X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=899fd64c22aa8cae29e54a5637e634a860b14ef5;hb=29089a540cfa30a834e3ee19a8b4c52ff2e331b2;hp=204972083d50b994fe8a7d840c35b4b582f1a95d;hpb=7f81a52a98bf166c5419363f12ad369a037eb1b7;p=sliver-openvswitch.git diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index 204972083..899fd64c2 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -1,4 +1,4 @@ -/* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira Networks +/* 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 @@ -36,11 +35,13 @@ #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" #include "sha1.h" #include "shash.h" +#include "smap.h" #include "socket-util.h" #include "stream.h" #include "stream-ssl.h" @@ -65,6 +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. */ @@ -144,19 +146,30 @@ 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; -/* Each time this timer expires, the bridge fetches systems and interface +/* Each time this timer expires, the bridge fetches interface and mirror * statistics and pushes them into the database. */ -#define STATS_INTERVAL (5 * 1000) /* In milliseconds. */ -static long long int stats_timer = LLONG_MIN; - -/* Stores the time after which rate limited statistics may be written to the - * database. Only updated when changes to the database require rate limiting. - */ -#define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */ -static long long int db_limiter = LLONG_MIN; +#define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */ +static long long int iface_stats_timer = LLONG_MIN; /* In some datapaths, creating and destroying OpenFlow ports can be extremely * expensive. This can cause bridge_reconfigure() to take a long time during @@ -185,10 +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); @@ -215,10 +230,12 @@ 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 *); +static void run_system_stats(void); + static void bridge_configure_mirrors(struct bridge *); static struct mirror *mirror_create(struct bridge *, const struct ovsrec_mirror *); @@ -227,8 +244,11 @@ 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 void iface_create(struct bridge *, struct if_cfg *, int ofp_port); -static void iface_refresh_type(struct iface *); +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 *); static struct iface *iface_lookup(const struct bridge *, const char *name); static struct iface *iface_find(const char *name); @@ -244,10 +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 void shash_from_ovs_idl_map(char **keys, char **values, size_t n, - struct shash *); -static void shash_to_ovs_idl_map(struct shash *, - char ***keys, char ***values, size_t *n); +static int64_t iface_pick_ofport(const struct ovsrec_interface *); /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.) * @@ -266,6 +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. */ @@ -276,9 +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); @@ -302,6 +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); @@ -310,6 +372,7 @@ bridge_init(const char *remote) ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status); 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_lacp_current); ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids); @@ -326,8 +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); @@ -425,7 +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 @@ -458,6 +522,8 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) iface_clear_db_record(if_cfg->cfg); } } + + reconfigure_system_stats(ovs_cfg); } static bool @@ -476,6 +542,11 @@ bridge_reconfigure_ofp(void) struct ofpp_garbage *garbage, *next; LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) { + /* It's a bit dangerous to call bridge_run_fast() here as ofproto's + * internal datastructures may not be consistent. Eventually, when + * port additions and deletions are cheaper, these calls should be + * removed. */ + bridge_run_fast(); ofproto_port_del(br->ofproto, garbage->ofp_port); list_remove(&garbage->list_node); free(garbage); @@ -484,6 +555,7 @@ bridge_reconfigure_ofp(void) if (time_msec() >= deadline) { return false; } + bridge_run_fast(); } } @@ -511,7 +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. */ @@ -538,22 +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; - } - return done; } @@ -686,8 +753,8 @@ port_configure(struct port *port) s.vlan_mode = PORT_VLAN_TRUNK; } } - s.use_priority_tags = !strcmp("true", ovsrec_port_get_other_config_value( - cfg, "priority-tags", "")); + s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags", + false); /* Get LACP settings. */ s.lacp = port_configure_lacp(port, &lacp_settings); @@ -705,12 +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); } @@ -723,7 +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'. */ @@ -750,13 +813,28 @@ bridge_configure_datapath_id(struct bridge *br) memcpy(br->ea, ea, ETH_ADDR_LEN); dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface); - ofproto_set_datapath_id(br->ofproto, dpid); + if (dpid != ofproto_get_datapath_id(br->ofproto)) { + VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid); + ofproto_set_datapath_id(br->ofproto, dpid); + } dpid_string = xasprintf("%016"PRIx64, dpid); ovsrec_bridge_set_datapath_id(br->cfg, dpid_string); 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) @@ -869,6 +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, @@ -877,9 +1028,7 @@ port_configure_stp(const struct ofproto *ofproto, struct port *port, const char *config_str; struct iface *iface; - config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-enable", - NULL); - if (config_str && !strcmp(config_str, "false")) { + if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) { port_s->enable = false; return; } else { @@ -911,8 +1060,7 @@ port_configure_stp(const struct ofproto *ofproto, struct port *port, return; } - config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-port-num", - NULL); + config_str = smap_get(&port->cfg->other_config, "stp-port-num"); if (config_str) { unsigned long int port_num = strtoul(config_str, NULL, 0); int port_idx = port_num - 1; @@ -932,7 +1080,7 @@ port_configure_stp(const struct ofproto *ofproto, struct port *port, bitmap_set1(port_num_bitmap, port_idx); port_s->port_num = port_idx; } else { - if (*port_num_counter > STP_MAX_PORTS) { + if (*port_num_counter >= STP_MAX_PORTS) { VLOG_ERR("port %s: too many STP ports, disabling", port->name); port_s->enable = false; return; @@ -941,27 +1089,19 @@ port_configure_stp(const struct ofproto *ofproto, struct port *port, port_s->port_num = (*port_num_counter)++; } - config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-path-cost", - NULL); + config_str = smap_get(&port->cfg->other_config, "stp-path-cost"); if (config_str) { 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 = ovsrec_port_get_other_config_value(port->cfg, - "stp-port-priority", - NULL); + config_str = smap_get(&port->cfg->other_config, "stp-port-priority"); if (config_str) { port_s->priority = strtoul(config_str, NULL, 0); } else { @@ -982,9 +1122,7 @@ bridge_configure_stp(struct bridge *br) int port_num_counter; unsigned long *port_num_bitmap; - config_str = ovsrec_bridge_get_other_config_value(br->cfg, - "stp-system-id", - NULL); + config_str = smap_get(&br->cfg->other_config, "stp-system-id"); if (config_str) { uint8_t ea[ETH_ADDR_LEN]; @@ -999,36 +1137,28 @@ bridge_configure_stp(struct bridge *br) br_s.system_id = eth_addr_to_uint64(br->ea); } - config_str = ovsrec_bridge_get_other_config_value(br->cfg, - "stp-priority", - NULL); + config_str = smap_get(&br->cfg->other_config, "stp-priority"); if (config_str) { br_s.priority = strtoul(config_str, NULL, 0); } else { br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY; } - config_str = ovsrec_bridge_get_other_config_value(br->cfg, - "stp-hello-time", - NULL); + config_str = smap_get(&br->cfg->other_config, "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 = ovsrec_bridge_get_other_config_value(br->cfg, - "stp-max-age", - NULL); + config_str = smap_get(&br->cfg->other_config, "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 = ovsrec_bridge_get_other_config_value(br->cfg, - "stp-forward-delay", - NULL); + config_str = smap_get(&br->cfg->other_config, "stp-forward-delay"); if (config_str) { br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000; } else { @@ -1137,22 +1267,84 @@ 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); } -static void -bridge_ofproto_port_del(struct bridge *br, struct ofproto_port ofproto_port) +/* Configures 'netdev' based on the "options" column in 'iface_cfg'. + * Returns 0 if successful, otherwise a positive errno value. */ +static int +iface_set_netdev_config(const struct ovsrec_interface *iface_cfg, + struct netdev *netdev) { - int error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port); + int error; + + error = netdev_set_config(netdev, &iface_cfg->options); if (error) { - VLOG_WARN("bridge %s: failed to remove %s interface (%s)", - br->name, ofproto_port.name, strerror(error)); + VLOG_WARN("could not configure network device %s (%s)", + iface_cfg->name, strerror(error)); + } + return error; +} + +/* This function determines whether 'ofproto_port', which is attached to + * br->ofproto's datapath, is one that we want in 'br'. + * + * If it is, it returns true, after creating an iface (if necessary), + * configuring the iface's netdev according to the iface's options, and setting + * iface's ofp_port member to 'ofproto_port->ofp_port'. + * + * If, on the other hand, 'port' should be removed, it returns false. The + * caller should later detach the port from br->ofproto. */ +static bool +bridge_refresh_one_ofp_port(struct bridge *br, + const struct ofproto_port *ofproto_port) +{ + const char *name = ofproto_port->name; + const char *type = ofproto_port->type; + uint16_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) { + VLOG_WARN("bridge %s: interface %s reported twice", + br->name, name); + return false; + } else if (iface_from_ofp_port(br, ofp_port)) { + VLOG_WARN("bridge %s: interface %"PRIu16" reported twice", + br->name, ofp_port); + return false; + } + + /* There's a configured interface named 'name'. */ + if (strcmp(type, iface->type) + || iface_set_netdev_config(iface->cfg, iface->netdev)) { + /* It's the wrong type, or it's the right type but can't be + * configured as the user requested, so we must destroy it. */ + return false; + } else { + /* It's the right type and configured correctly. Keep it. */ + iface_set_ofp_port(iface, ofp_port); + return true; + } + } else if (bridge_has_bond_fake_iface(br, name) + && !strcmp(type, "internal")) { + /* It's a bond fake iface. Keep it. */ + return true; } else { - VLOG_INFO("bridge %s: removed interface %s (%d)", br->name, - ofproto_port.name, ofproto_port.ofp_port); + /* There's no configured interface named 'name', but there might be an + * interface of that name queued to be created. + * + * If there is, and it has the correct type, then try to configure it + * and add it. If that's successful, we'll keep it. Otherwise, we'll + * delete it and later try to re-add it. */ + struct if_cfg *if_cfg = if_cfg_lookup(br, name); + return (if_cfg + && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg)) + && iface_create(br, if_cfg, ofp_port)); } } @@ -1179,33 +1371,10 @@ bridge_refresh_ofp_port(struct bridge *br) * already exist in the datapath and promote them to full fledged "struct * iface"s. Mark ports in the datapath which don't belong as garbage. */ OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) { - struct iface *iface = iface_lookup(br, ofproto_port.name); - if (iface) { - if (iface->ofp_port >= 0) { - VLOG_WARN("bridge %s: interface %s reported twice", - br->name, ofproto_port.name); - } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) { - VLOG_WARN("bridge %s: interface %"PRIu16" reported twice", - br->name, ofproto_port.ofp_port); - } else if (!strcmp(ofproto_port.type, iface->type)) { - iface_set_ofp_port(iface, ofproto_port.ofp_port); - } else { - /* Port has incorrect type so delete it later. */ - } - } else { - struct if_cfg *if_cfg = if_cfg_lookup(br, ofproto_port.name); - - if (if_cfg) { - iface_create(br, if_cfg, ofproto_port.ofp_port); - } else if (bridge_has_bond_fake_iface(br, ofproto_port.name) - && strcmp(ofproto_port.type, "internal")) { - /* Bond fake iface with the wrong type. */ - bridge_ofproto_port_del(br, ofproto_port); - } else { - struct ofpp_garbage *garbage = xmalloc(sizeof *garbage); - garbage->ofp_port = ofproto_port.ofp_port; - list_push_front(&br->ofpp_garbage, &garbage->list_node); - } + if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) { + struct ofpp_garbage *garbage = xmalloc(sizeof *garbage); + garbage->ofp_port = ofproto_port.ofp_port; + list_push_front(&br->ofpp_garbage, &garbage->list_node); } } @@ -1228,124 +1397,123 @@ bridge_refresh_ofp_port(struct bridge *br) } } -/* 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 - * automatically allocated for the iface. Takes ownership of and - * deallocates 'if_cfg'. */ -static void -iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port) +/* 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. + * + * If successful, returns 0 and stores the network device in '*netdevp'. On + * failure, returns a positive errno value and stores NULL in '*netdevp'. */ +static int +iface_do_create(const struct bridge *br, + const struct if_cfg *if_cfg, + int *ofp_portp, struct netdev **netdevp) { - struct iface *iface; - struct port *port; + const struct ovsrec_interface *iface_cfg = if_cfg->cfg; + const struct ovsrec_port *port_cfg = if_cfg->parent; + struct netdev *netdev; int error; - assert(!iface_lookup(br, if_cfg->cfg->name)); - - port = port_lookup(br, if_cfg->parent->name); - if (!port) { - port = port_create(br, if_cfg->parent); + 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)); + goto error; } - iface = xzalloc(sizeof *iface); - iface->port = port; - iface->name = xstrdup(if_cfg->cfg->name); - iface->ofp_port = -1; - iface->netdev = NULL; - iface->cfg = if_cfg->cfg; - hmap_insert(&br->iface_by_name, &iface->name_node, - hash_string(iface->name, 0)); - list_push_back(&port->ifaces, &iface->port_elem); - iface_refresh_type(iface); - if (ofp_port >= 0) { - iface_set_ofp_port(iface, ofp_port); + error = iface_set_netdev_config(iface_cfg, netdev); + if (error) { + goto error; } - hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node); - free(if_cfg); - if_cfg = NULL; + if (*ofp_portp < 0) { + uint16_t ofp_port = if_cfg->ofport; - error = netdev_open(iface->name, iface->type, &iface->netdev); - if (error) { - VLOG_WARN("could not open network device %s (%s)", iface->name, - strerror(error)); + error = ofproto_port_add(br->ofproto, netdev, &ofp_port); + if (error) { + goto error; + } + *ofp_portp = ofp_port; + + VLOG_INFO("bridge %s: added interface %s on port %d", + br->name, iface_cfg->name, *ofp_portp); + } else { + VLOG_DBG("bridge %s: interface %s is on port %d", + br->name, iface_cfg->name, *ofp_portp); } - if (iface->netdev - && port->cfg->vlan_mode - && !strcmp(port->cfg->vlan_mode, "splinter")) { - netdev_turn_flags_on(iface->netdev, NETDEV_UP, true); + 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); } - /* Configure the netdev. */ - if (iface->netdev) { - struct shash args; + *netdevp = netdev; + return 0; - 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); +error: + *netdevp = NULL; + netdev_close(netdev); + return error; +} - if (error) { - VLOG_WARN("could not configure network device %s (%s)", - iface->name, strerror(error)); - netdev_close(iface->netdev); - iface->netdev = NULL; - } - } +/* 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 + * 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) +{ + const struct ovsrec_interface *iface_cfg = if_cfg->cfg; + const struct ovsrec_port *port_cfg = if_cfg->parent; - /* Add the port, if necessary. */ - if (iface->netdev && iface->ofp_port < 0) { - uint16_t new_ofp_port; - int error; + struct netdev *netdev; + struct iface *iface; + struct port *port; + int error; + bool ok = true; - error = ofproto_port_add(br->ofproto, iface->netdev, &new_ofp_port); - if (!error) { - VLOG_INFO("bridge %s: added interface %s (%d)", br->name, - iface->name, new_ofp_port); - iface_set_ofp_port(iface, new_ofp_port); - } else { - netdev_close(iface->netdev); - iface->netdev = NULL; - } + /* Do the bits that can fail up front. + * + * It's a bit dangerous to call bridge_run_fast() here as ofproto's + * internal datastructures may not be consistent. Eventually, when port + * additions and deletions are cheaper, these calls should be removed. */ + bridge_run_fast(); + 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); + ok = false; + goto done; } - /* Initially populate stats columns. */ - if (iface->netdev) { - iface_refresh_stats(iface); - iface_refresh_status(iface); + /* Get or create the port structure. */ + port = port_lookup(br, port_cfg->name); + if (!port) { + port = port_create(br, port_cfg); } - /* 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); - } else { - struct ofproto_port ofproto_port; + /* Create the iface structure. */ + iface = xzalloc(sizeof *iface); + list_push_back(&port->ifaces, &iface->port_elem); + hmap_insert(&br->iface_by_name, &iface->name_node, + hash_string(iface_cfg->name, 0)); + iface->port = port; + iface->name = xstrdup(iface_cfg->name); + iface->ofp_port = -1; + iface->netdev = netdev; + iface->type = iface_get_type(iface_cfg, br->cfg); + iface->cfg = iface_cfg; - if (iface->netdev) { - VLOG_ERR("bridge %s: missing %s interface, dropping", - br->name, iface->name); - } else { - /* We already reported a related error, don't bother - * duplicating it. */ - } - if (!ofproto_port_query_by_name(br->ofproto, port->name, - &ofproto_port)) { - VLOG_INFO("bridge %s: removed interface %s (%d)", - br->name, port->name, ofproto_port.ofp_port); - bridge_ofproto_port_del(br, ofproto_port); - ofproto_port_destroy(&ofproto_port); - } - iface_clear_db_record(iface->cfg); - iface_destroy(iface); - } + iface_set_ofp_port(iface, ofp_port); - if (list_is_empty(&port->ifaces)) { - port_destroy(port); - return; - } + /* Populate initial status in database. */ + iface_refresh_stats(iface); + iface_refresh_status(iface); /* Add bond fake iface if necessary. */ if (port_is_bond_fake_iface(port)) { @@ -1358,7 +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)", @@ -1369,6 +1539,12 @@ iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port) ofproto_port_destroy(&ofproto_port); } } + +done: + hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node); + free(if_cfg); + + return ok; } /* Set Flow eviction threshold */ @@ -1378,14 +1554,12 @@ bridge_configure_flow_eviction_threshold(struct bridge *br) const char *threshold_str; unsigned threshold; - threshold_str = - ovsrec_bridge_get_other_config_value(br->cfg, - "flow-eviction-threshold", - NULL); + threshold_str = smap_get(&br->cfg->other_config, + "flow-eviction-threshold"); 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); } @@ -1394,32 +1568,33 @@ bridge_configure_flow_eviction_threshold(struct bridge *br) static void bridge_configure_forward_bpdu(struct bridge *br) { - const char *forward_bpdu_str; - bool forward_bpdu = false; - - forward_bpdu_str = ovsrec_bridge_get_other_config_value(br->cfg, - "forward-bpdu", - NULL); - if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) { - forward_bpdu = true; - } - ofproto_set_forward_bpdu(br->ofproto, forward_bpdu); + ofproto_set_forward_bpdu(br->ofproto, + smap_get_bool(&br->cfg->other_config, + "forward-bpdu", + 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; - idle_time_str = ovsrec_bridge_get_other_config_value(br->cfg, - "mac-aging-time", - NULL); + 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 @@ -1436,7 +1611,7 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], *hw_addr_iface = NULL; /* Did the user request a particular MAC? */ - hwaddr = ovsrec_bridge_get_other_config_value(br->cfg, "hwaddr", NULL); + hwaddr = smap_get(&br->cfg->other_config, "hwaddr"); if (hwaddr && eth_addr_from_string(hwaddr, ea)) { if (eth_addr_is_multicast(ea)) { VLOG_ERR("bridge %s: cannot set MAC address to multicast " @@ -1522,15 +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); @@ -1561,8 +1731,7 @@ bridge_pick_datapath_id(struct bridge *br, const char *datapath_id; uint64_t dpid; - datapath_id = ovsrec_bridge_get_other_config_value(br->cfg, "datapath-id", - NULL); + datapath_id = smap_get(&br->cfg->other_config, "datapath-id"); if (datapath_id && dpid_from_string(datapath_id, &dpid)) { return dpid; } @@ -1608,55 +1777,37 @@ dpid_from_hash(const void *data, size_t n) static void iface_refresh_status(struct iface *iface) { - struct shash sh; + struct smap smap; enum netdev_features current; - enum netdev_flags flags; int64_t bps; int mtu; int64_t mtu_64; + uint8_t mac[ETH_ADDR_LEN]; int error; if (iface_is_synthetic(iface)) { return; } - shash_init(&sh); - - if (!netdev_get_drv_info(iface->netdev, &sh)) { - size_t n; - char **keys, **values; + smap_init(&smap); - shash_to_ovs_idl_map(&sh, &keys, &values, &n); - ovsrec_interface_set_status(iface->cfg, keys, values, n); - - free(keys); - free(values); + if (!netdev_get_status(iface->netdev, &smap)) { + ovsrec_interface_set_status(iface->cfg, &smap); } else { - ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0); + ovsrec_interface_set_status(iface->cfg, NULL); } - shash_destroy_free_data(&sh); - - error = netdev_get_flags(iface->netdev, &flags); - if (!error) { - ovsrec_interface_set_admin_state(iface->cfg, - flags & NETDEV_UP ? "up" : "down"); - } - else { - ovsrec_interface_set_admin_state(iface->cfg, NULL); - } + 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); } @@ -1665,64 +1816,68 @@ 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. */ +/* Writes 'iface''s CFM statistics to the database. 'iface' must not be + * synthetic. */ static void iface_refresh_cfm_stats(struct iface *iface) { const struct ovsrec_interface *cfg = iface->cfg; - int fault, error; - const uint64_t *rmps; - size_t n_rmps; - int health; + struct ofproto_cfm_status status; - if (iface_is_synthetic(iface)) { - return; - } - - fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto, - iface->ofp_port); - if (fault >= 0) { + 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); - } - 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); + } } } @@ -1764,7 +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)); @@ -1774,32 +1929,27 @@ iface_refresh_stats(struct iface *iface) static void br_refresh_stp_status(struct bridge *br) { + struct smap smap = SMAP_INITIALIZER(&smap); 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 (!status.enabled) { - ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0); + ovsrec_bridge_set_status(br->cfg, NULL); 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)); + smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT, + STP_ID_ARGS(status.bridge_id)); + smap_add_format(&smap, "stp_designated_root", STP_ID_FMT, + STP_ID_ARGS(status.designated_root)); + smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost); - for (i = 0; i < ARRAY_SIZE(values); i++) { - free(values[i]); - } + ovsrec_bridge_set_status(br->cfg, &smap); + smap_destroy(&smap); } static void @@ -1808,10 +1958,9 @@ port_refresh_stp_status(struct port *port) struct ofproto *ofproto = port->bridge->ofproto; struct iface *iface; struct ofproto_port_stp_status status; - char *keys[4]; - char *str_values[4]; + char *keys[3]; int64_t int_values[3]; - size_t i; + struct smap smap; if (port_is_synthetic(port)) { return; @@ -1819,7 +1968,7 @@ port_refresh_stp_status(struct port *port) /* STP doesn't currently support bonds. */ if (!list_is_singleton(&port->ifaces)) { - ovsrec_port_set_status(port->cfg, NULL, NULL, 0); + ovsrec_port_set_status(port->cfg, NULL); return; } @@ -1830,27 +1979,19 @@ port_refresh_stp_status(struct port *port) } if (!status.enabled) { - ovsrec_port_set_status(port->cfg, NULL, NULL, 0); + ovsrec_port_set_status(port->cfg, NULL); 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]); - } + smap_init(&smap); + smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id); + smap_add(&smap, "stp_state", stp_state_name(status.state)); + smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state); + smap_add(&smap, "stp_role", stp_role_name(status.role)); + ovsrec_port_set_status(port->cfg, &smap); + smap_destroy(&smap); /* Set Statistics column. */ keys[0] = "stp_tx_count"; @@ -1867,46 +2008,53 @@ port_refresh_stp_status(struct port *port) static bool enable_system_stats(const struct ovsrec_open_vswitch *cfg) { - const char *enable; + return smap_get_bool(&cfg->other_config, "enable-statistics", false); +} - /* Use other-config:enable-system-stats by preference. */ - enable = ovsrec_open_vswitch_get_other_config_value(cfg, - "enable-statistics", - NULL); - if (enable) { - return !strcmp(enable, "true"); - } +static void +reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg) +{ + bool enable = enable_system_stats(cfg); - /* Disable by default. */ - return false; + system_stats_enable(enable); + if (!enable) { + ovsrec_open_vswitch_set_statistics(cfg, NULL); + } } static void -refresh_system_stats(const struct ovsrec_open_vswitch *cfg) +run_system_stats(void) { - struct ovsdb_datum datum; - struct shash stats; + const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl); + struct smap *stats; - shash_init(&stats); - if (enable_system_stats(cfg)) { - get_system_stats(&stats); - } + stats = system_stats_run(); + if (stats && cfg) { + struct ovsdb_idl_txn *txn; + struct ovsdb_datum datum; + + txn = ovsdb_idl_txn_create(idl); + ovsdb_datum_from_smap(&datum, stats); + ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics, + &datum); + ovsdb_idl_txn_commit(txn); + ovsdb_idl_txn_destroy(txn); - ovsdb_datum_from_shash(&datum, &stats); - ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics, - &datum); + free(stats); + } } 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 ***"; } @@ -1932,46 +2080,152 @@ refresh_controller_status(void) shash_find_data(&info, cfg->target); if (cinfo) { + struct smap smap = SMAP_INITIALIZER(&smap); + const char **values = cinfo->pairs.values; + const char **keys = cinfo->pairs.keys; + size_t i; + + for (i = 0; i < cinfo->pairs.n; i++) { + smap_add(&smap, keys[i], values[i]); + } + ovsrec_controller_set_is_connected(cfg, cinfo->is_connected); - ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role)); - ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys, - (char **) cinfo->pairs.values, - cinfo->pairs.n); + ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str( + cinfo->role)); + ovsrec_controller_set_status(cfg, &smap); + smap_destroy(&smap); } else { ovsrec_controller_set_is_connected(cfg, false); ovsrec_controller_set_role(cfg, NULL); - ovsrec_controller_set_status(cfg, NULL, NULL, 0); + ovsrec_controller_set_status(cfg, NULL); } } 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_cfm_stats(void) +instant_stats_run(void) { - static struct ovsdb_idl_txn *txn = NULL; + enum ovsdb_idl_txn_status status; - if (!txn) { + instant_stats_could_have_changed = true; + + 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; + + br_refresh_stp_status(br); + + HMAP_FOR_EACH (port, hmap_node, &br->ports) { + port_refresh_stp_status(port); + } HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) { + enum netdev_flags flags; + const char *link_state; + int64_t link_resets; + int current, error; + + if (iface_is_synthetic(iface)) { + continue; + } + + 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); + } + + error = netdev_get_flags(iface->netdev, &flags); + if (!error) { + const char *state = flags & NETDEV_UP ? "up" : "down"; + ovsrec_interface_set_admin_state(iface->cfg, state); + } else { + ovsrec_interface_set_admin_state(iface->cfg, NULL); + } + + 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); + iface_refresh_cfm_stats(iface); } } } - 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. * @@ -1981,8 +2235,17 @@ refresh_cfm_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); } @@ -1991,13 +2254,17 @@ 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(&null_cfg); + /* (Re)configure if necessary. */ if (!reconfiguring) { ovsdb_idl_run(idl); @@ -2019,6 +2286,20 @@ 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); @@ -2064,15 +2345,25 @@ 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; + } } } @@ -2082,8 +2373,22 @@ bridge_run(void) reconf_txn = NULL; } - /* Refresh system and interface stats if necessary. */ - if (time_msec() >= stats_timer) { + 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) { struct ovsdb_idl_txn *txn; @@ -2106,84 +2411,62 @@ bridge_run(void) } } - refresh_system_stats(cfg); refresh_controller_status(); ovsdb_idl_txn_commit(txn); ovsdb_idl_txn_destroy(txn); /* XXX */ } - stats_timer = time_msec() + STATS_INTERVAL; + iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL; } - if (time_msec() >= db_limiter) { - struct ovsdb_idl_txn *txn; - - 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) { - 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; - } - - 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 (ovsdb_idl_txn_commit(txn) != TXN_UNCHANGED) { - db_limiter = time_msec() + DB_LIMIT_INTERVAL; - } - ovsdb_idl_txn_destroy(txn); - } - - refresh_cfm_stats(); + run_system_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; HMAP_FOR_EACH (br, node, &all_bridges) { ofproto_wait(br->ofproto); } - poll_timer_wait_until(stats_timer); + poll_timer_wait_until(iface_stats_timer); + } - if (db_limiter > time_msec()) { - poll_timer_wait_until(db_limiter); - } + system_stats_wait(); + instant_stats_wait(); +} + +/* Adds some memory usage statistics for bridges into 'usage', for use with + * memory_report(). */ +void +bridge_get_memory_usage(struct simap *usage) +{ + struct bridge *br; + + HMAP_FOR_EACH (br, node, &all_bridges) { + ofproto_get_memory_usage(br->ofproto, usage); } } @@ -2196,14 +2479,14 @@ struct qos_unixctl_show_cbdata { static void qos_unixctl_show_cb(unsigned int queue_id, - const struct shash *details, + const struct smap *details, void *aux) { struct qos_unixctl_show_cbdata *data = aux; struct ds *ds = data->ds; struct iface *iface = data->iface; struct netdev_queue_stats stats; - struct shash_node *node; + struct smap_node *node; int error; ds_put_cstr(ds, "\n"); @@ -2213,8 +2496,8 @@ qos_unixctl_show_cb(unsigned int queue_id, ds_put_cstr(ds, "Default:\n"); } - SHASH_FOR_EACH (node, details) { - ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data); + SMAP_FOR_EACH (node, details) { + ds_put_format(ds, "\t%s: %s\n", node->key, node->value); } error = netdev_get_queue_stats(iface->netdev, queue_id, &stats); @@ -2241,10 +2524,10 @@ qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED, const char *argv[], void *aux OVS_UNUSED) { struct ds ds = DS_EMPTY_INITIALIZER; - struct shash sh = SHASH_INITIALIZER(&sh); + struct smap smap = SMAP_INITIALIZER(&smap); struct iface *iface; const char *type; - struct shash_node *node; + struct smap_node *node; struct qos_unixctl_show_cbdata data; int error; @@ -2254,13 +2537,13 @@ qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED, return; } - netdev_get_qos(iface->netdev, &type, &sh); + netdev_get_qos(iface->netdev, &type, &smap); if (*type != '\0') { ds_put_format(&ds, "QoS: %s %s\n", iface->name, type); - SHASH_FOR_EACH (node, &sh) { - ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data); + SMAP_FOR_EACH (node, &smap) { + ds_put_format(&ds, "%s: %s\n", node->key, node->value); } data.ds = &ds; @@ -2276,7 +2559,7 @@ qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED, unixctl_command_reply_error(conn, ds_cstr(&ds)); } - shash_destroy_free_data(&sh); + smap_destroy(&smap); ds_destroy(&ds); } @@ -2286,7 +2569,7 @@ 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); @@ -2435,6 +2718,7 @@ 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)); } @@ -2451,7 +2735,7 @@ 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); @@ -2467,6 +2751,9 @@ bridge_add_del_ports(struct bridge *br, VLOG_WARN("bridge %s: no port named %s, synthesizing one", br->name, br->name); + ovsrec_interface_init(&br->synth_local_iface); + ovsrec_port_init(&br->synth_local_port); + br->synth_local_port.interfaces = &br->synth_local_ifacep; br->synth_local_port.n_interfaces = 1; br->synth_local_port.name = br->name; @@ -2494,6 +2781,8 @@ bridge_add_del_ports(struct bridge *br, } } + /* Update iface->cfg and iface->type in interfaces that still exist. + * Add new interfaces to creation queue. */ SHASH_FOR_EACH (port_node, &new_ports) { const struct ovsrec_port *port = port_node->data; size_t i; @@ -2501,10 +2790,16 @@ bridge_add_del_ports(struct bridge *br, for (i = 0; i < port->n_interfaces; i++) { const struct ovsrec_interface *cfg = port->interfaces[i]; struct iface *iface = iface_lookup(br, cfg->name); + const char *type = iface_get_type(cfg, br->cfg); if (iface) { iface->cfg = cfg; - iface_refresh_type(iface); + iface->type = type; + } else if (!strcmp(type, "null")) { + VLOG_WARN_ONCE("%s: The null interface type is deprecated and" + " may be removed in February 2013. Please email" + " dev@openvswitch.org with concerns.", + cfg->name); } else { bridge_queue_if_cfg(br, cfg, port); } @@ -2536,7 +2831,7 @@ static void bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c, struct ofproto_controller *oc) { - const char *config_str; + int dscp; oc->target = c->target; oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8; @@ -2548,16 +2843,11 @@ bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c, ? *c->controller_burst_limit : 0); oc->enable_async_msgs = (!c->enable_async_messages || *c->enable_async_messages); - config_str = ovsrec_controller_get_other_config_value(c, "dscp", NULL); - - oc->dscp = DSCP_DEFAULT; - if (config_str) { - int dscp = atoi(config_str); - - if (dscp >= 0 && dscp <= 63) { - oc->dscp = dscp; - } + dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT); + if (dscp < 0 || dscp > 63) { + dscp = DSCP_DEFAULT; } + oc->dscp = dscp; } /* Configures the IP stack for 'br''s local interface properly according to the @@ -2590,7 +2880,7 @@ 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. */ @@ -2599,19 +2889,30 @@ 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') { @@ -2621,16 +2922,13 @@ equal_pathnames(const char *a, const char *b) b++; } } - return false; } static void bridge_configure_remotes(struct bridge *br, const struct sockaddr_in *managers, size_t n_managers) { - const char *disable_ib_str, *queue_id_str; - bool disable_in_band = false; - int queue_id; + bool disable_in_band; struct ovsrec_controller **controllers; size_t n_controllers; @@ -2642,19 +2940,13 @@ bridge_configure_remotes(struct bridge *br, size_t i; /* Check if we should disable in-band control on this bridge. */ - disable_ib_str = ovsrec_bridge_get_other_config_value(br->cfg, - "disable-in-band", - NULL); - if (disable_ib_str && !strcmp(disable_ib_str, "true")) { - disable_in_band = true; - } + disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band", + false); /* Set OpenFlow queue ID for in-band control. */ - queue_id_str = ovsrec_bridge_get_other_config_value(br->cfg, - "in-band-queue", - NULL); - queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1; - ofproto_set_in_band_queue(br->ofproto, queue_id); + ofproto_set_in_band_queue(br->ofproto, + smap_get_int(&br->cfg->other_config, + "in-band-queue", -1)); if (disable_in_band) { ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0); @@ -2676,21 +2968,42 @@ 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); @@ -2704,7 +3017,8 @@ 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); @@ -2786,6 +3100,13 @@ 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. */ @@ -2901,8 +3222,7 @@ port_configure_lacp(struct port *port, struct lacp_settings *s) s->name = port->name; - system_id = ovsrec_port_get_other_config_value(port->cfg, "lacp-system-id", - NULL); + system_id = smap_get(&port->cfg->other_config, "lacp-system-id"); if (system_id) { if (sscanf(system_id, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) { @@ -2920,16 +3240,14 @@ port_configure_lacp(struct port *port, struct lacp_settings *s) } /* Prefer bondable links if unspecified. */ - priority = atoi(ovsrec_port_get_other_config_value(port->cfg, - "lacp-system-priority", - "0")); + priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority", + 0); s->priority = (priority > 0 && priority <= UINT16_MAX ? priority : UINT16_MAX - !list_is_short(&port->ifaces)); - lacp_time = ovsrec_port_get_other_config_value(port->cfg, "lacp-time", - "slow"); - s->fast = !strcasecmp(lacp_time, "fast"); + lacp_time = smap_get(&port->cfg->other_config, "lacp-time"); + s->fast = lacp_time && !strcasecmp(lacp_time, "fast"); return s; } @@ -2938,16 +3256,10 @@ iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s) { int priority, portid, key; - portid = atoi(ovsrec_interface_get_other_config_value(iface->cfg, - "lacp-port-id", - "0")); - priority = - atoi(ovsrec_interface_get_other_config_value(iface->cfg, - "lacp-port-priority", - "0")); - key = atoi(ovsrec_interface_get_other_config_value(iface->cfg, - "lacp-aggregation-key", - "0")); + portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0); + priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority", + 0); + key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0); if (portid <= 0 || portid > UINT16_MAX) { portid = iface->ofp_port; @@ -2968,13 +3280,11 @@ 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; @@ -3000,17 +3310,14 @@ port_configure_bond(struct port *port, struct bond_settings *s, port->name); } - miimon_interval = - atoi(ovsrec_port_get_other_config_value(port->cfg, - "bond-miimon-interval", "0")); + miimon_interval = smap_get_int(&port->cfg->other_config, + "bond-miimon-interval", 0); if (miimon_interval <= 0) { miimon_interval = 200; } - detect_s = ovsrec_port_get_other_config_value(port->cfg, - "bond-detect-mode", - "carrier"); - if (!strcmp(detect_s, "carrier")) { + detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode"); + if (!detect_s || !strcmp(detect_s, "carrier")) { miimon_interval = 0; } else if (strcmp(detect_s, "miimon")) { VLOG_WARN("port %s: unsupported bond-detect-mode %s, " @@ -3020,32 +3327,16 @@ port_configure_bond(struct port *port, struct bond_settings *s, s->up_delay = MAX(0, port->cfg->bond_updelay); s->down_delay = MAX(0, port->cfg->bond_downdelay); - s->basis = atoi(ovsrec_port_get_other_config_value(port->cfg, - "bond-hash-basis", - "0")); - s->rebalance_interval = atoi( - ovsrec_port_get_other_config_value(port->cfg, - "bond-rebalance-interval", - "10000")); + s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0); + s->rebalance_interval = smap_get_int(&port->cfg->other_config, + "bond-rebalance-interval", 10000); if (s->rebalance_interval && s->rebalance_interval < 1000) { s->rebalance_interval = 1000; } 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(ovsrec_interface_get_other_config_value(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); } } @@ -3060,15 +3351,32 @@ port_is_synthetic(const struct port *port) /* Interface functions. */ -static void -iface_refresh_type(struct iface *iface) +static bool +iface_is_internal(const struct ovsrec_interface *iface, + const struct ovsrec_bridge *br) { - /* Determine interface type. The local port always has type - * "internal". Other ports take their type from the database and - * default to "system" if none is specified. */ - iface->type = (!strcmp(iface->name, iface->port->bridge->name) ? "internal" - : iface->cfg->type[0] ? iface->cfg->type - : "system"); + /* 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) +{ + 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 @@ -3199,12 +3507,12 @@ 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, NULL, 0); + 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); @@ -3214,54 +3522,6 @@ iface_clear_db_record(const struct ovsrec_interface *if_cfg) } } -/* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'. - * - * The value strings in '*shash' are taken directly from values[], not copied, - * so the caller should not modify or free them. */ -static void -shash_from_ovs_idl_map(char **keys, char **values, size_t n, - struct shash *shash) -{ - size_t i; - - shash_init(shash); - for (i = 0; i < n; i++) { - shash_add(shash, keys[i], values[i]); - } -} - -/* Creates 'keys' and 'values' arrays from 'shash'. - * - * Sets 'keys' and 'values' to heap allocated arrays representing the key-value - * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They - * are populated with with strings taken directly from 'shash' and thus have - * the same ownership of the key-value pairs in shash. - */ -static void -shash_to_ovs_idl_map(struct shash *shash, - char ***keys, char ***values, size_t *n) -{ - size_t i, count; - char **k, **v; - struct shash_node *sn; - - count = shash_count(shash); - - k = xmalloc(count * sizeof *k); - v = xmalloc(count * sizeof *v); - - i = 0; - SHASH_FOR_EACH(sn, shash) { - k[i] = sn->name; - v[i] = sn->data; - i++; - } - - *n = count; - *keys = k; - *values = v; -} - struct iface_delete_queues_cbdata { struct netdev *netdev; const struct ovsdb_datum *queues; @@ -3278,7 +3538,7 @@ queue_ids_include(const struct ovsdb_datum *queues, int64_t target) static void iface_delete_queues(unsigned int queue_id, - const struct shash *details OVS_UNUSED, void *cbdata_) + const struct smap *details OVS_UNUSED, void *cbdata_) { struct iface_delete_queues_cbdata *cbdata = cbdata_; @@ -3298,15 +3558,11 @@ iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) 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'. */ - shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config, - qos->n_other_config, &details); - netdev_set_qos(iface->netdev, qos->type, &details); - shash_destroy(&details); + netdev_set_qos(iface->netdev, qos->type, &qos->other_config); /* Deconfigure queues that were deleted. */ cbdata.netdev = iface->netdev; @@ -3333,16 +3589,14 @@ iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) port_queue->dscp = queue->dscp[0]; } - 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); + netdev_set_queue(iface->netdev, queue_id, &queue->other_config); } if (!queue_zero) { - shash_init(&details); + struct smap details; + + smap_init(&details); netdev_set_queue(iface->netdev, 0, &details); - shash_destroy(&details); + smap_destroy(&details); } } @@ -3365,30 +3619,39 @@ static void iface_configure_cfm(struct iface *iface) { const struct ovsrec_interface *cfg = iface->cfg; - const char *extended_str, *opstate_str; + 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 = atoi(ovsrec_interface_get_other_config_value(iface->cfg, - "cfm_interval", - "0")); - cfm_ccm_vlan = ovsrec_interface_get_other_config_value(iface->cfg, - "cfm_ccm_vlan", - "0"); - s.ccm_pcp = atoi(ovsrec_interface_get_other_config_value(iface->cfg, - "cfm_ccm_pcp", - "0")); + s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0); + cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan"); + s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0); + if (s.interval <= 0) { s.interval = 1000; } - if (!strcasecmp("random", cfm_ccm_vlan)) { + if (!cfm_ccm_vlan) { + s.ccm_vlan = 0; + } else if (!strcasecmp("random", cfm_ccm_vlan)) { s.ccm_vlan = CFM_RANDOM_VLAN; } else { s.ccm_vlan = atoi(cfm_ccm_vlan); @@ -3397,15 +3660,11 @@ iface_configure_cfm(struct iface *iface) } } - extended_str = ovsrec_interface_get_other_config_value(iface->cfg, - "cfm_extended", - "false"); - s.extended = !strcasecmp("true", extended_str); + s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended", + false); - opstate_str = ovsrec_interface_get_other_config_value(iface->cfg, - "cfm_opstate", - "up"); - s.opup = !strcasecmp("up", opstate_str); + opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate"); + s.opup = !opstate_str || !strcasecmp("up", opstate_str); ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s); } @@ -3418,6 +3677,13 @@ 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. */ @@ -3613,30 +3879,43 @@ mirror_configure(struct mirror *m) * devices are not used. When broken device drivers are no longer in * widespread use, we will delete these interfaces. */ -static void **blocks; -static size_t n_blocks, allocated_blocks; +static struct ovsrec_port **recs; +static size_t n_recs, allocated_recs; -/* Adds 'block' to a list of blocks that have to be freed with free() when the - * VLAN splinters are reconfigured. */ +/* Adds 'rec' to a list of recs that have to be destroyed when the VLAN + * splinters are reconfigured. */ static void -register_block(void *block) +register_rec(struct ovsrec_port *rec) { - if (n_blocks >= allocated_blocks) { - blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks); + if (n_recs >= allocated_recs) { + recs = x2nrealloc(recs, &allocated_recs, sizeof *recs); } - blocks[n_blocks++] = block; + recs[n_recs++] = rec; } -/* Frees all of the blocks registered with register_block(). */ +/* Frees all of the ports registered with register_reg(). */ static void -free_registered_blocks(void) +free_registered_recs(void) { size_t i; - for (i = 0; i < n_blocks; i++) { - free(blocks[i]); + for (i = 0; i < n_recs; i++) { + struct ovsrec_port *port = recs[i]; + size_t j; + + for (j = 0; j < port->n_interfaces; j++) { + struct ovsrec_interface *iface = port->interfaces[j]; + free(iface->name); + free(iface); + } + + smap_destroy(&port->other_config); + free(port->interfaces); + free(port->name); + free(port->tag); + free(port); } - n_blocks = 0; + n_recs = 0; } /* Returns true if VLAN splinters are enabled on 'iface_cfg', false @@ -3644,12 +3923,8 @@ free_registered_blocks(void) static bool vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg) { - const char *value; - - value = ovsrec_interface_get_other_config_value(iface_cfg, - "enable-vlan-splinters", - ""); - return !strcmp(value, "true"); + return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters", + false); } /* Figures out the set of VLANs that are in use for the purpose of VLAN @@ -3677,7 +3952,7 @@ collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg) /* Free space allocated for synthesized ports and interfaces, since we're * in the process of reconstructing all of them. */ - free_registered_blocks(); + free_registered_recs(); splinter_vlans = bitmap_allocate(4096); sset_init(&splinter_ifaces); @@ -3793,8 +4068,7 @@ configure_splinter_port(struct port *port) vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem); - realdev_name = ovsrec_port_get_other_config_value(port->cfg, - "realdev", NULL); + realdev_name = smap_get(&port->cfg->other_config, "realdev"); realdev = iface_lookup(port->bridge, realdev_name); realdev_ofp_port = realdev ? realdev->ofp_port : 0; @@ -3809,33 +4083,23 @@ synthesize_splinter_port(const char *real_dev_name, struct ovsrec_interface *iface; struct ovsrec_port *port; - iface = xzalloc(sizeof *iface); + iface = xmalloc(sizeof *iface); + ovsrec_interface_init(iface); iface->name = xstrdup(vlan_dev_name); iface->type = "system"; - port = xzalloc(sizeof *port); + port = xmalloc(sizeof *port); + ovsrec_port_init(port); port->interfaces = xmemdup(&iface, sizeof iface); port->n_interfaces = 1; port->name = xstrdup(vlan_dev_name); port->vlan_mode = "splinter"; port->tag = xmalloc(sizeof *port->tag); *port->tag = vid; - port->key_other_config = xmalloc(sizeof *port->key_other_config); - port->key_other_config[0] = "realdev"; - port->value_other_config = xmalloc(sizeof *port->value_other_config); - port->value_other_config[0] = xstrdup(real_dev_name); - port->n_other_config = 1; - - register_block(iface); - register_block(iface->name); - register_block(port); - register_block(port->interfaces); - register_block(port->name); - register_block(port->tag); - register_block(port->key_other_config); - register_block(port->value_other_config); - register_block(port->value_other_config[0]); + smap_add(&port->other_config, "realdev", real_dev_name); + + register_rec(port); return port; }