X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=eb1ab3c4b814451906cb2480b55380627cd19442;hb=4816a18f33380a33d381b77d41df39113c94500d;hp=940e5e7f5895ec8b49a9a6782b2ae3c9389a7419;hpb=91ae5bece01cfe69741f9e284d0ccf78fce87eb0;p=sliver-openvswitch.git diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index 940e5e7f5..eb1ab3c4b 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -1,4 +1,4 @@ -/* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc. +/* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -15,7 +15,6 @@ #include #include "bridge.h" -#include #include #include #include @@ -36,6 +35,7 @@ #include "meta-flow.h" #include "netdev.h" #include "ofp-print.h" +#include "ofp-util.h" #include "ofpbuf.h" #include "ofproto/ofproto.h" #include "poll-loop.h" @@ -66,6 +66,7 @@ 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. */ }; /* OpenFlow port slated for removal from ofproto. */ @@ -145,6 +146,23 @@ 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; @@ -180,10 +198,12 @@ 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_idle_time(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 *); static void bridge_configure_remotes(struct bridge *, const struct sockaddr_in *managers, size_t n_managers); @@ -210,8 +230,7 @@ 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 *, - uint32_t *bond_stable_ids); +static void port_configure_bond(struct port *, struct bond_settings *); static bool port_is_synthetic(const struct port *); static void reconfigure_system_stats(const struct ovsrec_open_vswitch *); @@ -226,6 +245,8 @@ 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_is_internal(const struct ovsrec_interface *iface, + const struct ovsrec_bridge *br); static const char *iface_get_type(const struct ovsrec_interface *, const struct ovsrec_bridge *); static void iface_destroy(struct iface *); @@ -243,6 +264,7 @@ 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 *); /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.) * @@ -261,6 +283,49 @@ static void configure_splinter_port(struct port *); static void add_vlan_splinter_ports(struct bridge *, const unsigned long int *splinter_vlans, struct shash *ports); + +static void +bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg) +{ + struct shash iface_hints; + static bool initialized = false; + int i; + + if (initialized) { + return; + } + + shash_init(&iface_hints); + + if (cfg) { + for (i = 0; i < cfg->n_bridges; i++) { + const struct ovsrec_bridge *br_cfg = cfg->bridges[i]; + int j; + + for (j = 0; j < br_cfg->n_ports; j++) { + struct ovsrec_port *port_cfg = br_cfg->ports[j]; + int k; + + for (k = 0; k < port_cfg->n_interfaces; k++) { + struct ovsrec_interface *if_cfg = port_cfg->interfaces[k]; + struct iface_hint *iface_hint; + + iface_hint = xmalloc(sizeof *iface_hint); + iface_hint->br_name = br_cfg->name; + iface_hint->br_type = br_cfg->datapath_type; + iface_hint->ofp_port = iface_pick_ofport(if_cfg); + + shash_add(&iface_hints, if_cfg->name, iface_hint); + } + } + } + } + + ofproto_init(&iface_hints); + + shash_destroy_free_data(&iface_hints); + initialized = true; +} /* Public functions. */ @@ -271,9 +336,10 @@ 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); ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg); ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics); @@ -297,6 +363,7 @@ 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_mtu); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics); @@ -322,8 +389,9 @@ 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); @@ -421,7 +489,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) COVERAGE_INC(bridge_reconfigure); - assert(!reconfiguring); + ovs_assert(!reconfiguring); reconfiguring = true; /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according @@ -515,7 +583,7 @@ bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg) struct bridge *br; bool done; - assert(reconfiguring); + ovs_assert(reconfiguring); done = bridge_reconfigure_ofp(); /* Complete the configuration. */ @@ -542,24 +610,17 @@ bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg) bridge_configure_mirrors(br); bridge_configure_flow_eviction_threshold(br); bridge_configure_forward_bpdu(br); - bridge_configure_mac_idle_time(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("%s (Open vSwitch) %s", program_name, VERSION); - } - return done; } @@ -711,12 +772,9 @@ port_configure(struct port *port) /* Get bond settings. */ if (s.n_slaves > 1) { 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); + port_configure_bond(port, &bond_settings); } else { s.bond = NULL; - s.bond_stable_ids = NULL; - LIST_FOR_EACH (iface, port_elem, &port->ifaces) { netdev_set_miimon_interval(iface->netdev, 0); } @@ -729,7 +787,6 @@ port_configure(struct port *port) 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'. */ @@ -766,6 +823,18 @@ bridge_configure_datapath_id(struct bridge *br) free(dpid_string); } +/* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with + * 'br'. */ +static uint32_t +bridge_get_allowed_versions(struct bridge *br) +{ + if (!br->cfg->n_protocols) + return 0; + + return ofputil_versions_from_strings(br->cfg->protocols, + br->cfg->n_protocols); +} + /* Set NetFlow configuration on 'br'. */ static void bridge_configure_netflow(struct bridge *br) @@ -878,6 +947,79 @@ 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 (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++; + } + } + } + + 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, @@ -952,16 +1094,11 @@ port_configure_stp(const struct ofproto *ofproto, struct port *port, port_s->path_cost = strtoul(config_str, NULL, 10); } else { enum netdev_features current; + unsigned int mbps; - 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); - } + netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL); + mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000; + port_s->path_cost = stp_convert_speed_to_cost(mbps); } config_str = smap_get(&port->cfg->other_config, "stp-port-priority"); @@ -1130,7 +1267,7 @@ iface_set_ofp_port(struct iface *iface, int ofp_port) { struct bridge *br = iface->port->bridge; - assert(iface->ofp_port < 0 && ofp_port >= 0); + ovs_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); @@ -1189,7 +1326,7 @@ bridge_refresh_one_ofp_port(struct bridge *br, * configured as the user requested, so we must destroy it. */ return false; } else { - /* It's the right type and configured correctly. keep it. */ + /* It's the right type and configured correctly. Keep it. */ iface_set_ofp_port(iface, ofp_port); return true; } @@ -1260,7 +1397,7 @@ bridge_refresh_ofp_port(struct bridge *br) } } -/* Opens a network device for 'iface_cfg' and configures it. If '*ofp_portp' +/* 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 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp' * untouched. @@ -1269,10 +1406,11 @@ bridge_refresh_ofp_port(struct bridge *br) * failure, returns a positive errno value and stores NULL in '*netdevp'. */ static int iface_do_create(const struct bridge *br, - const struct ovsrec_interface *iface_cfg, - const struct ovsrec_port *port_cfg, + const struct if_cfg *if_cfg, int *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; int error; @@ -1290,7 +1428,7 @@ iface_do_create(const struct bridge *br, } if (*ofp_portp < 0) { - uint16_t ofp_port; + uint16_t ofp_port = if_cfg->ofport; error = ofproto_port_add(br->ofproto, netdev, &ofp_port); if (error) { @@ -1305,7 +1443,8 @@ iface_do_create(const struct bridge *br, br->name, iface_cfg->name, *ofp_portp); } - if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) { + 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); } @@ -1334,11 +1473,7 @@ iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port) struct iface *iface; struct port *port; int error; - - /* Get rid of 'if_cfg' itself. We already copied out the interesting - * bits. */ - hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node); - free(if_cfg); + bool ok = true; /* Do the bits that can fail up front. * @@ -1346,12 +1481,14 @@ iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port) * internal datastructures may not be consistent. Eventually, when port * additions and deletions are cheaper, these calls should be removed. */ bridge_run_fast(); - assert(!iface_lookup(br, iface_cfg->name)); - error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev); + ovs_assert(!iface_lookup(br, iface_cfg->name)); + error = iface_do_create(br, if_cfg, &ofp_port, &netdev); bridge_run_fast(); if (error) { + iface_set_ofport(iface_cfg, -1); iface_clear_db_record(iface_cfg); - return false; + ok = false; + goto done; } /* Get or create the port structure. */ @@ -1389,7 +1526,9 @@ iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port) error = netdev_open(port->name, "internal", &netdev); if (!error) { - ofproto_port_add(br->ofproto, netdev, NULL); + uint16_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)", @@ -1401,7 +1540,11 @@ iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port) } } - return true; +done: + hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node); + free(if_cfg); + + return ok; } /* Set Flow eviction threshold */ @@ -1416,7 +1559,7 @@ bridge_configure_flow_eviction_threshold(struct bridge *br) if (threshold_str) { threshold = strtoul(threshold_str, NULL, 10); } else { - threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT; + threshold = OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT; } ofproto_set_flow_eviction_threshold(br->ofproto, threshold); } @@ -1431,18 +1574,27 @@ bridge_configure_forward_bpdu(struct bridge *br) false)); } -/* Set MAC aging time for 'br'. */ +/* Set MAC learning table configuration for 'br'. */ static void -bridge_configure_mac_idle_time(struct bridge *br) +bridge_configure_mac_table(struct bridge *br) { const char *idle_time_str; int idle_time; + const char *mac_table_size_str; + int mac_table_size; + idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time"); idle_time = (idle_time_str && atoi(idle_time_str) ? atoi(idle_time_str) : MAC_ENTRY_DEFAULT_IDLE_TIME); - ofproto_set_mac_idle_time(br->ofproto, idle_time); + + mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size"); + mac_table_size = (mac_table_size_str && atoi(mac_table_size_str) + ? atoi(mac_table_size_str) + : MAC_DEFAULT_MAX); + + ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size); } static void @@ -1545,15 +1697,10 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], found_addr = true; } } - if (found_addr) { - VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT, - br->name, ETH_ADDR_ARGS(ea)); - } else { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10); + + if (!found_addr) { memcpy(ea, br->default_ea, ETH_ADDR_LEN); *hw_addr_iface = NULL; - VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet " - "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea)); } hmapx_destroy(&mirror_output_ports); @@ -1636,6 +1783,7 @@ iface_refresh_status(struct iface *iface) int64_t bps; int mtu; int64_t mtu_64; + uint8_t mac[ETH_ADDR_LEN]; int error; if (iface_is_synthetic(iface)) { @@ -1644,7 +1792,7 @@ iface_refresh_status(struct iface *iface) smap_init(&smap); - if (!netdev_get_drv_info(iface->netdev, &smap)) { + if (!netdev_get_status(iface->netdev, &smap)) { ovsrec_interface_set_status(iface->cfg, &smap); } else { ovsrec_interface_set_status(iface->cfg, NULL); @@ -1653,15 +1801,13 @@ iface_refresh_status(struct iface *iface) smap_destroy(&smap); error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL); - if (!error) { + bps = !error ? netdev_features_to_bps(current, 0) : 0; + if (bps) { ovsrec_interface_set_duplex(iface->cfg, netdev_features_is_full_duplex(current) ? "full" : "half"); - /* warning: uint64_t -> int64_t conversion */ - bps = netdev_features_to_bps(current); ovsrec_interface_set_link_speed(iface->cfg, &bps, 1); - } - else { + } else { ovsrec_interface_set_duplex(iface->cfg, NULL); ovsrec_interface_set_link_speed(iface->cfg, NULL, 0); } @@ -1670,10 +1816,19 @@ iface_refresh_status(struct iface *iface) if (!error) { mtu_64 = mtu; ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1); - } - else { + } 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); + } } /* Writes 'iface''s CFM statistics to the database. 'iface' must not be @@ -1682,57 +1837,47 @@ static void iface_refresh_cfm_stats(struct iface *iface) { const struct ovsrec_interface *cfg = iface->cfg; - int fault, opup, error; - const uint64_t *rmps; - size_t n_rmps; - int health; - - fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto, - iface->ofp_port); - if (fault >= 0) { + struct ofproto_cfm_status status; + + if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto, + iface->ofp_port, &status)) { + ovsrec_interface_set_cfm_fault(cfg, NULL, 0); + ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0); + ovsrec_interface_set_cfm_remote_opstate(cfg, NULL); + ovsrec_interface_set_cfm_health(cfg, NULL, 0); + ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0); + } else { const char *reasons[CFM_FAULT_N_REASONS]; - bool fault_bool = fault; + int64_t cfm_health = status.health; + bool faulted = status.faults != 0; size_t i, j; + ovsrec_interface_set_cfm_fault(cfg, &faulted, 1); + j = 0; for (i = 0; i < CFM_FAULT_N_REASONS; i++) { int reason = 1 << i; - if (fault & reason) { + if (status.faults & reason) { reasons[j++] = cfm_fault_reason_to_str(reason); } } - - ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1); ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j); - } else { - ovsrec_interface_set_cfm_fault(cfg, NULL, 0); - ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0); - } - - opup = ofproto_port_get_cfm_opup(iface->port->bridge->ofproto, - iface->ofp_port); - if (opup >= 0) { - ovsrec_interface_set_cfm_remote_opstate(cfg, opup ? "up" : "down"); - } else { - ovsrec_interface_set_cfm_remote_opstate(cfg, NULL); - } - 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); - } + if (status.remote_opstate >= 0) { + const char *remote_opstate = status.remote_opstate ? "up" : "down"; + ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate); + } else { + ovsrec_interface_set_cfm_remote_opstate(cfg, NULL); + } - health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto, - iface->ofp_port); - if (health >= 0) { - int64_t cfm_health = health; - ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1); - } else { - ovsrec_interface_set_cfm_health(cfg, NULL, 0); + ovsrec_interface_set_cfm_remote_mpids(cfg, + (const int64_t *)status.rmps, + status.n_rmps); + if (cfm_health >= 0) { + ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1); + } else { + ovsrec_interface_set_cfm_health(cfg, NULL, 0); + } } } @@ -1774,7 +1919,7 @@ iface_refresh_stats(struct iface *iface) #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER; IFACE_STATS; #undef IFACE_STAT - assert(i == ARRAY_SIZE(keys)); + ovs_assert(i == ARRAY_SIZE(keys)); ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys)); @@ -1900,15 +2045,16 @@ run_system_stats(void) } static inline const char * -nx_role_to_str(enum nx_role role) +ofp12_controller_role_to_str(enum ofp12_controller_role role) { switch (role) { - case NX_ROLE_OTHER: + case OFPCR12_ROLE_EQUAL: return "other"; - case NX_ROLE_MASTER: + case OFPCR12_ROLE_MASTER: return "master"; - case NX_ROLE_SLAVE: + case OFPCR12_ROLE_SLAVE: return "slave"; + case OFPCR12_ROLE_NOCHANGE: default: return "*** INVALID ROLE ***"; } @@ -1944,7 +2090,8 @@ refresh_controller_status(void) } ovsrec_controller_set_is_connected(cfg, cinfo->is_connected); - ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role)); + ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str( + cinfo->role)); ovsrec_controller_set_status(cfg, &smap); smap_destroy(&smap); } else { @@ -1956,17 +2103,61 @@ 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; @@ -2015,12 +2206,26 @@ refresh_instant_stats(void) } } - 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. * @@ -2030,8 +2235,17 @@ refresh_instant_stats(void) void bridge_run_fast(void) { + struct sset types; + const char *type; struct bridge *br; + sset_init(&types); + ofproto_enumerate_types(&types); + SSET_FOR_EACH (type, &types) { + ofproto_type_run_fast(type); + } + sset_destroy(&types); + HMAP_FOR_EACH (br, node, &all_bridges) { ofproto_run_fast(br->ofproto); } @@ -2040,14 +2254,16 @@ bridge_run_fast(void) void bridge_run(void) { - static const struct ovsrec_open_vswitch null_cfg; + static struct ovsrec_open_vswitch null_cfg; const struct ovsrec_open_vswitch *cfg; struct ovsdb_idl_txn *reconf_txn = NULL; + struct sset types; + const char *type; bool vlan_splinters_changed; struct bridge *br; - ovsrec_open_vswitch_init((struct ovsrec_open_vswitch *) &null_cfg); + ovsrec_open_vswitch_init(&null_cfg); /* (Re)configure if necessary. */ if (!reconfiguring) { @@ -2058,11 +2274,16 @@ 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; @@ -2070,6 +2291,20 @@ bridge_run(void) } cfg = ovsrec_open_vswitch_first(idl); + /* Initialize the ofproto library. This only needs to run once, but + * it must be done after the configuration is set. If the + * initialization has already occurred, bridge_init_ofproto() + * returns immediately. */ + bridge_init_ofproto(cfg); + + /* Let each datapath type do the work that it needs to do. */ + sset_init(&types); + ofproto_enumerate_types(&types); + SSET_FOR_EACH (type, &types) { + ofproto_type_run(type); + } + sset_destroy(&types); + /* Let each bridge do the work that it needs to do. */ HMAP_FOR_EACH (br, node, &all_bridges) { ofproto_run(br->ofproto); @@ -2115,15 +2350,25 @@ 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; + } } } @@ -2133,6 +2378,20 @@ 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_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION); + } + } + /* Refresh interface and mirror stats if necessary. */ if (time_msec() >= iface_stats_timer) { if (cfg) { @@ -2166,18 +2425,31 @@ bridge_run(void) } run_system_stats(); - refresh_instant_stats(); + instant_stats_run(); } void bridge_wait(void) { + struct sset types; + const char *type; + ovsdb_idl_wait(idl); + if (daemonize_txn) { + ovsdb_idl_txn_wait(daemonize_txn); + } if (reconfiguring) { poll_immediate_wake(); } + sset_init(&types); + ofproto_enumerate_types(&types); + SSET_FOR_EACH (type, &types) { + ofproto_type_wait(type); + } + sset_destroy(&types); + if (!hmap_is_empty(&all_bridges)) { struct bridge *br; @@ -2188,6 +2460,7 @@ bridge_wait(void) } system_stats_wait(); + instant_stats_wait(); } /* Adds some memory usage statistics for bridges into 'usage', for use with @@ -2301,7 +2574,7 @@ bridge_create(const struct ovsrec_bridge *br_cfg) { struct bridge *br; - assert(!bridge_lookup(br_cfg->name)); + ovs_assert(!bridge_lookup(br_cfg->name)); br = xzalloc(sizeof *br); br->name = xstrdup(br_cfg->name); @@ -2450,6 +2723,7 @@ bridge_queue_if_cfg(struct bridge *br, if_cfg->cfg = cfg; if_cfg->parent = parent; + if_cfg->ofport = iface_pick_ofport(cfg); hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node, hash_string(if_cfg->cfg->name, 0)); } @@ -2466,7 +2740,7 @@ bridge_add_del_ports(struct bridge *br, struct shash new_ports; size_t i; - assert(hmap_is_empty(&br->if_cfg_todo)); + ovs_assert(hmap_is_empty(&br->if_cfg_todo)); /* Collect new ports. */ shash_init(&new_ports); @@ -2611,7 +2885,7 @@ bridge_configure_local_iface_netdev(struct bridge *br, } if (!netdev_set_in4(netdev, ip, mask)) { VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT, - br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr)); + br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr)); } /* Configure the default gateway. */ @@ -2620,19 +2894,30 @@ bridge_configure_local_iface_netdev(struct bridge *br, && gateway.s_addr) { if (!netdev_add_router(netdev, gateway)) { VLOG_INFO("bridge %s: configured gateway "IP_FMT, - br->name, IP_ARGS(&gateway.s_addr)); + br->name, IP_ARGS(gateway.s_addr)); } } } /* 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". */ + * e.g. "x///y" is equal to "x/y". + * + * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding + * bytes from 'a', the function considers this success. Specify 'b_stoplen' as + * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of + * 'b' against a prefix of 'a'. + */ static bool -equal_pathnames(const char *a, const char *b) +equal_pathnames(const char *a, const char *b, size_t b_stoplen) { - while (*a == *b) { - if (*a == '/') { + const char *b_start = b; + for (;;) { + if (b - b_start >= b_stoplen) { + return true; + } else if (*a != *b) { + return false; + } else if (*a == '/') { a += strspn(a, "/"); b += strspn(b, "/"); } else if (*a == '\0') { @@ -2642,7 +2927,6 @@ equal_pathnames(const char *a, const char *b) b++; } } - return false; } static void @@ -2689,21 +2973,42 @@ bridge_configure_remotes(struct bridge *br, static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); char *whitelist; - whitelist = xasprintf("unix:%s/%s.controller", + if (!strncmp(c->target, "unix:", 5)) { + /* Connect to a listening socket */ + whitelist = xasprintf("unix:%s/", ovs_rundir()); + if (strchr(c->target, '/') && + !equal_pathnames(c->target, whitelist, + strlen(whitelist))) { + /* Absolute path specified, but not in ovs_rundir */ + VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket " + "controller \"%s\" due to possibility for " + "remote exploit. Instead, specify socket " + "in 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); - 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; + free(whitelist); + continue; + } + } else { + whitelist = xasprintf("punix:%s/%s.controller", + ovs_rundir(), br->name); + if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) { + /* 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 of " + "overwriting local files. 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; + } } free(whitelist); @@ -2717,7 +3022,8 @@ bridge_configure_remotes(struct bridge *br, n_ocs++; } - ofproto_set_controllers(br->ofproto, ocs, n_ocs); + ofproto_set_controllers(br->ofproto, ocs, n_ocs, + bridge_get_allowed_versions(br)); free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */ free(ocs); @@ -2799,6 +3105,13 @@ bridge_configure_tables(struct bridge *br) br->cfg->key_flow_tables[j]); } } + +static void +bridge_configure_dp_desc(struct bridge *br) +{ + ofproto_set_dp_desc(br->ofproto, + smap_get(&br->cfg->other_config, "dp-desc")); +} /* Port functions. */ @@ -2972,13 +3285,11 @@ iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s) } static void -port_configure_bond(struct port *port, struct bond_settings *s, - uint32_t *bond_stable_ids) +port_configure_bond(struct port *port, struct bond_settings *s) { const char *detect_s; struct iface *iface; int miimon_interval; - size_t i; s->name = port->name; s->balance = BM_AB; @@ -3030,17 +3341,7 @@ 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 = smap_get_int(&iface->cfg->other_config, "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); } } @@ -3055,17 +3356,32 @@ port_is_synthetic(const struct port *port) /* Interface functions. */ +static bool +iface_is_internal(const struct ovsrec_interface *iface, + const struct ovsrec_bridge *br) +{ + /* The local port and "internal" ports are always "internal". */ + return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name); +} + /* Returns the correct network device type for interface 'iface' in bridge * 'br'. */ static const char * iface_get_type(const struct ovsrec_interface *iface, const struct ovsrec_bridge *br) { - /* The local port always has type "internal". Other ports take their type - * from the database and default to "system" if none is specified. */ - return (!strcmp(iface->name, br->name) ? "internal" - : iface->type[0] ? iface->type - : "system"); + const char *type; + + /* The local port always has type "internal". Other ports take + * their type from the database and default to "system" if none is + * specified. */ + if (iface_is_internal(iface, br)) { + type = "internal"; + } else { + type = iface->type[0] ? iface->type : "system"; + } + + return ofproto_port_open_type(br->datapath_type, type); } static void @@ -3196,12 +3512,12 @@ 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); 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_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); @@ -3311,12 +3627,24 @@ iface_configure_cfm(struct iface *iface) const char *opstate_str; const char *cfm_ccm_vlan; struct cfm_settings s; + struct smap netdev_args; if (!cfg->n_cfm_mpid) { ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port); return; } + s.check_tnl_key = false; + smap_init(&netdev_args); + if (!netdev_get_config(iface->netdev, &netdev_args)) { + const char *key = smap_get(&netdev_args, "key"); + const char *in_key = smap_get(&netdev_args, "in_key"); + + s.check_tnl_key = (key && !strcmp(key, "flow")) + || (in_key && !strcmp(in_key, "flow")); + } + smap_destroy(&netdev_args); + s.mpid = *cfg->cfm_mpid; s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0); cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan"); @@ -3354,6 +3682,13 @@ iface_is_synthetic(const struct iface *iface) return ovsdb_idl_row_is_synthetic(&iface->cfg->header_); } +static int64_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; +} + /* Port mirroring. */