X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=8ade873faf5133c7043374c7fb1dbef7ddf60b78;hb=cdee00fd635d1e0f1eeb5d9c009daeb59abd4777;hp=75cce3c8da315f1889e4bf641d6222b5051500a1;hpb=27bcf966b4057623f7b4d856c0348a1e0eb452e0;p=openvswitch diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index 75cce3c8..8ade873f 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -15,6 +15,7 @@ #include #include "bridge.h" +#include "byte-order.h" #include #include #include @@ -31,6 +32,8 @@ #include #include #include "bitmap.h" +#include "cfm.h" +#include "classifier.h" #include "coverage.h" #include "dirs.h" #include "dpif.h" @@ -42,6 +45,7 @@ #include "list.h" #include "mac-learning.h" #include "netdev.h" +#include "netlink.h" #include "odp-util.h" #include "ofp-print.h" #include "ofpbuf.h" @@ -65,10 +69,13 @@ #include "vswitchd/vswitch-idl.h" #include "xenserver.h" #include "vlog.h" -#include "xtoxll.h" #include "sflow_api.h" -VLOG_DEFINE_THIS_MODULE(bridge) +VLOG_DEFINE_THIS_MODULE(bridge); + +COVERAGE_DEFINE(bridge_flush); +COVERAGE_DEFINE(bridge_process_flow); +COVERAGE_DEFINE(bridge_reconfigure); struct dst { uint16_t vlan; @@ -90,6 +97,7 @@ struct iface { struct netdev *netdev; /* Network device. */ bool enabled; /* May be chosen for flows? */ const char *type; /* Usually same as cfg->type. */ + struct cfm *cfm; /* Connectivity Fault Management */ const struct ovsrec_interface *cfg; }; @@ -255,7 +263,11 @@ static struct iface *iface_lookup(const struct bridge *, const char *name); static struct iface *iface_from_dp_ifidx(const struct bridge *, uint16_t dp_ifidx); static void iface_set_mac(struct iface *); +static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport); static void iface_update_qos(struct iface *, const struct ovsrec_qos *); +static void iface_update_cfm(struct iface *); +static void iface_refresh_cfm_stats(struct iface *iface); +static void iface_send_packet(struct iface *, struct ofpbuf *packet); static void shash_from_ovs_idl_map(char **keys, char **values, size_t n, struct shash *); @@ -272,10 +284,10 @@ void bridge_init(const char *remote) { /* Create connection to database. */ - idl = ovsdb_idl_create(remote, &ovsrec_idl_class); + idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true); - ovsdb_idl_set_write_only(idl, &ovsrec_open_vswitch_col_cur_cfg); - ovsdb_idl_set_write_only(idl, &ovsrec_open_vswitch_col_statistics); + ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg); + ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics); ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids); ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids); @@ -283,8 +295,8 @@ bridge_init(const char *remote) ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids); ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge); - ovsdb_idl_set_write_only(idl, &ovsrec_interface_col_ofport); - ovsdb_idl_set_write_only(idl, &ovsrec_interface_col_statistics); + ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport); + ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics); ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids); /* Register unixctl commands. */ @@ -365,10 +377,7 @@ bridge_configure_once(const struct ovsrec_open_vswitch *cfg) /* Initializes 'options' and fills it with the options for 'if_cfg'. Merges * keys from "options" and "other_config", preferring "options" keys over - * "other_config" keys. - * - * The value strings in '*options' are taken directly from if_cfg, not copied, - * so the caller should not modify or free them. */ + * "other_config" keys. */ static void iface_get_options(const struct ovsrec_interface *if_cfg, struct shash *options) { @@ -390,68 +399,6 @@ iface_get_options(const struct ovsrec_interface *if_cfg, struct shash *options) } } -/* Returns the type of network device that 'iface' should have. (This is - * ordinarily the same type as the interface, but the network devices for - * "internal" ports have type "system".) */ -static const char * -iface_get_netdev_type(const struct iface *iface) -{ - return !strcmp(iface->type, "internal") ? "system" : iface->type; -} - -/* Attempt to create the network device for 'iface' through the netdev - * library. */ -static int -create_iface_netdev(struct iface *iface) -{ - struct netdev_options netdev_options; - struct shash options; - int error; - - memset(&netdev_options, 0, sizeof netdev_options); - netdev_options.name = iface->cfg->name; - netdev_options.type = iface_get_netdev_type(iface); - netdev_options.args = &options; - netdev_options.ethertype = NETDEV_ETH_TYPE_NONE; - - iface_get_options(iface->cfg, &options); - - error = netdev_open(&netdev_options, &iface->netdev); - - if (iface->netdev) { - netdev_get_carrier(iface->netdev, &iface->enabled); - } - - shash_destroy(&options); - - return error; -} - -static int -reconfigure_iface_netdev(struct iface *iface) -{ - const char *netdev_type, *iface_type; - struct shash options; - int error; - - /* Skip reconfiguration if the device has the wrong type. This shouldn't - * happen, but... */ - iface_type = iface_get_netdev_type(iface); - netdev_type = netdev_get_type(iface->netdev); - if (iface_type && strcmp(netdev_type, iface_type)) { - VLOG_WARN("%s: attempting change device type from %s to %s", - iface->cfg->name, netdev_type, iface_type); - return EINVAL; - } - - /* Reconfigure device. */ - iface_get_options(iface->cfg, &options); - error = netdev_reconfigure(iface->netdev, &options); - shash_destroy(&options); - - return error; -} - /* Callback for iterate_and_prune_ifaces(). */ static bool check_iface(struct bridge *br, struct iface *iface, void *aux OVS_UNUSED) @@ -509,6 +456,7 @@ iterate_and_prune_ifaces(struct bridge *br, if (cb(br, iface, aux)) { j++; } else { + iface_set_ofport(iface->cfg, -1); iface_destroy(iface); } } @@ -531,30 +479,52 @@ iterate_and_prune_ifaces(struct bridge *br, * should not be and in fact is not directly involved in that. But * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so * it has to tell in-band control where the managers are to enable that. + * (Thus, only managers connected in-band are collected.) */ static void -collect_managers(const struct ovsrec_open_vswitch *ovs_cfg, - struct sockaddr_in **managersp, size_t *n_managersp) +collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg, + struct sockaddr_in **managersp, size_t *n_managersp) { struct sockaddr_in *managers = NULL; size_t n_managers = 0; + struct shash targets; + size_t i; - if (ovs_cfg->n_managers > 0) { - 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. */ + 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]; - managers = xmalloc(ovs_cfg->n_managers * sizeof *managers); - for (i = 0; i < ovs_cfg->n_managers; i++) { - const char *name = ovs_cfg->managers[i]; - struct sockaddr_in *sin = &managers[i]; + if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) { + shash_find_and_delete(&targets, m->target); + } else { + shash_add_once(&targets, m->target, NULL); + } + } + + /* Now extract the targets' IP addresses. */ + if (!shash_is_empty(&targets)) { + struct shash_node *node; + + managers = xmalloc(shash_count(&targets) * sizeof *managers); + SHASH_FOR_EACH (node, &targets) { + const char *target = node->name; + struct sockaddr_in *sin = &managers[n_managers]; - if ((!strncmp(name, "tcp:", 4) - && inet_parse_active(name + 4, JSONRPC_TCP_PORT, sin)) || - (!strncmp(name, "ssl:", 4) - && inet_parse_active(name + 4, JSONRPC_SSL_PORT, sin))) { + if ((!strncmp(target, "tcp:", 4) + && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) || + (!strncmp(target, "ssl:", 4) + && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) { n_managers++; } } } + shash_destroy(&targets); *managersp = managers; *n_managersp = n_managers; @@ -573,7 +543,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) COVERAGE_INC(bridge_reconfigure); - collect_managers(ovs_cfg, &managers, &n_managers); + collect_in_band_managers(ovs_cfg, &managers, &n_managers); /* Collect old and new bridges. */ shash_init(&old_br); @@ -667,18 +637,16 @@ 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; - bool internal = !iface || !strcmp(iface->type, "internal"); struct odp_port *dpif_port = shash_find_data(&cur_ifaces, if_name); + const char *type = iface ? iface->type : "internal"; int error; /* If we have a port or a netdev already, and it's not the type we * want, then delete the port (if any) and close the netdev (if * any). */ - if (internal - ? dpif_port && !(dpif_port->flags & ODP_PORT_INTERNAL) - : (iface->netdev - && strcmp(iface->type, netdev_get_type(iface->netdev)))) - { + if ((dpif_port && strcmp(dpif_port->type, type)) + || (iface && iface->netdev + && strcmp(type, netdev_get_type(iface->netdev)))) { if (dpif_port) { error = ofproto_port_del(br->ofproto, dpif_port->port); if (error) { @@ -692,48 +660,62 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) } } - /* If it's not an internal port, open (possibly create) the - * netdev. */ - if (!internal) { - if (!iface->netdev) { - error = create_iface_netdev(iface); - if (error) { - VLOG_WARN("could not create iface %s: %s", iface->name, - strerror(error)); - continue; - } - } else { - reconfigure_iface_netdev(iface); + /* If the port doesn't exist or we don't have the netdev open, + * we need to do more work. */ + if (!dpif_port || (iface && !iface->netdev)) { + struct netdev_options options; + struct netdev *netdev; + struct shash args; + + /* First open the network device. */ + options.name = if_name; + options.type = type; + options.args = &args; + options.ethertype = NETDEV_ETH_TYPE_NONE; + + shash_init(&args); + if (iface) { + iface_get_options(iface->cfg, &args); } - } + error = netdev_open(&options, &netdev); + shash_destroy(&args); - /* If it's not part of the datapath, add it. */ - if (!dpif_port) { - error = dpif_port_add(br->dpif, if_name, - internal ? ODP_PORT_INTERNAL : 0, NULL); - if (error == EFBIG) { - 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)); + if (error) { + VLOG_WARN("could not open network device %s (%s)", + if_name, strerror(error)); continue; } - } - /* If it's an internal port, open the netdev. */ - if (internal) { - if (iface && !iface->netdev) { - error = create_iface_netdev(iface); + /* Then add the port if we haven't already. */ + if (!dpif_port) { + error = dpif_port_add(br->dpif, netdev, NULL); if (error) { - VLOG_WARN("could not create iface %s: %s", iface->name, - strerror(error)); - continue; + netdev_close(netdev); + if (error == EFBIG) { + VLOG_ERR("ran out of valid port numbers on %s", + dpif_name(br->dpif)); + break; + } else { + VLOG_ERR("failed to add %s interface to %s: %s", + if_name, dpif_name(br->dpif), + strerror(error)); + continue; + } } } - } else { - assert(iface->netdev != NULL); + + /* Update 'iface'. */ + if (iface) { + iface->netdev = netdev; + iface->enabled = netdev_get_carrier(iface->netdev); + } + } else if (iface && iface->netdev) { + struct shash args; + + shash_init(&args); + iface_get_options(iface->cfg, &args); + netdev_reconfigure(iface->netdev, &args); + shash_destroy(&args); } } free(dpif_ports); @@ -894,6 +876,13 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) iterate_and_prune_ifaces(br, set_iface_properties, NULL); } + LIST_FOR_EACH (br, node, &all_bridges) { + struct iface *iface; + HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) { + iface_update_cfm(iface); + } + } + free(managers); } @@ -1112,6 +1101,82 @@ dpid_from_hash(const void *data, size_t n) return eth_addr_to_uint64(hash); } +static void +iface_refresh_cfm_stats(struct iface *iface) +{ + size_t i; + struct cfm *cfm; + const struct ovsrec_monitor *mon; + + mon = iface->cfg->monitor; + cfm = iface->cfm; + + if (!cfm || !mon) { + return; + } + + for (i = 0; i < mon->n_remote_mps; i++) { + const struct ovsrec_maintenance_point *mp; + const struct remote_mp *rmp; + + mp = mon->remote_mps[i]; + rmp = cfm_get_remote_mp(cfm, mp->mpid); + + ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1); + } + + if (hmap_is_empty(&cfm->x_remote_mps)) { + ovsrec_monitor_set_unexpected_remote_mpids(mon, NULL, 0); + } else { + size_t length; + struct remote_mp *rmp; + int64_t *x_remote_mps; + + length = hmap_count(&cfm->x_remote_mps); + x_remote_mps = xzalloc(length * sizeof *x_remote_mps); + + i = 0; + HMAP_FOR_EACH (rmp, node, &cfm->x_remote_mps) { + x_remote_mps[i++] = rmp->mpid; + } + + ovsrec_monitor_set_unexpected_remote_mpids(mon, x_remote_mps, length); + free(x_remote_mps); + } + + if (hmap_is_empty(&cfm->x_remote_maids)) { + ovsrec_monitor_set_unexpected_remote_maids(mon, NULL, 0); + } else { + size_t length; + char **x_remote_maids; + struct remote_maid *rmaid; + + length = hmap_count(&cfm->x_remote_maids); + x_remote_maids = xzalloc(length * sizeof *x_remote_maids); + + i = 0; + HMAP_FOR_EACH (rmaid, node, &cfm->x_remote_maids) { + size_t j; + + x_remote_maids[i] = xzalloc(CCM_MAID_LEN * 2 + 1); + + for (j = 0; j < CCM_MAID_LEN; j++) { + snprintf(&x_remote_maids[i][j * 2], 3, "%02hhx", + rmaid->maid[j]); + } + i++; + } + ovsrec_monitor_set_unexpected_remote_maids(mon, x_remote_maids, length); + + for (i = 0; i < length; i++) { + free(x_remote_maids[i]); + } + free(x_remote_maids); + } + + ovsrec_monitor_set_fault(mon, &cfm->fault, 1); +} + static void iface_refresh_stats(struct iface *iface) { @@ -1244,6 +1309,7 @@ bridge_run(void) for (j = 0; j < port->n_ifaces; j++) { struct iface *iface = port->ifaces[j]; iface_refresh_stats(iface); + iface_refresh_cfm_stats(iface); } } } @@ -1260,6 +1326,7 @@ void bridge_wait(void) { struct bridge *br; + struct iface *iface; LIST_FOR_EACH (br, node, &all_bridges) { ofproto_wait(br->ofproto); @@ -1269,6 +1336,12 @@ bridge_wait(void) mac_learning_wait(br->ml); bond_wait(br); + + HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) { + if (iface->cfm) { + cfm_wait(iface->cfm); + } + } } ovsdb_idl_wait(idl); poll_timer_wait_until(stats_timer); @@ -1469,6 +1542,7 @@ static int bridge_run_one(struct bridge *br) { int error; + struct iface *iface; error = ofproto_run1(br->ofproto); if (error) { @@ -1481,6 +1555,21 @@ bridge_run_one(struct bridge *br) 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; } @@ -1598,7 +1687,7 @@ bridge_reconfigure_one(struct bridge *br) /* Configure OpenFlow controller connection snooping. */ svec_init(&snoops); svec_add_nocopy(&snoops, xasprintf("punix:%s/%s.snoop", - ovs_rundir, br->name)); + ovs_rundir(), br->name)); svec_init(&old_snoops); ofproto_get_snoops(br->ofproto, &old_snoops); if (!svec_equal(&snoops, &old_snoops)) { @@ -1618,7 +1707,7 @@ static void bridge_ofproto_controller_for_mgmt(const struct bridge *br, struct ofproto_controller *oc) { - oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir, br->name); + oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name); oc->max_backoff = 0; oc->probe_interval = 60; oc->band = OFPROTO_OUT_OF_BAND; @@ -1699,6 +1788,10 @@ bridge_reconfigure_remotes(struct bridge *br, const struct sockaddr_in *managers, size_t n_managers) { + const char *disable_ib_str, *queue_id_str; + bool disable_in_band = false; + int queue_id; + struct ovsrec_controller **controllers; size_t n_controllers; bool had_primary; @@ -1707,7 +1800,22 @@ bridge_reconfigure_remotes(struct bridge *br, size_t n_ocs; size_t i; - ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers); + /* Check if we should disable in-band control on this bridge. */ + disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band"); + if (disable_ib_str && !strcmp(disable_ib_str, "true")) { + disable_in_band = true; + } + + /* Set OpenFlow queue ID for in-band control. */ + queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue"); + queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1; + ofproto_set_in_band_queue(br->ofproto, queue_id); + + if (disable_in_band) { + ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0); + } else { + ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers); + } had_primary = ofproto_has_primary_controller(br->ofproto); n_controllers = bridge_get_controllers(br, &controllers); @@ -1732,7 +1840,11 @@ bridge_reconfigure_remotes(struct bridge *br, } bridge_configure_local_iface_netdev(br, c); - bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs++]); + bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]); + if (disable_in_band) { + ocs[n_ocs].band = OFPROTO_OUT_OF_BAND; + } + n_ocs++; } ofproto_set_controllers(br->ofproto, ocs, n_ocs); @@ -1751,14 +1863,14 @@ bridge_reconfigure_remotes(struct bridge *br, if (!n_controllers && ofproto_get_fail_mode(br->ofproto) == OFPROTO_FAIL_STANDALONE) { union ofp_action action; - struct flow flow; + struct cls_rule rule; memset(&action, 0, sizeof action); action.type = htons(OFPAT_OUTPUT); action.output.len = htons(sizeof action); action.output.port = htons(OFPP_NORMAL); - memset(&flow, 0, sizeof flow); - ofproto_add_flow(br->ofproto, &flow, OVSFW_ALL, 0, &action, 1, 0); + cls_rule_init_catchall(&rule, 0); + ofproto_add_flow(br->ofproto, &rule, &action, 1); } } @@ -1822,12 +1934,10 @@ bridge_fetch_dp_ifaces(struct bridge *br) hash_int(iface->dp_ifidx, 0)); } - if (iface->cfg) { - int64_t ofport = (iface->dp_ifidx >= 0 - ? odp_port_to_ofp_port(iface->dp_ifidx) - : -1); - ovsrec_interface_set_ofport(iface->cfg, &ofport, 1); - } + iface_set_ofport(iface->cfg, + (iface->dp_ifidx >= 0 + ? odp_port_to_ofp_port(iface->dp_ifidx) + : -1)); } } free(dpif_ports); @@ -1836,15 +1946,16 @@ bridge_fetch_dp_ifaces(struct bridge *br) /* Bridge packet processing functions. */ static int -bond_hash(const uint8_t mac[ETH_ADDR_LEN]) +bond_hash(const uint8_t mac[ETH_ADDR_LEN], uint16_t vlan) { - return hash_bytes(mac, ETH_ADDR_LEN, 0) & BOND_MASK; + return hash_bytes(mac, ETH_ADDR_LEN, vlan) & BOND_MASK; } static struct bond_entry * -lookup_bond_entry(const struct port *port, const uint8_t mac[ETH_ADDR_LEN]) +lookup_bond_entry(const struct port *port, const uint8_t mac[ETH_ADDR_LEN], + uint16_t vlan) { - return &port->bond_hash[bond_hash(mac)]; + return &port->bond_hash[bond_hash(mac, vlan)]; } static int @@ -1879,7 +1990,7 @@ bond_choose_iface(const struct port *port) static bool choose_output_iface(const struct port *port, const uint8_t *dl_src, - uint16_t *dp_ifidx, tag_type *tags) + uint16_t vlan, uint16_t *dp_ifidx, tag_type *tags) { struct iface *iface; @@ -1887,7 +1998,7 @@ choose_output_iface(const struct port *port, const uint8_t *dl_src, if (port->n_ifaces == 1) { iface = port->ifaces[0]; } else { - struct bond_entry *e = lookup_bond_entry(port, dl_src); + 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) { /* XXX select interface properly. The current interface selection @@ -2060,10 +2171,11 @@ bond_run(struct bridge *br) /* Track carrier going up and down on interfaces. */ while (!netdev_monitor_poll(port->monitor, &devname)) { struct iface *iface; - bool carrier; iface = port_lookup_iface(port, devname); - if (iface && !netdev_get_carrier(iface->netdev, &carrier)) { + if (iface) { + bool carrier = netdev_get_carrier(iface->netdev); + bond_link_status_update(iface, carrier); port_update_bond_compat(port); } @@ -2121,8 +2233,10 @@ set_dst(struct dst *p, const struct flow *flow, { p->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE : in_port->vlan >= 0 ? in_port->vlan - : ntohs(flow->dl_vlan)); - return choose_output_iface(out_port, flow->dl_src, &p->dp_ifidx, tags); + : flow->vlan_tci == 0 ? OFP_VLAN_NONE + : vlan_tci_to_vid(flow->vlan_tci)); + return choose_output_iface(out_port, flow->dl_src, p->vlan, + &p->dp_ifidx, tags); } static void @@ -2223,9 +2337,15 @@ compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan, struct dst dsts[], tag_type *tags, uint16_t *nf_output_iface) { mirror_mask_t mirrors = in_port->src_mirrors; + int flow_vlan; struct dst *dst = dsts; size_t i; + flow_vlan = vlan_tci_to_vid(flow->vlan_tci); + if (flow_vlan == 0) { + flow_vlan = OFP_VLAN_NONE; + } + if (out_port == FLOOD_PORT) { /* XXX use ODP_FLOOD if no vlans or bonding. */ /* XXX even better, define each VLAN as a datapath port group */ @@ -2261,7 +2381,6 @@ compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan, if (port_includes_vlan(port, m->out_vlan) && set_dst(dst, flow, in_port, port, tags)) { - int flow_vlan; if (port->vlan < 0) { dst->vlan = m->out_vlan; @@ -2276,10 +2395,6 @@ compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan, * tagging tags place. This is necessary because * dst->vlan is the final vlan, after removing implicit * tags. */ - flow_vlan = ntohs(flow->dl_vlan); - if (flow_vlan == 0) { - flow_vlan = OFP_VLAN_NONE; - } if (port == in_port && dst->vlan == flow_vlan) { /* Don't send out input port on same VLAN. */ continue; @@ -2292,7 +2407,7 @@ compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan, mirrors &= mirrors - 1; } - partition_dsts(dsts, dst - dsts, ntohs(flow->dl_vlan)); + partition_dsts(dsts, dst - dsts, flow_vlan); return dst - dsts; } @@ -2310,7 +2425,7 @@ print_dsts(const struct dst *dsts, size_t n) static void compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan, const struct port *in_port, const struct port *out_port, - tag_type *tags, struct odp_actions *actions, + tag_type *tags, struct ofpbuf *actions, uint16_t *nf_output_iface) { struct dst dsts[DP_MAX_PORTS * (MAX_MIRRORS + 1)]; @@ -2321,21 +2436,23 @@ compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan, n_dsts = compose_dsts(br, flow, vlan, in_port, out_port, dsts, tags, nf_output_iface); - cur_vlan = ntohs(flow->dl_vlan); + cur_vlan = vlan_tci_to_vid(flow->vlan_tci); + if (cur_vlan == 0) { + cur_vlan = OFP_VLAN_NONE; + } for (p = dsts; p < &dsts[n_dsts]; p++) { - union odp_action *a; if (p->vlan != cur_vlan) { if (p->vlan == OFP_VLAN_NONE) { - odp_actions_add(actions, ODPAT_STRIP_VLAN); + nl_msg_put_flag(actions, ODPAT_STRIP_VLAN); } else { - a = odp_actions_add(actions, ODPAT_SET_DL_TCI); - a->dl_tci.tci = htons(p->vlan & VLAN_VID_MASK); - a->dl_tci.tci |= htons(flow->dl_vlan_pcp << VLAN_PCP_SHIFT); + ovs_be16 tci; + tci = htons(p->vlan & VLAN_VID_MASK); + tci |= flow->vlan_tci & htons(VLAN_PCP_MASK); + nl_msg_put_be16(actions, ODPAT_SET_DL_TCI, tci); } cur_vlan = p->vlan; } - a = odp_actions_add(actions, ODPAT_OUTPUT); - a->output.port = p->dp_ifidx; + nl_msg_put_u32(actions, ODPAT_OUTPUT, p->dp_ifidx); } } @@ -2346,25 +2463,16 @@ compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan, static int flow_get_vlan(struct bridge *br, const struct flow *flow, struct port *in_port, bool have_packet) { - /* Note that dl_vlan of 0 and of OFP_VLAN_NONE both mean that the packet - * 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; - } + int vlan = vlan_tci_to_vid(flow->vlan_tci); 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 " + VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged " "packet received on port %s configured with " "implicit VLAN %"PRIu16, - br->name, ntohs(flow->dl_vlan), - in_port->name, in_port->vlan); + br->name, vlan, in_port->name, in_port->vlan); } return -1; } @@ -2537,7 +2645,7 @@ is_admissible(struct bridge *br, const struct flow *flow, bool have_packet, * not at all, if 'packet' was NULL. */ static bool process_flow(struct bridge *br, const struct flow *flow, - const struct ofpbuf *packet, struct odp_actions *actions, + const struct ofpbuf *packet, struct ofpbuf *actions, tag_type *tags, uint16_t *nf_output_iface) { struct port *in_port; @@ -2588,26 +2696,37 @@ done: static bool bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet, - struct odp_actions *actions, tag_type *tags, + 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); + iface = iface_from_dp_ifidx(br, flow->in_port); + + if (cfm_should_process_flow(flow)) { + if (packet && iface->cfm) { + cfm_process_heartbeat(iface->cfm, packet); + } + return false; + } + return process_flow(br, flow, packet, actions, tags, nf_output_iface); } static void bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags, - const union odp_action *actions, - size_t n_actions, unsigned long long int n_bytes, - void *br_) + const struct nlattr *actions, + unsigned int actions_len, + unsigned long long int n_bytes, void *br_) { struct bridge *br = br_; - const union odp_action *a; + const struct nlattr *a; struct port *in_port; tag_type dummy = 0; + unsigned int left; int vlan; /* Feed information from the active flows back into the learning table to @@ -2625,12 +2744,15 @@ bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags, if (!br->has_bonded_ports) { return; } - for (a = actions; a < &actions[n_actions]; a++) { - if (a->type == ODPAT_OUTPUT) { - struct port *out_port = port_from_dp_ifidx(br, a->output.port); + NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) { + 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) { + uint16_t vlan = (flow->vlan_tci + ? vlan_tci_to_vid(flow->vlan_tci) + : OFP_VLAN_NONE); struct bond_entry *e = lookup_bond_entry(out_port, - flow->dl_src); + flow->dl_src, vlan); e->tx_bytes += n_bytes; } } @@ -2984,7 +3106,7 @@ bond_send_learning_packets(struct port *port) int retval; if (e->port == port->port_idx - || !choose_output_iface(port, e->mac, &dp_ifidx, &tags)) { + || !choose_output_iface(port, e->mac, e->vlan, &dp_ifidx, &tags)) { continue; } @@ -3130,9 +3252,10 @@ bond_unixctl_show(struct unixctl_conn *conn, LIST_FOR_EACH (me, lru_node, &port->bridge->ml->lrus) { uint16_t dp_ifidx; tag_type tags = 0; - if (bond_hash(me->mac) == hash + if (bond_hash(me->mac, me->vlan) == hash && me->port != port->port_idx - && choose_output_iface(port, me->mac, &dp_ifidx, &tags) + && choose_output_iface(port, me->mac, me->vlan, + &dp_ifidx, &tags) && dp_ifidx == iface->dp_ifidx) { ds_put_format(&ds, "\t\t"ETH_ADDR_FMT"\n", @@ -3152,7 +3275,6 @@ bond_unixctl_migrate(struct unixctl_conn *conn, const char *args_, char *args = (char *) args_; char *save_ptr = NULL; char *bond_s, *hash_s, *slave_s; - uint8_t mac[ETH_ADDR_LEN]; struct port *port; struct iface *iface; struct bond_entry *entry; @@ -3173,10 +3295,7 @@ bond_unixctl_migrate(struct unixctl_conn *conn, const char *args_, return; } - if (sscanf(hash_s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(mac)) - == ETH_ADDR_SCAN_COUNT) { - hash = bond_hash(mac); - } else if (strspn(hash_s, "0123456789") == strlen(hash_s)) { + if (strspn(hash_s, "0123456789") == strlen(hash_s)) { hash = atoi(hash_s) & BOND_MASK; } else { unixctl_command_reply(conn, 501, "bad hash"); @@ -3298,23 +3417,39 @@ bond_unixctl_disable_slave(struct unixctl_conn *conn, const char *args, } static void -bond_unixctl_hash(struct unixctl_conn *conn, const char *args, +bond_unixctl_hash(struct unixctl_conn *conn, const char *args_, void *aux OVS_UNUSED) { - 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"); - } + char *args = (char *) args_; + uint8_t mac[ETH_ADDR_LEN]; + uint8_t hash; + char *hash_cstr; + unsigned int vlan; + char *mac_s, *vlan_s; + char *save_ptr = NULL; + + mac_s = strtok_r(args, " ", &save_ptr); + vlan_s = strtok_r(NULL, " ", &save_ptr); + + if (vlan_s) { + if (sscanf(vlan_s, "%u", &vlan) != 1) { + unixctl_command_reply(conn, 501, "invalid vlan"); + return; + } + } else { + vlan = OFP_VLAN_NONE; + } + + if (sscanf(mac_s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(mac)) + == ETH_ADDR_SCAN_COUNT) { + hash = bond_hash(mac, vlan); + + 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 @@ -3435,6 +3570,7 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) if (!shash_add_once(&new_ifaces, if_cfg->name, NULL)) { VLOG_WARN("port %s: %s specified twice as port interface", port->name, if_cfg->name); + iface_set_ofport(if_cfg, -1); continue; } @@ -3547,6 +3683,8 @@ port_destroy(struct port *port) del = br->ports[port->port_idx] = br->ports[--br->n_ports]; del->port_idx = port->port_idx; + VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name); + netdev_monitor_destroy(port->monitor); free(port->ifaces); bitmap_free(port->trunks); @@ -3732,6 +3870,26 @@ port_update_vlan_compat(struct port *port) /* Interface functions. */ +static void +iface_send_packet(struct iface *iface, struct ofpbuf *packet) +{ + struct flow flow; + union ofp_action action; + + memset(&action, 0, sizeof action); + action.output.type = htons(OFPAT_OUTPUT); + action.output.len = htons(sizeof action); + action.output.port = htons(odp_port_to_ofp_port(iface->dp_ifidx)); + + flow_extract(packet, 0, ODPP_NONE, &flow); + + if (ofproto_send_packet(iface->port->bridge->ofproto, &flow, &action, 1, + packet)) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + VLOG_WARN_RL(&rl, "interface %s: Failed to send packet.", iface->name); + } +} + static struct iface * iface_create(struct port *port, const struct ovsrec_interface *if_cfg) { @@ -3793,6 +3951,8 @@ iface_destroy(struct iface *iface) bond_send_learning_packets(port); } + cfm_destroy(iface->cfm); + free(iface->name); free(iface); @@ -3844,6 +4004,15 @@ iface_set_mac(struct iface *iface) } } +/* Sets the ofport column of 'if_cfg' to 'ofport'. */ +static void +iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport) +{ + if (if_cfg) { + ovsrec_interface_set_ofport(if_cfg, &ofport, 1); + } +} + /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'. * * The value strings in '*shash' are taken directly from values[], not copied, @@ -3920,6 +4089,56 @@ iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos) } } } + +static void +iface_update_cfm(struct iface *iface) +{ + size_t i; + struct cfm *cfm; + uint16_t *remote_mps; + struct ovsrec_monitor *mon; + uint8_t ea[ETH_ADDR_LEN], maid[CCM_MAID_LEN]; + + mon = iface->cfg->monitor; + + if (!mon) { + return; + } + + if (netdev_get_etheraddr(iface->netdev, ea)) { + VLOG_WARN("interface %s: Failed to get ethernet address. " + "Skipping Monitor.", iface->name); + return; + } + + if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) { + VLOG_WARN("interface %s: Failed to generate MAID.", iface->name); + return; + } + + if (!iface->cfm) { + iface->cfm = cfm_create(); + } + + cfm = iface->cfm; + cfm->mpid = mon->mpid; + cfm->interval = mon->interval ? *mon->interval : 1000; + + memcpy(cfm->eth_src, ea, sizeof cfm->eth_src); + memcpy(cfm->maid, maid, sizeof cfm->maid); + + remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps); + for(i = 0; i < mon->n_remote_mps; i++) { + remote_mps[i] = mon->remote_mps[i]->mpid; + } + cfm_update_remote_mps(cfm, remote_mps, mon->n_remote_mps); + free(remote_mps); + + if (!cfm_configure(iface->cfm)) { + cfm_destroy(iface->cfm); + iface->cfm = NULL; + } +} /* Port mirroring. */