From: Ethan Jackson Date: Wed, 19 Jan 2011 02:46:58 +0000 (-0800) Subject: vswitchd: Implement Link Aggregation Control Protocol. X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=c25c91fd5ed075a6e9f37ab38c3ea2302e9e8442;p=openvswitch vswitchd: Implement Link Aggregation Control Protocol. This commit implements LACP, a protocol which allows directly connected switches to automatically negotiate which links may participate in bonds. This commit disables LACP by default. Once sufficiently tested, LACP will be enabled in "active" mode on bonded ports, and "passive" mode on all others. Bug #4213. --- diff --git a/lib/packets.c b/lib/packets.c index 05148fe0..c1e1cdbb 100644 --- a/lib/packets.c +++ b/lib/packets.c @@ -203,3 +203,59 @@ ipv6_is_cidr(const struct in6_addr *netmask) return true; } + +/* Fills 'b' with a LACP packet whose source address is 'eth_src', LACP actor + * information is 'actor', and LACP partner information is 'partner'. */ +void +compose_lacp_packet(struct ofpbuf *b, struct lacp_info *actor, + struct lacp_info *partner, + const uint8_t eth_src[ETH_ADDR_LEN]) +{ + struct eth_header *eth; + struct lacp_pdu *pdu; + + ofpbuf_clear(b); + + ofpbuf_prealloc_tailroom(b, ETH_HEADER_LEN + LACP_PDU_LEN); + eth = ofpbuf_put_zeros(b, ETH_HEADER_LEN); + pdu = ofpbuf_put_zeros(b, LACP_PDU_LEN); + + memcpy(eth->eth_dst, eth_addr_lacp, ETH_ADDR_LEN); + memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN); + eth->eth_type = htons(ETH_TYPE_LACP); + + pdu->subtype = 1; + pdu->version = 1; + + pdu->actor_type = 1; + pdu->actor_len = 20; + pdu->actor = *actor; + + pdu->partner_type = 2; + pdu->partner_len = 20; + pdu->partner = *partner; + + pdu->collector_type = 3; + pdu->collector_len = 16; + pdu->collector_delay = htons(UINT16_MAX); +} + +/* Parses 'b' which represents a packet containing a LACP PDU. This function + * returns NULL if 'b' is malformed, or does not represent a LACP PDU format + * supported by OVS. Otherwise, it returns a pointer to the lacp_pdu contained + * within 'b'. */ +const struct lacp_pdu * +parse_lacp_packet(const struct ofpbuf *b) +{ + const struct lacp_pdu *pdu; + + pdu = ofpbuf_at(b, (uint8_t *)b->l3 - (uint8_t *)b->data, LACP_PDU_LEN); + + if (pdu && pdu->subtype == 1 + && pdu->actor_type == 1 && pdu->actor_len == 20 + && pdu->partner_type == 2 && pdu->partner_len == 20) { + return pdu; + } else { + return NULL; + } +} diff --git a/lib/packets.h b/lib/packets.h index 38f98fb0..0bc368b9 100644 --- a/lib/packets.h +++ b/lib/packets.h @@ -40,6 +40,9 @@ static const uint8_t eth_addr_broadcast[ETH_ADDR_LEN] OVS_UNUSED static const uint8_t eth_addr_stp[ETH_ADDR_LEN] OVS_UNUSED = { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x01 }; +static const uint8_t eth_addr_lacp[ETH_ADDR_LEN] OVS_UNUSED + = { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x02 }; + static inline bool eth_addr_is_broadcast(const uint8_t ea[6]) { return (ea[0] & ea[1] & ea[2] & ea[3] & ea[4] & ea[5]) == 0xff; @@ -161,6 +164,7 @@ void compose_benign_packet(struct ofpbuf *, const char *tag, #define ETH_TYPE_VLAN 0x8100 #define ETH_TYPE_IPV6 0x86dd #define ETH_TYPE_CFM 0x8902 +#define ETH_TYPE_LACP 0x8809 /* Minimum value for an Ethernet type. Values below this are IEEE 802.2 frame * lengths. */ @@ -408,4 +412,58 @@ struct in6_addr ipv6_create_mask(int mask); int ipv6_count_cidr_bits(const struct in6_addr *netmask); bool ipv6_is_cidr(const struct in6_addr *netmask); +/* Masks for lacp_info state member. */ +#define LACP_STATE_ACT 0x01 /* Activity. Active or passive? */ +#define LACP_STATE_TIME 0x02 /* Timeout. Short or long timeout? */ +#define LACP_STATE_AGG 0x04 /* Aggregation. Is the link is bondable? */ +#define LACP_STATE_SYNC 0x08 /* Synchronization. Is the link in up to date? */ +#define LACP_STATE_COL 0x10 /* Collecting. Is the link receiving frames? */ +#define LACP_STATE_DIST 0x20 /* Distributing. Is the link sending frames? */ +#define LACP_STATE_DEF 0x40 /* Defaulted. Using default partner info? */ +#define LACP_STATE_EXP 0x80 /* Expired. Using expired partner info? */ + +#define LACP_FAST_TIME_TX 1000 /* Fast transmission rate. */ +#define LACP_SLOW_TIME_TX 30000 /* Slow transmission rate. */ +#define LACP_FAST_TIME_RX (LACP_FAST_TIME_TX * 3) /* Fast receive rate. */ +#define LACP_SLOW_TIME_RX (LACP_SLOW_TIME_TX * 3) /* Slow receive rate. */ + +#define LACP_INFO_LEN 15 +struct lacp_info { + ovs_be16 sys_priority; /* System priority. */ + uint8_t sysid[ETH_ADDR_LEN]; /* System ID. */ + ovs_be16 key; /* Operational key. */ + ovs_be16 port_priority; /* Port priority. */ + ovs_be16 portid; /* Port ID. */ + uint8_t state; /* State mask. See LACP_STATE macros. */ +} __attribute__((packed)); +BUILD_ASSERT_DECL(LACP_INFO_LEN == sizeof(struct lacp_info)); + +#define LACP_PDU_LEN 110 +struct lacp_pdu { + uint8_t subtype; /* Always 1. */ + uint8_t version; /* Always 1. */ + + uint8_t actor_type; /* Always 1. */ + uint8_t actor_len; /* Always 20. */ + struct lacp_info actor; /* LACP actor information. */ + uint8_t z1[3]; /* Reserved. Always 0. */ + + uint8_t partner_type; /* Always 2. */ + uint8_t partner_len; /* Always 20. */ + struct lacp_info partner; /* LACP partner information. */ + uint8_t z2[3]; /* Reserved. Always 0. */ + + uint8_t collector_type; /* Always 3. */ + uint8_t collector_len; /* Always 16. */ + ovs_be16 collector_delay; /* Maximum collector delay. Set to UINT16_MAX. */ + uint8_t z3[64]; /* Combination of several fields. Always 0. */ +} __attribute__((packed)); +BUILD_ASSERT_DECL(LACP_PDU_LEN == sizeof(struct lacp_pdu)); + +void compose_lacp_packet(struct ofpbuf *, struct lacp_info *actor, + struct lacp_info *partner, + const uint8_t eth_src[ETH_ADDR_LEN]); + +const struct lacp_pdu *parse_lacp_packet(const struct ofpbuf *); + #endif /* packets.h */ diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index b2b997dc..39a94e12 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -77,6 +77,12 @@ COVERAGE_DEFINE(bridge_flush); COVERAGE_DEFINE(bridge_process_flow); COVERAGE_DEFINE(bridge_reconfigure); +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; @@ -110,6 +116,16 @@ struct iface { 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. */ }; #define BOND_MASK 0xff @@ -145,6 +161,11 @@ 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; @@ -174,6 +195,12 @@ struct port { 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. */ + 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. */ int bond_rebalance_interval; /* Interval between rebalances, in ms. */ @@ -253,6 +280,10 @@ 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 void bond_init(void); static void bond_run(struct bridge *); static void bond_wait(struct bridge *); @@ -271,6 +302,7 @@ static struct port *port_from_dp_ifidx(const struct bridge *, 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 *); static void mirror_create(struct bridge *, struct ovsrec_mirror *); static void mirror_destroy(struct mirror *); @@ -290,6 +322,10 @@ 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 shash_from_ovs_idl_map(char **keys, char **values, size_t n, struct shash *); @@ -877,6 +913,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) port_update_vlan_compat(port); port_update_bonding(port); + port_update_lacp(port); for (j = 0; j < port->n_ifaces; j++) { iface_update_qos(port->ifaces[j], port->cfg->qos); @@ -1459,6 +1496,7 @@ bridge_wait(void) } mac_learning_wait(br->ml); + lacp_wait(br); bond_wait(br); HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) { @@ -1674,6 +1712,7 @@ bridge_run_one(struct bridge *br) } mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto)); + lacp_run(br); bond_run(br); error = ofproto_run2(br->ofproto, br->flush); @@ -2092,7 +2131,9 @@ bond_choose_iface(const struct port *port) if (iface->enabled) { return i; - } else if (iface->delay_expires < next_delay_expiration) { + } 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; } @@ -2154,33 +2195,54 @@ 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 carrier = iface->up; + bool up = iface->up; + int updelay, downdelay; + + updelay = port->updelay; + downdelay = port->downdelay; + + if (iface->port->lacp & LACP_NEGOTIATED) { + 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 ((carrier == iface->enabled) == (iface->delay_expires == LLONG_MAX)) { + if ((up == iface->enabled) == (iface->delay_expires == LLONG_MAX)) { /* Nothing to do. */ return; } VLOG_INFO_RL(&rl, "interface %s: link state %s", - iface->name, carrier ? "up" : "down"); - if (carrier == iface->enabled) { + iface->name, up ? "up" : "down"); + if (up == iface->enabled) { iface->delay_expires = LLONG_MAX; VLOG_INFO_RL(&rl, "interface %s: will not be %s", - iface->name, carrier ? "disabled" : "enabled"); - } else if (carrier && port->active_iface < 0) { + iface->name, up ? "disabled" : "enabled"); + } else if (up && port->active_iface < 0) { bond_enable_slave(iface, true); - if (port->updelay) { + if (updelay) { VLOG_INFO_RL(&rl, "interface %s: skipping %d ms updelay since no " - "other interface is up", iface->name, port->updelay); + "other interface is up", iface->name, updelay); } } else { - int delay = carrier ? port->updelay : port->downdelay; + int delay = up ? updelay : downdelay; iface->delay_expires = time_msec() + delay; if (delay) { VLOG_INFO_RL(&rl, "interface %s: will be %s if it stays %s for %d ms", iface->name, - carrier ? "enabled" : "disabled", - carrier ? "up" : "down", + up ? "enabled" : "disabled", + up ? "up" : "down", delay); } } @@ -2288,6 +2350,22 @@ bond_update_fake_iface_stats(struct port *port) } } +static void +bond_link_carrier_update(struct iface *iface, bool carrier) +{ + if (carrier == iface->up) { + return; + } + + if (iface->lacp_status == LACP_STATUS_CURRENT) { + iface_set_lacp_expired(iface); + } + + iface->up = carrier; + iface->lacp_tx = 0; + iface->port->bond_compat_is_stale = true; +} + static void bond_run(struct bridge *br) { @@ -2309,11 +2387,7 @@ bond_run(struct bridge *br) iface = port_lookup_iface(port, devname); if (iface) { bool up = netdev_get_carrier(iface->netdev); - - if (up != iface->up) { - port->bond_compat_is_stale = true; - } - iface->up = up; + bond_link_carrier_update(iface, up); } free(devname); } @@ -2324,11 +2398,7 @@ bond_run(struct bridge *br) for (j = 0; j < port->n_ifaces; j++) { struct iface *iface = port->ifaces[j]; bool up = netdev_get_miimon(iface->netdev); - - if (up != iface->up) { - port->bond_compat_is_stale = true; - } - iface->up = up; + bond_link_carrier_update(iface, up); } port->bond_miimon_next_update = time_msec() + port->bond_miimon_interval; @@ -2811,8 +2881,14 @@ is_admissible(struct bridge *br, const struct flow *flow, bool have_packet, return false; } - /* Packets received on bonds need special attention to avoid duplicates. */ - if (in_port->n_ifaces > 1) { + /* When using LACP, do not accept packets from disabled interfaces. */ + if (in_port->lacp & LACP_NEGOTIATED && !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)) { int src_idx; bool is_grat_arp_locked; @@ -2913,6 +2989,11 @@ bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet, cfm_process_heartbeat(iface->cfm, packet); } return false; + } else if (flow->dl_type == htons(ETH_TYPE_LACP)) { + if (packet) { + lacp_process_packet(packet, iface); + } + return false; } return process_flow(br, flow, packet, actions, tags, nf_output_iface); @@ -2990,6 +3071,192 @@ static struct ofhooks bridge_ofhooks = { bridge_account_checkpoint_ofhook_cb, }; +/* LACP functions. */ + +static void +lacp_process_packet(const struct ofpbuf *packet, struct iface *iface) +{ + const struct lacp_pdu *pdu; + + if (!iface->port->lacp) { + return; + } + + pdu = parse_lacp_packet(packet); + if (!pdu) { + return; + } + + iface->lacp_status = LACP_STATUS_CURRENT; + iface->lacp_rx = time_msec() + LACP_SLOW_TIME_RX; + + iface->lacp_actor.state = iface_get_lacp_state(iface); + if (memcmp(&iface->lacp_actor, &pdu->partner, sizeof pdu->partner)) { + iface->lacp_tx = 0; + } + + if (memcmp(&iface->lacp_partner, &pdu->actor, sizeof pdu->actor)) { + iface->port->lacp_need_update = true; + iface->lacp_partner = pdu->actor; + } +} + +static void +lacp_update_ifaces(struct port *port) +{ + size_t i; + struct iface *lead; + struct lacp_info lead_pri; + + port->lacp_need_update = false; + + if (!port->lacp) { + return; + } + + 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)) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10); + 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 @@ -3429,6 +3696,42 @@ 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) @@ -3445,6 +3748,14 @@ bond_unixctl_show(struct unixctl_conn *conn, ds_put_format(&ds, "bond_mode: %s\n", bond_mode_to_string(port->bond_mode)); + + if (port->lacp) { + ds_put_format(&ds, "\tlacp: %s\n", + port->lacp & LACP_ACTIVE ? "active" : "passive"); + } else { + ds_put_cstr(&ds, "\tlacp: off\n"); + } + ds_put_format(&ds, "bond-detect-mode: %s\n", port->miimon ? "miimon" : "carrier"); @@ -3477,6 +3788,66 @@ 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\n"); + } + if (port->bond_mode != BM_SLB) { continue; } @@ -3756,6 +4127,17 @@ get_port_other_config(const struct ovsrec_port *port, const char *key, return value ? value : default_value; } +static const char * +get_interface_other_config(const struct ovsrec_interface *iface, + const char *key, const char *default_value) +{ + const char *value; + + value = get_ovsrec_key_value(&iface->header_, + &ovsrec_interface_col_other_config, key); + return value ? value : default_value; +} + static void port_del_ifaces(struct port *port, const struct ovsrec_port *cfg) { @@ -3786,7 +4168,7 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) { const char *detect_mode; struct shash new_ifaces; - long long int next_rebalance, miimon_next_update; + long long int next_rebalance, miimon_next_update, lacp_priority; unsigned long *trunks; int vlan; size_t i; @@ -3879,9 +4261,45 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal" : if_cfg->type[0] ? if_cfg->type : "system"); + + lacp_priority = + atoi(get_interface_other_config(if_cfg, "lacp-port-priority", + "0")); + + if (lacp_priority <= 0 || lacp_priority > UINT16_MAX) { + iface->lacp_priority = UINT16_MAX; + } else { + iface->lacp_priority = lacp_priority; + } } shash_destroy(&new_ifaces); + lacp_priority = + atoi(get_port_other_config(cfg, "lacp-system-priority", "0")); + + if (lacp_priority <= 0 || lacp_priority > UINT16_MAX) { + /* Prefer bondable links if unspecified. */ + port->lacp_priority = port->n_ifaces > 1 ? UINT16_MAX - 1 : UINT16_MAX; + } else { + port->lacp_priority = lacp_priority; + } + + if (!port->cfg->lacp) { + /* XXX when LACP implementation has been sufficiently tested, enable by + * default and make active on bonded ports. */ + port->lacp = 0; + } else if (!strcmp(port->cfg->lacp, "off")) { + port->lacp = 0; + } else if (!strcmp(port->cfg->lacp, "active")) { + port->lacp = LACP_ACTIVE; + } else if (!strcmp(port->cfg->lacp, "passive")) { + port->lacp = LACP_PASSIVE; + } else { + VLOG_WARN("port %s: unknown LACP mode %s", + port->name, port->cfg->lacp); + port->lacp = 0; + } + /* Get VLAN tag. */ vlan = -1; if (cfg->tag) { @@ -4000,6 +4418,49 @@ port_lookup_iface(const struct port *port, const char *name) return iface && iface->port == port ? iface : NULL; } +static void +port_update_lacp(struct port *port) +{ + size_t i; + bool key_changed; + + if (!port->lacp || port->n_ifaces < 1) { + return; + } + + key_changed = true; + for (i = 0; i < port->n_ifaces; i++) { + struct iface *iface = port->ifaces[i]; + + if (iface->dp_ifidx <= 0 || iface->dp_ifidx > UINT16_MAX) { + port->lacp = 0; + return; + } + + if (iface->dp_ifidx == port->lacp_key) { + key_changed = false; + } + } + + 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) { @@ -4162,6 +4623,87 @@ port_update_vlan_compat(struct port *port) /* Interface functions. */ +static void +iface_set_lacp_defaulted(struct iface *iface) +{ + memset(&iface->lacp_partner, 0xff, sizeof iface->lacp_partner); + iface->lacp_partner.state = 0; + + 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) { @@ -4198,6 +4740,11 @@ iface_create(struct port *port, const struct ovsrec_interface *if_cfg) 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); diff --git a/vswitchd/vswitch.ovsschema b/vswitchd/vswitch.ovsschema index 85534997..6282a825 100644 --- a/vswitchd/vswitch.ovsschema +++ b/vswitchd/vswitch.ovsschema @@ -1,6 +1,6 @@ {"name": "Open_vSwitch", - "version": "1.2.0", - "cksum": "2368100022 15573", + "version": "1.3.0", + "cksum": "1230201059 15724", "tables": { "Open_vSwitch": { "columns": { @@ -124,6 +124,10 @@ "type": {"key": {"type": "string", "enum": ["set", ["balance-slb", "active-backup"]]}, "min": 0, "max": 1}}, + "lacp": { + "type": {"key": {"type": "string", + "enum": ["set", ["active", "passive", "off"]]}, + "min": 0, "max": 1}}, "bond_updelay": { "type": "integer"}, "bond_downdelay": { diff --git a/vswitchd/vswitch.xml b/vswitchd/vswitch.xml index a02ea534..d9e71b33 100644 --- a/vswitchd/vswitch.xml +++ b/vswitchd/vswitch.xml @@ -546,6 +546,18 @@ name of the port. Use only for compatibility with legacy software that requires this. + + +

Configures LACP on this port. LACP allows directly connected + switches to negotiate which links may be bonded. LACP may be enabled + on non-bonded ports for the benefit of any switches they may be + connected to. active ports are allowed to initiate LACP + negotiations. passive ports are allowed to participate + in LACP negotiations initiated by a remote switch, but not allowed to + initiate such negotiations themselves. If unset Open vSwitch will + choose a reasonable default.

+ + @@ -607,6 +619,11 @@
The number of milliseconds between successive attempts to poll each interface's MII. Only relevant on ports which use miimon to detect failures.
+
lacp-system-priority
+
The LACP system priority of this . In + LACP negotiations, link status decisions are made by the system + with the numerically lower priority. Must be a number between 1 + and 65535.
 @@ -1194,8 +1211,14 @@ - Key-value pairs for rarely used interface features. Currently, - there are none defined. + Key-value pairs for rarely used interface features. +
+
lacp-port-priority
+
The LACP port priority of this . In + LACP negotiations s with numerically lower + priorities are preferred for aggregation. Must be a number between + 1 and 65535.
+