X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=3b7ec51f44fee1ae14326184d4d46b5e38143fe6;hb=bb60fa749da7f725f2f000a90ea317ad4b1383d2;hp=f4b2fb2c3030e4e3dff6fe884ff8ea302595e1ec;hpb=2886875a385df582354ac916431b278f2fe698e0;p=openvswitch diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index f4b2fb2c..3b7ec51f 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -43,13 +43,14 @@ #include "odp-util.h" #include "ofp-print.h" #include "ofpbuf.h" +#include "ofproto/netflow.h" +#include "ofproto/ofproto.h" #include "packets.h" #include "poll-loop.h" #include "port-array.h" #include "proc-net-compat.h" #include "process.h" -#include "secchan/netflow.h" -#include "secchan/ofproto.h" +#include "shash.h" #include "socket-util.h" #include "stp.h" #include "svec.h" @@ -60,6 +61,7 @@ #include "vconn-ssl.h" #include "xenserver.h" #include "xtoxll.h" +#include "sflow_api.h" #define THIS_MODULE VLM_bridge #include "vlog.h" @@ -72,17 +74,18 @@ struct dst { extern uint64_t mgmt_id; struct iface { + /* These members are always valid. */ struct port *port; /* Containing port. */ size_t port_ifidx; /* Index within containing port. */ - char *name; /* Host network device name. */ - int dp_ifidx; /* Index within kernel datapath. */ - - uint8_t mac[ETH_ADDR_LEN]; /* Ethernet address (all zeros if unknowns). */ - tag_type tag; /* Tag associated with this interface. */ - bool enabled; /* May be chosen for flows? */ long long delay_expires; /* Time after which 'enabled' may change. */ + + /* These members are valid only after bridge_reconfigure() causes them to + * be initialized.*/ + int dp_ifidx; /* Index within kernel datapath. */ + struct netdev *netdev; /* Network device. */ + bool enabled; /* May be chosen for flows? */ }; #define BOND_MASK 0xff @@ -147,7 +150,7 @@ struct port { struct bridge { struct list node; /* Node in global list of bridges. */ char *name; /* User-specified arbitrary name. */ - struct mac_learning *ml; /* MAC learning table, or null not to learn. */ + struct mac_learning *ml; /* MAC learning table. */ bool sent_config_request; /* Successfully sent config request? */ uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */ @@ -160,7 +163,7 @@ struct bridge { struct ofproto *ofproto; /* OpenFlow switch. */ /* Kernel datapath information. */ - struct dpif dpif; /* Kernel datapath. */ + struct dpif *dpif; /* Datapath. */ struct port_array ifaces; /* Indexed by kernel datapath port number. */ /* Bridge ports. */ @@ -203,10 +206,12 @@ static void bridge_fetch_dp_ifaces(struct bridge *); static void bridge_flush(struct bridge *); static void bridge_pick_local_hw_addr(struct bridge *, uint8_t ea[ETH_ADDR_LEN], - const char **devname); + struct iface **hw_addr_iface); static uint64_t bridge_pick_datapath_id(struct bridge *, const uint8_t bridge_ea[ETH_ADDR_LEN], - const char *devname); + struct iface *hw_addr_iface); +static struct iface *bridge_get_local_iface(struct bridge *); +static const char *bridge_get_controller(const struct bridge *br); static uint64_t dpid_from_hash(const void *, size_t nbytes); static void bridge_unixctl_fdb_show(struct unixctl_conn *, const char *args); @@ -216,6 +221,7 @@ static void bond_run(struct bridge *); static void bond_wait(struct bridge *); static void bond_rebalance_port(struct port *); static void bond_send_learning_packets(struct port *); +static void bond_enable_slave(struct iface *iface, bool enable); static void port_create(struct bridge *, const char *name); static void port_reconfigure(struct port *); @@ -226,6 +232,7 @@ static struct port *port_from_dp_ifidx(const struct bridge *, uint16_t dp_ifidx); static void port_update_bond_compat(struct port *); static void port_update_vlan_compat(struct port *); +static void port_update_bonding(struct port *); static void mirror_create(struct bridge *, const char *name); static void mirror_destroy(struct mirror *); @@ -266,8 +273,8 @@ bridge_get_ifaces(struct svec *svec) for (j = 0; j < port->n_ifaces; j++) { struct iface *iface = port->ifaces[j]; if (iface->dp_ifidx < 0) { - VLOG_ERR("%s interface not in dp%u, ignoring", - iface->name, dpif_id(&br->dpif)); + VLOG_ERR("%s interface not in datapath %s, ignoring", + iface->name, dpif_name(br->dpif)); } else { if (iface->dp_ifidx != ODPP_LOCAL) { svec_add(svec, iface->name); @@ -282,34 +289,41 @@ bridge_get_ifaces(struct svec *svec) void bridge_init(void) { - int retval; - int i; - - bond_init(); + struct svec dpif_names; + size_t i; unixctl_command_register("fdb/show", bridge_unixctl_fdb_show); - for (i = 0; i < DP_MAX; i++) { - struct dpif dpif; - char devname[16]; + svec_init(&dpif_names); + dp_enumerate(&dpif_names); + for (i = 0; i < dpif_names.n; i++) { + const char *dpif_name = dpif_names.names[i]; + struct dpif *dpif; + int retval; - sprintf(devname, "dp%d", i); - retval = dpif_open(devname, &dpif); + retval = dpif_open(dpif_name, &dpif); if (!retval) { - char dpif_name[IF_NAMESIZE]; - if (dpif_get_name(&dpif, dpif_name, sizeof dpif_name) - || !cfg_has("bridge.%s.port", dpif_name)) { - dpif_delete(&dpif); + struct svec all_names; + size_t j; + + svec_init(&all_names); + dpif_get_all_names(dpif, &all_names); + for (j = 0; j < all_names.n; j++) { + if (cfg_has("bridge.%s.port", all_names.names[j])) { + goto found; + } } - dpif_close(&dpif); - } else if (retval != ENODEV) { - VLOG_ERR("failed to delete datapath dp%d: %s", - i, strerror(retval)); + dpif_delete(dpif); + found: + svec_destroy(&all_names); + dpif_close(dpif); } } + svec_destroy(&dpif_names); unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows); + bond_init(); bridge_reconfigure(); } @@ -351,43 +365,181 @@ bridge_configure_ssl(void) * the old certificate will still be trusted until vSwitch is * restarted. We may want to address this in vconn's SSL library. */ if (config_string_change("ssl.ca-cert", &cacert_file) - || (stat(cacert_file, &s) && errno == ENOENT)) { + || (cacert_file && stat(cacert_file, &s) && errno == ENOENT)) { vconn_ssl_set_ca_cert_file(cacert_file, cfg_get_bool(0, "ssl.bootstrap-ca-cert")); } } #endif +/* Attempt to create the network device 'iface_name' through the netdev + * library. */ +static int +set_up_iface(const char *iface_name, bool create) +{ + const char *type; + const char *arg; + struct svec arg_svec; + struct shash args; + int error; + size_t i; + + /* If a type is not explicitly declared, then assume it's an existing + * "system" device. */ + type = cfg_get_string(0, "iface.%s.type", iface_name); + if (!type || !strcmp(type, "system")) { + return 0; + } + + svec_init(&arg_svec); + cfg_get_subsections(&arg_svec, "iface.%s.args", iface_name); + + shash_init(&args); + SVEC_FOR_EACH (i, arg, &arg_svec) { + const char *value; + + value = cfg_get_string(0, "iface.%s.args.%s", iface_name, arg); + if (value) { + shash_add(&args, arg, xstrdup(value)); + } + } + + if (create) { + error = netdev_create(iface_name, type, &args); + } else { + /* xxx Check to make sure that the type hasn't changed. */ + error = netdev_reconfigure(iface_name, &args); + } + + svec_destroy(&arg_svec); + shash_destroy(&args); + + return error; +} + +static int +create_iface(const char *iface_name) +{ + return set_up_iface(iface_name, true); +} + +static int +reconfigure_iface(const char *iface_name) +{ + return set_up_iface(iface_name, false); +} + +static void +destroy_iface(const char *iface_name) +{ + netdev_destroy(iface_name); +} + + +/* iterate_and_prune_ifaces() callback function that opens the network device + * for 'iface', if it is not already open, and retrieves the interface's MAC + * address and carrier status. */ +static bool +init_iface_netdev(struct bridge *br UNUSED, struct iface *iface, + void *aux UNUSED) +{ + if (iface->netdev) { + return true; + } else if (!netdev_open(iface->name, NETDEV_ETH_TYPE_NONE, + &iface->netdev)) { + netdev_get_carrier(iface->netdev, &iface->enabled); + return true; + } else { + /* If the network device can't be opened, then we're not going to try + * to do anything with this interface. */ + return false; + } +} + +static bool +check_iface_dp_ifidx(struct bridge *br, struct iface *iface, void *aux 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 { + VLOG_ERR("%s interface not in %s, dropping", + iface->name, dpif_name(br->dpif)); + return false; + } +} + +static bool +set_iface_properties(struct bridge *br UNUSED, struct iface *iface, + void *aux UNUSED) +{ + int rate, burst; + + /* Set policing attributes. */ + rate = cfg_get_int(0, "port.%s.ingress.policing-rate", iface->name); + burst = cfg_get_int(0, "port.%s.ingress.policing-burst", iface->name); + netdev_set_policing(iface->netdev, rate, burst); + + /* Set MAC address of internal interfaces other than the local + * interface. */ + if (iface->dp_ifidx != ODPP_LOCAL + && iface_is_internal(br, iface->name)) { + iface_set_mac(iface); + } + + return true; +} + +/* Calls 'cb' for each interfaces in 'br', passing along the 'aux' argument. + * Deletes from 'br' all the interfaces for which 'cb' returns false, and then + * deletes from 'br' any ports that no longer have any interfaces. */ +static void +iterate_and_prune_ifaces(struct bridge *br, + bool (*cb)(struct bridge *, struct iface *, + void *aux), + void *aux) +{ + size_t i, j; + + for (i = 0; i < br->n_ports; ) { + struct port *port = br->ports[i]; + for (j = 0; j < port->n_ifaces; ) { + struct iface *iface = port->ifaces[j]; + if (cb(br, iface, aux)) { + j++; + } else { + iface_destroy(iface); + } + } + + if (port->n_ifaces) { + i++; + } else { + VLOG_ERR("%s port has no interfaces, dropping", port->name); + port_destroy(port); + } + } +} + void bridge_reconfigure(void) { - struct svec old_br, new_br, raw_new_br; + struct svec old_br, new_br; struct bridge *br, *next; - size_t i, j; + size_t i; + int sflow_bridge_number; COVERAGE_INC(bridge_reconfigure); - /* Collect old bridges. */ + /* Collect old and new bridges. */ svec_init(&old_br); + svec_init(&new_br); LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { svec_add(&old_br, br->name); } - - /* Collect new bridges. */ - svec_init(&raw_new_br); - cfg_get_subsections(&raw_new_br, "bridge"); - svec_init(&new_br); - for (i = 0; i < raw_new_br.n; i++) { - const char *name = raw_new_br.names[i]; - if ((!strncmp(name, "dp", 2) && isdigit(name[2])) || - (!strncmp(name, "nl:", 3) && isdigit(name[3]))) { - VLOG_ERR("%s is not a valid bridge name (bridges may not be " - "named \"dp\" or \"nl:\" followed by a digit)", name); - } else { - svec_add(&new_br, name); - } - } - svec_destroy(&raw_new_br); + cfg_get_subsections(&new_br, "bridge"); /* Get rid of deleted bridges and add new bridges. */ svec_sort(&old_br); @@ -428,17 +580,19 @@ bridge_reconfigure(void) size_t n_dpif_ports; struct svec want_ifaces; - dpif_port_list(&br->dpif, &dpif_ports, &n_dpif_ports); + dpif_port_list(br->dpif, &dpif_ports, &n_dpif_ports); bridge_get_all_ifaces(br, &want_ifaces); for (i = 0; i < n_dpif_ports; i++) { const struct odp_port *p = &dpif_ports[i]; if (!svec_contains(&want_ifaces, p->devname) && strcmp(p->devname, br->name)) { - int retval = dpif_port_del(&br->dpif, p->port); + int retval = dpif_port_del(br->dpif, p->port); if (retval) { - VLOG_ERR("failed to remove %s interface from dp%u: %s", - p->devname, dpif_id(&br->dpif), strerror(retval)); + VLOG_ERR("failed to remove %s interface from %s: %s", + p->devname, dpif_name(br->dpif), + strerror(retval)); } + destroy_iface(p->devname); } } svec_destroy(&want_ifaces); @@ -448,9 +602,8 @@ bridge_reconfigure(void) struct odp_port *dpif_ports; size_t n_dpif_ports; struct svec cur_ifaces, want_ifaces, add_ifaces; - int next_port_no; - dpif_port_list(&br->dpif, &dpif_ports, &n_dpif_ports); + dpif_port_list(br->dpif, &dpif_ports, &n_dpif_ports); svec_init(&cur_ifaces); for (i = 0; i < n_dpif_ports; i++) { svec_add(&cur_ifaces, dpif_ports[i].devname); @@ -460,74 +613,60 @@ bridge_reconfigure(void) bridge_get_all_ifaces(br, &want_ifaces); svec_diff(&want_ifaces, &cur_ifaces, &add_ifaces, NULL, NULL); - next_port_no = 1; + for (i = 0; i < cur_ifaces.n; i++) { + const char *if_name = cur_ifaces.names[i]; + reconfigure_iface(if_name); + } + for (i = 0; i < add_ifaces.n; i++) { const char *if_name = add_ifaces.names[i]; - for (;;) { - bool internal; - int error; - - /* Add to datapath. */ - internal = iface_is_internal(br, if_name); - error = dpif_port_add(&br->dpif, if_name, next_port_no++, - internal ? ODP_PORT_INTERNAL : 0); - if (error != EEXIST) { - if (next_port_no >= 256) { - VLOG_ERR("ran out of valid port numbers on dp%u", - dpif_id(&br->dpif)); - goto out; - } - if (error) { - VLOG_ERR("failed to add %s interface to dp%u: %s", - if_name, dpif_id(&br->dpif), strerror(error)); - } - break; - } + bool internal; + int error; + + /* Attempt to create the network interface in case it + * doesn't exist yet. */ + error = create_iface(if_name); + if (error) { + VLOG_WARN("could not create iface %s: %s\n", if_name, + strerror(error)); + continue; + } + + /* Add to datapath. */ + internal = iface_is_internal(br, if_name); + error = dpif_port_add(br->dpif, if_name, + internal ? ODP_PORT_INTERNAL : 0, NULL); + if (error == EFBIG) { + VLOG_ERR("ran out of valid port numbers on %s", + dpif_name(br->dpif)); + break; + } else if (error) { + VLOG_ERR("failed to add %s interface to %s: %s", + if_name, dpif_name(br->dpif), strerror(error)); } } - out: svec_destroy(&cur_ifaces); svec_destroy(&want_ifaces); svec_destroy(&add_ifaces); } + sflow_bridge_number = 0; LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { uint8_t ea[8]; uint64_t dpid; - struct iface *local_iface = NULL; - const char *devname; + struct iface *local_iface; + struct iface *hw_addr_iface; struct netflow_options nf_options; bridge_fetch_dp_ifaces(br); - for (i = 0; i < br->n_ports; ) { - struct port *port = br->ports[i]; + iterate_and_prune_ifaces(br, init_iface_netdev, NULL); - for (j = 0; j < port->n_ifaces; ) { - struct iface *iface = port->ifaces[j]; - if (iface->dp_ifidx < 0) { - VLOG_ERR("%s interface not in dp%u, dropping", - iface->name, dpif_id(&br->dpif)); - iface_destroy(iface); - } else { - if (iface->dp_ifidx == ODPP_LOCAL) { - local_iface = iface; - } - VLOG_DBG("dp%u has interface %s on port %d", - dpif_id(&br->dpif), iface->name, iface->dp_ifidx); - j++; - } - } - if (!port->n_ifaces) { - VLOG_ERR("%s port has no interfaces, dropping", port->name); - port_destroy(port); - continue; - } - i++; - } + iterate_and_prune_ifaces(br, check_iface_dp_ifidx, NULL); /* Pick local port hardware address, datapath ID. */ - bridge_pick_local_hw_addr(br, ea, &devname); + bridge_pick_local_hw_addr(br, ea, &hw_addr_iface); + local_iface = bridge_get_local_iface(br); if (local_iface) { - int error = netdev_nodev_set_etheraddr(local_iface->name, ea); + int error = netdev_set_etheraddr(local_iface->netdev, ea); if (error) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge " @@ -536,13 +675,13 @@ bridge_reconfigure(void) } } - dpid = bridge_pick_datapath_id(br, ea, devname); + dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface); ofproto_set_datapath_id(br->ofproto, dpid); /* Set NetFlow configuration on this bridge. */ memset(&nf_options, 0, sizeof nf_options); - nf_options.engine_type = br->dpif.minor; - nf_options.engine_id = br->dpif.minor; + dpif_get_netflow_ids(br->dpif, &nf_options.engine_type, + &nf_options.engine_id); nf_options.active_timeout = -1; if (cfg_has("netflow.%s.engine-type", br->name)) { @@ -581,6 +720,42 @@ bridge_reconfigure(void) } svec_destroy(&nf_options.collectors); + if (cfg_has("sflow.%s.host", br->name)) { + struct ofproto_sflow_options oso; + + svec_init(&oso.targets); + cfg_get_all_keys(&oso.targets, "sflow.%s.host", br->name); + + oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE; + if (cfg_has("sflow.%s.sampling", br->name)) { + oso.sampling_rate = cfg_get_int(0, "sflow.%s.sampling", + br->name); + } + + oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL; + if (cfg_has("sflow.%s.polling", br->name)) { + oso.polling_interval = cfg_get_int(0, "sflow.%s.polling", + br->name); + } + + oso.header_len = SFL_DEFAULT_HEADER_SIZE; + if (cfg_has("sflow.%s.header", br->name)) { + oso.header_len = cfg_get_int(0, "sflow.%s.header", br->name); + } + + oso.sub_id = sflow_bridge_number++; + oso.agent_device = (char *) cfg_get_string(0, "sflow.%s.agent", + br->name); + oso.control_ip = (char *) cfg_get_string(0, + "bridge.%s.controller.ip", + br->name); + ofproto_set_sflow(br->ofproto, &oso); + + svec_destroy(&oso.targets); + } else { + ofproto_set_sflow(br->ofproto, NULL); + } + /* Update the controller and related settings. It would be more * straightforward to call this from bridge_reconfigure_one(), but we * can't do it there for two reasons. First, and most importantly, at @@ -597,30 +772,24 @@ bridge_reconfigure(void) struct port *port = br->ports[i]; port_update_vlan_compat(port); - - for (j = 0; j < port->n_ifaces; j++) { - struct iface *iface = port->ifaces[j]; - if (iface->dp_ifidx != ODPP_LOCAL - && iface_is_internal(br, iface->name)) { - iface_set_mac(iface); - } - } + port_update_bonding(port); } } LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { brstp_reconfigure(br); + iterate_and_prune_ifaces(br, set_iface_properties, NULL); } } static void bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], - const char **devname) + struct iface **hw_addr_iface) { uint64_t requested_ea; size_t i, j; int error; - *devname = NULL; + *hw_addr_iface = NULL; /* Did the user request a particular MAC? */ requested_ea = cfg_get_mac(0, "bridge.%s.mac", br->name); @@ -663,7 +832,7 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], for (j = 0; j < port->n_ifaces; j++) { struct iface *candidate = port->ifaces[j]; uint8_t candidate_ea[ETH_ADDR_LEN]; - if (!netdev_nodev_get_etheraddr(candidate->name, candidate_ea) + if (!netdev_get_etheraddr(candidate->netdev, candidate_ea) && eth_addr_equals(iface_ea, candidate_ea)) { iface = candidate; } @@ -693,7 +862,7 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], } /* Grab MAC. */ - error = netdev_nodev_get_etheraddr(iface->name, iface_ea); + error = netdev_get_etheraddr(iface->netdev, iface_ea); if (error) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s", @@ -709,12 +878,12 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], memcmp(iface_ea, ea, ETH_ADDR_LEN) < 0) { memcpy(ea, iface_ea, ETH_ADDR_LEN); - *devname = iface ? iface->name : NULL; + *hw_addr_iface = iface; } } if (eth_addr_is_multicast(ea) || eth_addr_is_vif(ea)) { memcpy(ea, br->default_ea, ETH_ADDR_LEN); - *devname = NULL; + *hw_addr_iface = NULL; VLOG_WARN("bridge %s: using default bridge Ethernet " "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea)); } else { @@ -725,13 +894,13 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], /* Choose and returns the datapath ID for bridge 'br' given that the bridge * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of - * a network device, then that network device's name must be passed in as - * 'devname'; if 'bridge_ea' was derived some other way, then 'devname' must be - * passed in as a null pointer. */ + * an interface on 'br', then that interface must be passed in as + * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then + * 'hw_addr_iface' must be passed in as a null pointer. */ static uint64_t bridge_pick_datapath_id(struct bridge *br, const uint8_t bridge_ea[ETH_ADDR_LEN], - const char *devname) + struct iface *hw_addr_iface) { /* * The procedure for choosing a bridge MAC address will, in the most @@ -752,9 +921,9 @@ bridge_pick_datapath_id(struct bridge *br, return dpid; } - if (devname) { + if (hw_addr_iface) { int vlan; - if (!netdev_get_vlan_vid(devname, &vlan)) { + if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) { /* * A bridge whose MAC address is taken from a VLAN network device * (that is, a network device created with vconfig(8) or similar @@ -845,9 +1014,7 @@ bridge_wait(void) continue; } - if (br->ml) { - mac_learning_wait(br->ml); - } + mac_learning_wait(br->ml); bond_wait(br); brstp_wait(br); } @@ -860,9 +1027,27 @@ bridge_flush(struct bridge *br) { COVERAGE_INC(bridge_flush); br->flush = true; - if (br->ml) { - mac_learning_flush(br->ml); + mac_learning_flush(br->ml); +} + +/* Returns the 'br' interface for the ODPP_LOCAL port, or null if 'br' has no + * such interface. */ +static struct iface * +bridge_get_local_iface(struct bridge *br) +{ + size_t i, j; + + for (i = 0; i < br->n_ports; i++) { + struct port *port = br->ports[i]; + for (j = 0; j < port->n_ifaces; j++) { + struct iface *iface = port->ifaces[j]; + if (iface->dp_ifidx == ODPP_LOCAL) { + return iface; + } + } } + + return NULL; } /* Bridge unixctl user interface functions. */ @@ -871,6 +1056,7 @@ bridge_unixctl_fdb_show(struct unixctl_conn *conn, const char *args) { struct ds ds = DS_EMPTY_INITIALIZER; const struct bridge *br; + const struct mac_entry *e; br = bridge_lookup(args); if (!br) { @@ -879,16 +1065,13 @@ bridge_unixctl_fdb_show(struct unixctl_conn *conn, const char *args) } ds_put_cstr(&ds, " port VLAN MAC Age\n"); - if (br->ml) { - const struct mac_entry *e; - LIST_FOR_EACH (e, struct mac_entry, lru_node, &br->ml->lrus) { - if (e->port < 0 || e->port >= br->n_ports) { - continue; - } - ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n", - br->ports[e->port]->ifaces[0]->dp_ifidx, - e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e)); + LIST_FOR_EACH (e, struct mac_entry, lru_node, &br->ml->lrus) { + if (e->port < 0 || e->port >= br->n_ports) { + continue; } + ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n", + br->ports[e->port]->ifaces[0]->dp_ifidx, + e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e)); } unixctl_command_reply(conn, 200, ds_cstr(&ds)); ds_destroy(&ds); @@ -905,27 +1088,18 @@ bridge_create(const char *name) assert(!bridge_lookup(name)); br = xcalloc(1, sizeof *br); - error = dpif_create(name, &br->dpif); - if (error == EEXIST) { - error = dpif_open(name, &br->dpif); - if (error) { - VLOG_ERR("datapath %s already exists but cannot be opened: %s", - name, strerror(error)); - free(br); - return NULL; - } - dpif_flow_flush(&br->dpif); - } else if (error) { - VLOG_ERR("failed to create datapath %s: %s", name, strerror(error)); + error = dpif_create_and_open(name, &br->dpif); + if (error) { free(br); return NULL; } + dpif_flow_flush(br->dpif); error = ofproto_create(name, &bridge_ofhooks, br, &br->ofproto); if (error) { VLOG_ERR("failed to create switch %s: %s", name, strerror(error)); - dpif_delete(&br->dpif); - dpif_close(&br->dpif); + dpif_delete(br->dpif); + dpif_close(br->dpif); free(br); return NULL; } @@ -942,7 +1116,7 @@ bridge_create(const char *name) list_push_back(&all_bridges, &br->node); - VLOG_INFO("created bridge %s on dp%u", br->name, dpif_id(&br->dpif)); + VLOG_INFO("created bridge %s on %s", br->name, dpif_name(br->dpif)); return br; } @@ -957,12 +1131,12 @@ bridge_destroy(struct bridge *br) port_destroy(br->ports[br->n_ports - 1]); } list_remove(&br->node); - error = dpif_delete(&br->dpif); + error = dpif_delete(br->dpif); if (error && error != ENOENT) { - VLOG_ERR("failed to delete dp%u: %s", - dpif_id(&br->dpif), strerror(error)); + VLOG_ERR("failed to delete %s: %s", + dpif_name(br->dpif), strerror(error)); } - dpif_close(&br->dpif); + dpif_close(br->dpif); ofproto_destroy(br->ofproto); free(br->controller); mac_learning_destroy(br->ml); @@ -1030,9 +1204,7 @@ bridge_run_one(struct bridge *br) return error; } - if (br->ml) { - mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto)); - } + mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto)); bond_run(br); brstp_run(br); @@ -1054,13 +1226,29 @@ bridge_get_controller(const struct bridge *br) return controller && controller[0] ? controller : NULL; } +static bool +check_duplicate_ifaces(struct bridge *br, struct iface *iface, void *ifaces_) +{ + struct svec *ifaces = ifaces_; + if (!svec_contains(ifaces, iface->name)) { + svec_add(ifaces, iface->name); + svec_sort(ifaces); + return true; + } else { + VLOG_ERR("bridge %s: %s interface is on multiple ports, " + "removing from %s", + br->name, iface->name, iface->port->name); + return false; + } +} + static void bridge_reconfigure_one(struct bridge *br) { struct svec old_ports, new_ports, ifaces; struct svec listeners, old_listeners; struct svec snoops, old_snoops; - size_t i, j; + size_t i; /* Collect old ports. */ svec_init(&old_ports); @@ -1074,9 +1262,16 @@ bridge_reconfigure_one(struct bridge *br) svec_init(&new_ports); cfg_get_all_keys(&new_ports, "bridge.%s.port", br->name); svec_sort(&new_ports); - if (bridge_get_controller(br) && !svec_contains(&new_ports, br->name)) { - svec_add(&new_ports, br->name); - svec_sort(&new_ports); + if (bridge_get_controller(br)) { + char local_name[IF_NAMESIZE]; + int error; + + error = dpif_port_get_name(br->dpif, ODPP_LOCAL, + local_name, sizeof local_name); + if (!error && !svec_contains(&new_ports, local_name)) { + svec_add(&new_ports, local_name); + svec_sort(&new_ports); + } } if (!svec_is_unique(&new_ports)) { VLOG_WARN("bridge %s: %s specified twice as bridge port", @@ -1111,28 +1306,7 @@ bridge_reconfigure_one(struct bridge *br) /* Check and delete duplicate interfaces. */ svec_init(&ifaces); - for (i = 0; i < br->n_ports; ) { - struct port *port = br->ports[i]; - for (j = 0; j < port->n_ifaces; ) { - struct iface *iface = port->ifaces[j]; - if (svec_contains(&ifaces, iface->name)) { - VLOG_ERR("bridge %s: %s interface is on multiple ports, " - "removing from %s", - br->name, iface->name, port->name); - iface_destroy(iface); - } else { - svec_add(&ifaces, iface->name); - svec_sort(&ifaces); - j++; - } - } - if (!port->n_ifaces) { - VLOG_ERR("%s port has no interfaces, dropping", port->name); - port_destroy(port); - } else { - i++; - } - } + iterate_and_prune_ifaces(br, check_duplicate_ifaces, &ifaces); svec_destroy(&ifaces); /* Delete all flows if we're switching from connected to standalone or vice @@ -1213,9 +1387,8 @@ bridge_reconfigure_controller(struct bridge *br) cfg_get_string(0, "%s.accept-regex", pfx), update_resolv_conf); } else { - struct netdev *netdev; + struct iface *local_iface; bool in_band; - int error; in_band = (!cfg_is_valid(CFG_BOOL | CFG_REQUIRED, "%s.in-band", pfx) @@ -1223,33 +1396,32 @@ bridge_reconfigure_controller(struct bridge *br) ofproto_set_discovery(br->ofproto, false, NULL, NULL); ofproto_set_in_band(br->ofproto, in_band); - error = netdev_open(br->name, NETDEV_ETH_TYPE_NONE, &netdev); - if (!error) { - if (cfg_is_valid(CFG_IP | CFG_REQUIRED, "%s.ip", pfx)) { - struct in_addr ip, mask, gateway; - ip.s_addr = cfg_get_ip(0, "%s.ip", pfx); - mask.s_addr = cfg_get_ip(0, "%s.netmask", pfx); - gateway.s_addr = cfg_get_ip(0, "%s.gateway", pfx); - - 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)); - } + local_iface = bridge_get_local_iface(br); + if (local_iface + && cfg_is_valid(CFG_IP | CFG_REQUIRED, "%s.ip", pfx)) { + struct netdev *netdev = local_iface->netdev; + struct in_addr ip, mask, gateway; + ip.s_addr = cfg_get_ip(0, "%s.ip", pfx); + mask.s_addr = cfg_get_ip(0, "%s.netmask", pfx); + gateway.s_addr = cfg_get_ip(0, "%s.gateway", pfx); + + 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(gateway)) { - VLOG_INFO("bridge %s: configured gateway "IP_FMT, - br->name, IP_ARGS(&gateway.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)); } } - netdev_close(netdev); } } @@ -1379,17 +1551,17 @@ bridge_fetch_dp_ifaces(struct bridge *br) } port_array_clear(&br->ifaces); - dpif_port_list(&br->dpif, &dpif_ports, &n_dpif_ports); + dpif_port_list(br->dpif, &dpif_ports, &n_dpif_ports); for (i = 0; i < n_dpif_ports; i++) { struct odp_port *p = &dpif_ports[i]; struct iface *iface = iface_lookup(br, p->devname); if (iface) { if (iface->dp_ifidx >= 0) { - VLOG_WARN("dp%u reported interface %s twice", - dpif_id(&br->dpif), p->devname); + VLOG_WARN("%s reported interface %s twice", + dpif_name(br->dpif), p->devname); } else if (iface_from_dp_ifidx(br, p->port)) { - VLOG_WARN("dp%u reported interface %"PRIu16" twice", - dpif_id(&br->dpif), p->port); + 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; @@ -1416,13 +1588,31 @@ lookup_bond_entry(const struct port *port, const uint8_t mac[ETH_ADDR_LEN]) static int bond_choose_iface(const struct port *port) { - size_t i; + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); + size_t i, best_down_slave = -1; + long long next_delay_expiration = LLONG_MAX; + for (i = 0; i < port->n_ifaces; i++) { - if (port->ifaces[i]->enabled) { + struct iface *iface = port->ifaces[i]; + + if (iface->enabled) { return i; + } else if (iface->delay_expires < next_delay_expiration) { + best_down_slave = i; + next_delay_expiration = iface->delay_expires; } } - return -1; + + if (best_down_slave != -1) { + struct iface *iface = port->ifaces[best_down_slave]; + + VLOG_INFO_RL(&rl, "interface %s: skipping remaining %lli ms updelay " + "since no other interface is up", iface->name, + iface->delay_expires - time_msec()); + bond_enable_slave(iface, true); + } + + return best_down_slave; } static bool @@ -1472,10 +1662,12 @@ bond_link_status_update(struct iface *iface, bool carrier) iface->delay_expires = LLONG_MAX; VLOG_INFO_RL(&rl, "interface %s: will not be %s", iface->name, carrier ? "disabled" : "enabled"); - } else if (carrier && port->updelay && port->active_iface < 0) { - iface->delay_expires = time_msec(); - VLOG_INFO_RL(&rl, "interface %s: skipping %d ms updelay since no " - "other interface is up", iface->name, port->updelay); + } else if (carrier && port->active_iface < 0) { + bond_enable_slave(iface, true); + if (port->updelay) { + VLOG_INFO_RL(&rl, "interface %s: skipping %d ms updelay since no " + "other interface is up", iface->name, port->updelay); + } } else { int delay = carrier ? port->updelay : port->downdelay; iface->delay_expires = time_msec() + delay; @@ -1512,6 +1704,12 @@ bond_enable_slave(struct iface *iface, bool enable) struct port *port = iface->port; struct bridge *br = port->bridge; + /* This acts as a recursion check. If the act of disabling a slave + * causes a different slave to be enabled, the flag will allow us to + * skip redundant work when we reenter this function. It must be + * cleared on exit to keep things safe with multiple bonds. */ + static bool moving_active_iface = false; + iface->delay_expires = LLONG_MAX; if (enable == iface->enabled) { return; @@ -1524,19 +1722,29 @@ bond_enable_slave(struct iface *iface, bool enable) if (iface->port_ifidx == port->active_iface) { ofproto_revalidate(br->ofproto, port->active_iface_tag); + + /* Disabling a slave can lead to another slave being immediately + * enabled if there will be no active slaves but one is waiting + * on an updelay. In this case we do not need to run most of the + * code for the newly enabled slave since there was no period + * without an active slave and it is redundant with the disabling + * path. */ + moving_active_iface = true; bond_choose_active_iface(port); } bond_send_learning_packets(port); } else { VLOG_WARN("interface %s: enabled", iface->name); - if (port->active_iface < 0) { + if (port->active_iface < 0 && !moving_active_iface) { ofproto_revalidate(br->ofproto, port->no_ifaces_tag); bond_choose_active_iface(port); bond_send_learning_packets(port); } iface->tag = tag_create_random(); } - port_update_bond_compat(port); + + moving_active_iface = false; + port->bond_compat_is_stale = true; } static void @@ -1547,20 +1755,19 @@ bond_run(struct bridge *br) for (i = 0; i < br->n_ports; i++) { struct port *port = br->ports[i]; + if (port->n_ifaces >= 2) { + for (j = 0; j < port->n_ifaces; j++) { + struct iface *iface = port->ifaces[j]; + if (time_msec() >= iface->delay_expires) { + bond_enable_slave(iface, !iface->enabled); + } + } + } + if (port->bond_compat_is_stale) { port->bond_compat_is_stale = false; port_update_bond_compat(port); } - - if (port->n_ifaces < 2) { - continue; - } - for (j = 0; j < port->n_ifaces; j++) { - struct iface *iface = port->ifaces[j]; - if (time_msec() >= iface->delay_expires) { - bond_enable_slave(iface, !iface->enabled); - } - } } } @@ -1803,6 +2010,71 @@ compose_actions(struct bridge *br, const flow_t *flow, uint16_t vlan, } } +/* Returns the effective vlan of a packet, taking into account both the + * 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, + struct port *in_port, bool have_packet) +{ + /* Note that dl_vlan of 0 and of OFP_VLAN_NONE both mean that the packet + * belongs to VLAN 0, so we should treat both cases identically. (In the + * former case, the packet has an 802.1Q header that specifies VLAN 0, + * presumably to allow a priority to be specified. In the latter case, the + * packet does not have any 802.1Q header.) */ + int vlan = ntohs(flow->dl_vlan); + if (vlan == OFP_VLAN_NONE) { + vlan = 0; + } + if (in_port->vlan >= 0) { + if (vlan) { + /* XXX support double tagging? */ + if (have_packet) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %"PRIu16" tagged " + "packet received on port %s configured with " + "implicit VLAN %"PRIu16, + br->name, ntohs(flow->dl_vlan), + in_port->name, in_port->vlan); + } + return -1; + } + vlan = in_port->vlan; + } else { + if (!port_includes_vlan(in_port, vlan)) { + if (have_packet) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged " + "packet received on port %s not configured for " + "trunking VLAN %d", + br->name, vlan, in_port->name, vlan); + } + return -1; + } + } + + return vlan; +} + +static void +update_learning_table(struct bridge *br, const flow_t *flow, int vlan, + struct port *in_port) +{ + tag_type rev_tag = mac_learning_learn(br->ml, flow->dl_src, + vlan, in_port->port_idx); + if (rev_tag) { + /* The log messages here could actually be useful in debugging, + * so keep the rate limit relatively high. */ + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, + 300); + VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is " + "on port %s in VLAN %d", + br->name, ETH_ADDR_ARGS(flow->dl_src), + in_port->name, vlan); + ofproto_revalidate(br->ofproto, rev_tag); + } +} + static bool is_bcast_arp_reply(const flow_t *flow) { @@ -1823,6 +2095,7 @@ process_flow(struct bridge *br, const flow_t *flow, struct port *in_port; struct port *out_port = NULL; /* By default, drop the packet/flow. */ int vlan; + int out_port_idx; /* Find the interface and port structure for the received packet. */ in_iface = iface_from_dp_ifidx(br, flow->in_port); @@ -1850,41 +2123,9 @@ process_flow(struct bridge *br, const flow_t *flow, return true; } in_port = in_iface->port; - - /* Figure out what VLAN this packet belongs to. - * - * Note that dl_vlan of 0 and of OFP_VLAN_NONE both mean that the packet - * belongs to VLAN 0, so we should treat both cases identically. (In the - * former case, the packet has an 802.1Q header that specifies VLAN 0, - * presumably to allow a priority to be specified. In the latter case, the - * packet does not have any 802.1Q header.) */ - vlan = ntohs(flow->dl_vlan); - if (vlan == OFP_VLAN_NONE) { - vlan = 0; - } - if (in_port->vlan >= 0) { - if (vlan) { - /* XXX support double tagging? */ - if (packet != NULL) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); - VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %"PRIu16" tagged " - "packet received on port %s configured with " - "implicit VLAN %"PRIu16, - br->name, ntohs(flow->dl_vlan), - in_port->name, in_port->vlan); - } - goto done; - } - vlan = in_port->vlan; - } else { - if (!port_includes_vlan(in_port, vlan)) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); - VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged " - "packet received on port %s not configured for " - "trunking VLAN %d", - br->name, vlan, in_port->name, vlan); - goto done; - } + vlan = flow_get_vlan(br, flow, in_port, !!packet); + if (vlan < 0) { + goto done; } /* Drop frames for ports that STP wants entirely killed (both for @@ -1933,37 +2174,23 @@ process_flow(struct bridge *br, const flow_t *flow, /* MAC learning. */ out_port = FLOOD_PORT; - if (br->ml) { - int out_port_idx; - - /* Learn source MAC (but don't try to learn from revalidation). */ - if (packet) { - tag_type rev_tag = mac_learning_learn(br->ml, flow->dl_src, - vlan, in_port->port_idx); - if (rev_tag) { - /* The log messages here could actually be useful in debugging, - * so keep the rate limit relatively high. */ - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, - 300); - VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is " - "on port %s in VLAN %d", - br->name, ETH_ADDR_ARGS(flow->dl_src), - in_port->name, vlan); - ofproto_revalidate(br->ofproto, rev_tag); - } - } - - /* Determine output port. */ - out_port_idx = mac_learning_lookup_tag(br->ml, flow->dl_dst, vlan, - tags); - if (out_port_idx >= 0 && out_port_idx < br->n_ports) { - out_port = br->ports[out_port_idx]; - } else if (!packet) { - /* If we are revalidating but don't have a learning entry then - * eject the flow. Installing a flow that floods packets will - * prevent us from seeing future packets and learning properly. */ - return false; - } + /* Learn source MAC (but don't try to learn from revalidation). */ + if (packet) { + update_learning_table(br, flow, vlan, in_port); + } + + /* Determine output port. */ + out_port_idx = mac_learning_lookup_tag(br->ml, flow->dl_dst, vlan, + tags); + if (out_port_idx >= 0 && out_port_idx < br->n_ports) { + out_port = br->ports[out_port_idx]; + } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) { + /* If we are revalidating but don't have a learning entry then + * eject the flow. Installing a flow that floods packets opens + * up a window of time where we could learn from a packet reflected + * on a bond and blackhole packets before the learning table is + * updated to reflect the correct port. */ + return false; } /* Don't send packets out their input ports. Don't forward frames that STP @@ -2008,7 +2235,6 @@ bridge_port_changed_ofhook_cb(enum ofp_port_reason reason, bridge_flush(br); } else { - memcpy(iface->mac, opp->hw_addr, ETH_ADDR_LEN); if (port->n_ifaces > 1) { bool up = !(opp->state & OFPPS_LINK_DOWN); bond_link_status_update(iface, up); @@ -2043,17 +2269,30 @@ bridge_account_flow_ofhook_cb(const flow_t *flow, void *br_) { struct bridge *br = br_; + struct port *in_port; const union odp_action *a; + /* 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. */ + in_port = port_from_dp_ifidx(br, flow->in_port); + if (in_port) { + int vlan = flow_get_vlan(br, flow, in_port, false); + if (vlan >= 0) { + update_learning_table(br, flow, vlan, in_port); + } + } + if (!br->has_bonded_ports) { return; } for (a = actions; a < &actions[n_actions]; a++) { if (a->type == ODPAT_OUTPUT) { - struct port *port = port_from_dp_ifidx(br, a->output.port); - if (port && port->n_ifaces >= 2) { - struct bond_entry *e = lookup_bond_entry(port, flow->dl_src); + struct port *out_port = port_from_dp_ifidx(br, a->output.port); + if (out_port && out_port->n_ifaces >= 2) { + struct bond_entry *e = lookup_bond_entry(out_port, + flow->dl_src); e->tx_bytes += n_bytes; } } @@ -2399,7 +2638,7 @@ bond_send_learning_packets(struct port *port) struct ofpbuf packet; int error, n_packets, n_errors; - if (!port->n_ifaces || port->active_iface < 0 || !br->ml) { + if (!port->n_ifaces || port->active_iface < 0) { return; } @@ -2554,10 +2793,6 @@ bond_unixctl_show(struct unixctl_conn *conn, const char *args) hash, be->tx_bytes / 1024); /* MACs. */ - if (!port->bridge->ml) { - break; - } - LIST_FOR_EACH (me, struct mac_entry, lru_node, &port->bridge->ml->lrus) { uint16_t dp_ifidx; @@ -2725,6 +2960,25 @@ bond_unixctl_disable_slave(struct unixctl_conn *conn, const char *args) enable_slave(conn, args, false); } +static void +bond_unixctl_hash(struct unixctl_conn *conn, const char *args) +{ + uint8_t mac[ETH_ADDR_LEN]; + uint8_t hash; + char *hash_cstr; + + if (sscanf(args, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(mac)) + == ETH_ADDR_SCAN_COUNT) { + hash = bond_hash(mac); + + hash_cstr = xasprintf("%u", hash); + unixctl_command_reply(conn, 200, hash_cstr); + free(hash_cstr); + } else { + unixctl_command_reply(conn, 501, "invalid mac"); + } +} + static void bond_init(void) { @@ -2735,6 +2989,7 @@ bond_init(void) bond_unixctl_set_active_slave); unixctl_command_register("bond/enable-slave", bond_unixctl_enable_slave); unixctl_command_register("bond/disable-slave", bond_unixctl_disable_slave); + unixctl_command_register("bond/hash", bond_unixctl_hash); } /* Port functions. */ @@ -3033,7 +3288,7 @@ port_update_bond_compat(struct port *port) if (slave->up) { bond.up = true; } - memcpy(slave->mac, iface->mac, ETH_ADDR_LEN); + netdev_get_etheraddr(iface->netdev, slave->mac); } if (cfg_get_bool(0, "bonding.%s.fake-iface", port->name)) { @@ -3077,7 +3332,8 @@ port_update_vlan_compat(struct port *port) && p->n_ifaces && (!vlandev_name || strcmp(p->name, vlandev_name) <= 0)) { - const uint8_t *ea = p->ifaces[0]->mac; + uint8_t ea[ETH_ADDR_LEN]; + netdev_get_etheraddr(p->ifaces[0]->netdev, ea); if (!eth_addr_is_multicast(ea) && !eth_addr_is_reserved(ea) && !eth_addr_is_zero(ea)) { @@ -3103,18 +3359,7 @@ iface_create(struct port *port, const char *name) iface->dp_ifidx = -1; iface->tag = tag_create_random(); iface->delay_expires = LLONG_MAX; - - if (!cfg_get_bool(0, "iface.%s.internal", iface->name)) { - netdev_nodev_get_etheraddr(name, iface->mac); - netdev_nodev_get_carrier(name, &iface->enabled); - } else { - /* Internal interfaces are created later by the call to dpif_port_add() - * in bridge_reconfigure(). Until then, we can't obtain any - * information about them. (There's no real value in doing so, anyway, - * because the 'mac' and 'enabled' values are only used for interfaces - * that are bond slaves, and it doesn't normally make sense to bond an - * internal interface.) */ - } + iface->netdev = NULL; if (port->n_ifaces >= port->allocated_ifaces) { port->ifaces = x2nrealloc(port->ifaces, &port->allocated_ifaces, @@ -3127,7 +3372,6 @@ iface_create(struct port *port, const char *name) VLOG_DBG("attached network device %s to port %s", iface->name, port->name); - port_update_bonding(port); bridge_flush(port->bridge); } @@ -3147,6 +3391,7 @@ iface_destroy(struct iface *iface) del = port->ifaces[iface->port_ifidx] = port->ifaces[--port->n_ifaces]; del->port_ifidx = iface->port_ifidx; + netdev_close(iface->netdev); free(iface->name); free(iface); @@ -3156,7 +3401,6 @@ iface_destroy(struct iface *iface) bond_send_learning_packets(port); } - port_update_bonding(port); bridge_flush(port->bridge); } } @@ -3229,7 +3473,7 @@ iface_set_mac(struct iface *iface) VLOG_ERR("ignoring iface.%s.mac; use bridge.%s.mac instead", iface->name, iface->name); } else { - int error = netdev_nodev_set_etheraddr(iface->name, ea); + int error = netdev_set_etheraddr(iface->netdev, ea); if (error) { VLOG_ERR("interface %s: setting MAC failed (%s)", iface->name, strerror(error)); @@ -3244,7 +3488,8 @@ static void mirror_reconfigure(struct bridge *br) { struct svec old_mirrors, new_mirrors; - size_t i; + size_t i, n_rspan_vlans; + unsigned long *rspan_vlans; /* Collect old and new mirrors. */ svec_init(&old_mirrors); @@ -3293,6 +3538,29 @@ mirror_reconfigure(struct bridge *br) m->out_port->is_mirror_output_port = true; } } + + /* Update learning disabled vlans (for RSPAN). */ + rspan_vlans = NULL; + n_rspan_vlans = cfg_count("vlan.%s.disable-learning", br->name); + if (n_rspan_vlans) { + rspan_vlans = bitmap_allocate(4096); + + for (i = 0; i < n_rspan_vlans; i++) { + int vlan = cfg_get_vlan(i, "vlan.%s.disable-learning", br->name); + if (vlan >= 0) { + bitmap_set1(rspan_vlans, vlan); + VLOG_INFO("bridge %s: disabling learning on vlan %d\n", + br->name, vlan); + } else { + VLOG_ERR("bridge %s: invalid value '%s' for learning disabled " + "VLAN", br->name, + cfg_get_string(i, "vlan.%s.disable-learning", br->name)); + } + } + } + if (mac_learning_set_disabled_vlans(br->ml, rspan_vlans)) { + bridge_flush(br); + } } static void @@ -3558,23 +3826,25 @@ brstp_send_bpdu(struct ofpbuf *pkt, int port_no, void *br_) if (!iface) { VLOG_WARN_RL(&rl, "%s: cannot send BPDU on unknown port %d", br->name, port_no); - } else if (eth_addr_is_zero(iface->mac)) { - VLOG_WARN_RL(&rl, "%s: cannot send BPDU on port %d with unknown MAC", - br->name, port_no); } else { - union ofp_action action; struct eth_header *eth = pkt->l2; - flow_t flow; - memcpy(eth->eth_src, iface->mac, ETH_ADDR_LEN); + netdev_get_etheraddr(iface->netdev, eth->eth_src); + if (eth_addr_is_zero(eth->eth_src)) { + VLOG_WARN_RL(&rl, "%s: cannot send BPDU on port %d " + "with unknown MAC", br->name, port_no); + } else { + union ofp_action action; + flow_t flow; - memset(&action, 0, sizeof action); - action.type = htons(OFPAT_OUTPUT); - action.output.len = htons(sizeof action); - action.output.port = htons(port_no); + memset(&action, 0, sizeof action); + action.type = htons(OFPAT_OUTPUT); + action.output.len = htons(sizeof action); + action.output.port = htons(port_no); - flow_extract(pkt, ODPP_NONE, &flow); - ofproto_send_packet(br->ofproto, &flow, &action, 1, pkt); + flow_extract(pkt, ODPP_NONE, &flow); + ofproto_send_packet(br->ofproto, &flow, &action, 1, pkt); + } } ofpbuf_delete(pkt); }