X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=53a24e36e1180c0e505342492bfa97e78e5cdfb0;hb=8828e5c714ed2b8af3158b0149d47cda1bdc69dc;hp=fc03a1226565c836b10eb8c12200d1e504b741dd;hpb=1a048029a0902439c3f17c49a7f71e4d09fecf70;p=sliver-openvswitch.git diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index fc03a1226..53a24e36e 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -15,6 +15,7 @@ #include #include "bridge.h" +#include "byte-order.h" #include #include #include @@ -37,6 +38,7 @@ #include "dynamic-string.h" #include "flow.h" #include "hash.h" +#include "hmap.h" #include "jsonrpc.h" #include "list.h" #include "mac-learning.h" @@ -49,7 +51,6 @@ #include "ovsdb-data.h" #include "packets.h" #include "poll-loop.h" -#include "port-array.h" #include "proc-net-compat.h" #include "process.h" #include "sha1.h" @@ -57,6 +58,7 @@ #include "socket-util.h" #include "stream-ssl.h" #include "svec.h" +#include "system-stats.h" #include "timeval.h" #include "util.h" #include "unixctl.h" @@ -64,10 +66,9 @@ #include "vswitchd/vswitch-idl.h" #include "xenserver.h" #include "vlog.h" -#include "xtoxll.h" #include "sflow_api.h" -VLOG_DEFINE_THIS_MODULE(bridge) +VLOG_DEFINE_THIS_MODULE(bridge); struct dst { uint16_t vlan; @@ -84,9 +85,11 @@ struct iface { /* These members are valid only after bridge_reconfigure() causes them to * be initialized. */ + struct hmap_node dp_ifidx_node; /* In struct bridge's "ifaces" hmap. */ int dp_ifidx; /* Index within kernel datapath. */ struct netdev *netdev; /* Network device. */ bool enabled; /* May be chosen for flows? */ + const char *type; /* Usually same as cfg->type. */ const struct ovsrec_interface *cfg; }; @@ -141,9 +144,10 @@ struct port { int updelay, downdelay; /* Delay before iface goes up/down, in ms. */ bool bond_compat_is_stale; /* Need to call port_update_bond_compat()? */ bool bond_fake_iface; /* Fake a bond interface for legacy compat? */ - long bond_next_fake_iface_update; /* Next update to fake bond stats. */ + long long int bond_next_fake_iface_update; /* Time of next update. */ int bond_rebalance_interval; /* Interval between rebalances, in ms. */ long long int bond_next_rebalance; /* Next rebalancing time. */ + struct netdev_monitor *monitor; /* Tracks carrier up/down status. */ /* Port mirroring info. */ mirror_mask_t src_mirrors; /* Mirrors triggered when packet received. */ @@ -164,7 +168,7 @@ struct bridge { /* Kernel datapath information. */ struct dpif *dpif; /* Datapath. */ - struct port_array ifaces; /* Indexed by kernel datapath port number. */ + struct hmap ifaces; /* Contains "struct iface"s. */ /* Bridge ports. */ struct port **ports; @@ -188,10 +192,10 @@ static struct list all_bridges = LIST_INITIALIZER(&all_bridges); /* OVSDB IDL used to obtain configuration. */ static struct ovsdb_idl *idl; -/* Each time this timer expires, the bridge fetches statistics for every - * interface and pushes them into the database. */ -#define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */ -static long long int iface_stats_timer = LLONG_MIN; +/* Each time this timer expires, the bridge fetches systems and interface + * statistics and pushes them into the database. */ +#define STATS_INTERVAL (5 * 1000) /* In milliseconds. */ +static long long int stats_timer = LLONG_MIN; static struct bridge *bridge_create(const struct ovsrec_bridge *br_cfg); static void bridge_destroy(struct bridge *); @@ -244,16 +248,19 @@ static void mirror_reconfigure(struct bridge *); static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *); static bool vlan_is_mirrored(const struct mirror *, int vlan); -static struct iface *iface_create(struct port *port, +static struct iface *iface_create(struct port *port, const struct ovsrec_interface *if_cfg); static void iface_destroy(struct iface *); static struct iface *iface_lookup(const struct bridge *, const char *name); static struct iface *iface_from_dp_ifidx(const struct bridge *, uint16_t dp_ifidx); -static bool iface_is_internal(const struct bridge *, const char *name); static void iface_set_mac(struct iface *); +static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport); static void iface_update_qos(struct iface *, const struct ovsrec_qos *); +static void shash_from_ovs_idl_map(char **keys, char **values, size_t n, + struct shash *); + /* Hooks into ofproto processing. */ static struct ofhooks bridge_ofhooks; @@ -268,6 +275,19 @@ bridge_init(const char *remote) /* Create connection to database. */ idl = ovsdb_idl_create(remote, &ovsrec_idl_class); + ovsdb_idl_set_write_only(idl, &ovsrec_open_vswitch_col_cur_cfg); + ovsdb_idl_set_write_only(idl, &ovsrec_open_vswitch_col_statistics); + ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids); + + ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids); + + ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids); + ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge); + + ovsdb_idl_set_write_only(idl, &ovsrec_interface_col_ofport); + ovsdb_idl_set_write_only(idl, &ovsrec_interface_col_statistics); + ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids); + /* Register unixctl commands. */ unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL); unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows, @@ -293,7 +313,7 @@ bridge_configure_once(const struct ovsrec_open_vswitch *cfg) } already_configured_once = true; - iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL; + stats_timer = time_msec() + STATS_INTERVAL; /* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */ svec_init(&bridge_names); @@ -344,115 +364,116 @@ bridge_configure_once(const struct ovsrec_open_vswitch *cfg) svec_destroy(&dpif_types); } -#ifdef HAVE_OPENSSL +/* Initializes 'options' and fills it with the options for 'if_cfg'. Merges + * keys from "options" and "other_config", preferring "options" keys over + * "other_config" keys. + * + * The value strings in '*options' are taken directly from if_cfg, not copied, + * so the caller should not modify or free them. */ static void -bridge_configure_ssl(const struct ovsrec_ssl *ssl) +iface_get_options(const struct ovsrec_interface *if_cfg, struct shash *options) { - /* XXX SSL should be configurable on a per-bridge basis. */ - if (ssl) { - stream_ssl_set_private_key_file(ssl->private_key); - stream_ssl_set_certificate_file(ssl->certificate); - stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert); + size_t i; + + shash_from_ovs_idl_map(if_cfg->key_options, if_cfg->value_options, + if_cfg->n_options, options); + + for (i = 0; i < if_cfg->n_other_config; i++) { + char *key = if_cfg->key_other_config[i]; + char *value = if_cfg->value_other_config[i]; + + if (!shash_find_data(options, key)) { + shash_add(options, key, value); + } else { + VLOG_WARN("%s: ignoring \"other_config\" key %s that conflicts " + "with \"options\" key %s", if_cfg->name, key, key); + } } } -#endif -/* Attempt to create the network device 'iface_name' through the netdev +/* Returns the type of network device that 'iface' should have. (This is + * ordinarily the same type as the interface, but the network devices for + * "internal" ports have type "system".) */ +static const char * +iface_get_netdev_type(const struct iface *iface) +{ + return !strcmp(iface->type, "internal") ? "system" : iface->type; +} + +/* Attempt to create the network device for 'iface' through the netdev * library. */ static int -set_up_iface(const struct ovsrec_interface *iface_cfg, struct iface *iface, - bool create) +create_iface_netdev(struct iface *iface) { + struct netdev_options netdev_options; struct shash options; - int error = 0; - size_t i; - - shash_init(&options); - for (i = 0; i < iface_cfg->n_options; i++) { - shash_add(&options, iface_cfg->key_options[i], - xstrdup(iface_cfg->value_options[i])); - } - - if (create) { - struct netdev_options netdev_options; + int error; - memset(&netdev_options, 0, sizeof netdev_options); - netdev_options.name = iface_cfg->name; - if (!strcmp(iface_cfg->type, "internal")) { - /* An "internal" config type maps to a netdev "system" type. */ - netdev_options.type = "system"; - } else { - netdev_options.type = iface_cfg->type; - } - netdev_options.args = &options; - netdev_options.ethertype = NETDEV_ETH_TYPE_NONE; + memset(&netdev_options, 0, sizeof netdev_options); + netdev_options.name = iface->cfg->name; + netdev_options.type = iface_get_netdev_type(iface); + netdev_options.args = &options; + netdev_options.ethertype = NETDEV_ETH_TYPE_NONE; - error = netdev_open(&netdev_options, &iface->netdev); + iface_get_options(iface->cfg, &options); - if (iface->netdev) { - netdev_get_carrier(iface->netdev, &iface->enabled); - } - } else if (iface->netdev) { - const char *netdev_type = netdev_get_type(iface->netdev); - const char *iface_type = iface_cfg->type && strlen(iface_cfg->type) - ? iface_cfg->type : NULL; - - /* An "internal" config type maps to a netdev "system" type. */ - if (iface_type && !strcmp(iface_type, "internal")) { - iface_type = "system"; - } + error = netdev_open(&netdev_options, &iface->netdev); - if (!iface_type || !strcmp(netdev_type, iface_type)) { - error = netdev_reconfigure(iface->netdev, &options); - } else { - VLOG_WARN("%s: attempting change device type from %s to %s", - iface_cfg->name, netdev_type, iface_type); - error = EINVAL; - } + if (iface->netdev) { + iface->enabled = netdev_get_carrier(iface->netdev); } - shash_destroy_free_data(&options); + + shash_destroy(&options); return error; } static int -reconfigure_iface(const struct ovsrec_interface *iface_cfg, struct iface *iface) +reconfigure_iface_netdev(struct iface *iface) { - return set_up_iface(iface_cfg, iface, false); -} + const char *netdev_type, *iface_type; + struct shash options; + int error; -static bool -check_iface_netdev(struct bridge *br OVS_UNUSED, struct iface *iface, - void *aux OVS_UNUSED) -{ - if (!iface->netdev) { - int error = set_up_iface(iface->cfg, iface, true); - if (error) { - VLOG_WARN("could not open netdev on %s, dropping: %s", iface->name, - strerror(error)); - return false; - } + /* Skip reconfiguration if the device has the wrong type. This shouldn't + * happen, but... */ + iface_type = iface_get_netdev_type(iface); + netdev_type = netdev_get_type(iface->netdev); + if (iface_type && strcmp(netdev_type, iface_type)) { + VLOG_WARN("%s: attempting change device type from %s to %s", + iface->cfg->name, netdev_type, iface_type); + return EINVAL; } - return true; + /* Reconfigure device. */ + iface_get_options(iface->cfg, &options); + error = netdev_reconfigure(iface->netdev, &options); + shash_destroy(&options); + + return error; } +/* Callback for iterate_and_prune_ifaces(). */ static bool -check_iface_dp_ifidx(struct bridge *br, struct iface *iface, - void *aux OVS_UNUSED) +check_iface(struct bridge *br, struct iface *iface, void *aux OVS_UNUSED) { - if (iface->dp_ifidx >= 0) { - VLOG_DBG("%s has interface %s on port %d", - dpif_name(br->dpif), - iface->name, iface->dp_ifidx); - return true; - } else { + if (!iface->netdev) { + /* We already reported a related error, don't bother duplicating it. */ + return false; + } + + if (iface->dp_ifidx < 0) { VLOG_ERR("%s interface not in %s, dropping", iface->name, dpif_name(br->dpif)); return false; } + + VLOG_DBG("%s has interface %s on port %d", dpif_name(br->dpif), + iface->name, iface->dp_ifidx); + return true; } +/* Callback for iterate_and_prune_ifaces(). */ static bool set_iface_properties(struct bridge *br OVS_UNUSED, struct iface *iface, void *aux OVS_UNUSED) @@ -464,8 +485,7 @@ set_iface_properties(struct bridge *br OVS_UNUSED, struct iface *iface, /* Set MAC address of internal interfaces other than the local * interface. */ - if (iface->dp_ifidx != ODPP_LOCAL - && iface_is_internal(br, iface->name)) { + if (iface->dp_ifidx != ODPP_LOCAL && !strcmp(iface->type, "internal")) { iface_set_mac(iface); } @@ -490,6 +510,7 @@ iterate_and_prune_ifaces(struct bridge *br, if (cb(br, iface, aux)) { j++; } else { + iface_set_ofport(iface->cfg, -1); iface_destroy(iface); } } @@ -559,7 +580,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) /* Collect old and new bridges. */ shash_init(&old_br); shash_init(&new_br); - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { shash_add(&old_br, br->name, br); } for (i = 0; i < ovs_cfg->n_bridges; i++) { @@ -570,7 +591,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) } /* Get rid of deleted bridges and add new bridges. */ - LIST_FOR_EACH_SAFE (br, next, struct bridge, node, &all_bridges) { + LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) { struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name); if (br_cfg) { br->cfg = br_cfg; @@ -596,13 +617,8 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) shash_destroy(&old_br); shash_destroy(&new_br); -#ifdef HAVE_OPENSSL - /* Configure SSL. */ - bridge_configure_ssl(ovs_cfg->ssl); -#endif - /* Reconfigure all bridges. */ - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { bridge_reconfigure_one(br); } @@ -611,7 +627,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) * The kernel will reject any attempt to add a given port to a datapath if * that port already belongs to a different datapath, so we must do all * port deletions before any port additions. */ - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { struct odp_port *dpif_ports; size_t n_dpif_ports; struct shash want_ifaces; @@ -633,40 +649,68 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) shash_destroy(&want_ifaces); free(dpif_ports); } - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { struct odp_port *dpif_ports; size_t n_dpif_ports; struct shash cur_ifaces, want_ifaces; - struct shash_node *node; /* Get the set of interfaces currently in this datapath. */ dpif_port_list(br->dpif, &dpif_ports, &n_dpif_ports); shash_init(&cur_ifaces); for (i = 0; i < n_dpif_ports; i++) { const char *name = dpif_ports[i].devname; - shash_add_once(&cur_ifaces, name, NULL); + shash_add_once(&cur_ifaces, name, &dpif_ports[i]); } - free(dpif_ports); /* Get the set of interfaces we want on this datapath. */ bridge_get_all_ifaces(br, &want_ifaces); + hmap_clear(&br->ifaces); SHASH_FOR_EACH (node, &want_ifaces) { const char *if_name = node->name; struct iface *iface = node->data; - - if (shash_find(&cur_ifaces, if_name)) { - /* Already exists, just reconfigure it. */ + bool internal = !iface || !strcmp(iface->type, "internal"); + struct odp_port *dpif_port = shash_find_data(&cur_ifaces, if_name); + int error; + + /* If we have a port or a netdev already, and it's not the type we + * want, then delete the port (if any) and close the netdev (if + * any). */ + if (internal + ? dpif_port && !(dpif_port->flags & ODP_PORT_INTERNAL) + : (iface->netdev + && strcmp(iface->type, netdev_get_type(iface->netdev)))) + { + if (dpif_port) { + error = ofproto_port_del(br->ofproto, dpif_port->port); + if (error) { + continue; + } + dpif_port = NULL; + } if (iface) { - reconfigure_iface(iface->cfg, iface); + netdev_close(iface->netdev); + iface->netdev = NULL; } - } else { - /* Need to add to datapath. */ - bool internal; - int error; + } - /* Add to datapath. */ - internal = iface_is_internal(br, if_name); + /* If it's not an internal port, open (possibly create) the + * netdev. */ + if (!internal) { + if (!iface->netdev) { + error = create_iface_netdev(iface); + if (error) { + VLOG_WARN("could not create iface %s: %s", iface->name, + strerror(error)); + continue; + } + } else { + reconfigure_iface_netdev(iface); + } + } + + /* If it's not part of the datapath, add it. */ + if (!dpif_port) { error = dpif_port_add(br->dpif, if_name, internal ? ODP_PORT_INTERNAL : 0, NULL); if (error == EFBIG) { @@ -676,14 +720,30 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) } else if (error) { VLOG_ERR("failed to add %s interface to %s: %s", if_name, dpif_name(br->dpif), strerror(error)); + continue; + } + } + + /* If it's an internal port, open the netdev. */ + if (internal) { + if (iface && !iface->netdev) { + error = create_iface_netdev(iface); + if (error) { + VLOG_WARN("could not create iface %s: %s", iface->name, + strerror(error)); + continue; + } } + } else { + assert(iface->netdev != NULL); } } + free(dpif_ports); shash_destroy(&cur_ifaces); shash_destroy(&want_ifaces); } sflow_bridge_number = 0; - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { uint8_t ea[8]; uint64_t dpid; struct iface *local_iface; @@ -692,8 +752,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) bridge_fetch_dp_ifaces(br); - iterate_and_prune_ifaces(br, check_iface_netdev, NULL); - iterate_and_prune_ifaces(br, check_iface_dp_ifidx, NULL); + iterate_and_prune_ifaces(br, check_iface, NULL); /* Pick local port hardware address, datapath ID. */ bridge_pick_local_hw_addr(br, ea, &hw_addr_iface); @@ -757,7 +816,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) opts.collectors.n = nf_cfg->n_targets; opts.collectors.names = nf_cfg->targets; if (ofproto_set_netflow(br->ofproto, &opts)) { - VLOG_ERR("bridge %s: problem setting netflow collectors", + VLOG_ERR("bridge %s: problem setting netflow collectors", br->name); } } else { @@ -770,7 +829,6 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) struct ovsrec_controller **controllers; struct ofproto_sflow_options oso; size_t n_controllers; - size_t i; memset(&oso, 0, sizeof oso); @@ -821,7 +879,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) * the datapath ID before the controller. */ bridge_reconfigure_remotes(br, managers, n_managers); } - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { for (i = 0; i < br->n_ports; i++) { struct port *port = br->ports[i]; int j; @@ -834,7 +892,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) } } } - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { iterate_and_prune_ifaces(br, set_iface_properties, NULL); } @@ -1103,15 +1161,32 @@ iface_refresh_stats(struct iface *iface) ovsrec_interface_set_statistics(iface->cfg, keys, values, n); } +static void +refresh_system_stats(const struct ovsrec_open_vswitch *cfg) +{ + struct ovsdb_datum datum; + struct shash stats; + + shash_init(&stats); + get_system_stats(&stats); + + ovsdb_datum_from_shash(&datum, &stats); + ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics, + &datum); +} + void bridge_run(void) { + const struct ovsrec_open_vswitch *cfg; + bool datapath_destroyed; + bool database_changed; struct bridge *br; /* Let each bridge do the work that it needs to do. */ datapath_destroyed = false; - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { int error = bridge_run_one(br); if (error) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); @@ -1122,8 +1197,9 @@ bridge_run(void) } /* (Re)configure if necessary. */ - if (ovsdb_idl_run(idl) || datapath_destroyed) { - const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl); + database_changed = ovsdb_idl_run(idl); + cfg = ovsrec_open_vswitch_first(idl); + if (database_changed || datapath_destroyed) { if (cfg) { struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl); @@ -1142,28 +1218,43 @@ bridge_run(void) } } - /* Refresh interface stats if necessary. */ - if (time_msec() >= iface_stats_timer) { - struct ovsdb_idl_txn *txn; +#ifdef HAVE_OPENSSL + /* Re-configure SSL. We do this on every trip through the main loop, + * instead of just when the database changes, because the contents of the + * key and certificate files can change without the database changing. */ + if (cfg && cfg->ssl) { + const struct ovsrec_ssl *ssl = cfg->ssl; - txn = ovsdb_idl_txn_create(idl); - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { - size_t i; + stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate); + stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert); + } +#endif - for (i = 0; i < br->n_ports; i++) { - struct port *port = br->ports[i]; - size_t j; + /* Refresh system and interface stats if necessary. */ + if (time_msec() >= stats_timer) { + if (cfg) { + struct ovsdb_idl_txn *txn; - for (j = 0; j < port->n_ifaces; j++) { - struct iface *iface = port->ifaces[j]; - iface_refresh_stats(iface); + txn = ovsdb_idl_txn_create(idl); + LIST_FOR_EACH (br, node, &all_bridges) { + size_t i; + + for (i = 0; i < br->n_ports; i++) { + struct port *port = br->ports[i]; + size_t j; + + for (j = 0; j < port->n_ifaces; j++) { + struct iface *iface = port->ifaces[j]; + iface_refresh_stats(iface); + } } } + refresh_system_stats(cfg); + ovsdb_idl_txn_commit(txn); + ovsdb_idl_txn_destroy(txn); /* XXX */ } - ovsdb_idl_txn_commit(txn); - ovsdb_idl_txn_destroy(txn); /* XXX */ - iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL; + stats_timer = time_msec() + STATS_INTERVAL; } } @@ -1172,9 +1263,9 @@ bridge_wait(void) { struct bridge *br; - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { ofproto_wait(br->ofproto); - if (ofproto_has_controller(br->ofproto)) { + if (ofproto_has_primary_controller(br->ofproto)) { continue; } @@ -1182,7 +1273,7 @@ bridge_wait(void) bond_wait(br); } ovsdb_idl_wait(idl); - poll_timer_wait_until(iface_stats_timer); + poll_timer_wait_until(stats_timer); } /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s @@ -1231,7 +1322,7 @@ bridge_unixctl_fdb_show(struct unixctl_conn *conn, } ds_put_cstr(&ds, " port VLAN MAC Age\n"); - LIST_FOR_EACH (e, struct mac_entry, lru_node, &br->ml->lrus) { + LIST_FOR_EACH (e, lru_node, &br->ml->lrus) { if (e->port < 0 || e->port >= br->n_ports) { continue; } @@ -1277,7 +1368,7 @@ bridge_create(const struct ovsrec_bridge *br_cfg) br->ml = mac_learning_create(); eth_addr_nicira_random(br->default_ea); - port_array_init(&br->ifaces); + hmap_init(&br->ifaces); shash_init(&br->port_by_name); shash_init(&br->iface_by_name); @@ -1309,7 +1400,7 @@ bridge_destroy(struct bridge *br) dpif_close(br->dpif); ofproto_destroy(br->ofproto); mac_learning_destroy(br->ml); - port_array_destroy(&br->ifaces); + hmap_destroy(&br->ifaces); shash_destroy(&br->port_by_name); shash_destroy(&br->iface_by_name); free(br->ports); @@ -1323,7 +1414,7 @@ bridge_lookup(const char *name) { struct bridge *br; - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { if (!strcmp(br->name, name)) { return br; } @@ -1339,7 +1430,7 @@ bridge_unixctl_dump_flows(struct unixctl_conn *conn, { struct bridge *br; struct ds results; - + br = bridge_lookup(args); if (!br) { unixctl_command_reply(conn, 501, "Unknown bridge"); @@ -1369,7 +1460,7 @@ bridge_unixctl_reconnect(struct unixctl_conn *conn, } ofproto_reconnect_controllers(br->ofproto); } else { - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { ofproto_reconnect_controllers(br->ofproto); } } @@ -1420,9 +1511,9 @@ static void bridge_reconfigure_one(struct bridge *br) { struct shash old_ports, new_ports; - struct svec listeners, old_listeners; struct svec snoops, old_snoops; struct shash_node *node; + enum ofproto_fail_mode fail_mode; size_t i; /* Collect old ports. */ @@ -1491,22 +1582,21 @@ bridge_reconfigure_one(struct bridge *br) shash_destroy(&old_ports); shash_destroy(&new_ports); + /* Set the fail-mode */ + fail_mode = !br->cfg->fail_mode + || !strcmp(br->cfg->fail_mode, "standalone") + ? OFPROTO_FAIL_STANDALONE + : OFPROTO_FAIL_SECURE; + if (ofproto_get_fail_mode(br->ofproto) != fail_mode + && !ofproto_has_primary_controller(br->ofproto)) { + ofproto_flush_flows(br->ofproto); + } + ofproto_set_fail_mode(br->ofproto, fail_mode); + /* Delete all flows if we're switching from connected to standalone or vice * versa. (XXX Should we delete all flows if we are switching from one * controller to another?) */ - /* Configure OpenFlow management listener. */ - svec_init(&listeners); - svec_add_nocopy(&listeners, xasprintf("punix:%s/%s.mgmt", - ovs_rundir, br->name)); - svec_init(&old_listeners); - ofproto_get_listeners(br->ofproto, &old_listeners); - if (!svec_equal(&listeners, &old_listeners)) { - ofproto_set_listeners(br->ofproto, &listeners); - } - svec_destroy(&listeners); - svec_destroy(&old_listeners); - /* Configure OpenFlow controller connection snooping. */ svec_init(&snoops); svec_add_nocopy(&snoops, xasprintf("punix:%s/%s.snoop", @@ -1522,6 +1612,90 @@ bridge_reconfigure_one(struct bridge *br) mirror_reconfigure(br); } +/* Initializes 'oc' appropriately as a management service controller for + * 'br'. + * + * The caller must free oc->target when it is no longer needed. */ +static void +bridge_ofproto_controller_for_mgmt(const struct bridge *br, + struct ofproto_controller *oc) +{ + oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir, br->name); + oc->max_backoff = 0; + oc->probe_interval = 60; + oc->band = OFPROTO_OUT_OF_BAND; + oc->accept_re = NULL; + oc->update_resolv_conf = false; + oc->rate_limit = 0; + oc->burst_limit = 0; +} + +/* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */ +static void +bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c, + struct ofproto_controller *oc) +{ + oc->target = c->target; + oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8; + oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5; + oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band") + ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND); + oc->accept_re = c->discover_accept_regex; + oc->update_resolv_conf = c->discover_update_resolv_conf; + oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0; + oc->burst_limit = (c->controller_burst_limit + ? *c->controller_burst_limit : 0); +} + +/* Configures the IP stack for 'br''s local interface properly according to the + * configuration in 'c'. */ +static void +bridge_configure_local_iface_netdev(struct bridge *br, + struct ovsrec_controller *c) +{ + struct netdev *netdev; + struct in_addr mask, gateway; + + struct iface *local_iface; + struct in_addr ip; + + /* Controller discovery does its own TCP/IP configuration later. */ + if (strcmp(c->target, "discover")) { + return; + } + + /* If there's no local interface or no IP address, give up. */ + local_iface = bridge_get_local_iface(br); + if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) { + return; + } + + /* Bring up the local interface. */ + netdev = local_iface->netdev; + netdev_turn_flags_on(netdev, NETDEV_UP, true); + + /* Configure the IP address and netmask. */ + if (!c->local_netmask + || !inet_aton(c->local_netmask, &mask) + || !mask.s_addr) { + mask.s_addr = guess_netmask(ip.s_addr); + } + 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)); + } + + /* Configure the default gateway. */ + if (c->local_gateway + && inet_aton(c->local_gateway, &gateway) + && gateway.s_addr) { + if (!netdev_add_router(netdev, gateway)) { + VLOG_INFO("bridge %s: configured gateway "IP_FMT, + br->name, IP_ARGS(&gateway.s_addr)); + } + } +} + static void bridge_reconfigure_remotes(struct bridge *br, const struct sockaddr_in *managers, @@ -1529,99 +1703,81 @@ bridge_reconfigure_remotes(struct bridge *br, { struct ovsrec_controller **controllers; size_t n_controllers; + bool had_primary; + const char *disable_ib_str; + bool disable_in_band = false; + + struct ofproto_controller *ocs; + size_t n_ocs; + size_t i; + + + /* Check if we should disable in-band control on this bridge. */ + disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band"); + if (disable_ib_str && !strcmp(disable_ib_str, "true")) { + disable_in_band = true; + } - ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers); + if (disable_in_band) { + ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0); + } else { + ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers); + } + had_primary = ofproto_has_primary_controller(br->ofproto); n_controllers = bridge_get_controllers(br, &controllers); - if (ofproto_has_controller(br->ofproto) != (n_controllers != 0)) { + + ocs = xmalloc((n_controllers + 1) * sizeof *ocs); + n_ocs = 0; + + bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]); + for (i = 0; i < n_controllers; i++) { + struct ovsrec_controller *c = controllers[i]; + + if (!strncmp(c->target, "punix:", 6) + || !strncmp(c->target, "unix:", 5)) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + + /* Prevent remote ovsdb-server users from accessing arbitrary Unix + * domain sockets and overwriting arbitrary local files. */ + VLOG_ERR_RL(&rl, "%s: not adding Unix domain socket controller " + "\"%s\" due to possibility for remote exploit", + dpif_name(br->dpif), c->target); + continue; + } + + bridge_configure_local_iface_netdev(br, c); + bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]); + if (disable_in_band) { + ocs[n_ocs].band = OFPROTO_OUT_OF_BAND; + } + n_ocs++; + } + + ofproto_set_controllers(br->ofproto, ocs, n_ocs); + free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */ + free(ocs); + + if (had_primary != ofproto_has_primary_controller(br->ofproto)) { ofproto_flush_flows(br->ofproto); } - if (!n_controllers) { + /* If there are no controllers and the bridge is in standalone + * mode, set up a flow that matches every packet and directs + * them to OFPP_NORMAL (which goes to us). Otherwise, the + * switch is in secure mode and we won't pass any traffic until + * a controller has been defined and it tells us to do so. */ + if (!n_controllers + && ofproto_get_fail_mode(br->ofproto) == OFPROTO_FAIL_STANDALONE) { union ofp_action action; - flow_t flow; - - /* Clear out controllers. */ - ofproto_set_controllers(br->ofproto, NULL, 0); + struct flow flow; - /* Set up a flow that matches every packet and directs them to - * OFPP_NORMAL (which goes to us). */ memset(&action, 0, sizeof action); action.type = htons(OFPAT_OUTPUT); action.output.len = htons(sizeof action); action.output.port = htons(OFPP_NORMAL); memset(&flow, 0, sizeof flow); ofproto_add_flow(br->ofproto, &flow, OVSFW_ALL, 0, &action, 1, 0); - } else { - struct ofproto_controller *ocs; - size_t i; - - ocs = xmalloc(n_controllers * sizeof *ocs); - for (i = 0; i < n_controllers; i++) { - struct ovsrec_controller *c = controllers[i]; - struct ofproto_controller *oc = &ocs[i]; - - if (strcmp(c->target, "discover")) { - struct iface *local_iface; - struct in_addr ip; - - local_iface = bridge_get_local_iface(br); - if (local_iface && c->local_ip - && inet_aton(c->local_ip, &ip)) { - struct netdev *netdev = local_iface->netdev; - struct in_addr mask, gateway; - - if (!c->local_netmask - || !inet_aton(c->local_netmask, &mask)) { - mask.s_addr = 0; - } - if (!c->local_gateway - || !inet_aton(c->local_gateway, &gateway)) { - gateway.s_addr = 0; - } - - netdev_turn_flags_on(netdev, NETDEV_UP, true); - if (!mask.s_addr) { - mask.s_addr = guess_netmask(ip.s_addr); - } - 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)); - } - - if (gateway.s_addr) { - if (!netdev_add_router(netdev, gateway)) { - VLOG_INFO("bridge %s: configured gateway "IP_FMT, - br->name, IP_ARGS(&gateway.s_addr)); - } - } - } - } - - oc->target = c->target; - oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8; - oc->probe_interval = (c->inactivity_probe - ? *c->inactivity_probe / 1000 : 5); - oc->fail = (!c->fail_mode - || !strcmp(c->fail_mode, "standalone") - || !strcmp(c->fail_mode, "open") - ? OFPROTO_FAIL_STANDALONE - : OFPROTO_FAIL_SECURE); - oc->band = (!c->connection_mode - || !strcmp(c->connection_mode, "in-band") - ? OFPROTO_IN_BAND - : OFPROTO_OUT_OF_BAND); - oc->accept_re = c->discover_accept_regex; - oc->update_resolv_conf = c->discover_update_resolv_conf; - oc->rate_limit = (c->controller_rate_limit - ? *c->controller_rate_limit : 0); - oc->burst_limit = (c->controller_burst_limit - ? *c->controller_burst_limit : 0); - } - ofproto_set_controllers(br->ofproto, ocs, n_controllers); - free(ocs); } } @@ -1666,7 +1822,7 @@ bridge_fetch_dp_ifaces(struct bridge *br) iface->dp_ifidx = -1; } } - port_array_clear(&br->ifaces); + hmap_clear(&br->ifaces); dpif_port_list(br->dpif, &dpif_ports, &n_dpif_ports); for (i = 0; i < n_dpif_ports; i++) { @@ -1680,16 +1836,15 @@ bridge_fetch_dp_ifaces(struct bridge *br) VLOG_WARN("%s reported interface %"PRIu16" twice", dpif_name(br->dpif), p->port); } else { - port_array_set(&br->ifaces, p->port, iface); iface->dp_ifidx = p->port; + hmap_insert(&br->ifaces, &iface->dp_ifidx_node, + hash_int(iface->dp_ifidx, 0)); } - if (iface->cfg) { - int64_t ofport = (iface->dp_ifidx >= 0 - ? odp_port_to_ofp_port(iface->dp_ifidx) - : -1); - ovsrec_interface_set_ofport(iface->cfg, &ofport, 1); - } + iface_set_ofport(iface->cfg, + (iface->dp_ifidx >= 0 + ? odp_port_to_ofp_port(iface->dp_ifidx) + : -1)); } } free(dpif_ports); @@ -1917,6 +2072,22 @@ bond_run(struct bridge *br) struct port *port = br->ports[i]; if (port->n_ifaces >= 2) { + char *devname; + + /* Track carrier going up and down on interfaces. */ + while (!netdev_monitor_poll(port->monitor, &devname)) { + struct iface *iface; + + iface = port_lookup_iface(port, devname); + if (iface) { + bool carrier = netdev_get_carrier(iface->netdev); + + bond_link_status_update(iface, carrier); + port_update_bond_compat(port); + } + free(devname); + } + for (j = 0; j < port->n_ifaces; j++) { struct iface *iface = port->ifaces[j]; if (time_msec() >= iface->delay_expires) { @@ -1948,6 +2119,7 @@ bond_wait(struct bridge *br) if (port->n_ifaces < 2) { continue; } + netdev_monitor_poll_wait(port->monitor); for (j = 0; j < port->n_ifaces; j++) { struct iface *iface = port->ifaces[j]; if (iface->delay_expires != LLONG_MAX) { @@ -1961,7 +2133,7 @@ bond_wait(struct bridge *br) } static bool -set_dst(struct dst *p, const flow_t *flow, +set_dst(struct dst *p, const struct flow *flow, const struct port *in_port, const struct port *out_port, tag_type *tags) { @@ -2049,8 +2221,22 @@ port_includes_vlan(const struct port *port, uint16_t vlan) return vlan == port->vlan || port_trunks_vlan(port, vlan); } +static bool +port_is_floodable(const struct port *port) +{ + int i; + + for (i = 0; i < port->n_ifaces; i++) { + if (!ofproto_port_is_floodable(port->bridge->ofproto, + port->ifaces[i]->dp_ifidx)) { + return false; + } + } + return true; +} + static size_t -compose_dsts(const struct bridge *br, const flow_t *flow, uint16_t vlan, +compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan, const struct port *in_port, const struct port *out_port, struct dst dsts[], tag_type *tags, uint16_t *nf_output_iface) { @@ -2063,7 +2249,9 @@ compose_dsts(const struct bridge *br, const flow_t *flow, uint16_t vlan, /* XXX even better, define each VLAN as a datapath port group */ for (i = 0; i < br->n_ports; i++) { struct port *port = br->ports[i]; - if (port != in_port && port_includes_vlan(port, vlan) + if (port != in_port + && port_is_floodable(port) + && port_includes_vlan(port, vlan) && !port->is_mirror_output_port && set_dst(dst, flow, in_port, port, tags)) { mirrors |= port->dst_mirrors; @@ -2138,7 +2326,7 @@ print_dsts(const struct dst *dsts, size_t n) } static void -compose_actions(struct bridge *br, const flow_t *flow, uint16_t vlan, +compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan, const struct port *in_port, const struct port *out_port, tag_type *tags, struct odp_actions *actions, uint16_t *nf_output_iface) @@ -2158,8 +2346,9 @@ compose_actions(struct bridge *br, const flow_t *flow, uint16_t vlan, if (p->vlan == OFP_VLAN_NONE) { odp_actions_add(actions, ODPAT_STRIP_VLAN); } else { - a = odp_actions_add(actions, ODPAT_SET_VLAN_VID); - a->vlan_vid.vlan_vid = htons(p->vlan); + a = odp_actions_add(actions, ODPAT_SET_DL_TCI); + a->dl_tci.tci = htons(p->vlan & VLAN_VID_MASK); + a->dl_tci.tci |= htons(flow->dl_vlan_pcp << VLAN_PCP_SHIFT); } cur_vlan = p->vlan; } @@ -2172,7 +2361,7 @@ compose_actions(struct bridge *br, const flow_t *flow, uint16_t vlan, * 802.1Q header and implicitly tagged ports. A value of 0 indicates that * the packet is untagged and -1 indicates it has an invalid header and * should be dropped. */ -static int flow_get_vlan(struct bridge *br, const flow_t *flow, +static int flow_get_vlan(struct bridge *br, const struct flow *flow, struct port *in_port, bool have_packet) { /* Note that dl_vlan of 0 and of OFP_VLAN_NONE both mean that the packet @@ -2218,7 +2407,7 @@ static int flow_get_vlan(struct bridge *br, const flow_t *flow, * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to * indicate this; newer upstream kernels use gratuitous ARP requests. */ static bool -is_gratuitous_arp(const flow_t *flow) +is_gratuitous_arp(const struct flow *flow) { return (flow->dl_type == htons(ETH_TYPE_ARP) && eth_addr_is_broadcast(flow->dl_dst) @@ -2228,7 +2417,7 @@ is_gratuitous_arp(const flow_t *flow) } static void -update_learning_table(struct bridge *br, const flow_t *flow, int vlan, +update_learning_table(struct bridge *br, const struct flow *flow, int vlan, struct port *in_port) { enum grat_arp_lock_type lock_type; @@ -2276,7 +2465,7 @@ update_learning_table(struct bridge *br, const flow_t *flow, int vlan, * so in one special case. */ static bool -is_admissible(struct bridge *br, const flow_t *flow, bool have_packet, +is_admissible(struct bridge *br, const struct flow *flow, bool have_packet, tag_type *tags, int *vlanp, struct port **in_portp) { struct iface *in_iface; @@ -2302,7 +2491,7 @@ is_admissible(struct bridge *br, const flow_t *flow, bool have_packet, static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown " - "interface %"PRIu16, br->name, flow->in_port); + "interface %"PRIu16, br->name, flow->in_port); } *in_portp = NULL; @@ -2365,7 +2554,7 @@ is_admissible(struct bridge *br, const flow_t *flow, bool have_packet, * returns true. Otherwise, the actions should only be applied to 'packet', or * not at all, if 'packet' was NULL. */ static bool -process_flow(struct bridge *br, const flow_t *flow, +process_flow(struct bridge *br, const struct flow *flow, const struct ofpbuf *packet, struct odp_actions *actions, tag_type *tags, uint16_t *nf_output_iface) { @@ -2415,56 +2604,20 @@ done: return true; } -/* Careful: 'opp' is in host byte order and opp->port_no is an OFP port - * number. */ -static void -bridge_port_changed_ofhook_cb(enum ofp_port_reason reason, - const struct ofp_phy_port *opp, - void *br_) -{ - struct bridge *br = br_; - struct iface *iface; - struct port *port; - - iface = iface_from_dp_ifidx(br, ofp_port_to_odp_port(opp->port_no)); - if (!iface) { - return; - } - port = iface->port; - - if (reason == OFPPR_DELETE) { - VLOG_WARN("bridge %s: interface %s deleted unexpectedly", - br->name, iface->name); - iface_destroy(iface); - if (!port->n_ifaces) { - VLOG_WARN("bridge %s: port %s has no interfaces, dropping", - br->name, port->name); - port_destroy(port); - } - - bridge_flush(br); - } else { - if (port->n_ifaces > 1) { - bool up = !(opp->state & OFPPS_LINK_DOWN); - bond_link_status_update(iface, up); - port_update_bond_compat(port); - } - } -} - static bool -bridge_normal_ofhook_cb(const flow_t *flow, const struct ofpbuf *packet, +bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet, struct odp_actions *actions, tag_type *tags, uint16_t *nf_output_iface, void *br_) { struct bridge *br = br_; COVERAGE_INC(bridge_process_flow); + return process_flow(br, flow, packet, actions, tags, nf_output_iface); } static void -bridge_account_flow_ofhook_cb(const flow_t *flow, +bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags, const union odp_action *actions, size_t n_actions, unsigned long long int n_bytes, void *br_) @@ -2472,20 +2625,24 @@ bridge_account_flow_ofhook_cb(const flow_t *flow, struct bridge *br = br_; const union odp_action *a; struct port *in_port; - tag_type tags = 0; + tag_type dummy = 0; int vlan; - /* Feed information from the active flows back into the learning table - * to ensure that table is always in sync with what is actually flowing - * through the datapath. */ - if (is_admissible(br, flow, false, &tags, &vlan, &in_port)) { + /* Feed information from the active flows back into the learning table to + * ensure that table is always in sync with what is actually flowing + * through the datapath. + * + * We test that 'tags' is nonzero to ensure that only flows that include an + * OFPP_NORMAL action are used for learning. This works because + * bridge_normal_ofhook_cb() always sets a nonzero tag value. */ + if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) { update_learning_table(br, flow, vlan, in_port); } + /* Account for bond slave utilization. */ if (!br->has_bonded_ports) { return; } - for (a = actions; a < &actions[n_actions]; a++) { if (a->type == ODPAT_OUTPUT) { struct port *out_port = port_from_dp_ifidx(br, a->output.port); @@ -2520,7 +2677,6 @@ bridge_account_checkpoint_ofhook_cb(void *br_) } static struct ofhooks bridge_ofhooks = { - bridge_port_changed_ofhook_cb, bridge_normal_ofhook_cb, bridge_account_flow_ofhook_cb, bridge_account_checkpoint_ofhook_cb, @@ -2758,7 +2914,6 @@ bond_rebalance_port(struct port *port) * smallest hashes instead of the biggest ones. There is little * reason behind this decision; we could use the opposite sort * order to shift away big hashes ahead of small ones. */ - size_t i; bool order_swapped; for (i = 0; i < from->n_hashes; i++) { @@ -2839,11 +2994,11 @@ bond_send_learning_packets(struct port *port) ofpbuf_init(&packet, 128); error = n_packets = n_errors = 0; - LIST_FOR_EACH (e, struct mac_entry, lru_node, &br->ml->lrus) { + LIST_FOR_EACH (e, lru_node, &br->ml->lrus) { union ofp_action actions[2], *a; uint16_t dp_ifidx; tag_type tags = 0; - flow_t flow; + struct flow flow; int retval; if (e->port == port->port_idx @@ -2901,7 +3056,7 @@ bond_unixctl_list(struct unixctl_conn *conn, ds_put_cstr(&ds, "bridge\tbond\tslaves\n"); - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { size_t i; for (i = 0; i < br->n_ports; i++) { @@ -2930,7 +3085,7 @@ bond_find(const char *name) { const struct bridge *br; - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { size_t i; for (i = 0; i < br->n_ports; i++) { @@ -2990,8 +3145,7 @@ bond_unixctl_show(struct unixctl_conn *conn, hash, be->tx_bytes / 1024); /* MACs. */ - LIST_FOR_EACH (me, struct mac_entry, lru_node, - &port->bridge->ml->lrus) { + LIST_FOR_EACH (me, lru_node, &port->bridge->ml->lrus) { uint16_t dp_ifidx; tag_type tags = 0; if (bond_hash(me->mac) == hash @@ -3276,7 +3430,7 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) if (port->updelay < 0) { port->updelay = 0; } - port->updelay = cfg->bond_downdelay; + port->downdelay = cfg->bond_downdelay; if (port->downdelay < 0) { port->downdelay = 0; } @@ -3299,6 +3453,7 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) if (!shash_add_once(&new_ifaces, if_cfg->name, NULL)) { VLOG_WARN("port %s: %s specified twice as port interface", port->name, if_cfg->name); + iface_set_ofport(if_cfg, -1); continue; } @@ -3312,8 +3467,15 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) } iface->cfg = if_cfg; } else { - iface_create(port, if_cfg); + iface = iface_create(port, if_cfg); } + + /* 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(if_cfg->name, port->bridge->name) ? "internal" + : if_cfg->type[0] ? if_cfg->type + : "system"); } shash_destroy(&new_ifaces); @@ -3343,7 +3505,6 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) trunks = NULL; if (vlan < 0 && cfg->n_trunks) { size_t n_errors; - size_t i; trunks = bitmap_allocate(4096); n_errors = 0; @@ -3405,6 +3566,9 @@ port_destroy(struct port *port) del = br->ports[port->port_idx] = br->ports[--br->n_ports]; del->port_idx = port->port_idx; + VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name); + + netdev_monitor_destroy(port->monitor); free(port->ifaces); bitmap_free(port->trunks); free(port->name); @@ -3436,6 +3600,10 @@ port_lookup_iface(const struct port *port, const char *name) static void port_update_bonding(struct port *port) { + if (port->monitor) { + netdev_monitor_destroy(port->monitor); + port->monitor = NULL; + } if (port->n_ifaces < 2) { /* Not a bonded port. */ if (port->bond_hash) { @@ -3445,9 +3613,9 @@ port_update_bonding(struct port *port) port->bond_fake_iface = false; } } else { - if (!port->bond_hash) { - size_t i; + size_t i; + if (!port->bond_hash) { port->bond_hash = xcalloc(BOND_MASK + 1, sizeof *port->bond_hash); for (i = 0; i <= BOND_MASK; i++) { struct bond_entry *e = &port->bond_hash[i]; @@ -3465,6 +3633,11 @@ port_update_bonding(struct port *port) } port->bond_compat_is_stale = true; port->bond_fake_iface = port->cfg->bond_fake_iface; + + port->monitor = netdev_monitor_create(); + for (i = 0; i < port->n_ifaces; i++) { + netdev_monitor_add(port->monitor, port->ifaces[i]->netdev); + } } } @@ -3506,15 +3679,15 @@ port_update_bond_compat(struct port *port) /* We need to make the same determination as the Linux bonding * code to determine whether a slave should be consider "up". - * The Linux function bond_miimon_inspect() supports four + * The Linux function bond_miimon_inspect() supports four * BOND_LINK_* states: - * + * * - BOND_LINK_UP: carrier detected, updelay has passed. * - BOND_LINK_FAIL: carrier lost, downdelay in progress. * - BOND_LINK_DOWN: carrier lost, downdelay has passed. * - BOND_LINK_BACK: carrier detected, updelay in progress. * - * The function bond_info_show_slave() only considers BOND_LINK_UP + * The function bond_info_show_slave() only considers BOND_LINK_UP * to be "up" and anything else to be "down". */ slave->up = iface->enabled && iface->delay_expires == LLONG_MAX; @@ -3586,7 +3759,6 @@ iface_create(struct port *port, const struct ovsrec_interface *if_cfg) struct bridge *br = port->bridge; struct iface *iface; char *name = if_cfg->name; - int error; iface = xzalloc(sizeof *iface); iface->port = port; @@ -3600,20 +3772,6 @@ iface_create(struct port *port, const struct ovsrec_interface *if_cfg) shash_add_assert(&br->iface_by_name, iface->name, iface); - /* Attempt to create the network interface in case it doesn't exist yet. */ - if (!iface_is_internal(br, iface->name)) { - error = set_up_iface(if_cfg, iface, true); - if (error) { - VLOG_WARN("could not create iface %s: %s", iface->name, - strerror(error)); - - shash_find_and_delete_assert(&br->iface_by_name, iface->name); - free(iface->name); - free(iface); - return NULL; - } - } - if (port->n_ifaces >= port->allocated_ifaces) { port->ifaces = x2nrealloc(port->ifaces, &port->allocated_ifaces, sizeof *port->ifaces); @@ -3642,7 +3800,7 @@ iface_destroy(struct iface *iface) shash_find_and_delete_assert(&br->iface_by_name, iface->name); if (iface->dp_ifidx >= 0) { - port_array_set(&br->ifaces, iface->dp_ifidx, NULL); + hmap_remove(&br->ifaces, &iface->dp_ifidx_node); } del = port->ifaces[iface->port_ifidx] = port->ifaces[--port->n_ifaces]; @@ -3671,40 +3829,16 @@ iface_lookup(const struct bridge *br, const char *name) static struct iface * iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx) -{ - return port_array_get(&br->ifaces, dp_ifidx); -} - -/* Returns true if 'iface' is the name of an "internal" interface on bridge - * 'br', that is, an interface that is entirely simulated within the datapath. - * The local port (ODPP_LOCAL) is always an internal interface. Other local - * interfaces are created by setting "iface..internal = true". - * - * In addition, we have a kluge-y feature that creates an internal port with - * the name of a bonded port if "bonding..fake-iface = true" is set. - * This feature needs to go away in the long term. Until then, this is one - * reason why this function takes a name instead of a struct iface: the fake - * interfaces created this way do not have a struct iface. */ -static bool -iface_is_internal(const struct bridge *br, const char *if_name) { struct iface *iface; - struct port *port; - - if (!strcmp(if_name, br->name)) { - return true; - } - iface = iface_lookup(br, if_name); - if (iface && !strcmp(iface->cfg->type, "internal")) { - return true; - } - - port = port_lookup(br, if_name); - if (port && port->n_ifaces > 1 && port->cfg->bond_fake_iface) { - return true; + HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node, + hash_int(dp_ifidx, 0), &br->ifaces) { + if (iface->dp_ifidx == dp_ifidx) { + return iface; + } } - return false; + return NULL; } /* Set Ethernet address of 'iface', if one is specified in the configuration @@ -3731,6 +3865,19 @@ iface_set_mac(struct iface *iface) } } +/* Sets the ofport column of 'if_cfg' to 'ofport'. */ +static void +iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport) +{ + if (if_cfg) { + ovsrec_interface_set_ofport(if_cfg, &ofport, 1); + } +} + +/* 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)