struct iface *iface) } else { ovsrec_interface_set_mtu(iface->cfg, NULL, 0); } + + error = netdev_get_etheraddr(iface->netdev, mac); + if (!error) { + char mac_string[32]; + + sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac)); + ovsrec_interface_set_mac_in_use(iface->cfg, mac_string); + } else { + ovsrec_interface_set_mac_in_use(iface->cfg, NULL); + } + + /* The netdev may return a negative number (such as -EOPNOTSUPP) + * if there is no valid ifindex number. */ + ifindex64 = netdev_get_ifindex(iface->netdev); + if (ifindex64 < 0) { + ifindex64 = 0; + } + ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1); } /* Writes 'iface''s CFM statistics to the database. 'iface' must not be @@ -1782,6 +1925,8 @@ iface_refresh_cfm_stats(struct iface *iface) } else { ovsrec_interface_set_cfm_health(cfg, NULL, 0); } + + free(status.rmps); } } @@ -1802,11 +1947,12 @@ iface_refresh_stats(struct iface *iface) IFACE_STAT(rx_crc_errors, "rx_crc_err") \ IFACE_STAT(collisions, "collisions") -#define IFACE_STAT(MEMBER, NAME) NAME, - static char *keys[] = { IFACE_STATS }; +#define IFACE_STAT(MEMBER, NAME) + 1 + enum { N_IFACE_STATS = IFACE_STATS }; #undef IFACE_STAT - int64_t values[ARRAY_SIZE(keys)]; - int i; + int64_t values[N_IFACE_STATS]; + char *keys[N_IFACE_STATS]; + int n; struct netdev_stats stats; @@ -1818,15 +1964,19 @@ iface_refresh_stats(struct iface *iface) * 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; + /* Copy statistics into keys[] and values[]. */ + n = 0; +#define IFACE_STAT(MEMBER, NAME) \ + if (stats.MEMBER != UINT64_MAX) { \ + keys[n] = NAME; \ + values[n] = stats.MEMBER; \ + n++; \ + } IFACE_STATS; #undef IFACE_STAT - ovs_assert(i == ARRAY_SIZE(keys)); + ovs_assert(n <= N_IFACE_STATS); - ovsrec_interface_set_statistics(iface->cfg, keys, values, - ARRAY_SIZE(keys)); + ovsrec_interface_set_statistics(iface->cfg, keys, values, n); #undef IFACE_STATS } @@ -2007,17 +2157,61 @@ refresh_controller_status(void) ofproto_free_ofproto_controller_info(&info); } + +/* "Instant" stats. + * + * Some information in the database must be kept as up-to-date as possible to + * allow controllers to respond rapidly to network outages. We call these + * statistics "instant" stats. + * + * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds, + * assuming that they've changed. The only means we have to determine whether + * they have changed are: + * + * - Try to commit changes to the database. If nothing changed, then + * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other + * value. + * + * - instant_stats_run() is called late in the run loop, after anything that + * might change any of the instant stats. + * + * We use these two facts together to avoid waking the process up every + * INSTANT_INTERVAL_MSEC whether there is any change or not. + */ + +/* Minimum interval between writing updates to the instant stats to the + * database. */ +#define INSTANT_INTERVAL_MSEC 100 + +/* Current instant stats database transaction, NULL if there is no ongoing + * transaction. */ +static struct ovsdb_idl_txn *instant_txn; + +/* Next time (in msec on monotonic clock) at which we will update the instant + * stats. */ +static long long int instant_next_txn = LLONG_MIN; + +/* True if the run loop has run since we last saw that the instant stats were + * unchanged, that is, this is true if we need to wake up at 'instant_next_txn' + * to refresh the instant stats. */ +static bool instant_stats_could_have_changed; static void -refresh_instant_stats(void) +instant_stats_run(void) { - static struct ovsdb_idl_txn *txn = NULL; + enum ovsdb_idl_txn_status status; + + instant_stats_could_have_changed = true; - if (!txn) { + if (!instant_txn) { struct bridge *br; - txn = ovsdb_idl_txn_create(idl); + if (time_msec() < instant_next_txn) { + return; + } + instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC; + instant_txn = ovsdb_idl_txn_create(idl); HMAP_FOR_EACH (br, node, &all_bridges) { struct iface *iface; struct port *port; @@ -2030,6 +2224,7 @@ refresh_instant_stats(void) HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) { enum netdev_flags flags; + struct smap smap; const char *link_state; int64_t link_resets; int current, error; @@ -2062,16 +2257,37 @@ refresh_instant_stats(void) ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1); iface_refresh_cfm_stats(iface); + + smap_init(&smap); + if (!ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port, + &smap)) { + ovsrec_interface_set_bfd_status(iface->cfg, &smap); + } + smap_destroy(&smap); } } } - if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) { - ovsdb_idl_txn_destroy(txn); - txn = NULL; + status = ovsdb_idl_txn_commit(instant_txn); + if (status != TXN_INCOMPLETE) { + ovsdb_idl_txn_destroy(instant_txn); + instant_txn = NULL; + } + if (status == TXN_UNCHANGED) { + instant_stats_could_have_changed = false; } } +static void +instant_stats_wait(void) +{ + if (instant_txn) { + ovsdb_idl_txn_wait(instant_txn); + } else if (instant_stats_could_have_changed) { + poll_timer_wait_until(instant_next_txn); + } +} + /* Performs periodic activity required by bridges that needs to be done with * the least possible latency. * @@ -2120,11 +2336,16 @@ bridge_run(void) struct bridge *br, *next_br; VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, " - "disabling this process until it goes away"); + "disabling this process (pid %ld) until it goes away", + (long int) getpid()); HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) { bridge_destroy(br); } + /* Since we will not be running system_stats_run() in this process + * with the current situation of multiple ovs-vswitchd daemons, + * disable system stats collection. */ + system_stats_enable(false); return; } else if (!ovsdb_idl_has_lock(idl)) { return; @@ -2138,6 +2359,13 @@ bridge_run(void) * returns immediately. */ bridge_init_ofproto(cfg); + /* Once the value of flow-restore-wait is false, we no longer should + * check its value from the database. */ + if (cfg && ofproto_get_flow_restore_wait()) { + ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config, + "flow-restore-wait", false)); + } + /* Let each datapath type do the work that it needs to do. */ sset_init(&types); ofproto_enumerate_types(&types); @@ -2191,15 +2419,25 @@ bridge_run(void) } if (reconfiguring) { - if (cfg) { - if (!reconf_txn) { - reconf_txn = ovsdb_idl_txn_create(idl); - } - if (bridge_reconfigure_continue(cfg)) { + if (!reconf_txn) { + reconf_txn = ovsdb_idl_txn_create(idl); + } + + if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) { + reconfiguring = false; + + if (cfg) { ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg); } - } else { - bridge_reconfigure_continue(&null_cfg); + + /* If we are completing our initial configuration for this run + * of ovs-vswitchd, then keep the transaction around to monitor + * it for completion. */ + if (!initial_config_done) { + initial_config_done = true; + daemonize_txn = reconf_txn; + reconf_txn = NULL; + } } } @@ -2209,6 +2447,22 @@ bridge_run(void) reconf_txn = NULL; } + if (daemonize_txn) { + enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn); + if (status != TXN_INCOMPLETE) { + ovsdb_idl_txn_destroy(daemonize_txn); + daemonize_txn = NULL; + + /* ovs-vswitchd has completed initialization, so allow the + * process that forked us to exit successfully. */ + daemonize_complete(); + + vlog_enable_async(); + + VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION); + } + } + /* Refresh interface and mirror stats if necessary. */ if (time_msec() >= iface_stats_timer) { if (cfg) { @@ -2242,7 +2496,7 @@ bridge_run(void) } run_system_stats(); - refresh_instant_stats(); + instant_stats_run(); } void @@ -2252,6 +2506,9 @@ bridge_wait(void) const char *type; ovsdb_idl_wait(idl); + if (daemonize_txn) { + ovsdb_idl_txn_wait(daemonize_txn); + } if (reconfiguring) { poll_immediate_wake(); @@ -2274,6 +2531,7 @@ bridge_wait(void) } system_stats_wait(); + instant_stats_wait(); } /* Adds some memory usage statistics for bridges into 'usage', for use with @@ -2296,13 +2554,11 @@ struct qos_unixctl_show_cbdata { }; static void -qos_unixctl_show_cb(unsigned int queue_id, - const struct smap *details, - void *aux) +qos_unixctl_show_queue(unsigned int queue_id, + const struct smap *details, + struct iface *iface, + struct ds *ds) { - struct qos_unixctl_show_cbdata *data = aux; - struct ds *ds = data->ds; - struct iface *iface = data->iface; struct netdev_queue_stats stats; struct smap_node *node; int error; @@ -2333,7 +2589,7 @@ qos_unixctl_show_cb(unsigned int queue_id, } } else { ds_put_format(ds, "\tFailed to get statistics for queue %u: %s", - queue_id, strerror(error)); + queue_id, ovs_strerror(error)); } } @@ -2346,8 +2602,6 @@ qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED, struct iface *iface; const char *type; struct smap_node *node; - struct qos_unixctl_show_cbdata data; - int error; iface = iface_find(argv[1]); if (!iface) { @@ -2358,19 +2612,22 @@ qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED, netdev_get_qos(iface->netdev, &type, &smap); if (*type != '\0') { + struct netdev_queue_dump dump; + struct smap details; + unsigned int queue_id; + ds_put_format(&ds, "QoS: %s %s\n", iface->name, type); SMAP_FOR_EACH (node, &smap) { ds_put_format(&ds, "%s: %s\n", node->key, node->value); } - data.ds = &ds; - data.iface = iface; - error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data); - - if (error) { - ds_put_format(&ds, "failed to dump queues: %s", strerror(error)); + smap_init(&details); + NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) { + qos_unixctl_show_queue(queue_id, &details, iface, &ds); } + smap_destroy(&details); + unixctl_command_reply(conn, ds_cstr(&ds)); } else { ds_put_format(&ds, "QoS not configured on %s\n", iface->name); @@ -2688,7 +2945,7 @@ bridge_configure_local_iface_netdev(struct bridge *br, /* Bring up the local interface. */ netdev = local_iface->netdev; - netdev_turn_flags_on(netdev, NETDEV_UP, true); + netdev_turn_flags_on(netdev, NETDEV_UP, NULL); /* Configure the IP address and netmask. */ if (!c->local_netmask @@ -3080,7 +3337,7 @@ iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s) key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0); if (portid <= 0 || portid > UINT16_MAX) { - portid = iface->ofp_port; + portid = ofp_to_u16(iface->ofp_port); } if (priority <= 0 || priority > UINT16_MAX) { @@ -3204,11 +3461,11 @@ iface_destroy(struct iface *iface) struct port *port = iface->port; struct bridge *br = port->bridge; - if (br->ofproto && iface->ofp_port >= 0) { + if (br->ofproto && iface->ofp_port != OFPP_NONE) { ofproto_port_unregister(br->ofproto, iface->ofp_port); } - if (iface->ofp_port >= 0) { + if (iface->ofp_port != OFPP_NONE) { hmap_remove(&br->ifaces, &iface->ofp_port_node); } @@ -3268,12 +3525,12 @@ if_cfg_lookup(const struct bridge *br, const char *name) } static struct iface * -iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port) +iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port) { struct iface *iface; - HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, - hash_int(ofp_port, 0), &br->ifaces) { + HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port), + &br->ifaces) { if (iface->ofp_port == ofp_port) { return iface; } @@ -3301,7 +3558,7 @@ iface_set_mac(struct iface *iface) int error = netdev_set_etheraddr(iface->netdev, ea); if (error) { VLOG_ERR("interface %s: setting MAC failed (%s)", - iface->name, strerror(error)); + iface->name, ovs_strerror(error)); } } } @@ -3309,10 +3566,12 @@ iface_set_mac(struct iface *iface) /* Sets the ofport column of 'if_cfg' to 'ofport'. */ static void -iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport) +iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport) { + int64_t port_; + port_ = (ofport == OFPP_NONE) ? -1 : ofp_to_u16(ofport); if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) { - ovsrec_interface_set_ofport(if_cfg, &ofport, 1); + ovsrec_interface_set_ofport(if_cfg, &port_, 1); } } @@ -3330,20 +3589,17 @@ iface_clear_db_record(const struct ovsrec_interface *if_cfg) 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_mac_in_use(if_cfg, NULL); ovsrec_interface_set_mtu(if_cfg, NULL, 0); ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0); ovsrec_interface_set_cfm_fault_status(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); + ovsrec_interface_set_ifindex(if_cfg, NULL, 0); } } -struct iface_delete_queues_cbdata { - struct netdev *netdev; - const struct ovsdb_datum *queues; -}; - static bool queue_ids_include(const struct ovsdb_datum *queues, int64_t target) { @@ -3353,17 +3609,6 @@ queue_ids_include(const struct ovsdb_datum *queues, int64_t target) return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX; } -static void -iface_delete_queues(unsigned int queue_id, - const struct smap *details OVS_UNUSED, void *cbdata_) -{ - struct iface_delete_queues_cbdata *cbdata = cbdata_; - - if (!queue_ids_include(cbdata->queues, queue_id)) { - netdev_delete_queue(cbdata->netdev, queue_id); - } -} - static void iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) { @@ -3374,7 +3619,10 @@ iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) { netdev_set_qos(iface->netdev, NULL, NULL); } else { - struct iface_delete_queues_cbdata cbdata; + const struct ovsdb_datum *queues; + struct netdev_queue_dump dump; + unsigned int queue_id; + struct smap details; bool queue_zero; size_t i; @@ -3382,10 +3630,15 @@ iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) netdev_set_qos(iface->netdev, qos->type, &qos->other_config); /* Deconfigure queues that were deleted. */ - cbdata.netdev = iface->netdev; - cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER, - OVSDB_TYPE_UUID); - netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata); + queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER, + OVSDB_TYPE_UUID); + smap_init(&details); + NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) { + if (!queue_ids_include(queues, queue_id)) { + netdev_delete_queue(iface->netdev, queue_id); + } + } + smap_destroy(&details); /* Configure queues for 'iface'. */ queue_zero = false; @@ -3417,7 +3670,7 @@ iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) } } - if (iface->ofp_port >= 0) { + if (iface->ofp_port != OFPP_NONE) { const struct ofproto_port_queue *port_queues = queues_buf.data; size_t n_queues = queues_buf.size / sizeof *port_queues; @@ -3479,6 +3732,7 @@ iface_configure_cfm(struct iface *iface) s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended", false); + s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false); opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate"); s.opup = !opstate_str || !strcasecmp("up", opstate_str); @@ -3494,11 +3748,13 @@ iface_is_synthetic(const struct iface *iface) return ovsdb_idl_row_is_synthetic(&iface->cfg->header_); } -static int64_t +static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *cfg) { - int64_t ofport = cfg->n_ofport ? *cfg->ofport : OFPP_NONE; - return cfg->n_ofport_request ? *cfg->ofport_request : ofport; + ofp_port_t ofport = cfg->n_ofport ? u16_to_ofp(*cfg->ofport) + : OFPP_NONE; + return cfg->n_ofport_request ? u16_to_ofp(*cfg->ofport_request) + : ofport; } @@ -3836,10 +4092,10 @@ collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg) if (!netdev_open(vlan_dev->name, "system", &netdev)) { if (!netdev_get_in4(netdev, NULL, NULL) || !netdev_get_in6(netdev, NULL)) { - vlandev_del(vlan_dev->name); - } else { /* It has an IP address configured, so we don't own * it. Don't delete it. */ + } else { + vlandev_del(vlan_dev->name); } netdev_close(netdev); } @@ -3876,7 +4132,7 @@ static void configure_splinter_port(struct port *port) { struct ofproto *ofproto = port->bridge->ofproto; - uint16_t realdev_ofp_port; + ofp_port_t realdev_ofp_port; const char *realdev_name; struct iface *vlandev, *realdev;