X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;ds=inline;f=vswitchd%2Fbridge.c;h=0a6501a3675a5a5a741b7735cc31a7dd76164281;hb=83db796889bf98f8699fbd7305ffe920a0b527a1;hp=b4f7a2ef90ab8e4b6414ec2da19fe693b08dfd7b;hpb=750d79a51bec1e9319892ba52b36a9880ba9d8fa;p=openvswitch diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index b4f7a2ef..0a6501a3 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" @@ -42,6 +43,7 @@ #include "hash.h" #include "hmap.h" #include "jsonrpc.h" +#include "lacp.h" #include "list.h" #include "mac-learning.h" #include "netdev.h" @@ -54,7 +56,6 @@ #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 +76,11 @@ 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 */ - LACP_STATUS_EXPIRED, /* Partner is out of date.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; @@ -101,8 +98,8 @@ static void dst_set_free(struct dst_set *); struct iface { /* These members are always valid. */ + struct list port_elem; /* Element in struct port's "ifaces" list. */ struct port *port; /* Containing port. */ - size_t port_ifidx; /* Index within containing port. */ char *name; /* Host network device name. */ tag_type tag; /* Tag associated with this interface. */ long long delay_expires; /* Time after which 'enabled' may change. */ @@ -115,25 +112,17 @@ struct iface { bool enabled; /* May be chosen for flows? */ bool up; /* Is the interface up? */ const char *type; /* Usually same as cfg->type. */ - struct cfm *cfm; /* Connectivity Fault Management */ const struct ovsrec_interface *cfg; /* LACP information. */ - enum lacp_status lacp_status; /* LACP state machine 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. */ + uint16_t lacp_priority; /* LACP port priority. */ }; #define BOND_MASK 0xff struct bond_entry { - int iface_idx; /* Index of assigned iface, or -1 if none. */ + struct iface *iface; /* Assigned iface, or NULL if none. */ uint64_t tx_bytes; /* Count of bytes recently transmitted. */ - tag_type iface_tag; /* Tag associated with iface_idx. */ + tag_type tag; /* Tag for bond_entry<->iface association. */ }; enum bond_mode { @@ -163,11 +152,6 @@ struct mirror { int out_vlan; }; -/* Flags for a port's lacp member. */ -#define LACP_ACTIVE 0x01 /* LACP is in active mode. */ -#define LACP_PASSIVE 0x02 /* LACP is in passive mode. */ -#define LACP_NEGOTIATED 0x04 /* LACP has successfully negotiated. */ - #define FLOOD_PORT ((struct port *) 1) /* The 'flood' output port. */ struct port { struct bridge *bridge; @@ -178,30 +162,29 @@ struct port { const struct ovsrec_port *cfg; char *name; + /* Monitoring. */ + struct netdev_monitor *monitor; /* Tracks carrier. NULL if miimon. */ + long long int miimon_interval; /* Miimon status refresh interval. */ + long long int miimon_next_update; /* Time of next miimon update. */ + /* An ordinary bridge port has 1 interface. * A bridge port for bonding has at least 2 interfaces. */ - struct iface **ifaces; - size_t n_ifaces, allocated_ifaces; + struct list ifaces; /* List of "struct iface"s. */ + size_t n_ifaces; /* list_size(ifaces). */ /* Bonding info. */ 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. */ + struct iface *active_iface; /* iface on which bcasts accepted, or NULL. */ 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. */ /* LACP information. */ - int lacp; /* LACP status flags. 0 if LACP is off. */ - uint16_t lacp_key; /* LACP aggregation key. */ + struct lacp *lacp; /* LACP object. NULL if LACP is disabled. */ + bool lacp_active; /* True if LACP is active */ + bool lacp_fast; /* True if LACP is in fast mode. */ uint16_t lacp_priority; /* LACP system priority. */ - bool lacp_need_update; /* Need to update attached interfaces? */ /* SLB specific bonding info. */ struct bond_entry *bond_hash; /* An array of (BOND_MASK + 1) elements. */ @@ -277,32 +260,29 @@ static void bridge_pick_local_hw_addr(struct bridge *, static uint64_t bridge_pick_datapath_id(struct bridge *, const uint8_t bridge_ea[ETH_ADDR_LEN], struct iface *hw_addr_iface); -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 void lacp_run(struct bridge *); -static void lacp_wait(struct bridge *); -static void lacp_process_packet(const struct ofpbuf *, struct iface *); +static unixctl_cb_func qos_unixctl_show; 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 *); static void port_destroy(struct port *); 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 iface *port_get_an_iface(const struct port *); 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 +296,7 @@ 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 *); @@ -323,18 +304,14 @@ 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 uint8_t iface_get_lacp_state(const struct iface *); -static void iface_get_lacp_priority(struct iface *, struct lacp_info *); -static void iface_set_lacp_defaulted(struct iface *); -static void iface_set_lacp_expired(struct iface *); +static void iface_update_carrier(struct iface *); +static bool iface_get_carrier(const struct iface *); static void shash_from_ovs_idl_map(char **keys, char **values, size_t n, struct shash *); static void shash_to_ovs_idl_map(struct shash *, char ***keys, char ***values, size_t *n); - /* Hooks into ofproto processing. */ static struct ofhooks bridge_ofhooks; @@ -364,10 +341,12 @@ 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, NULL); + lacp_init(); bond_init(); } @@ -485,15 +464,14 @@ iterate_and_prune_ifaces(struct bridge *br, void *aux), void *aux) { - size_t i, j; + size_t i; 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 { + struct iface *iface, *next; + + LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) { + if (!cb(br, iface, aux)) { iface_set_ofport(iface->cfg, -1); iface_destroy(iface); } @@ -502,7 +480,7 @@ 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 +506,10 @@ 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 +619,9 @@ 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 +711,9 @@ 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; } } @@ -747,7 +722,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) /* Update 'iface'. */ if (iface) { iface->netdev = netdev; - iface->enabled = netdev_get_carrier(iface->netdev); + iface->enabled = iface_get_carrier(iface); iface->up = iface->enabled; } } else if (iface && iface->netdev) { @@ -784,7 +759,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) /* Pick local port hardware address, datapath ID. */ bridge_pick_local_hw_addr(br, ea, &hw_addr_iface); - local_iface = bridge_get_local_iface(br); + local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL); if (local_iface) { int error = netdev_set_etheraddr(local_iface->netdev, ea); if (error) { @@ -911,14 +886,21 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) LIST_FOR_EACH (br, node, &all_bridges) { for (i = 0; i < br->n_ports; i++) { struct port *port = br->ports[i]; - int j; + struct iface *iface; + + if (port->monitor) { + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + netdev_monitor_add(port->monitor, iface->netdev); + } + } else { + port->miimon_next_update = 0; + } - port_update_vlan_compat(port); - port_update_bonding(port); port_update_lacp(port); + port_update_bonding(port); - for (j = 0; j < port->n_ifaces; j++) { - iface_update_qos(port->ifaces[j], port->cfg->qos); + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + iface_update_qos(iface, port->cfg->qos); } } } @@ -934,6 +916,10 @@ 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 * @@ -963,7 +949,7 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], struct iface **hw_addr_iface) { const char *hwaddr; - size_t i, j; + size_t i; int error; *hw_addr_iface = NULL; @@ -983,10 +969,11 @@ 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]; + struct iface *candidate; struct iface *iface; /* Mirror output ports don't participate. */ @@ -995,12 +982,11 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], } /* Choose the MAC address to represent the port. */ + iface = NULL; if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) { /* Find the interface with this Ethernet address (if any) so that * we can provide the correct devname to the caller. */ - iface = NULL; - for (j = 0; j < port->n_ifaces; j++) { - struct iface *candidate = port->ifaces[j]; + LIST_FOR_EACH (candidate, port_elem, &port->ifaces) { uint8_t candidate_ea[ETH_ADDR_LEN]; if (!netdev_get_etheraddr(candidate->netdev, candidate_ea) && eth_addr_equals(iface_ea, candidate_ea)) { @@ -1014,10 +1000,8 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], * scripts always add slaves to a bond in alphabetical order, so * for compatibility we choose the interface with the name that is * first in alphabetical order. */ - iface = port->ifaces[0]; - for (j = 1; j < port->n_ifaces; j++) { - struct iface *candidate = port->ifaces[j]; - if (strcmp(candidate->name, iface->name) < 0) { + LIST_FOR_EACH (candidate, port_elem, &port->ifaces) { + if (!iface || strcmp(candidate->name, iface->name) < 0) { iface = candidate; } } @@ -1204,8 +1188,7 @@ iface_refresh_status(struct iface *iface) ovsrec_interface_set_link_state(iface->cfg, - netdev_get_carrier(iface->netdev) - ? "up" : "down"); + iface_get_carrier(iface) ? "up" : "down"); error = netdev_get_mtu(iface->netdev, &mtu); if (!error && mtu != INT_MAX) { @@ -1220,12 +1203,12 @@ iface_refresh_status(struct iface *iface) static void iface_refresh_cfm_stats(struct iface *iface) { - size_t i; - struct cfm *cfm; const struct ovsrec_monitor *mon; + const struct cfm *cfm; + size_t i; mon = iface->cfg->monitor; - cfm = iface->cfm; + cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx); if (!cfm || !mon) { return; @@ -1465,10 +1448,9 @@ bridge_run(void) for (i = 0; i < br->n_ports; i++) { struct port *port = br->ports[i]; - size_t j; + struct iface *iface; - for (j = 0; j < port->n_ifaces; j++) { - struct iface *iface = port->ifaces[j]; + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { iface_refresh_stats(iface); iface_refresh_cfm_stats(iface); iface_refresh_status(iface); @@ -1489,22 +1471,19 @@ 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); @@ -1520,26 +1499,6 @@ bridge_flush(struct bridge *br) br->flush = true; 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. */ static void @@ -1562,13 +1521,106 @@ bridge_unixctl_fdb_show(struct unixctl_conn *conn, continue; } ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n", - br->ports[e->port]->ifaces[0]->dp_ifidx, + port_get_an_iface(br->ports[e->port])->dp_ifidx, e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e)); } unixctl_command_reply(conn, 200, ds_cstr(&ds)); 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 +1757,8 @@ 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 +1766,14 @@ 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 +1915,6 @@ 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 +1929,6 @@ 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); @@ -1911,13 +1946,8 @@ 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); + local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL); if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) { return; } @@ -2042,13 +2072,14 @@ bridge_reconfigure_remotes(struct bridge *br, static void bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces) { - size_t i, j; + size_t i; shash_init(ifaces); 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]; + struct iface *iface; + + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { shash_add_once(ifaces, iface->name, iface); } if (port->n_ifaces > 1 && port->cfg->bond_fake_iface) { @@ -2070,13 +2101,14 @@ bridge_fetch_dp_ifaces(struct bridge *br) { struct dpif_port_dump dump; struct dpif_port dpif_port; - size_t i, j; + size_t i; /* Reset all interface numbers. */ 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]; + struct iface *iface; + + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { iface->dp_ifidx = -1; } } @@ -2110,7 +2142,7 @@ bridge_fetch_dp_ifaces(struct bridge *br) static bool bond_is_tcp_hash(const struct port *port) { - return port->bond_mode == BM_TCP && port->lacp & LACP_NEGOTIATED; + return port->bond_mode == BM_TCP && lacp_negotiated(port->lacp); } static int @@ -2145,33 +2177,30 @@ lookup_bond_entry(const struct port *port, const struct flow *flow, } } -static int +static struct iface * bond_choose_iface(const struct port *port) { 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++) { - struct iface *iface = port->ifaces[i]; + struct iface *best_down_slave; + struct iface *iface; + best_down_slave = NULL; + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { if (iface->enabled) { - return i; - } else if (iface->delay_expires < next_delay_expiration - && (iface->lacp_attached - || !(port->lacp & LACP_NEGOTIATED))) { - best_down_slave = i; - next_delay_expiration = iface->delay_expires; + return iface; + } else if ((!best_down_slave + || iface->delay_expires < best_down_slave->delay_expires) + && lacp_slave_may_enable(port->lacp, iface)) { + best_down_slave = iface; } } - if (best_down_slave != -1) { - struct iface *iface = port->ifaces[best_down_slave]; - + if (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); + "since no other interface is up", + best_down_slave->name, + best_down_slave->delay_expires - time_msec()); + bond_enable_slave(best_down_slave, true); } return best_down_slave; @@ -2185,29 +2214,27 @@ choose_output_iface(const struct port *port, const struct flow *flow, assert(port->n_ifaces); if (port->n_ifaces == 1) { - iface = port->ifaces[0]; + iface = port_get_an_iface(port); } else if (port->bond_mode == BM_AB) { - if (port->active_iface < 0) { + iface = port->active_iface; + if (!iface) { *tags |= port->no_ifaces_tag; return false; } - iface = port->ifaces[port->active_iface]; } else { struct bond_entry *e = lookup_bond_entry(port, flow, vlan); - if (e->iface_idx < 0 || e->iface_idx >= port->n_ifaces - || !port->ifaces[e->iface_idx]->enabled) { + if (!e->iface || !e->iface->enabled) { /* XXX select interface properly. The current interface selection * is only good for testing the rebalancing code. */ - e->iface_idx = bond_choose_iface(port); - if (e->iface_idx < 0) { + e->iface = bond_choose_iface(port); + if (!e->iface) { *tags |= port->no_ifaces_tag; return false; } - e->iface_tag = tag_create_random(); - ((struct port *) port)->bond_compat_is_stale = true; + e->tag = tag_create_random(); } - *tags |= e->iface_tag; - iface = port->ifaces[e->iface_idx]; + *tags |= e->tag; + iface = e->iface; } *dp_ifidx = iface->dp_ifidx; *tags |= iface->tag; /* Currently only used for bonding. */ @@ -2219,29 +2246,17 @@ bond_link_status_update(struct iface *iface) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); struct port *port = iface->port; - bool up = iface->up; + bool up = iface->up && lacp_slave_may_enable(port->lacp, iface); int updelay, downdelay; updelay = port->updelay; downdelay = port->downdelay; - if (iface->port->lacp & LACP_NEGOTIATED) { + if (lacp_negotiated(port->lacp)) { downdelay = 0; updelay = 0; } - if (iface->port->lacp && up) { - /* The interface is up if it's attached to an aggregator and its - * partner is synchronized. The only exception is defaulted links. - * 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 - && (iface->lacp_partner.state & LACP_STATE_SYNC - || iface->lacp_status == LACP_STATUS_DEFAULTED); - } - - if ((up == iface->enabled) == (iface->delay_expires == LLONG_MAX)) { /* Nothing to do. */ return; @@ -2252,7 +2267,7 @@ bond_link_status_update(struct iface *iface) iface->delay_expires = LLONG_MAX; VLOG_INFO_RL(&rl, "interface %s: will not be %s", iface->name, up ? "disabled" : "enabled"); - } else if (up && port->active_iface < 0) { + } else if (up && !port->active_iface) { bond_enable_slave(iface, true); if (updelay) { VLOG_INFO_RL(&rl, "interface %s: skipping %d ms updelay since no " @@ -2278,10 +2293,9 @@ bond_choose_active_iface(struct port *port) static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); port->active_iface = bond_choose_iface(port); - port->active_iface_tag = tag_create_random(); - if (port->active_iface >= 0) { + if (port->active_iface) { VLOG_INFO_RL(&rl, "port %s: active interface is now %s", - port->name, port->ifaces[port->active_iface]->name); + port->name, port->active_iface->name); } else { VLOG_WARN_RL(&rl, "port %s: all ports disabled, no active interface", port->name); @@ -2309,10 +2323,7 @@ bond_enable_slave(struct iface *iface, bool enable) if (!iface->enabled) { VLOG_WARN("interface %s: disabled", iface->name); ofproto_revalidate(br->ofproto, iface->tag); - if (iface->port_ifidx == port->active_iface) { - ofproto_revalidate(br->ofproto, - port->active_iface_tag); - + if (iface == port->active_iface) { /* 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 @@ -2325,7 +2336,7 @@ bond_enable_slave(struct iface *iface, bool enable) bond_send_learning_packets(port); } else { VLOG_WARN("interface %s: enabled", iface->name); - if (port->active_iface < 0 && !moving_active_iface) { + if (!port->active_iface && !moving_active_iface) { ofproto_revalidate(br->ofproto, port->no_ifaces_tag); bond_choose_active_iface(port); bond_send_learning_packets(port); @@ -2334,7 +2345,6 @@ 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 @@ -2344,14 +2354,14 @@ bond_update_fake_iface_stats(struct port *port) { struct netdev_stats bond_stats; struct netdev *bond_dev; - size_t i; + struct iface *iface; memset(&bond_stats, 0, sizeof bond_stats); - for (i = 0; i < port->n_ifaces; i++) { + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { struct netdev_stats slave_stats; - if (!netdev_get_stats(port->ifaces[i]->netdev, &slave_stats)) { + if (!netdev_get_stats(iface->netdev, &slave_stats)) { /* XXX: We swap the stats here because they are swapped back when * reported by the internal device. The reason for this is * internal devices normally represent packets going into the system @@ -2375,113 +2385,48 @@ bond_update_fake_iface_stats(struct port *port) } static void -bond_link_carrier_update(struct iface *iface, bool carrier) +bond_run(struct port *port) { - if (carrier == iface->up) { + struct iface *iface; + + if (port->n_ifaces < 2) { return; } - if (iface->lacp_status == LACP_STATUS_CURRENT) { - iface_set_lacp_expired(iface); + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + bond_link_status_update(iface); } - iface->up = carrier; - iface->lacp_tx = 0; - iface->port->bond_compat_is_stale = true; -} - -static void -bond_run(struct bridge *br) -{ - 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; - - if (port->monitor) { - assert(!port->miimon); - - /* 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_carrier_update(iface, up); - } - 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_carrier_update(iface, up); - } - port->bond_miimon_next_update = time_msec() + - port->bond_miimon_interval; - } - } - - for (j = 0; j < port->n_ifaces; j++) { - bond_link_status_update(port->ifaces[j]); - } - - 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_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; - } + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + 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; - - for (i = 0; i < br->n_ports; i++) { - struct port *port = br->ports[i]; - if (port->n_ifaces < 2) { - continue; - } + struct iface *iface; - if (port->monitor) { - netdev_monitor_poll_wait(port->monitor); - } + if (port->n_ifaces < 2) { + return; + } - if (port->miimon) { - poll_timer_wait_until(port->bond_miimon_next_update); + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + if (iface->delay_expires != LLONG_MAX) { + poll_timer_wait_until(iface->delay_expires); } + } - 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); - } + if (port->bond_fake_iface) { + poll_timer_wait_until(port->bond_next_fake_iface_update); } } @@ -2614,17 +2559,36 @@ port_includes_vlan(const struct port *port, uint16_t vlan) static bool port_is_floodable(const struct port *port) { - int i; + struct iface *iface; - for (i = 0; i < port->n_ifaces; i++) { + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { if (!ofproto_port_is_floodable(port->bridge->ofproto, - port->ifaces[i]->dp_ifidx)) { + iface->dp_ifidx)) { return false; } } return true; } +/* Returns the tag for 'port''s active iface, or 'port''s no_ifaces_tag if + * there is no active iface. */ +static tag_type +port_get_active_iface_tag(const struct port *port) +{ + return (port->active_iface + ? port->active_iface->tag + : port->no_ifaces_tag); +} + +/* Returns an arbitrary interface within 'port'. + * + * 'port' must have at least one interface. */ +static struct iface * +port_get_an_iface(const struct port *port) +{ + return CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem); +} + static void compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan, const struct port *in_port, const struct port *out_port, @@ -2906,19 +2870,19 @@ is_admissible(struct bridge *br, const struct flow *flow, bool have_packet, } /* When using LACP, do not accept packets from disabled interfaces. */ - if (in_port->lacp & LACP_NEGOTIATED && !in_iface->enabled) { + if (lacp_negotiated(in_port->lacp) && !in_iface->enabled) { return false; } /* Packets received on non-LACP bonds need special attention to avoid * duplicates. */ - if (in_port->n_ifaces > 1 && !(in_port->lacp & LACP_NEGOTIATED)) { + if (in_port->n_ifaces > 1 && !lacp_negotiated(in_port->lacp)) { int src_idx; bool is_grat_arp_locked; if (eth_addr_is_multicast(flow->dl_dst)) { - *tags |= in_port->active_iface_tag; - if (in_port->active_iface != in_iface->port_ifidx) { + *tags |= port_get_active_iface_tag(in_port); + if (in_port->active_iface != in_iface) { /* Drop all multicast packets on inactive slaves. */ return false; } @@ -3001,33 +2965,42 @@ 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) { - cfm_process_heartbeat(iface->cfm, packet); - } - return false; - } else if (flow->dl_type == htons(ETH_TYPE_LACP)) { - if (packet) { - lacp_process_packet(packet, iface); + if (flow->dl_type == htons(ETH_TYPE_LACP)) { + + if (iface && iface->port->lacp && packet) { + const struct lacp_pdu *pdu = parse_lacp_packet(packet); + + if (pdu) { + COVERAGE_INC(bridge_process_lacp); + lacp_process_pdu(iface->port->lacp, iface, pdu); + } } 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 +3028,7 @@ 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 +3053,7 @@ 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); @@ -3090,247 +3063,59 @@ 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, }; -/* LACP functions. */ - -static void -lacp_process_packet(const struct ofpbuf *packet, struct iface *iface) -{ - const struct lacp_pdu *pdu; +/* Bonding functions. */ - if (!iface->port->lacp) { - return; - } +/* Statistics for a single interface on a bonded port, used for load-based + * bond rebalancing. */ +struct slave_balance { + struct iface *iface; /* The interface. */ + uint64_t tx_bytes; /* Sum of hashes[*]->tx_bytes. */ - pdu = parse_lacp_packet(packet); - if (!pdu) { - return; - } + /* All the "bond_entry"s that are assigned to this interface, in order of + * increasing tx_bytes. */ + struct bond_entry **hashes; + size_t n_hashes; +}; - iface->lacp_status = LACP_STATUS_CURRENT; - iface->lacp_rx = time_msec() + LACP_SLOW_TIME_RX; +static const char * +bond_mode_to_string(enum bond_mode bm) { + static char *bm_slb = "balance-slb"; + static char *bm_ab = "active-backup"; + static char *bm_tcp = "balance-tcp"; - iface->lacp_actor.state = iface_get_lacp_state(iface); - if (memcmp(&iface->lacp_actor, &pdu->partner, sizeof pdu->partner)) { - iface->lacp_tx = 0; + switch (bm) { + case BM_SLB: return bm_slb; + case BM_AB: return bm_ab; + case BM_TCP: return bm_tcp; } - if (memcmp(&iface->lacp_partner, &pdu->actor, sizeof pdu->actor)) { - iface->port->lacp_need_update = true; - iface->lacp_partner = pdu->actor; - } + NOT_REACHED(); + return NULL; } -static void -lacp_update_ifaces(struct port *port) +/* Sorts pointers to pointers to bond_entries in ascending order by the + * interface to which they are assigned, and within a single interface in + * ascending order of bytes transmitted. */ +static int +compare_bond_entries(const void *a_, const void *b_) { - size_t i; - struct iface *lead; - struct lacp_info lead_pri; - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10); - - port->lacp_need_update = false; - COVERAGE_INC(bridge_lacp_update); - - if (!port->lacp) { - return; + const struct bond_entry *const *ap = a_; + const struct bond_entry *const *bp = b_; + const struct bond_entry *a = *ap; + const struct bond_entry *b = *bp; + if (a->iface != b->iface) { + return a->iface > b->iface ? 1 : -1; + } else if (a->tx_bytes != b->tx_bytes) { + return a->tx_bytes > b->tx_bytes ? 1 : -1; + } else { + return 0; } - - VLOG_DBG_RL(&rl, "port %s: re-evaluating LACP link status", port->name); - - lead = NULL; - for (i = 0; i < port->n_ifaces; i++) { - struct iface *iface = port->ifaces[i]; - struct lacp_info pri; - - iface->lacp_attached = true; - ofproto_revalidate(port->bridge->ofproto, iface->tag); - - /* Don't allow loopback interfaces to send traffic or lead. */ - if (eth_addr_equals(iface->lacp_partner.sysid, - 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; - continue; - } - - if (iface->lacp_status == LACP_STATUS_DEFAULTED) { - continue; - } - - iface_get_lacp_priority(iface, &pri); - - if (!lead || memcmp(&pri, &lead_pri, sizeof pri) < 0) { - lead = iface; - lead_pri = pri; - } - } - - if (!lead) { - port->lacp &= ~LACP_NEGOTIATED; - return; - } - - port->lacp |= LACP_NEGOTIATED; - - for (i = 0; i < port->n_ifaces; i++) { - struct iface *iface = port->ifaces[i]; - - if (iface->lacp_status == LACP_STATUS_DEFAULTED - || lead->lacp_partner.key != iface->lacp_partner.key - || !eth_addr_equals(lead->lacp_partner.sysid, - iface->lacp_partner.sysid)) { - iface->lacp_attached = false; - } - } -} - -static bool -lacp_iface_may_tx(const struct iface *iface) -{ - return iface->port->lacp & LACP_ACTIVE - || iface->lacp_status != LACP_STATUS_DEFAULTED; -} - -static void -lacp_run(struct bridge *br) -{ - size_t i, j; - 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) { - continue; - } - - for (j = 0; j < port->n_ifaces; j++) { - struct iface *iface = port->ifaces[j]; - - 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 (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 (time_msec() < iface->lacp_tx || !lacp_iface_may_tx(iface)) { - continue; - } - - 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) -{ - size_t i, j; - - for (i = 0; i < br->n_ports; i++) { - struct port *port = br->ports[i]; - - if (!port->lacp) { - continue; - } - - for (j = 0; j < port->n_ifaces; j++) { - struct iface *iface = port->ifaces[j]; - - 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); - } - } - } -} - -/* Bonding functions. */ - -/* Statistics for a single interface on a bonded port, used for load-based - * bond rebalancing. */ -struct slave_balance { - struct iface *iface; /* The interface. */ - uint64_t tx_bytes; /* Sum of hashes[*]->tx_bytes. */ - - /* All the "bond_entry"s that are assigned to this interface, in order of - * increasing tx_bytes. */ - struct bond_entry **hashes; - size_t n_hashes; -}; - -static const char * -bond_mode_to_string(enum bond_mode bm) { - static char *bm_slb = "balance-slb"; - static char *bm_ab = "active-backup"; - static char *bm_tcp = "balance-tcp"; - - switch (bm) { - case BM_SLB: return bm_slb; - case BM_AB: return bm_ab; - case BM_TCP: return bm_tcp; - } - - NOT_REACHED(); - return NULL; -} - -/* Sorts pointers to pointers to bond_entries in ascending order by the - * interface to which they are assigned, and within a single interface in - * ascending order of bytes transmitted. */ -static int -compare_bond_entries(const void *a_, const void *b_) -{ - const struct bond_entry *const *ap = a_; - const struct bond_entry *const *bp = b_; - const struct bond_entry *a = *ap; - const struct bond_entry *b = *bp; - if (a->iface_idx != b->iface_idx) { - return a->iface_idx > b->iface_idx ? 1 : -1; - } else if (a->tx_bytes != b->tx_bytes) { - return a->tx_bytes > b->tx_bytes ? 1 : -1; - } else { - return 0; - } -} +} /* Sorts slave_balances so that enabled ports come first, and otherwise in * *descending* order by number of bytes transmitted. */ @@ -3421,7 +3206,7 @@ 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 " @@ -3450,9 +3235,9 @@ bond_shift_load(struct slave_balance *from, struct slave_balance *to, to->tx_bytes += delta; /* Arrange for flows to be revalidated. */ - ofproto_revalidate(port->bridge->ofproto, hash->iface_tag); - hash->iface_idx = to->iface->port_ifidx; - hash->iface_tag = tag_create_random(); + ofproto_revalidate(port->bridge->ofproto, hash->tag); + hash->iface = to->iface; + hash->tag = tag_create_random(); } static void @@ -3463,6 +3248,7 @@ bond_rebalance_port(struct port *port) struct bond_entry *hashes[BOND_MASK + 1]; struct slave_balance *b, *from, *to; struct bond_entry *e; + struct iface *iface; size_t i; assert(port->bond_mode != BM_AB); @@ -3478,26 +3264,34 @@ bond_rebalance_port(struct port *port) * become contiguous in memory, and then we point each 'hashes' members of * a slave_balance structure to the start of a contiguous group. */ n_bals = port->n_ifaces; - bals = xmalloc(n_bals * sizeof *bals); - for (b = bals; b < &bals[n_bals]; b++) { - b->iface = port->ifaces[b - bals]; + b = bals = xmalloc(n_bals * sizeof *bals); + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + b->iface = iface; b->tx_bytes = 0; b->hashes = NULL; b->n_hashes = 0; + b++; } + assert(b == &bals[n_bals]); for (i = 0; i <= BOND_MASK; i++) { hashes[i] = &port->bond_hash[i]; } qsort(hashes, BOND_MASK + 1, sizeof *hashes, compare_bond_entries); for (i = 0; i <= BOND_MASK; i++) { e = hashes[i]; - if (e->iface_idx >= 0 && e->iface_idx < port->n_ifaces) { - b = &bals[e->iface_idx]; - b->tx_bytes += e->tx_bytes; - if (!b->hashes) { - b->hashes = &hashes[i]; + if (!e->iface) { + continue; + } + + for (b = bals; b < &bals[n_bals]; b++) { + if (b->iface == e->iface) { + b->tx_bytes += e->tx_bytes; + if (!b->hashes) { + b->hashes = &hashes[i]; + } + b->n_hashes++; + break; } - b->n_hashes++; } } qsort(bals, n_bals, sizeof *bals, compare_slave_balance); @@ -3575,7 +3369,6 @@ 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. */ @@ -3599,6 +3392,9 @@ bond_rebalance_port(struct port *port) * historical data to decay to <1% in 7 rebalancing runs. */ for (e = &port->bond_hash[0]; e <= &port->bond_hash[BOND_MASK]; e++) { e->tx_bytes /= 2; + if (!e->tx_bytes) { + e->iface = NULL; + } } exit: @@ -3613,16 +3409,15 @@ bond_send_learning_packets(struct port *port) struct ofpbuf packet; int error, n_packets, n_errors; - if (!port->n_ifaces || port->active_iface < 0 || bond_is_tcp_hash(port)) { + if (!port->n_ifaces || !port->active_iface || bond_is_tcp_hash(port)) { return; } ofpbuf_init(&packet, 128); error = n_packets = n_errors = 0; LIST_FOR_EACH (e, lru_node, &br->ml->lrus) { - union ofp_action actions[2], *a; - uint16_t dp_ifidx; tag_type tags = 0; + uint16_t dp_ifidx; struct flow flow; int retval; @@ -3638,24 +3433,9 @@ bond_send_learning_packets(struct port *port) continue; } - /* Compose actions. */ - memset(actions, 0, sizeof actions); - a = actions; - if (e->vlan) { - a->vlan_vid.type = htons(OFPAT_SET_VLAN_VID); - a->vlan_vid.len = htons(sizeof *a); - a->vlan_vid.vlan_vid = htons(e->vlan); - a++; - } - a->output.type = htons(OFPAT_OUTPUT); - a->output.len = htons(sizeof *a); - a->output.port = htons(odp_port_to_ofp_port(dp_ifidx)); - a++; - /* Send packet. */ n_packets++; - retval = ofproto_send_packet(br->ofproto, &flow, actions, a - actions, - &packet); + retval = ofproto_send_packet(br->ofproto, dp_ifidx, e->vlan, &packet); if (retval) { error = retval; n_errors++; @@ -3691,13 +3471,12 @@ bond_unixctl_list(struct unixctl_conn *conn, for (i = 0; i < br->n_ports; i++) { const struct port *port = br->ports[i]; if (port->n_ifaces > 1) { - size_t j; + struct iface *iface; ds_put_format(&ds, "%s\t%s\t%s\t", br->name, port->name, bond_mode_to_string(port->bond_mode)); - for (j = 0; j < port->n_ifaces; j++) { - const struct iface *iface = port->ifaces[j]; - if (j) { + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + if (&iface->port_elem != list_front(&port->ifaces)) { ds_put_cstr(&ds, ", "); } ds_put_cstr(&ds, iface->name); @@ -3728,49 +3507,13 @@ bond_find(const char *name) return NULL; } -static void -ds_put_lacp_state(struct ds *ds, uint8_t state) -{ - if (state & LACP_STATE_ACT) { - ds_put_cstr(ds, "activity "); - } - - if (state & LACP_STATE_TIME) { - ds_put_cstr(ds, "timeout "); - } - - if (state & LACP_STATE_AGG) { - ds_put_cstr(ds, "aggregation "); - } - - if (state & LACP_STATE_SYNC) { - ds_put_cstr(ds, "synchronized "); - } - - if (state & LACP_STATE_COL) { - ds_put_cstr(ds, "collecting "); - } - - if (state & LACP_STATE_DIST) { - ds_put_cstr(ds, "distributing "); - } - - if (state & LACP_STATE_DEF) { - ds_put_cstr(ds, "defaulted "); - } - - if (state & LACP_STATE_EXP) { - ds_put_cstr(ds, "expired "); - } -} - static void bond_unixctl_show(struct unixctl_conn *conn, const char *args, void *aux OVS_UNUSED) { struct ds ds = DS_EMPTY_INITIALIZER; const struct port *port; - size_t j; + struct iface *iface; port = bond_find(args); if (!port) { @@ -3783,7 +3526,7 @@ bond_unixctl_show(struct unixctl_conn *conn, if (port->lacp) { ds_put_format(&ds, "lacp: %s\n", - port->lacp & LACP_ACTIVE ? "active" : "passive"); + port->lacp_active ? "active" : "passive"); } else { ds_put_cstr(&ds, "lacp: off\n"); } @@ -3795,11 +3538,11 @@ bond_unixctl_show(struct unixctl_conn *conn, ds_put_format(&ds, "bond-detect-mode: %s\n", - port->miimon ? "miimon" : "carrier"); + port->monitor ? "carrier" : "miimon"); - if (port->miimon) { + if (!port->monitor) { ds_put_format(&ds, "bond-miimon-interval: %lld\n", - port->bond_miimon_interval); + port->miimon_interval); } ds_put_format(&ds, "updelay: %d ms\n", port->updelay); @@ -3810,15 +3553,14 @@ bond_unixctl_show(struct unixctl_conn *conn, port->bond_next_rebalance - time_msec()); } - for (j = 0; j < port->n_ifaces; j++) { - const struct iface *iface = port->ifaces[j]; + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { struct bond_entry *be; struct flow flow; /* Basic info. */ ds_put_format(&ds, "\nslave %s: %s\n", iface->name, iface->enabled ? "enabled" : "disabled"); - if (j == port->active_iface) { + if (iface == port->active_iface) { ds_put_cstr(&ds, "\tactive slave\n"); } if (iface->delay_expires != LLONG_MAX) { @@ -3827,66 +3569,6 @@ bond_unixctl_show(struct unixctl_conn *conn, iface->delay_expires - time_msec()); } - if (port->lacp) { - ds_put_cstr(&ds, "\tstatus: "); - - if (iface->lacp_status == LACP_STATUS_CURRENT) { - ds_put_cstr(&ds, "current "); - } else if (iface->lacp_status == LACP_STATUS_EXPIRED) { - ds_put_cstr(&ds, "expired "); - } else { - ds_put_cstr(&ds, "defaulted "); - } - - if (iface->lacp_attached) { - ds_put_cstr(&ds, "attached "); - } - - ds_put_cstr(&ds, "\n"); - - ds_put_cstr(&ds, "\n\tactor sysid: "); - ds_put_format(&ds, ETH_ADDR_FMT, - ETH_ADDR_ARGS(iface->lacp_actor.sysid)); - ds_put_cstr(&ds, "\n"); - - ds_put_format(&ds, "\tactor sys_priority: %u\n", - ntohs(iface->lacp_actor.sys_priority)); - - ds_put_format(&ds, "\tactor portid: %u\n", - ntohs(iface->lacp_actor.portid)); - - ds_put_format(&ds, "\tactor port_priority: %u\n", - ntohs(iface->lacp_actor.port_priority)); - - ds_put_format(&ds, "\tactor key: %u\n", - ntohs(iface->lacp_actor.key)); - - ds_put_cstr(&ds, "\tactor state: "); - ds_put_lacp_state(&ds, iface_get_lacp_state(iface)); - ds_put_cstr(&ds, "\n\n"); - - ds_put_cstr(&ds, "\tpartner sysid: "); - ds_put_format(&ds, ETH_ADDR_FMT, - ETH_ADDR_ARGS(iface->lacp_partner.sysid)); - ds_put_cstr(&ds, "\n"); - - ds_put_format(&ds, "\tpartner sys_priority: %u\n", - ntohs(iface->lacp_partner.sys_priority)); - - ds_put_format(&ds, "\tpartner portid: %u\n", - ntohs(iface->lacp_partner.portid)); - - ds_put_format(&ds, "\tpartner port_priority: %u\n", - ntohs(iface->lacp_partner.port_priority)); - - ds_put_format(&ds, "\tpartner key: %u\n", - ntohs(iface->lacp_partner.key)); - - ds_put_cstr(&ds, "\tpartner state: "); - ds_put_lacp_state(&ds, iface->lacp_partner.state); - ds_put_cstr(&ds, "\n"); - } - if (port->bond_mode == BM_AB) { continue; } @@ -3897,7 +3579,7 @@ bond_unixctl_show(struct unixctl_conn *conn, int hash = be - port->bond_hash; struct mac_entry *me; - if (be->iface_idx != j) { + if (be->iface != iface) { continue; } @@ -3981,10 +3663,9 @@ bond_unixctl_migrate(struct unixctl_conn *conn, const char *args_, } entry = &port->bond_hash[hash]; - 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; + ofproto_revalidate(port->bridge->ofproto, entry->tag); + entry->iface = iface; + entry->tag = tag_create_random(); unixctl_command_reply(conn, 200, "migrated"); } @@ -4023,10 +3704,10 @@ bond_unixctl_set_active_slave(struct unixctl_conn *conn, const char *args_, return; } - if (port->active_iface != iface->port_ifidx) { - ofproto_revalidate(port->bridge->ofproto, port->active_iface_tag); - port->active_iface = iface->port_ifidx; - port->active_iface_tag = tag_create_random(); + if (port->active_iface != iface) { + ofproto_revalidate(port->bridge->ofproto, + port_get_active_iface_tag(port)); + port->active_iface = iface; VLOG_INFO("port %s: active interface is now %s", port->name, iface->name); bond_send_learning_packets(port); @@ -4136,6 +3817,83 @@ bond_init(void) /* Port functions. */ +static void +lacp_send_pdu_cb(void *aux, const struct lacp_pdu *pdu) +{ + struct iface *iface = aux; + uint8_t ea[ETH_ADDR_LEN]; + int error; + + error = netdev_get_etheraddr(iface->netdev, ea); + if (!error) { + struct ofpbuf packet; + + ofpbuf_init(&packet, ETH_HEADER_LEN + LACP_PDU_LEN); + compose_lacp_packet(&packet, ea, pdu); + ofproto_send_packet(iface->port->bridge->ofproto, + iface->dp_ifidx, 0, &packet); + ofpbuf_uninit(&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)); + } +} + +static void +port_run(struct port *port) +{ + if (port->monitor) { + char *devname; + + /* Track carrier going up and down on interfaces. */ + while (!netdev_monitor_poll(port->monitor, &devname)) { + struct iface *iface; + + iface = port_lookup_iface(port, devname); + if (iface) { + iface_update_carrier(iface); + } + free(devname); + } + } else if (time_msec() >= port->miimon_next_update) { + struct iface *iface; + + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + iface_update_carrier(iface); + } + port->miimon_next_update = time_msec() + port->miimon_interval; + } + + if (port->lacp) { + struct iface *iface; + + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + lacp_slave_enable(port->lacp, iface, iface->enabled); + } + + lacp_run(port->lacp, lacp_send_pdu_cb); + } + + bond_run(port); +} + +static void +port_wait(struct port *port) +{ + if (port->monitor) { + netdev_monitor_poll_wait(port->monitor); + } else { + poll_timer_wait_until(port->miimon_next_update); + } + + if (port->lacp) { + lacp_wait(port->lacp); + } + + bond_wait(port); +} + static struct port * port_create(struct bridge *br, const char *name) { @@ -4147,7 +3905,8 @@ port_create(struct bridge *br, const char *name) port->vlan = -1; port->trunks = NULL; port->name = xstrdup(name); - port->active_iface = -1; + port->active_iface = NULL; + list_init(&port->ifaces); if (br->n_ports >= br->allocated_ports) { br->ports = x2nrealloc(br->ports, &br->allocated_ports, @@ -4187,6 +3946,7 @@ get_interface_other_config(const struct ovsrec_interface *iface, static void port_del_ifaces(struct port *port, const struct ovsrec_port *cfg) { + struct iface *iface, *next; struct shash new_ifaces; size_t i; @@ -4198,11 +3958,9 @@ port_del_ifaces(struct port *port, const struct ovsrec_port *cfg) } /* Get rid of deleted interfaces. */ - for (i = 0; i < port->n_ifaces; ) { - if (!shash_find(&new_ifaces, cfg->interfaces[i]->name)) { - iface_destroy(port->ifaces[i]); - } else { - i++; + LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) { + if (!shash_find(&new_ifaces, iface->name)) { + iface_destroy(iface); } } @@ -4243,24 +4001,26 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) 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); + netdev_monitor_destroy(port->monitor); + port->monitor = NULL; + + if (strcmp(detect_mode, "miimon")) { + port->monitor = netdev_monitor_create(); + + if (strcmp(detect_mode, "carrier")) { + VLOG_WARN("port %s: unsupported bond-detect-mode %s, " + "defaulting to carrier", port->name, detect_mode); + } } - port->bond_miimon_interval = atoi( + port->miimon_interval = atoi( get_port_other_config(cfg, "bond-miimon-interval", "200")); - if (port->bond_miimon_interval < 100) { - port->bond_miimon_interval = 100; + if (port->miimon_interval < 100) { + port->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; + miimon_next_update = time_msec() + port->miimon_interval; + if (port->miimon_next_update > miimon_next_update) { + port->miimon_next_update = miimon_next_update; } if (!port->cfg->bond_mode || @@ -4322,6 +4082,9 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) } shash_destroy(&new_ifaces); + port->lacp_fast = !strcmp(get_port_other_config(cfg, "lacp-time", "slow"), + "fast"); + lacp_priority = atoi(get_port_other_config(cfg, "lacp-system-priority", "0")); @@ -4335,17 +4098,26 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) if (!port->cfg->lacp) { /* XXX when LACP implementation has been sufficiently tested, enable by * default and make active on bonded ports. */ - port->lacp = 0; + lacp_destroy(port->lacp); + port->lacp = NULL; } else if (!strcmp(port->cfg->lacp, "off")) { - port->lacp = 0; + lacp_destroy(port->lacp); + port->lacp = NULL; } else if (!strcmp(port->cfg->lacp, "active")) { - port->lacp = LACP_ACTIVE; + if (!port->lacp) { + port->lacp = lacp_create(); + } + port->lacp_active = true; } else if (!strcmp(port->cfg->lacp, "passive")) { - port->lacp = LACP_PASSIVE; + if (!port->lacp) { + port->lacp = lacp_create(); + } + port->lacp_active = false; } else { VLOG_WARN("port %s: unknown LACP mode %s", port->name, port->cfg->lacp); - port->lacp = 0; + lacp_destroy(port->lacp); + port->lacp = NULL; } /* Get VLAN tag. */ @@ -4413,12 +4185,10 @@ port_destroy(struct port *port) { if (port) { struct bridge *br = port->bridge; + struct iface *iface, *next; 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) { @@ -4426,8 +4196,8 @@ port_destroy(struct port *port) } } - while (port->n_ifaces > 0) { - iface_destroy(port->ifaces[port->n_ifaces - 1]); + LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) { + iface_destroy(iface); } shash_find_and_delete_assert(&br->port_by_name, port->name); @@ -4438,7 +4208,6 @@ port_destroy(struct port *port) VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name); netdev_monitor_destroy(port->monitor); - free(port->ifaces); bitmap_free(port->trunks); free(port->name); free(port); @@ -4469,65 +4238,30 @@ port_lookup_iface(const struct port *port, const char *name) static void port_update_lacp(struct port *port) { - size_t i; - bool key_changed; - - if (!port->lacp || port->n_ifaces < 1) { - for (i = 0; i < port->n_ifaces; i++) { - iface_set_lacp_defaulted(port->ifaces[i]); - } - return; - } - - key_changed = true; - for (i = 0; i < port->n_ifaces; i++) { - struct iface *iface = port->ifaces[i]; + if (port->lacp) { + struct iface *iface; - if (iface->dp_ifidx <= 0 || iface->dp_ifidx > UINT16_MAX) { - port->lacp = 0; - return; - } + lacp_configure(port->lacp, port->name, + port->bridge->ea, port->lacp_priority, + port->lacp_active, port->lacp_fast); - if (iface->dp_ifidx == port->lacp_key) { - key_changed = false; + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + lacp_slave_register(port->lacp, iface, iface->name, + iface->dp_ifidx, iface->lacp_priority); } } - - if (key_changed) { - port->lacp_key = port->ifaces[0]->dp_ifidx; - } - - for (i = 0; i < port->n_ifaces; i++) { - struct iface *iface = port->ifaces[i]; - - iface->lacp_actor.sys_priority = htons(port->lacp_priority); - memcpy(&iface->lacp_actor.sysid, port->bridge->ea, ETH_ADDR_LEN); - - iface->lacp_actor.port_priority = htons(iface->lacp_priority); - iface->lacp_actor.portid = htons(iface->dp_ifidx); - iface->lacp_actor.key = htons(port->lacp_key); - - iface->lacp_tx = 0; - } - port->lacp_need_update = true; } static void port_update_bonding(struct port *port) { - if (port->monitor) { - netdev_monitor_destroy(port->monitor); - port->monitor = NULL; - } if (port->n_ifaces < 2) { /* Not a bonded port. */ - if (port->bond_hash) { - 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 = NULL; + port->no_ifaces_tag = 0; } else { size_t i; @@ -4535,245 +4269,34 @@ port_update_bonding(struct port *port) 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]; - e->iface_idx = -1; + e->iface = NULL; 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) { + 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; + port->bond_fake_iface = port->cfg->bond_fake_iface; + if (port->bond_fake_iface) { + port->bond_next_fake_iface_update = time_msec(); } - 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); - } } - - 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; - } - } - } - } - proc_net_compat_update_vlan(port->name, vlandev_name, port->vlan); } /* Interface functions. */ -static void -iface_set_lacp_defaulted(struct iface *iface) -{ - memset(&iface->lacp_partner, 0, sizeof iface->lacp_partner); - - iface->lacp_status = LACP_STATUS_DEFAULTED; - iface->lacp_tx = 0; - iface->port->lacp_need_update = true; -} - -static void -iface_set_lacp_expired(struct iface *iface) -{ - iface->lacp_status = LACP_STATUS_EXPIRED; - iface->lacp_partner.state |= LACP_STATE_TIME; - iface->lacp_partner.state &= ~LACP_STATE_SYNC; - - iface->lacp_rx = time_msec() + LACP_FAST_TIME_RX; - iface->lacp_tx = 0; -} - -static uint8_t -iface_get_lacp_state(const struct iface *iface) -{ - uint8_t state = 0; - - if (iface->port->lacp & LACP_ACTIVE) { - state |= LACP_STATE_ACT; - } - - if (iface->lacp_status == LACP_STATUS_DEFAULTED) { - state |= LACP_STATE_DEF; - } else if (iface->lacp_attached) { - state |= LACP_STATE_SYNC; - } - - if (iface->lacp_status == LACP_STATUS_EXPIRED) { - state |= LACP_STATE_EXP; - } - - if (iface->port->n_ifaces > 1) { - state |= LACP_STATE_AGG; - } - - if (iface->enabled) { - state |= LACP_STATE_COL | LACP_STATE_DIST; - } - - return state; -} - -/* Given 'iface', populates 'priority' with data representing its LACP link - * priority. If two priority objects populated by this function are compared - * using memcmp, the higher priority link will be less than the lower priority - * link. */ -static void -iface_get_lacp_priority(struct iface *iface, struct lacp_info *priority) -{ - uint16_t partner_priority, actor_priority; - - /* Choose the lacp_info of the higher priority system by comparing their - * system priorities and mac addresses. */ - actor_priority = ntohs(iface->lacp_actor.sys_priority); - partner_priority = ntohs(iface->lacp_partner.sys_priority); - if (actor_priority < partner_priority) { - *priority = iface->lacp_actor; - } else if (partner_priority < actor_priority) { - *priority = iface->lacp_partner; - } else if (eth_addr_compare_3way(iface->lacp_actor.sysid, - iface->lacp_partner.sysid) < 0) { - *priority = iface->lacp_actor; - } else { - *priority = iface->lacp_partner; - } - - /* Key and state are not used in priority comparisons. */ - priority->key = 0; - priority->state = 0; -} - -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) { @@ -4783,26 +4306,18 @@ iface_create(struct port *port, const struct ovsrec_interface *if_cfg) iface = xzalloc(sizeof *iface); iface->port = port; - iface->port_ifidx = port->n_ifaces; iface->name = xstrdup(name); iface->dp_ifidx = -1; iface->tag = tag_create_random(); iface->delay_expires = LLONG_MAX; iface->netdev = NULL; iface->cfg = if_cfg; - iface_set_lacp_defaulted(iface); - - if (port->lacp & LACP_ACTIVE) { - iface_set_lacp_expired(iface); - } shash_add_assert(&br->iface_by_name, iface->name, iface); - if (port->n_ifaces >= port->allocated_ifaces) { - port->ifaces = x2nrealloc(port->ifaces, &port->allocated_ifaces, - sizeof *port->ifaces); - } - port->ifaces[port->n_ifaces++] = iface; + list_push_back(&port->ifaces, &iface->port_elem); + port->n_ifaces++; + if (port->n_ifaces > 1) { br->has_bonded_ports = true; } @@ -4820,10 +4335,22 @@ iface_destroy(struct iface *iface) if (iface) { struct port *port = iface->port; struct bridge *br = port->bridge; - bool del_active = port->active_iface == iface->port_ifidx; - struct iface *del; + bool del_active = port->active_iface == iface; + + if (port->bond_hash) { + struct bond_entry *e; + for (e = port->bond_hash; e <= &port->bond_hash[BOND_MASK]; e++) { + if (e->iface == iface) { + e->iface = NULL; + } + } + } - if (port->monitor) { + if (iface->port->lacp) { + lacp_slave_unregister(iface->port->lacp, iface); + } + + if (port->monitor && iface->netdev) { netdev_monitor_remove(port->monitor, iface->netdev); } @@ -4833,19 +4360,16 @@ iface_destroy(struct iface *iface) hmap_remove(&br->ifaces, &iface->dp_ifidx_node); } - del = port->ifaces[iface->port_ifidx] = port->ifaces[--port->n_ifaces]; - del->port_ifidx = iface->port_ifidx; + list_remove(&iface->port_elem); + port->n_ifaces--; netdev_close(iface->netdev); if (del_active) { - ofproto_revalidate(port->bridge->ofproto, port->active_iface_tag); bond_choose_active_iface(port); bond_send_learning_packets(port); } - cfm_destroy(iface->cfm); - free(iface->name); free(iface); @@ -4859,6 +4383,21 @@ 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) { @@ -4979,6 +4518,20 @@ iface_delete_queues(unsigned int queue_id, } } +static void +iface_update_carrier(struct iface *iface) +{ + bool carrier = iface_get_carrier(iface); + if (carrier == iface->up) { + return; + } + + iface->up = carrier; + if (iface->port->lacp) { + lacp_slave_carrier_changed(iface->port->lacp, iface); + } +} + static void iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos) { @@ -5019,7 +4572,7 @@ static void iface_update_cfm(struct iface *iface) { size_t i; - struct cfm *cfm; + struct cfm cfm; uint16_t *remote_mps; struct ovsrec_monitor *mon; uint8_t ea[ETH_ADDR_LEN], maid[CCM_MAID_LEN]; @@ -5027,8 +4580,7 @@ iface_update_cfm(struct iface *iface) mon = iface->cfg->monitor; if (!mon) { - cfm_destroy(iface->cfm); - iface->cfm = NULL; + ofproto_iface_clear_cfm(iface->port->bridge->ofproto, iface->dp_ifidx); return; } @@ -5043,28 +4595,32 @@ iface_update_cfm(struct iface *iface) return; } - if (!iface->cfm) { - iface->cfm = cfm_create(); - } - - cfm = iface->cfm; - cfm->mpid = mon->mpid; - cfm->interval = mon->interval ? *mon->interval : 1000; + 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); + 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); + + ofproto_iface_set_cfm(iface->port->bridge->ofproto, iface->dp_ifidx, + &cfm, remote_mps, mon->n_remote_mps); free(remote_mps); +} - if (!cfm_configure(iface->cfm)) { - cfm_destroy(iface->cfm); - iface->cfm = NULL; - } +/* Read carrier or miimon status directly from 'iface''s netdev, according to + * how 'iface''s port is configured. + * + * Returns true if 'iface' is up, false otherwise. */ +static bool +iface_get_carrier(const struct iface *iface) +{ + return (iface->port->monitor + ? netdev_get_carrier(iface->netdev) + : netdev_get_miimon(iface->netdev)); } /* Port mirroring. */