X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=eadafb0abed69d7c10ea1796aae82581b626845e;hb=16472256190e083258bfcf6a4ba4f74cb412fcd4;hp=0111040809f9c8658524c82012fe825e569234ec;hpb=8834f25e9983910258cd0cd1e400a14d9a664479;p=openvswitch diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index 01110408..eadafb0a 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -35,6 +35,7 @@ #include "cfm.h" #include "classifier.h" #include "coverage.h" +#include "daemon.h" #include "dirs.h" #include "dpif.h" #include "dynamic-string.h" @@ -54,7 +55,6 @@ #include "ovsdb-data.h" #include "packets.h" #include "poll-loop.h" -#include "proc-net-compat.h" #include "process.h" #include "sha1.h" #include "shash.h" @@ -75,13 +75,10 @@ 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); - -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 */ - LACP_STATUS_EXPIRED, /* Partner is out of date.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. */ -}; +COVERAGE_DEFINE(bridge_lacp_update); struct dst { uint16_t vlan; @@ -98,6 +95,13 @@ static void dst_set_init(struct dst_set *); static void dst_set_add(struct dst_set *, const struct dst *); static void dst_set_free(struct dst_set *); +enum lacp_status { + LACP_CURRENT = 0x01, /* Current State. */ + LACP_EXPIRED = 0x02, /* Expired State. */ + LACP_DEFAULTED = 0x04, /* Partner is defaulted. */ + LACP_ATTACHED = 0x08, /* Attached. Interface may be choosen for flows. */ +}; + struct iface { /* These members are always valid. */ struct port *port; /* Containing port. */ @@ -118,14 +122,12 @@ struct iface { const struct ovsrec_interface *cfg; /* LACP information. */ - enum lacp_status lacp_status; /* LACP state machine status. */ + enum lacp_status lacp_status; /* LACP status. */ uint16_t lacp_priority; /* LACP port priority. */ struct lacp_info lacp_actor; /* LACP actor information. */ struct lacp_info lacp_partner; /* LACP partner information. */ long long int lacp_tx; /* Next LACP message transmission time. */ long long int lacp_rx; /* Next LACP message receive time. */ - bool lacp_attached; /* Attached to its aggregator? LACP allows - this link to be chosen for flows. */ }; #define BOND_MASK 0xff @@ -188,7 +190,6 @@ struct port { tag_type active_iface_tag; /* Tag for bcast flows. */ tag_type no_ifaces_tag; /* Tag for flows when all ifaces disabled. */ int updelay, downdelay; /* Delay before iface goes up/down, in ms. */ - bool bond_compat_is_stale; /* Need to call port_update_bond_compat()? */ bool bond_fake_iface; /* Fake a bond interface for legacy compat? */ bool miimon; /* Use miimon instead of carrier? */ long long int bond_miimon_interval; /* Miimon status refresh interval. */ @@ -300,8 +301,6 @@ static struct port *port_lookup(const struct bridge *, const char *name); static struct iface *port_lookup_iface(const struct port *, const char *name); static struct port *port_from_dp_ifidx(const struct bridge *, uint16_t dp_ifidx); -static void port_update_bond_compat(struct port *); -static void port_update_vlan_compat(struct port *); static void port_update_bonding(struct port *); static void port_update_lacp(struct port *); @@ -527,13 +526,10 @@ 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]; @@ -912,7 +908,6 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) struct port *port = br->ports[i]; int j; - port_update_vlan_compat(port); port_update_bonding(port); port_update_lacp(port); @@ -933,6 +928,10 @@ 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 * @@ -982,7 +981,7 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], /* Otherwise choose the minimum non-local MAC address among all of the * interfaces. */ - memset(ea, 0xff, sizeof ea); + memset(ea, 0xff, ETH_ADDR_LEN); for (i = 0; i < br->n_ports; i++) { struct port *port = br->ports[i]; uint8_t iface_ea[ETH_ADDR_LEN]; @@ -2157,7 +2156,7 @@ bond_choose_iface(const struct port *port) if (iface->enabled) { return i; } else if (iface->delay_expires < next_delay_expiration - && (iface->lacp_attached + && (iface->lacp_status & LACP_ATTACHED || !(port->lacp & LACP_NEGOTIATED))) { best_down_slave = i; next_delay_expiration = iface->delay_expires; @@ -2203,7 +2202,6 @@ choose_output_iface(const struct port *port, const struct flow *flow, return false; } e->iface_tag = tag_create_random(); - ((struct port *) port)->bond_compat_is_stale = true; } *tags |= e->iface_tag; iface = port->ifaces[e->iface_idx]; @@ -2235,9 +2233,9 @@ bond_link_status_update(struct iface *iface) * They are not required to have synchronized partners because they * have no partners at all. However, they will only be attached if * negotiations failed on all interfaces in the bond. */ - up = iface->lacp_attached + up = iface->lacp_status & LACP_ATTACHED && (iface->lacp_partner.state & LACP_STATE_SYNC - || iface->lacp_status == LACP_STATUS_DEFAULTED); + || iface->lacp_status & LACP_DEFAULTED); } @@ -2333,7 +2331,6 @@ 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 @@ -2380,13 +2377,12 @@ bond_link_carrier_update(struct iface *iface, bool carrier) return; } - if (iface->lacp_status == LACP_STATUS_CURRENT) { + if (iface->lacp_status & LACP_CURRENT) { iface_set_lacp_expired(iface); } iface->up = carrier; iface->lacp_tx = 0; - iface->port->bond_compat_is_stale = true; } static void @@ -2445,11 +2441,6 @@ bond_run(struct bridge *br) port->bond_next_fake_iface_update = time_msec() + 1000; } } - - if (port->bond_compat_is_stale) { - port->bond_compat_is_stale = false; - port_update_bond_compat(port); - } } } @@ -3000,33 +2991,45 @@ bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet, struct ofpbuf *actions, tag_type *tags, uint16_t *nf_output_iface, void *br_) { - struct iface *iface; struct bridge *br = br_; COVERAGE_INC(bridge_process_flow); + return process_flow(br, flow, packet, actions, tags, nf_output_iface); +} + +static bool +bridge_special_ofhook_cb(const struct flow *flow, + const struct ofpbuf *packet, void *br_) +{ + struct iface *iface; + struct bridge *br = br_; iface = iface_from_dp_ifidx(br, flow->in_port); if (cfm_should_process_flow(flow)) { - if (packet && iface->cfm) { + + if (iface && packet && iface->cfm) { + COVERAGE_INC(bridge_process_cfm); cfm_process_heartbeat(iface->cfm, packet); } return false; } else if (flow->dl_type == htons(ETH_TYPE_LACP)) { - if (packet) { + + if (iface && packet) { + COVERAGE_INC(bridge_process_lacp); lacp_process_packet(packet, iface); } return false; } - return process_flow(br, flow, packet, actions, tags, nf_output_iface); + return true; } static void bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags, const struct nlattr *actions, size_t actions_len, - unsigned long long int n_bytes, void *br_) + uint64_t n_bytes, void *br_) { struct bridge *br = br_; const struct nlattr *a; @@ -3054,7 +3057,7 @@ 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); @@ -3079,7 +3082,7 @@ 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); @@ -3089,6 +3092,7 @@ bridge_account_checkpoint_ofhook_cb(void *br_) static struct ofhooks bridge_ofhooks = { bridge_normal_ofhook_cb, + bridge_special_ofhook_cb, bridge_account_flow_ofhook_cb, bridge_account_checkpoint_ofhook_cb, }; @@ -3109,7 +3113,8 @@ lacp_process_packet(const struct ofpbuf *packet, struct iface *iface) return; } - iface->lacp_status = LACP_STATUS_CURRENT; + iface->lacp_status |= LACP_CURRENT; + iface->lacp_status &= ~(LACP_EXPIRED | LACP_DEFAULTED); iface->lacp_rx = time_msec() + LACP_SLOW_TIME_RX; iface->lacp_actor.state = iface_get_lacp_state(iface); @@ -3129,32 +3134,35 @@ lacp_update_ifaces(struct port *port) 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; } + 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; + iface->lacp_status |= LACP_ATTACHED; 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; + iface->lacp_status &= ~LACP_ATTACHED; continue; } - if (iface->lacp_status == LACP_STATUS_DEFAULTED) { + if (iface->lacp_status & LACP_DEFAULTED) { continue; } @@ -3176,11 +3184,11 @@ lacp_update_ifaces(struct port *port) for (i = 0; i < port->n_ifaces; i++) { struct iface *iface = port->ifaces[i]; - if (iface->lacp_status == LACP_STATUS_DEFAULTED + if (iface->lacp_status & LACP_DEFAULTED || lead->lacp_partner.key != iface->lacp_partner.key || !eth_addr_equals(lead->lacp_partner.sysid, iface->lacp_partner.sysid)) { - iface->lacp_attached = false; + iface->lacp_status &= ~LACP_ATTACHED; } } } @@ -3189,7 +3197,7 @@ static bool lacp_iface_may_tx(const struct iface *iface) { return iface->port->lacp & LACP_ACTIVE - || iface->lacp_status != LACP_STATUS_DEFAULTED; + || iface->lacp_status & (LACP_CURRENT | LACP_EXPIRED); } static void @@ -3211,9 +3219,9 @@ lacp_run(struct bridge *br) struct iface *iface = port->ifaces[j]; if (time_msec() > iface->lacp_rx) { - if (iface->lacp_status == LACP_STATUS_CURRENT) { + if (iface->lacp_status & LACP_CURRENT) { iface_set_lacp_expired(iface); - } else if (iface->lacp_status == LACP_STATUS_EXPIRED) { + } else if (iface->lacp_status & LACP_EXPIRED) { iface_set_lacp_defaulted(iface); } } @@ -3272,7 +3280,7 @@ lacp_wait(struct bridge *br) poll_timer_wait_until(iface->lacp_tx); } - if (iface->lacp_status != LACP_STATUS_DEFAULTED) { + if (iface->lacp_status & (LACP_CURRENT | LACP_EXPIRED)) { poll_timer_wait_until(iface->lacp_rx); } } @@ -3417,7 +3425,7 @@ 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 " @@ -3571,7 +3579,6 @@ 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. */ @@ -3778,10 +3785,10 @@ bond_unixctl_show(struct unixctl_conn *conn, bond_mode_to_string(port->bond_mode)); if (port->lacp) { - ds_put_format(&ds, "\tlacp: %s\n", + ds_put_format(&ds, "lacp: %s\n", port->lacp & LACP_ACTIVE ? "active" : "passive"); } else { - ds_put_cstr(&ds, "\tlacp: off\n"); + ds_put_cstr(&ds, "lacp: off\n"); } if (port->bond_mode != BM_AB) { @@ -3812,7 +3819,7 @@ bond_unixctl_show(struct unixctl_conn *conn, struct flow flow; /* Basic info. */ - ds_put_format(&ds, "slave %s: %s\n", + ds_put_format(&ds, "\nslave %s: %s\n", iface->name, iface->enabled ? "enabled" : "disabled"); if (j == port->active_iface) { ds_put_cstr(&ds, "\tactive slave\n"); @@ -3826,15 +3833,19 @@ bond_unixctl_show(struct unixctl_conn *conn, if (port->lacp) { ds_put_cstr(&ds, "\tstatus: "); - if (iface->lacp_status == LACP_STATUS_CURRENT) { + if (iface->lacp_status & LACP_CURRENT) { ds_put_cstr(&ds, "current "); - } else if (iface->lacp_status == LACP_STATUS_EXPIRED) { + } + + if (iface->lacp_status & LACP_EXPIRED) { ds_put_cstr(&ds, "expired "); - } else { + } + + if (iface->lacp_status & LACP_DEFAULTED) { ds_put_cstr(&ds, "defaulted "); } - if (iface->lacp_attached) { + if (iface->lacp_status & LACP_ATTACHED) { ds_put_cstr(&ds, "attached "); } @@ -3880,7 +3891,7 @@ bond_unixctl_show(struct unixctl_conn *conn, ds_put_cstr(&ds, "\tpartner state: "); ds_put_lacp_state(&ds, iface->lacp_partner.state); - ds_put_cstr(&ds, "\n\n"); + ds_put_cstr(&ds, "\n"); } if (port->bond_mode == BM_AB) { @@ -3980,7 +3991,6 @@ bond_unixctl_migrate(struct unixctl_conn *conn, const char *args_, ofproto_revalidate(port->bridge->ofproto, entry->iface_tag); entry->iface_idx = iface->port_ifidx; entry->iface_tag = tag_create_random(); - port->bond_compat_is_stale = true; unixctl_command_reply(conn, 200, "migrated"); } @@ -4412,9 +4422,6 @@ port_destroy(struct port *port) struct port *del; int i; - proc_net_compat_update_vlan(port->name, NULL, 0); - proc_net_compat_update_bond(port->name, NULL); - for (i = 0; i < MAX_MIRRORS; i++) { struct mirror *m = br->mirrors[i]; if (m && m->out_port == port) { @@ -4517,12 +4524,8 @@ port_update_bonding(struct port *port) } if (port->n_ifaces < 2) { /* Not a bonded port. */ - if (port->bond_hash) { - free(port->bond_hash); - port->bond_hash = NULL; - port->bond_compat_is_stale = true; - } - + free(port->bond_hash); + port->bond_hash = NULL; port->bond_fake_iface = false; } else { size_t i; @@ -4546,7 +4549,6 @@ port_update_bonding(struct port *port) 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) { @@ -4557,116 +4559,6 @@ port_update_bonding(struct port *port) } } } - -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; - } - } - } - - bond.n_slaves = port->n_ifaces; - bond.slaves = xmalloc(port->n_ifaces * sizeof *bond.slaves); - for (i = 0; i < port->n_ifaces; i++) { - struct iface *iface = port->ifaces[i]; - struct compat_bond_slave *slave = &bond.slaves[i]; - slave->name = iface->name; - - /* We need to make the same determination as the Linux bonding - * code to determine whether a slave should be consider "up". - * The Linux function bond_miimon_inspect() supports four - * BOND_LINK_* states: - * - * - BOND_LINK_UP: carrier detected, updelay has passed. - * - BOND_LINK_FAIL: carrier lost, downdelay in progress. - * - BOND_LINK_DOWN: carrier lost, downdelay has passed. - * - BOND_LINK_BACK: carrier detected, updelay in progress. - * - * The function bond_info_show_slave() only considers BOND_LINK_UP - * to be "up" and anything else to be "down". - */ - slave->up = iface->enabled && iface->delay_expires == LLONG_MAX; - if (slave->up) { - bond.up = true; - } - netdev_get_etheraddr(iface->netdev, slave->mac); - } - - if (port->bond_fake_iface) { - struct netdev *bond_netdev; - - if (!netdev_open_default(port->name, &bond_netdev)) { - if (bond.up) { - netdev_turn_flags_on(bond_netdev, NETDEV_UP, true); - } else { - netdev_turn_flags_off(bond_netdev, NETDEV_UP, true); - } - netdev_close(bond_netdev); - } - } - - 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. */ @@ -4675,7 +4567,8 @@ iface_set_lacp_defaulted(struct iface *iface) { memset(&iface->lacp_partner, 0, sizeof iface->lacp_partner); - iface->lacp_status = LACP_STATUS_DEFAULTED; + iface->lacp_status |= LACP_DEFAULTED; + iface->lacp_status &= ~(LACP_CURRENT | LACP_EXPIRED); iface->lacp_tx = 0; iface->port->lacp_need_update = true; } @@ -4683,7 +4576,8 @@ iface_set_lacp_defaulted(struct iface *iface) static void iface_set_lacp_expired(struct iface *iface) { - iface->lacp_status = LACP_STATUS_EXPIRED; + iface->lacp_status &= ~LACP_CURRENT; + iface->lacp_status |= LACP_EXPIRED; iface->lacp_partner.state |= LACP_STATE_TIME; iface->lacp_partner.state &= ~LACP_STATE_SYNC; @@ -4700,13 +4594,15 @@ iface_get_lacp_state(const struct iface *iface) state |= LACP_STATE_ACT; } - if (iface->lacp_status == LACP_STATUS_DEFAULTED) { - state |= LACP_STATE_DEF; - } else if (iface->lacp_attached) { + if (iface->lacp_status & LACP_ATTACHED) { state |= LACP_STATE_SYNC; } - if (iface->lacp_status == LACP_STATUS_EXPIRED) { + if (iface->lacp_status & LACP_DEFAULTED) { + state |= LACP_STATE_DEF; + } + + if (iface->lacp_status & LACP_EXPIRED) { state |= LACP_STATE_EXP; }