X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=9fb257954227585a84552c7e1a82ce8d767ed3eb;hb=195c8086244e33ec42fd9fc8354eaedfd849bbba;hp=b4f7a2ef90ab8e4b6414ec2da19fe693b08dfd7b;hpb=750d79a51bec1e9319892ba52b36a9880ba9d8fa;p=openvswitch diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index b4f7a2ef..9fb25795 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -35,6 +35,7 @@ #include "cfm.h" #include "classifier.h" #include "coverage.h" +#include "daemon.h" #include "dirs.h" #include "dpif.h" #include "dynamic-string.h" @@ -54,7 +55,6 @@ #include "ovsdb-data.h" #include "packets.h" #include "poll-loop.h" -#include "proc-net-compat.h" #include "process.h" #include "sha1.h" #include "shash.h" @@ -75,15 +75,11 @@ VLOG_DEFINE_THIS_MODULE(bridge); COVERAGE_DEFINE(bridge_flush); COVERAGE_DEFINE(bridge_process_flow); +COVERAGE_DEFINE(bridge_process_cfm); +COVERAGE_DEFINE(bridge_process_lacp); COVERAGE_DEFINE(bridge_reconfigure); COVERAGE_DEFINE(bridge_lacp_update); -enum lacp_status { - LACP_STATUS_CURRENT, /* Partner is up to date. */ - LACP_STATUS_EXPIRED, /* Partner is out of date. Attempt to re-sync. */ - LACP_STATUS_DEFAULTED /* Partner information is unknown. */ -}; - struct dst { uint16_t vlan; uint16_t dp_ifidx; @@ -99,6 +95,13 @@ static void dst_set_init(struct dst_set *); static void dst_set_add(struct dst_set *, const struct dst *); static void dst_set_free(struct dst_set *); +enum lacp_status { + LACP_CURRENT = 0x01, /* Current State. */ + LACP_EXPIRED = 0x02, /* Expired State. */ + LACP_DEFAULTED = 0x04, /* Partner is defaulted. */ + LACP_ATTACHED = 0x08, /* Attached. Interface may be choosen for flows. */ +}; + struct iface { /* These members are always valid. */ struct port *port; /* Containing port. */ @@ -119,14 +122,12 @@ struct iface { const struct ovsrec_interface *cfg; /* LACP information. */ - enum lacp_status lacp_status; /* LACP state machine status. */ + enum lacp_status lacp_status; /* LACP status. */ uint16_t lacp_priority; /* LACP port priority. */ struct lacp_info lacp_actor; /* LACP actor information. */ struct lacp_info lacp_partner; /* LACP partner information. */ long long int lacp_tx; /* Next LACP message transmission time. */ long long int lacp_rx; /* Next LACP message receive time. */ - bool lacp_attached; /* Attached to its aggregator? LACP allows - this link to be chosen for flows. */ }; #define BOND_MASK 0xff @@ -189,7 +190,6 @@ struct port { tag_type active_iface_tag; /* Tag for bcast flows. */ tag_type no_ifaces_tag; /* Tag for flows when all ifaces disabled. */ 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? */ bool miimon; /* Use miimon instead of carrier? */ long long int bond_miimon_interval; /* Miimon status refresh interval. */ @@ -281,18 +281,21 @@ static struct iface *bridge_get_local_iface(struct bridge *); static uint64_t dpid_from_hash(const void *, size_t nbytes); static unixctl_cb_func bridge_unixctl_fdb_show; +static unixctl_cb_func qos_unixctl_show; -static void lacp_run(struct bridge *); -static void lacp_wait(struct bridge *); +static void lacp_run(struct port *); +static void lacp_wait(struct port *); static void lacp_process_packet(const struct ofpbuf *, struct iface *); static void bond_init(void); -static void bond_run(struct bridge *); -static void bond_wait(struct bridge *); +static void bond_run(struct port *); +static void bond_wait(struct port *); 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_run(struct port *); +static void port_wait(struct port *); static struct port *port_create(struct bridge *, const char *name); static void port_reconfigure(struct port *, const struct ovsrec_port *); static void port_del_ifaces(struct port *, const struct ovsrec_port *); @@ -301,8 +304,6 @@ static struct port *port_lookup(const struct bridge *, const char *name); static struct iface *port_lookup_iface(const struct port *, const char *name); 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 port_update_lacp(struct port *); @@ -316,6 +317,7 @@ 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_find(const char *name); static struct iface *iface_from_dp_ifidx(const struct bridge *, uint16_t dp_ifidx); static void iface_set_mac(struct iface *); @@ -364,6 +366,7 @@ bridge_init(const char *remote) /* Register unixctl commands. */ unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL); + unixctl_command_register("qos/show", qos_unixctl_show, NULL); unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows, NULL); unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect, @@ -502,7 +505,7 @@ iterate_and_prune_ifaces(struct bridge *br, if (port->n_ifaces) { i++; } else { - VLOG_ERR("%s port has no interfaces, dropping", port->name); + VLOG_WARN("%s port has no interfaces, dropping", port->name); port_destroy(port); } } @@ -528,13 +531,10 @@ collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg, struct shash targets; size_t i; - /* Collect all of the potential targets, as the union of the "managers" - * column and the "targets" columns of the rows pointed to by - * "manager_options", excluding any that are out-of-band. */ + /* Collect all of the potential targets from the "targets" columns of the + * rows pointed to by "manager_options", excluding any that are + * out-of-band. */ shash_init(&targets); - for (i = 0; i < ovs_cfg->n_managers; i++) { - shash_add_once(&targets, ovs_cfg->managers[i], NULL); - } for (i = 0; i < ovs_cfg->n_manager_options; i++) { struct ovsrec_manager *m = ovs_cfg->manager_options[i]; @@ -644,9 +644,9 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) && strcmp(dpif_port.name, br->name)) { int retval = dpif_port_del(br->dpif, dpif_port.port_no); if (retval) { - VLOG_ERR("failed to remove %s interface from %s: %s", - dpif_port.name, dpif_name(br->dpif), - strerror(retval)); + VLOG_WARN("failed to remove %s interface from %s: %s", + dpif_port.name, dpif_name(br->dpif), + strerror(retval)); } } } @@ -736,9 +736,9 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) dpif_name(br->dpif)); break; } else { - VLOG_ERR("failed to add %s interface to %s: %s", - if_name, dpif_name(br->dpif), - strerror(error)); + VLOG_WARN("failed to add %s interface to %s: %s", + if_name, dpif_name(br->dpif), + strerror(error)); continue; } } @@ -913,7 +913,6 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) struct port *port = br->ports[i]; int j; - port_update_vlan_compat(port); port_update_bonding(port); port_update_lacp(port); @@ -934,6 +933,10 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) } free(managers); + + /* ovs-vswitchd has completed initialization, so allow the process that + * forked us to exit successfully. */ + daemonize_complete(); } static const char * @@ -983,7 +986,7 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], /* Otherwise choose the minimum non-local MAC address among all of the * interfaces. */ - memset(ea, 0xff, sizeof ea); + memset(ea, 0xff, ETH_ADDR_LEN); for (i = 0; i < br->n_ports; i++) { struct port *port = br->ports[i]; uint8_t iface_ea[ETH_ADDR_LEN]; @@ -1489,22 +1492,19 @@ void bridge_wait(void) { struct bridge *br; - struct iface *iface; LIST_FOR_EACH (br, node, &all_bridges) { + size_t i; + ofproto_wait(br->ofproto); if (ofproto_has_primary_controller(br->ofproto)) { continue; } mac_learning_wait(br->ml); - lacp_wait(br); - bond_wait(br); - HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) { - if (iface->cfm) { - cfm_wait(iface->cfm); - } + for (i = 0; i < br->n_ports; i++) { + port_wait(br->ports[i]); } } ovsdb_idl_wait(idl); @@ -1569,6 +1569,99 @@ bridge_unixctl_fdb_show(struct unixctl_conn *conn, ds_destroy(&ds); } +/* QoS unixctl user interface functions. */ + +struct qos_unixctl_show_cbdata { + struct ds *ds; + struct iface *iface; +}; + +static void +qos_unixctl_show_cb(unsigned int queue_id, + const struct shash *details, + void *aux) +{ + struct qos_unixctl_show_cbdata *data = aux; + struct ds *ds = data->ds; + struct iface *iface = data->iface; + struct netdev_queue_stats stats; + struct shash_node *node; + int error; + + ds_put_cstr(ds, "\n"); + if (queue_id) { + ds_put_format(ds, "Queue %u:\n", queue_id); + } else { + ds_put_cstr(ds, "Default:\n"); + } + + SHASH_FOR_EACH (node, details) { + ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data); + } + + error = netdev_get_queue_stats(iface->netdev, queue_id, &stats); + if (!error) { + if (stats.tx_packets != UINT64_MAX) { + ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets); + } + + if (stats.tx_bytes != UINT64_MAX) { + ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes); + } + + if (stats.tx_errors != UINT64_MAX) { + ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors); + } + } else { + ds_put_format(ds, "\tFailed to get statistics for queue %u: %s", + queue_id, strerror(error)); + } +} + +static void +qos_unixctl_show(struct unixctl_conn *conn, + const char *args, void *aux OVS_UNUSED) +{ + struct ds ds = DS_EMPTY_INITIALIZER; + struct shash sh = SHASH_INITIALIZER(&sh); + struct iface *iface; + const char *type; + struct shash_node *node; + struct qos_unixctl_show_cbdata data; + int error; + + iface = iface_find(args); + if (!iface) { + unixctl_command_reply(conn, 501, "no such interface"); + return; + } + + netdev_get_qos(iface->netdev, &type, &sh); + + if (*type != '\0') { + ds_put_format(&ds, "QoS: %s %s\n", iface->name, type); + + SHASH_FOR_EACH (node, &sh) { + ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data); + } + + data.ds = &ds; + data.iface = iface; + error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data); + + if (error) { + ds_put_format(&ds, "failed to dump queues: %s", strerror(error)); + } + unixctl_command_reply(conn, 200, ds_cstr(&ds)); + } else { + ds_put_format(&ds, "QoS not configured on %s\n", iface->name); + unixctl_command_reply(conn, 501, ds_cstr(&ds)); + } + + shash_destroy_free_data(&sh); + ds_destroy(&ds); +} + /* Bridge reconfiguration functions. */ static struct bridge * bridge_create(const struct ovsrec_bridge *br_cfg) @@ -1705,8 +1798,8 @@ bridge_unixctl_reconnect(struct unixctl_conn *conn, static int bridge_run_one(struct bridge *br) { + size_t i; int error; - struct iface *iface; error = ofproto_run1(br->ofproto); if (error) { @@ -1714,27 +1807,14 @@ bridge_run_one(struct bridge *br) } mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto)); - lacp_run(br); - bond_run(br); + + for (i = 0; i < br->n_ports; i++) { + port_run(br->ports[i]); + } error = ofproto_run2(br->ofproto, br->flush); br->flush = false; - HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) { - struct ofpbuf *packet; - - if (!iface->cfm) { - continue; - } - - packet = cfm_run(iface->cfm); - if (packet) { - iface_send_packet(iface, packet); - ofpbuf_uninit(packet); - free(packet); - } - } - return error; } @@ -1876,8 +1956,6 @@ bridge_ofproto_controller_for_mgmt(const struct bridge *br, 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; } @@ -1892,8 +1970,6 @@ bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c, 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); @@ -1911,11 +1987,6 @@ bridge_configure_local_iface_netdev(struct bridge *br, 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)) { @@ -2158,7 +2229,7 @@ bond_choose_iface(const struct port *port) if (iface->enabled) { return i; } else if (iface->delay_expires < next_delay_expiration - && (iface->lacp_attached + && (iface->lacp_status & LACP_ATTACHED || !(port->lacp & LACP_NEGOTIATED))) { best_down_slave = i; next_delay_expiration = iface->delay_expires; @@ -2204,7 +2275,6 @@ choose_output_iface(const struct port *port, const struct flow *flow, return false; } e->iface_tag = tag_create_random(); - ((struct port *) port)->bond_compat_is_stale = true; } *tags |= e->iface_tag; iface = port->ifaces[e->iface_idx]; @@ -2236,9 +2306,9 @@ bond_link_status_update(struct iface *iface) * They are not required to have synchronized partners because they * have no partners at all. However, they will only be attached if * negotiations failed on all interfaces in the bond. */ - up = iface->lacp_attached + up = iface->lacp_status & LACP_ATTACHED && (iface->lacp_partner.state & LACP_STATE_SYNC - || iface->lacp_status == LACP_STATUS_DEFAULTED); + || iface->lacp_status & LACP_DEFAULTED); } @@ -2334,7 +2404,6 @@ bond_enable_slave(struct iface *iface, bool enable) } moving_active_iface = false; - port->bond_compat_is_stale = true; } /* Attempts to make the sum of the bond slaves' statistics appear on the fake @@ -2381,108 +2450,97 @@ bond_link_carrier_update(struct iface *iface, bool carrier) return; } - if (iface->lacp_status == LACP_STATUS_CURRENT) { + if (iface->lacp_status & LACP_CURRENT) { iface_set_lacp_expired(iface); } iface->up = carrier; iface->lacp_tx = 0; - iface->port->bond_compat_is_stale = true; } static void -bond_run(struct bridge *br) +bond_run(struct port *port) { - size_t i, j; - - for (i = 0; i < br->n_ports; i++) { - struct port *port = br->ports[i]; - - if (port->n_ifaces >= 2) { - char *devname; + size_t i; + char *devname; - if (port->monitor) { - assert(!port->miimon); + if (port->n_ifaces < 2) { + return; + } - /* Track carrier going up and down on interfaces. */ - while (!netdev_monitor_poll(port->monitor, &devname)) { - struct iface *iface; + if (port->monitor) { + assert(!port->miimon); - iface = port_lookup_iface(port, devname); - if (iface) { - bool up = netdev_get_carrier(iface->netdev); - bond_link_carrier_update(iface, up); - } - free(devname); - } - } else { - assert(port->miimon); + /* Track carrier going up and down on interfaces. */ + while (!netdev_monitor_poll(port->monitor, &devname)) { + struct iface *iface; - if (time_msec() >= port->bond_miimon_next_update) { - for (j = 0; j < port->n_ifaces; j++) { - struct iface *iface = port->ifaces[j]; - bool up = netdev_get_miimon(iface->netdev); - bond_link_carrier_update(iface, up); - } - port->bond_miimon_next_update = time_msec() + - port->bond_miimon_interval; - } + iface = port_lookup_iface(port, devname); + if (iface) { + bool up = netdev_get_carrier(iface->netdev); + bond_link_carrier_update(iface, up); } + free(devname); + } + } else { + assert(port->miimon); - for (j = 0; j < port->n_ifaces; j++) { - bond_link_status_update(port->ifaces[j]); + if (time_msec() >= port->bond_miimon_next_update) { + for (i = 0; i < port->n_ifaces; i++) { + struct iface *iface = port->ifaces[i]; + bool up = netdev_get_miimon(iface->netdev); + bond_link_carrier_update(iface, up); } + port->bond_miimon_next_update = time_msec() + + port->bond_miimon_interval; + } + } - 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); - } - } + for (i = 0; i < port->n_ifaces; i++) { + bond_link_status_update(port->ifaces[i]); + } - if (port->bond_fake_iface - && time_msec() >= port->bond_next_fake_iface_update) { - bond_update_fake_iface_stats(port); - port->bond_next_fake_iface_update = time_msec() + 1000; - } + for (i = 0; i < port->n_ifaces; i++) { + struct iface *iface = port->ifaces[i]; + 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->bond_fake_iface + && time_msec() >= port->bond_next_fake_iface_update) { + bond_update_fake_iface_stats(port); + port->bond_next_fake_iface_update = time_msec() + 1000; } } static void -bond_wait(struct bridge *br) +bond_wait(struct port *port) { - size_t i, j; + size_t i; - for (i = 0; i < br->n_ports; i++) { - struct port *port = br->ports[i]; - if (port->n_ifaces < 2) { - continue; - } + if (port->n_ifaces < 2) { + return; + } - if (port->monitor) { - netdev_monitor_poll_wait(port->monitor); - } + if (port->monitor) { + netdev_monitor_poll_wait(port->monitor); + } - if (port->miimon) { - poll_timer_wait_until(port->bond_miimon_next_update); - } + if (port->miimon) { + poll_timer_wait_until(port->bond_miimon_next_update); + } - for (j = 0; j < port->n_ifaces; j++) { - struct iface *iface = port->ifaces[j]; - if (iface->delay_expires != LLONG_MAX) { - poll_timer_wait_until(iface->delay_expires); - } - } - if (port->bond_fake_iface) { - poll_timer_wait_until(port->bond_next_fake_iface_update); + for (i = 0; i < port->n_ifaces; i++) { + struct iface *iface = port->ifaces[i]; + if (iface->delay_expires != LLONG_MAX) { + poll_timer_wait_until(iface->delay_expires); } } + + if (port->bond_fake_iface) { + poll_timer_wait_until(port->bond_next_fake_iface_update); + } } static bool @@ -3001,33 +3059,45 @@ bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet, struct ofpbuf *actions, tag_type *tags, uint16_t *nf_output_iface, void *br_) { - struct iface *iface; struct bridge *br = br_; COVERAGE_INC(bridge_process_flow); + return process_flow(br, flow, packet, actions, tags, nf_output_iface); +} + +static bool +bridge_special_ofhook_cb(const struct flow *flow, + const struct ofpbuf *packet, void *br_) +{ + struct iface *iface; + struct bridge *br = br_; iface = iface_from_dp_ifidx(br, flow->in_port); if (cfm_should_process_flow(flow)) { - if (packet && iface->cfm) { + + if (iface && packet && iface->cfm) { + COVERAGE_INC(bridge_process_cfm); cfm_process_heartbeat(iface->cfm, packet); } return false; } else if (flow->dl_type == htons(ETH_TYPE_LACP)) { - if (packet) { + + if (iface && packet) { + COVERAGE_INC(bridge_process_lacp); lacp_process_packet(packet, iface); } return false; } - return process_flow(br, flow, packet, actions, tags, nf_output_iface); + return true; } static void bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags, const struct nlattr *actions, size_t actions_len, - unsigned long long int n_bytes, void *br_) + uint64_t n_bytes, void *br_) { struct bridge *br = br_; const struct nlattr *a; @@ -3055,7 +3125,7 @@ bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags, if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) { struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a)); if (out_port && out_port->n_ifaces >= 2 && - out_port->bond_mode == BM_SLB) { + out_port->bond_mode != BM_AB) { uint16_t vlan = (flow->vlan_tci ? vlan_tci_to_vid(flow->vlan_tci) : OFP_VLAN_NONE); @@ -3080,7 +3150,7 @@ bridge_account_checkpoint_ofhook_cb(void *br_) now = time_msec(); for (i = 0; i < br->n_ports; i++) { struct port *port = br->ports[i]; - if (port->n_ifaces > 1 && port->bond_mode == BM_SLB + if (port->n_ifaces > 1 && port->bond_mode != BM_AB && now >= port->bond_next_rebalance) { port->bond_next_rebalance = now + port->bond_rebalance_interval; bond_rebalance_port(port); @@ -3090,6 +3160,7 @@ bridge_account_checkpoint_ofhook_cb(void *br_) static struct ofhooks bridge_ofhooks = { bridge_normal_ofhook_cb, + bridge_special_ofhook_cb, bridge_account_flow_ofhook_cb, bridge_account_checkpoint_ofhook_cb, }; @@ -3110,7 +3181,8 @@ lacp_process_packet(const struct ofpbuf *packet, struct iface *iface) return; } - iface->lacp_status = LACP_STATUS_CURRENT; + iface->lacp_status |= LACP_CURRENT; + iface->lacp_status &= ~(LACP_EXPIRED | LACP_DEFAULTED); iface->lacp_rx = time_msec() + LACP_SLOW_TIME_RX; iface->lacp_actor.state = iface_get_lacp_state(iface); @@ -3146,7 +3218,7 @@ lacp_update_ifaces(struct port *port) struct iface *iface = port->ifaces[i]; struct lacp_info pri; - iface->lacp_attached = true; + iface->lacp_status |= LACP_ATTACHED; ofproto_revalidate(port->bridge->ofproto, iface->tag); /* Don't allow loopback interfaces to send traffic or lead. */ @@ -3154,11 +3226,11 @@ lacp_update_ifaces(struct port *port) iface->lacp_actor.sysid)) { VLOG_WARN_RL(&rl, "iface %s: Loopback detected. Interface is " "connected to its own bridge", iface->name); - iface->lacp_attached = false; + iface->lacp_status &= ~LACP_ATTACHED; continue; } - if (iface->lacp_status == LACP_STATUS_DEFAULTED) { + if (iface->lacp_status & LACP_DEFAULTED) { continue; } @@ -3180,11 +3252,11 @@ lacp_update_ifaces(struct port *port) for (i = 0; i < port->n_ifaces; i++) { struct iface *iface = port->ifaces[i]; - if (iface->lacp_status == LACP_STATUS_DEFAULTED + if (iface->lacp_status & LACP_DEFAULTED || lead->lacp_partner.key != iface->lacp_partner.key || !eth_addr_equals(lead->lacp_partner.sysid, iface->lacp_partner.sysid)) { - iface->lacp_attached = false; + iface->lacp_status &= ~LACP_ATTACHED; } } } @@ -3193,92 +3265,84 @@ static bool lacp_iface_may_tx(const struct iface *iface) { return iface->port->lacp & LACP_ACTIVE - || iface->lacp_status != LACP_STATUS_DEFAULTED; + || iface->lacp_status & (LACP_CURRENT | LACP_EXPIRED); } static void -lacp_run(struct bridge *br) +lacp_run(struct port *port) { - size_t i, j; + size_t i; struct ofpbuf packet; - ofpbuf_init(&packet, ETH_HEADER_LEN + LACP_PDU_LEN); - - for (i = 0; i < br->n_ports; i++) { - struct port *port = br->ports[i]; + if (!port->lacp) { + return; + } - if (!port->lacp) { - continue; - } + ofpbuf_init(&packet, ETH_HEADER_LEN + LACP_PDU_LEN); - for (j = 0; j < port->n_ifaces; j++) { - struct iface *iface = port->ifaces[j]; + for (i = 0; i < port->n_ifaces; i++) { + struct iface *iface = port->ifaces[i]; - if (time_msec() > iface->lacp_rx) { - if (iface->lacp_status == LACP_STATUS_CURRENT) { - iface_set_lacp_expired(iface); - } else if (iface->lacp_status == LACP_STATUS_EXPIRED) { - iface_set_lacp_defaulted(iface); - } + if (time_msec() > iface->lacp_rx) { + if (iface->lacp_status & LACP_CURRENT) { + iface_set_lacp_expired(iface); + } else if (iface->lacp_status & LACP_EXPIRED) { + iface_set_lacp_defaulted(iface); } } + } - if (port->lacp_need_update) { - lacp_update_ifaces(port); - } - - for (j = 0; j < port->n_ifaces; j++) { - struct iface *iface = port->ifaces[j]; - uint8_t ea[ETH_ADDR_LEN]; - int error; + if (port->lacp_need_update) { + lacp_update_ifaces(port); + } - if (time_msec() < iface->lacp_tx || !lacp_iface_may_tx(iface)) { - continue; - } + for (i = 0; i < port->n_ifaces; i++) { + struct iface *iface = port->ifaces[i]; + uint8_t ea[ETH_ADDR_LEN]; + int error; - error = netdev_get_etheraddr(iface->netdev, ea); - if (!error) { - iface->lacp_actor.state = iface_get_lacp_state(iface); - compose_lacp_packet(&packet, &iface->lacp_actor, - &iface->lacp_partner, ea); - iface_send_packet(iface, &packet); - } else { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10); - VLOG_ERR_RL(&rl, "iface %s: failed to obtain Ethernet address " - "(%s)", iface->name, strerror(error)); - } + if (time_msec() < iface->lacp_tx || !lacp_iface_may_tx(iface)) { + continue; + } - iface->lacp_tx = time_msec() + - (iface->lacp_partner.state & LACP_STATE_TIME - ? LACP_FAST_TIME_TX - : LACP_SLOW_TIME_TX); + error = netdev_get_etheraddr(iface->netdev, ea); + if (!error) { + iface->lacp_actor.state = iface_get_lacp_state(iface); + compose_lacp_packet(&packet, &iface->lacp_actor, + &iface->lacp_partner, ea); + iface_send_packet(iface, &packet); + } else { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10); + VLOG_ERR_RL(&rl, "iface %s: failed to obtain Ethernet address " + "(%s)", iface->name, strerror(error)); } + + iface->lacp_tx = time_msec() + + (iface->lacp_partner.state & LACP_STATE_TIME + ? LACP_FAST_TIME_TX + : LACP_SLOW_TIME_TX); } ofpbuf_uninit(&packet); } static void -lacp_wait(struct bridge *br) +lacp_wait(struct port *port) { - size_t i, j; - - for (i = 0; i < br->n_ports; i++) { - struct port *port = br->ports[i]; + size_t i; - if (!port->lacp) { - continue; - } + if (!port->lacp) { + return; + } - for (j = 0; j < port->n_ifaces; j++) { - struct iface *iface = port->ifaces[j]; + for (i = 0; i < port->n_ifaces; i++) { + struct iface *iface = port->ifaces[i]; - if (lacp_iface_may_tx(iface)) { - poll_timer_wait_until(iface->lacp_tx); - } + if (lacp_iface_may_tx(iface)) { + poll_timer_wait_until(iface->lacp_tx); + } - if (iface->lacp_status != LACP_STATUS_DEFAULTED) { - poll_timer_wait_until(iface->lacp_rx); - } + if (iface->lacp_status & (LACP_CURRENT | LACP_EXPIRED)) { + poll_timer_wait_until(iface->lacp_rx); } } } @@ -3421,7 +3485,7 @@ bond_shift_load(struct slave_balance *from, struct slave_balance *to, struct port *port = from->iface->port; uint64_t delta = hash->tx_bytes; - assert(port->bond_mode == BM_SLB); + assert(port->bond_mode != BM_AB); VLOG_INFO("bond %s: shift %"PRIu64"kB of load (with hash %td) " "from %s to %s (now carrying %"PRIu64"kB and " @@ -3575,7 +3639,6 @@ bond_rebalance_port(struct port *port) } if (i < from->n_hashes) { bond_shift_load(from, to, i); - port->bond_compat_is_stale = true; /* If the result of the migration changed the relative order of * 'from' and 'to' swap them back to maintain invariants. */ @@ -3830,15 +3893,19 @@ bond_unixctl_show(struct unixctl_conn *conn, if (port->lacp) { ds_put_cstr(&ds, "\tstatus: "); - if (iface->lacp_status == LACP_STATUS_CURRENT) { + if (iface->lacp_status & LACP_CURRENT) { ds_put_cstr(&ds, "current "); - } else if (iface->lacp_status == LACP_STATUS_EXPIRED) { + } + + if (iface->lacp_status & LACP_EXPIRED) { ds_put_cstr(&ds, "expired "); - } else { + } + + if (iface->lacp_status & LACP_DEFAULTED) { ds_put_cstr(&ds, "defaulted "); } - if (iface->lacp_attached) { + if (iface->lacp_status & LACP_ATTACHED) { ds_put_cstr(&ds, "attached "); } @@ -3984,7 +4051,6 @@ bond_unixctl_migrate(struct unixctl_conn *conn, const char *args_, ofproto_revalidate(port->bridge->ofproto, entry->iface_tag); entry->iface_idx = iface->port_ifidx; entry->iface_tag = tag_create_random(); - port->bond_compat_is_stale = true; unixctl_command_reply(conn, 200, "migrated"); } @@ -4136,6 +4202,44 @@ bond_init(void) /* Port functions. */ +static void +port_run(struct port *port) +{ + size_t i; + + lacp_run(port); + bond_run(port); + + for (i = 0; i < port->n_ifaces; i++) { + struct iface *iface = port->ifaces[i]; + + if (iface->cfm) { + struct ofpbuf *packet = cfm_run(iface->cfm); + if (packet) { + iface_send_packet(iface, packet); + ofpbuf_uninit(packet); + free(packet); + } + } + } +} + +static void +port_wait(struct port *port) +{ + size_t i; + + lacp_wait(port); + bond_wait(port); + + for (i = 0; i < port->n_ifaces; i++) { + struct iface *iface = port->ifaces[i]; + if (iface->cfm) { + cfm_wait(iface->cfm); + } + } +} + static struct port * port_create(struct bridge *br, const char *name) { @@ -4416,9 +4520,6 @@ port_destroy(struct port *port) struct port *del; int i; - proc_net_compat_update_vlan(port->name, NULL, 0); - proc_net_compat_update_bond(port->name, NULL); - for (i = 0; i < MAX_MIRRORS; i++) { struct mirror *m = br->mirrors[i]; if (m && m->out_port == port) { @@ -4521,13 +4622,11 @@ port_update_bonding(struct port *port) } if (port->n_ifaces < 2) { /* Not a bonded port. */ - if (port->bond_hash) { - free(port->bond_hash); - port->bond_hash = NULL; - port->bond_compat_is_stale = true; - } - + free(port->bond_hash); + port->bond_hash = NULL; port->bond_fake_iface = false; + port->active_iface = -1; + port->no_ifaces_tag = 0; } else { size_t i; @@ -4538,138 +4637,33 @@ port_update_bonding(struct port *port) e->iface_idx = -1; e->tx_bytes = 0; } - port->no_ifaces_tag = tag_create_random(); - bond_choose_active_iface(port); port->bond_next_rebalance = time_msec() + port->bond_rebalance_interval; - - if (port->cfg->bond_fake_iface) { - port->bond_next_fake_iface_update = time_msec(); - } } else if (port->bond_mode == BM_AB) { free(port->bond_hash); port->bond_hash = NULL; } - port->bond_compat_is_stale = true; - port->bond_fake_iface = port->cfg->bond_fake_iface; - if (!port->miimon) { - port->monitor = netdev_monitor_create(); - for (i = 0; i < port->n_ifaces; i++) { - netdev_monitor_add(port->monitor, port->ifaces[i]->netdev); - } + if (!port->no_ifaces_tag) { + port->no_ifaces_tag = tag_create_random(); } - } -} -static void -port_update_bond_compat(struct port *port) -{ - struct compat_bond_hash compat_hashes[BOND_MASK + 1]; - struct compat_bond bond; - size_t i; - - if (port->n_ifaces < 2 || port->bond_mode != BM_SLB) { - proc_net_compat_update_bond(port->name, NULL); - return; - } - - bond.up = false; - bond.updelay = port->updelay; - bond.downdelay = port->downdelay; - - bond.n_hashes = 0; - bond.hashes = compat_hashes; - if (port->bond_hash) { - const struct bond_entry *e; - for (e = port->bond_hash; e <= &port->bond_hash[BOND_MASK]; e++) { - if (e->iface_idx >= 0 && e->iface_idx < port->n_ifaces) { - struct compat_bond_hash *cbh = &bond.hashes[bond.n_hashes++]; - cbh->hash = e - port->bond_hash; - cbh->netdev_name = port->ifaces[e->iface_idx]->name; - } + if (port->active_iface < 0) { + bond_choose_active_iface(port); } - } - bond.n_slaves = port->n_ifaces; - bond.slaves = xmalloc(port->n_ifaces * sizeof *bond.slaves); - for (i = 0; i < port->n_ifaces; i++) { - struct iface *iface = port->ifaces[i]; - struct compat_bond_slave *slave = &bond.slaves[i]; - slave->name = iface->name; - - /* 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 - * 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 - * to be "up" and anything else to be "down". - */ - slave->up = iface->enabled && iface->delay_expires == LLONG_MAX; - if (slave->up) { - bond.up = true; - } - netdev_get_etheraddr(iface->netdev, slave->mac); - } - - if (port->bond_fake_iface) { - struct netdev *bond_netdev; - - if (!netdev_open_default(port->name, &bond_netdev)) { - if (bond.up) { - netdev_turn_flags_on(bond_netdev, NETDEV_UP, true); - } else { - netdev_turn_flags_off(bond_netdev, NETDEV_UP, true); - } - netdev_close(bond_netdev); + port->bond_fake_iface = port->cfg->bond_fake_iface; + if (port->bond_fake_iface) { + port->bond_next_fake_iface_update = time_msec(); } - } - - proc_net_compat_update_bond(port->name, &bond); - free(bond.slaves); -} -static void -port_update_vlan_compat(struct port *port) -{ - struct bridge *br = port->bridge; - char *vlandev_name = NULL; - - if (port->vlan > 0) { - /* Figure out the name that the VLAN device should actually have, if it - * existed. This takes some work because the VLAN device would not - * have port->name in its name; rather, it would have the trunk port's - * name, and 'port' would be attached to a bridge that also had the - * VLAN device one of its ports. So we need to find a trunk port that - * includes port->vlan. - * - * There might be more than one candidate. This doesn't happen on - * XenServer, so if it happens we just pick the first choice in - * alphabetical order instead of creating multiple VLAN devices. */ - size_t i; - for (i = 0; i < br->n_ports; i++) { - struct port *p = br->ports[i]; - if (port_trunks_vlan(p, port->vlan) - && p->n_ifaces - && (!vlandev_name || strcmp(p->name, vlandev_name) <= 0)) - { - 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)) { - vlandev_name = p->name; - } + if (!port->miimon) { + port->monitor = netdev_monitor_create(); + for (i = 0; i < port->n_ifaces; i++) { + netdev_monitor_add(port->monitor, port->ifaces[i]->netdev); } } } - proc_net_compat_update_vlan(port->name, vlandev_name, port->vlan); } /* Interface functions. */ @@ -4679,7 +4673,8 @@ iface_set_lacp_defaulted(struct iface *iface) { memset(&iface->lacp_partner, 0, sizeof iface->lacp_partner); - iface->lacp_status = LACP_STATUS_DEFAULTED; + iface->lacp_status |= LACP_DEFAULTED; + iface->lacp_status &= ~(LACP_CURRENT | LACP_EXPIRED); iface->lacp_tx = 0; iface->port->lacp_need_update = true; } @@ -4687,7 +4682,8 @@ iface_set_lacp_defaulted(struct iface *iface) static void iface_set_lacp_expired(struct iface *iface) { - iface->lacp_status = LACP_STATUS_EXPIRED; + iface->lacp_status &= ~LACP_CURRENT; + iface->lacp_status |= LACP_EXPIRED; iface->lacp_partner.state |= LACP_STATE_TIME; iface->lacp_partner.state &= ~LACP_STATE_SYNC; @@ -4704,13 +4700,15 @@ iface_get_lacp_state(const struct iface *iface) state |= LACP_STATE_ACT; } - if (iface->lacp_status == LACP_STATUS_DEFAULTED) { - state |= LACP_STATE_DEF; - } else if (iface->lacp_attached) { + if (iface->lacp_status & LACP_ATTACHED) { state |= LACP_STATE_SYNC; } - if (iface->lacp_status == LACP_STATUS_EXPIRED) { + if (iface->lacp_status & LACP_DEFAULTED) { + state |= LACP_STATE_DEF; + } + + if (iface->lacp_status & LACP_EXPIRED) { state |= LACP_STATE_EXP; } @@ -4859,6 +4857,21 @@ iface_lookup(const struct bridge *br, const char *name) return shash_find_data(&br->iface_by_name, name); } +static struct iface * +iface_find(const char *name) +{ + const struct bridge *br; + + LIST_FOR_EACH (br, node, &all_bridges) { + struct iface *iface = iface_lookup(br, name); + + if (iface) { + return iface; + } + } + return NULL; +} + static struct iface * iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx) {