X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=c69557bd835186b4d5d2447f93195a211ea90165;hb=c97fb13280c565f55ed0de7cf0bf06ffe8320b70;hp=20cbee5c33cddff35bc1a1c706c4bee5c5718e00;hpb=ea763e0e2828c35660310f8b7791781b17b30cbd;p=openvswitch diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index 20cbee5c..c69557bd 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -1,4 +1,4 @@ -/* Copyright (c) 2008, 2009, 2010 Nicira Networks +/* Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -118,9 +118,9 @@ struct bond_entry { tag_type iface_tag; /* Tag associated with iface_idx. */ }; -enum bond_type { - BT_SLB, /* Source Load Balance (Default). */ - BT_AB /* Active Backup. */ +enum bond_mode { + BM_SLB, /* Source Load Balance (Default). */ + BM_AB /* Active Backup. */ }; #define MAX_MIRRORS 32 @@ -160,13 +160,16 @@ struct port { size_t n_ifaces, allocated_ifaces; /* Bonding info. */ - enum bond_type bond_type; /* Type of the bond. BT_SLB is the default. */ + enum bond_mode bond_mode; /* Type of the bond. BM_SLB is the default. */ int active_iface; /* Ifidx on which bcasts accepted, or -1. */ 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. */ + long long int bond_miimon_next_update; /* Time of next miimon update. */ long long int bond_next_fake_iface_update; /* Time of next update. */ struct netdev_monitor *monitor; /* Tracks carrier up/down status. */ @@ -603,38 +606,35 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) * that port already belongs to a different datapath, so we must do all * port deletions before any port additions. */ LIST_FOR_EACH (br, node, &all_bridges) { - struct odp_port *dpif_ports; - size_t n_dpif_ports; + struct dpif_port_dump dump; struct shash want_ifaces; + struct dpif_port dpif_port; - 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 (!shash_find(&want_ifaces, p->devname) - && strcmp(p->devname, br->name)) { - int retval = dpif_port_del(br->dpif, p->port); + DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) { + if (!shash_find(&want_ifaces, dpif_port.name) + && 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", - p->devname, dpif_name(br->dpif), + dpif_port.name, dpif_name(br->dpif), strerror(retval)); } } } shash_destroy(&want_ifaces); - free(dpif_ports); } LIST_FOR_EACH (br, node, &all_bridges) { - struct odp_port *dpif_ports; - size_t n_dpif_ports; struct shash cur_ifaces, want_ifaces; + struct dpif_port_dump dump; + struct dpif_port dpif_port; /* 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, &dpif_ports[i]); + DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) { + struct dpif_port *port_info = xmalloc(sizeof *port_info); + dpif_port_clone(port_info, &dpif_port); + shash_add(&cur_ifaces, dpif_port.name, port_info); } /* Get the set of interfaces we want on this datapath. */ @@ -644,10 +644,13 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) SHASH_FOR_EACH (node, &want_ifaces) { const char *if_name = node->name; struct iface *iface = node->data; - struct odp_port *dpif_port = shash_find_data(&cur_ifaces, if_name); - const char *type = iface ? iface->type : "internal"; + struct dpif_port *dpif_port; + const char *type; int error; + type = iface ? iface->type : "internal"; + dpif_port = shash_find_data(&cur_ifaces, if_name); + /* 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). */ @@ -655,7 +658,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) || (iface && iface->netdev && strcmp(type, netdev_get_type(iface->netdev)))) { if (dpif_port) { - error = ofproto_port_del(br->ofproto, dpif_port->port); + error = ofproto_port_del(br->ofproto, dpif_port->port_no); if (error) { continue; } @@ -725,13 +728,18 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) shash_from_ovs_idl_map(iface->cfg->key_options, iface->cfg->value_options, iface->cfg->n_options, &args); - netdev_reconfigure(iface->netdev, &args); + netdev_set_config(iface->netdev, &args); shash_destroy(&args); } } - free(dpif_ports); - shash_destroy(&cur_ifaces); shash_destroy(&want_ifaces); + + SHASH_FOR_EACH (node, &cur_ifaces) { + struct dpif_port *port_info = node->data; + dpif_port_destroy(port_info); + free(port_info); + } + shash_destroy(&cur_ifaces); } sflow_bridge_number = 0; LIST_FOR_EACH (br, node, &all_bridges) { @@ -1117,6 +1125,13 @@ iface_refresh_status(struct iface *iface) { struct shash sh; + enum netdev_flags flags; + uint32_t current; + int64_t bps; + int mtu; + int64_t mtu_64; + int error; + shash_init(&sh); if (!netdev_get_status(iface->netdev, &sh)) { @@ -1133,6 +1148,42 @@ iface_refresh_status(struct iface *iface) } shash_destroy_free_data(&sh); + + error = netdev_get_flags(iface->netdev, &flags); + if (!error) { + ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down"); + } + else { + ovsrec_interface_set_admin_state(iface->cfg, NULL); + } + + error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL); + if (!error) { + ovsrec_interface_set_duplex(iface->cfg, + netdev_features_is_full_duplex(current) + ? "full" : "half"); + /* warning: uint64_t -> int64_t conversion */ + bps = netdev_features_to_bps(current); + ovsrec_interface_set_link_speed(iface->cfg, &bps, 1); + } + else { + ovsrec_interface_set_duplex(iface->cfg, NULL); + ovsrec_interface_set_link_speed(iface->cfg, NULL, 0); + } + + + ovsrec_interface_set_link_state(iface->cfg, + netdev_get_carrier(iface->netdev) + ? "up" : "down"); + + error = netdev_get_mtu(iface->netdev, &mtu); + if (!error) { + mtu_64 = mtu; + ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1); + } + else { + ovsrec_interface_set_mtu(iface->cfg, NULL, 0); + } } static void @@ -1272,6 +1323,49 @@ refresh_system_stats(const struct ovsrec_open_vswitch *cfg) &datum); } +static inline const char * +nx_role_to_str(enum nx_role role) +{ + switch (role) { + case NX_ROLE_OTHER: + return "other"; + case NX_ROLE_MASTER: + return "master"; + case NX_ROLE_SLAVE: + return "slave"; + default: + return "*** INVALID ROLE ***"; + } +} + +static void +bridge_refresh_controller_status(const struct bridge *br) +{ + struct shash info; + const struct ovsrec_controller *cfg; + + ofproto_get_ofproto_controller_info(br->ofproto, &info); + + OVSREC_CONTROLLER_FOR_EACH(cfg, idl) { + struct ofproto_controller_info *cinfo = + shash_find_data(&info, cfg->target); + + if (cinfo) { + ovsrec_controller_set_is_connected(cfg, cinfo->is_connected); + ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role)); + ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys, + (char **) cinfo->pairs.values, + cinfo->pairs.n); + } else { + ovsrec_controller_set_is_connected(cfg, false); + ovsrec_controller_set_role(cfg, NULL); + ovsrec_controller_set_status(cfg, NULL, NULL, 0); + } + } + + ofproto_free_ofproto_controller_info(&info); +} + void bridge_run(void) { @@ -1296,6 +1390,20 @@ bridge_run(void) /* (Re)configure if necessary. */ database_changed = ovsdb_idl_run(idl); cfg = ovsrec_open_vswitch_first(idl); +#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. + * + * We do this before bridge_reconfigure() because that function might + * initiate SSL connections and thus requires SSL to be configured. */ + if (cfg && cfg->ssl) { + const struct ovsrec_ssl *ssl = cfg->ssl; + + 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 if (database_changed || datapath_destroyed) { if (cfg) { struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl); @@ -1315,18 +1423,6 @@ bridge_run(void) } } -#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; - - 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 - /* Refresh system and interface stats if necessary. */ if (time_msec() >= stats_timer) { if (cfg) { @@ -1347,6 +1443,7 @@ bridge_run(void) iface_refresh_status(iface); } } + bridge_refresh_controller_status(br); } refresh_system_stats(cfg); ovsdb_idl_txn_commit(txn); @@ -1938,8 +2035,8 @@ bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces) static void bridge_fetch_dp_ifaces(struct bridge *br) { - struct odp_port *dpif_ports; - size_t n_dpif_ports; + struct dpif_port_dump dump; + struct dpif_port dpif_port; size_t i, j; /* Reset all interface numbers. */ @@ -1952,19 +2049,17 @@ bridge_fetch_dp_ifaces(struct bridge *br) } hmap_clear(&br->ifaces); - 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); + DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) { + struct iface *iface = iface_lookup(br, dpif_port.name); if (iface) { if (iface->dp_ifidx >= 0) { VLOG_WARN("%s reported interface %s twice", - dpif_name(br->dpif), p->devname); - } else if (iface_from_dp_ifidx(br, p->port)) { + dpif_name(br->dpif), dpif_port.name); + } else if (iface_from_dp_ifidx(br, dpif_port.port_no)) { VLOG_WARN("%s reported interface %"PRIu16" twice", - dpif_name(br->dpif), p->port); + dpif_name(br->dpif), dpif_port.port_no); } else { - iface->dp_ifidx = p->port; + iface->dp_ifidx = dpif_port.port_no; hmap_insert(&br->ifaces, &iface->dp_ifidx_node, hash_int(iface->dp_ifidx, 0)); } @@ -1975,7 +2070,6 @@ bridge_fetch_dp_ifaces(struct bridge *br) : -1)); } } - free(dpif_ports); } /* Bridge packet processing functions. */ @@ -1990,7 +2084,7 @@ static struct bond_entry * lookup_bond_entry(const struct port *port, const uint8_t mac[ETH_ADDR_LEN], uint16_t vlan) { - assert(port->bond_type == BT_SLB); + assert(port->bond_mode == BM_SLB); return &port->bond_hash[bond_hash(mac, vlan)]; } @@ -2033,13 +2127,13 @@ choose_output_iface(const struct port *port, const uint8_t *dl_src, assert(port->n_ifaces); if (port->n_ifaces == 1) { iface = port->ifaces[0]; - } else if (port->bond_type == BT_AB) { + } else if (port->bond_mode == BM_AB) { if (port->active_iface < 0) { *tags |= port->no_ifaces_tag; return false; } iface = port->ifaces[port->active_iface]; - } else if (port->bond_type == BT_SLB){ + } else if (port->bond_mode == BM_SLB){ struct bond_entry *e = lookup_bond_entry(port, dl_src, vlan); if (e->iface_idx < 0 || e->iface_idx >= port->n_ifaces || !port->ifaces[e->iface_idx]->enabled) { @@ -2073,8 +2167,8 @@ bond_link_status_update(struct iface *iface, bool carrier) /* Nothing to do. */ return; } - VLOG_INFO_RL(&rl, "interface %s: carrier %s", - iface->name, carrier ? "detected" : "dropped"); + VLOG_INFO_RL(&rl, "interface %s: link state %s", + iface->name, carrier ? "up" : "down"); if (carrier == iface->enabled) { iface->delay_expires = LLONG_MAX; VLOG_INFO_RL(&rl, "interface %s: will not be %s", @@ -2212,18 +2306,36 @@ bond_run(struct bridge *br) 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; + if (port->monitor) { + assert(!port->miimon); - iface = port_lookup_iface(port, devname); - if (iface) { - bool carrier = netdev_get_carrier(iface->netdev); + /* 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 up = netdev_get_carrier(iface->netdev); - bond_link_status_update(iface, carrier); - port_update_bond_compat(port); + bond_link_status_update(iface, up); + port_update_bond_compat(port); + } + free(devname); + } + } else { + assert(port->miimon); + + 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_status_update(iface, up); + port_update_bond_compat(port); + } + port->bond_miimon_next_update = time_msec() + + port->bond_miimon_interval; } - free(devname); } for (j = 0; j < port->n_ifaces; j++) { @@ -2257,7 +2369,15 @@ bond_wait(struct bridge *br) if (port->n_ifaces < 2) { continue; } - 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); + } + for (j = 0; j < port->n_ifaces; j++) { struct iface *iface = port->ifaces[j]; if (iface->delay_expires != LLONG_MAX) { @@ -2829,7 +2949,7 @@ bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags, if (nl_attr_type(a) == ODPAT_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_type == BT_SLB) { + out_port->bond_mode == BM_SLB) { uint16_t vlan = (flow->vlan_tci ? vlan_tci_to_vid(flow->vlan_tci) : OFP_VLAN_NONE); @@ -2855,7 +2975,7 @@ 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_type == BT_SLB + if (port->n_ifaces > 1 && port->bond_mode == BM_SLB && now >= port->bond_next_rebalance) { port->bond_next_rebalance = now + port->bond_rebalance_interval; bond_rebalance_port(port); @@ -2884,13 +3004,13 @@ struct slave_balance { }; static const char * -bond_type_to_string(enum bond_type bt) { - static char *bt_slb = "slb"; - static char *bt_ab = "active-backup"; +bond_mode_to_string(enum bond_mode bm) { + static char *bm_slb = "balance-slb"; + static char *bm_ab = "active-backup"; - switch (bt) { - case BT_SLB: return bt_slb; - case BT_AB: return bt_ab; + switch (bm) { + case BM_SLB: return bm_slb; + case BM_AB: return bm_ab; } NOT_REACHED(); @@ -3005,7 +3125,7 @@ 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_type == BT_SLB); + assert(port->bond_mode == BM_SLB); VLOG_INFO("bond %s: shift %"PRIu64"kB of load (with hash %td) " "from %s to %s (now carrying %"PRIu64"kB and " @@ -3049,7 +3169,7 @@ bond_rebalance_port(struct port *port) struct bond_entry *e; size_t i; - assert(port->bond_type == BT_SLB); + assert(port->bond_mode == BM_SLB); /* Sets up 'bals' to describe each of the port's interfaces, sorted in * descending order of tx_bytes, so that bals[0] represents the most @@ -3274,7 +3394,7 @@ bond_unixctl_list(struct unixctl_conn *conn, size_t j; ds_put_format(&ds, "%s\t%s\t%s\t", br->name, port->name, - bond_type_to_string(port->bond_type)); + bond_mode_to_string(port->bond_mode)); for (j = 0; j < port->n_ifaces; j++) { const struct iface *iface = port->ifaces[j]; if (j) { @@ -3322,12 +3442,20 @@ bond_unixctl_show(struct unixctl_conn *conn, return; } - ds_put_format(&ds, "bond_type: %s\n", - bond_type_to_string(port->bond_type)); + ds_put_format(&ds, "bond_mode: %s\n", + bond_mode_to_string(port->bond_mode)); + ds_put_format(&ds, "bond-detect-mode: %s\n", + port->miimon ? "miimon" : "carrier"); + + if (port->miimon) { + ds_put_format(&ds, "bond-miimon-interval: %lld\n", + port->bond_miimon_interval); + } + ds_put_format(&ds, "updelay: %d ms\n", port->updelay); ds_put_format(&ds, "downdelay: %d ms\n", port->downdelay); - if (port->bond_type == BT_SLB) { + if (port->bond_mode == BM_SLB) { ds_put_format(&ds, "next rebalance: %lld ms\n", port->bond_next_rebalance - time_msec()); } @@ -3348,7 +3476,7 @@ bond_unixctl_show(struct unixctl_conn *conn, iface->delay_expires - time_msec()); } - if (port->bond_type != BT_SLB) { + if (port->bond_mode != BM_SLB) { continue; } @@ -3411,7 +3539,7 @@ bond_unixctl_migrate(struct unixctl_conn *conn, const char *args_, return; } - if (port->bond_type != BT_SLB) { + if (port->bond_mode != BM_SLB) { unixctl_command_reply(conn, 501, "not an SLB bond"); return; } @@ -3655,8 +3783,9 @@ port_del_ifaces(struct port *port, const struct ovsrec_port *cfg) static void port_reconfigure(struct port *port, const struct ovsrec_port *cfg) { + const char *detect_mode; struct shash new_ifaces; - long long int next_rebalance; + long long int next_rebalance, miimon_next_update; unsigned long *trunks; int vlan; size_t i; @@ -3682,16 +3811,39 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) port->bond_next_rebalance = next_rebalance; } - if (!port->cfg->bond_type || - !strcmp(port->cfg->bond_type, bond_type_to_string(BT_SLB))) { - port->bond_type = BT_SLB; - } else if (!strcmp(port->cfg->bond_type, bond_type_to_string(BT_AB))) { - port->bond_type = BT_AB; + detect_mode = get_port_other_config(cfg, "bond-detect-mode", + "carrier"); + + if (!strcmp(detect_mode, "carrier")) { + port->miimon = false; + } else if (!strcmp(detect_mode, "miimon")) { + port->miimon = true; + } else { + port->miimon = false; + VLOG_WARN("port %s: unsupported bond-detect-mode %s, defaulting to " + "carrier", port->name, detect_mode); + } + + port->bond_miimon_interval = atoi( + get_port_other_config(cfg, "bond-miimon-interval", "200")); + if (port->bond_miimon_interval < 100) { + port->bond_miimon_interval = 100; + } + miimon_next_update = time_msec() + port->bond_miimon_interval; + if (port->bond_miimon_next_update > miimon_next_update) { + port->bond_miimon_next_update = miimon_next_update; + } + + if (!port->cfg->bond_mode || + !strcmp(port->cfg->bond_mode, bond_mode_to_string(BM_SLB))) { + port->bond_mode = BM_SLB; + } else if (!strcmp(port->cfg->bond_mode, bond_mode_to_string(BM_AB))) { + port->bond_mode = BM_AB; } else { - port->bond_type = BT_SLB; - VLOG_WARN("port %s: unknown bond_type %s, defaulting to %s", - port->name, port->cfg->bond_type, - bond_type_to_string(port->bond_type)); + port->bond_mode = BM_SLB; + VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s", + port->name, port->cfg->bond_mode, + bond_mode_to_string(port->bond_mode)); } /* Add new interfaces and update 'cfg' member of existing ones. */ @@ -3866,7 +4018,7 @@ port_update_bonding(struct port *port) } else { size_t i; - if (port->bond_type == BT_SLB && !port->bond_hash) { + if (port->bond_mode == BM_SLB && !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]; @@ -3881,16 +4033,18 @@ port_update_bonding(struct port *port) if (port->cfg->bond_fake_iface) { port->bond_next_fake_iface_update = time_msec(); } - } else if (port->bond_type != BT_SLB) { + } else if (port->bond_mode != BM_SLB) { free(port->bond_hash); port->bond_hash = NULL; } 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); + 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); + } } } } @@ -3902,7 +4056,7 @@ port_update_bond_compat(struct port *port) struct compat_bond bond; size_t i; - if (port->n_ifaces < 2 || port->bond_type != BT_SLB) { + if (port->n_ifaces < 2 || port->bond_mode != BM_SLB) { proc_net_compat_update_bond(port->name, NULL); return; }