X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;ds=sidebyside;f=vswitchd%2Fbridge.c;h=c175c58f2b73ce30f343918314bdc86b29ee8bd5;hb=380cbf380247322ae7d18117fc9dd30137cc460c;hp=32885ebcfec538c73a399a6e14d5ed323028c3db;hpb=52a90c29ab472076fb8f20fba4f847350268e01e;p=openvswitch diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index 32885ebc..c175c58f 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -1,4 +1,4 @@ -/* Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks +/* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira Networks * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -28,9 +28,12 @@ #include "dynamic-string.h" #include "hash.h" #include "hmap.h" +#include "hmapx.h" #include "jsonrpc.h" #include "lacp.h" #include "list.h" +#include "mac-learning.h" +#include "meta-flow.h" #include "netdev.h" #include "ofp-print.h" #include "ofpbuf.h" @@ -39,6 +42,7 @@ #include "sha1.h" #include "shash.h" #include "socket-util.h" +#include "stream.h" #include "stream-ssl.h" #include "sset.h" #include "system-stats.h" @@ -78,6 +82,7 @@ struct mirror { struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */ struct bridge *bridge; char *name; + const struct ovsrec_mirror *cfg; }; struct port { @@ -153,8 +158,10 @@ static void bridge_configure_datapath_id(struct bridge *); static void bridge_configure_flow_eviction_threshold(struct bridge *); static void bridge_configure_netflow(struct bridge *); static void bridge_configure_forward_bpdu(struct bridge *); +static void bridge_configure_mac_idle_time(struct bridge *); static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number); static void bridge_configure_stp(struct bridge *); +static void bridge_configure_tables(struct bridge *); static void bridge_configure_remotes(struct bridge *, const struct sockaddr_in *managers, size_t n_managers); @@ -192,7 +199,8 @@ static void bridge_configure_mirrors(struct bridge *); static struct mirror *mirror_create(struct bridge *, const struct ovsrec_mirror *); static void mirror_destroy(struct mirror *); -static bool mirror_configure(struct mirror *, const struct ovsrec_mirror *); +static bool mirror_configure(struct mirror *); +static void mirror_refresh_stats(struct mirror *); static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *); static struct iface *iface_create(struct port *port, @@ -291,6 +299,7 @@ bridge_init(const char *remote) ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids); ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids); + ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics); ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids); @@ -305,14 +314,16 @@ bridge_init(const char *remote) ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids); /* Register unixctl commands. */ - unixctl_command_register("qos/show", "interface", qos_unixctl_show, NULL); - unixctl_command_register("bridge/dump-flows", "bridge", + unixctl_command_register("qos/show", "interface", 1, 1, + qos_unixctl_show, NULL); + unixctl_command_register("bridge/dump-flows", "bridge", 1, 1, bridge_unixctl_dump_flows, NULL); - unixctl_command_register("bridge/reconnect", "[bridge]", + unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1, bridge_unixctl_reconnect, NULL); lacp_init(); bond_init(); cfm_init(); + stp_init(); } void @@ -368,10 +379,10 @@ collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg, SSET_FOR_EACH (target, &targets) { struct sockaddr_in *sin = &managers[n_managers]; - if ((!strncmp(target, "tcp:", 4) - && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) || - (!strncmp(target, "ssl:", 4) - && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) { + if (stream_parse_target_with_default_ports(target, + JSONRPC_TCP_PORT, + JSONRPC_SSL_PORT, + sin)) { n_managers++; } } @@ -441,6 +452,10 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) HMAP_FOR_EACH (br, node, &all_bridges) { struct port *port; + /* We need the datapath ID early to allow LACP ports to use it as the + * default system ID. */ + bridge_configure_datapath_id(br); + HMAP_FOR_EACH (port, hmap_node, &br->ports) { struct iface *iface; @@ -453,13 +468,14 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) } } bridge_configure_mirrors(br); - bridge_configure_datapath_id(br); bridge_configure_flow_eviction_threshold(br); bridge_configure_forward_bpdu(br); + bridge_configure_mac_idle_time(br); bridge_configure_remotes(br, managers, n_managers); bridge_configure_netflow(br); bridge_configure_sflow(br, &sflow_bridge_number); bridge_configure_stp(br); + bridge_configure_tables(br); } free(managers); @@ -532,17 +548,8 @@ port_configure(struct port *port) /* Get VLAN tag. */ s.vlan = -1; - if (cfg->tag) { - if (list_is_short(&port->ifaces)) { - if (*cfg->tag >= 0 && *cfg->tag <= 4095) { - s.vlan = *cfg->tag; - } - } else { - /* It's possible that bonded, VLAN-tagged ports make sense. Maybe - * they even work as-is. But they have not been tested. */ - VLOG_WARN("port %s: VLAN tags not supported on bonded ports", - port->name); - } + if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) { + s.vlan = *cfg->tag; } /* Get VLAN trunks. */ @@ -1271,14 +1278,30 @@ bridge_configure_forward_bpdu(struct bridge *br) ofproto_set_forward_bpdu(br->ofproto, forward_bpdu); } +/* Set MAC aging time for 'br'. */ +static void +bridge_configure_mac_idle_time(struct bridge *br) +{ + const char *idle_time_str; + int idle_time; + + idle_time_str = bridge_get_other_config(br->cfg, "mac-aging-time"); + idle_time = (idle_time_str && atoi(idle_time_str) + ? atoi(idle_time_str) + : MAC_ENTRY_DEFAULT_IDLE_TIME); + ofproto_set_mac_idle_time(br->ofproto, idle_time); +} + static void bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], struct iface **hw_addr_iface) { + struct hmapx mirror_output_ports; const char *hwaddr; struct port *port; bool found_addr = false; int error; + int i; *hw_addr_iface = NULL; @@ -1295,6 +1318,18 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], } } + /* Mirror output ports don't participate in picking the local hardware + * address. ofproto can't help us find out whether a given port is a + * mirror output because we haven't configured mirrors yet, so we need to + * accumulate them ourselves. */ + hmapx_init(&mirror_output_ports); + for (i = 0; i < br->cfg->n_mirrors; i++) { + struct ovsrec_mirror *m = br->cfg->mirrors[i]; + if (m->output_port) { + hmapx_add(&mirror_output_ports, m->output_port); + } + } + /* Otherwise choose the minimum non-local MAC address among all of the * interfaces. */ HMAP_FOR_EACH (port, hmap_node, &br->ports) { @@ -1303,7 +1338,7 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], struct iface *iface; /* Mirror output ports don't participate. */ - if (ofproto_is_mirror_output_bundle(br->ofproto, port)) { + if (hmapx_contains(&mirror_output_ports, port->cfg)) { continue; } @@ -1366,6 +1401,8 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], VLOG_WARN("bridge %s: using default bridge Ethernet " "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea)); } + + hmapx_destroy(&mirror_output_ports); } /* Choose and returns the datapath ID for bridge 'br' given that the bridge @@ -1779,13 +1816,28 @@ refresh_cfm_stats(void) } } +/* Performs periodic activity required by bridges that needs to be done with + * the least possible latency. + * + * It makes sense to call this function a couple of times per poll loop, to + * provide a significant performance boost on some benchmarks with ofprotos + * that use the ofproto-dpif implementation. */ +void +bridge_run_fast(void) +{ + struct bridge *br; + + HMAP_FOR_EACH (br, node, &all_bridges) { + ofproto_run_fast(br->ofproto); + } +} + void bridge_run(void) { const struct ovsrec_open_vswitch *cfg; bool vlan_splinters_changed; - bool datapath_destroyed; bool database_changed; struct bridge *br; @@ -1808,15 +1860,8 @@ bridge_run(void) cfg = ovsrec_open_vswitch_first(idl); /* Let each bridge do the work that it needs to do. */ - datapath_destroyed = false; HMAP_FOR_EACH (br, node, &all_bridges) { - int error = ofproto_run(br->ofproto); - if (error) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); - VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, " - "forcing reconfiguration", br->name); - datapath_destroyed = true; - } + ofproto_run(br->ofproto); } /* Re-configure SSL. We do this on every trip through the main loop, @@ -1844,7 +1889,7 @@ bridge_run(void) } } - if (database_changed || datapath_destroyed || vlan_splinters_changed) { + if (database_changed || vlan_splinters_changed) { if (cfg) { struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl); @@ -1870,6 +1915,7 @@ bridge_run(void) txn = ovsdb_idl_txn_create(idl); HMAP_FOR_EACH (br, node, &all_bridges) { struct port *port; + struct mirror *m; HMAP_FOR_EACH (port, hmap_node, &br->ports) { struct iface *iface; @@ -1879,6 +1925,11 @@ bridge_run(void) iface_refresh_status(iface); } } + + HMAP_FOR_EACH (m, hmap_node, &br->mirrors) { + mirror_refresh_stats(m); + } + } refresh_system_stats(cfg); refresh_controller_status(); @@ -2006,8 +2057,8 @@ qos_unixctl_show_cb(unsigned int queue_id, } static void -qos_unixctl_show(struct unixctl_conn *conn, - const char *args, void *aux OVS_UNUSED) +qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED, + const char *argv[], void *aux OVS_UNUSED) { struct ds ds = DS_EMPTY_INITIALIZER; struct shash sh = SHASH_INITIALIZER(&sh); @@ -2017,7 +2068,7 @@ qos_unixctl_show(struct unixctl_conn *conn, struct qos_unixctl_show_cbdata data; int error; - iface = iface_find(args); + iface = iface_find(argv[1]); if (!iface) { unixctl_command_reply(conn, 501, "no such interface"); return; @@ -2116,13 +2167,13 @@ bridge_lookup(const char *name) /* Handle requests for a listing of all flows known by the OpenFlow * stack, including those normally hidden. */ static void -bridge_unixctl_dump_flows(struct unixctl_conn *conn, - const char *args, void *aux OVS_UNUSED) +bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED, + const char *argv[], void *aux OVS_UNUSED) { struct bridge *br; struct ds results; - br = bridge_lookup(args); + br = bridge_lookup(argv[1]); if (!br) { unixctl_command_reply(conn, 501, "Unknown bridge"); return; @@ -2139,12 +2190,12 @@ bridge_unixctl_dump_flows(struct unixctl_conn *conn, * connections and reconnect. If BRIDGE is not specified, then all bridges * drop their controller connections and reconnect. */ static void -bridge_unixctl_reconnect(struct unixctl_conn *conn, - const char *args, void *aux OVS_UNUSED) +bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc, + const char *argv[], void *aux OVS_UNUSED) { struct bridge *br; - if (args[0] != '\0') { - br = bridge_lookup(args); + if (argc > 1) { + br = bridge_lookup(argv[1]); if (!br) { unixctl_command_reply(conn, 501, "Unknown bridge"); return; @@ -2442,6 +2493,66 @@ bridge_configure_remotes(struct bridge *br, sset_destroy(&snoops); } } + +static void +bridge_configure_tables(struct bridge *br) +{ + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + int n_tables; + int i, j; + + n_tables = ofproto_get_n_tables(br->ofproto); + j = 0; + for (i = 0; i < n_tables; i++) { + struct ofproto_table_settings s; + + s.name = NULL; + s.max_flows = UINT_MAX; + s.groups = NULL; + s.n_groups = 0; + + if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) { + struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++]; + + s.name = cfg->name; + if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) { + s.max_flows = *cfg->flow_limit; + } + if (cfg->overflow_policy + && !strcmp(cfg->overflow_policy, "evict")) { + size_t k; + + s.groups = xmalloc(cfg->n_groups * sizeof *s.groups); + for (k = 0; k < cfg->n_groups; k++) { + const char *string = cfg->groups[k]; + char *msg; + + msg = mf_parse_subfield__(&s.groups[k], &string); + if (msg) { + VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing " + "'groups' (%s)", br->name, i, msg); + free(msg); + } else if (*string) { + VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' " + "element '%s' contains trailing garbage", + br->name, i, cfg->groups[k]); + } else { + s.n_groups++; + } + } + } + } + + ofproto_configure_table(br->ofproto, i, &s); + + free(s.groups); + } + for (; j < br->cfg->n_flow_tables; j++) { + VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table " + "%"PRId64" not supported by this datapath", br->name, + br->cfg->key_flow_tables[j]); + } +} /* Port functions. */ @@ -2623,7 +2734,7 @@ enable_lacp(struct port *port, bool *activep) static struct lacp_settings * port_configure_lacp(struct port *port, struct lacp_settings *s) { - const char *lacp_time; + const char *lacp_time, *system_id; long long int custom_time; int priority; @@ -2632,7 +2743,23 @@ port_configure_lacp(struct port *port, struct lacp_settings *s) } s->name = port->name; - memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN); + + system_id = get_port_other_config(port->cfg, "lacp-system-id", NULL); + if (system_id) { + if (sscanf(system_id, ETH_ADDR_SCAN_FMT, + ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) { + VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet" + " address.", port->name, system_id); + return NULL; + } + } else { + memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN); + } + + if (eth_addr_is_zero(s->id)) { + VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name); + return NULL; + } /* Prefer bondable links if unspecified. */ priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority", @@ -2645,7 +2772,6 @@ port_configure_lacp(struct port *port, struct lacp_settings *s) "lacp-heartbeat", "false"), "true"); - lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow"); custom_time = atoi(lacp_time); if (!strcmp(lacp_time, "fast")) { @@ -2701,12 +2827,22 @@ port_configure_bond(struct port *port, struct bond_settings *s, size_t i; s->name = port->name; - s->balance = BM_SLB; - if (port->cfg->bond_mode - && !bond_mode_from_string(&s->balance, port->cfg->bond_mode)) { - VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s", - port->name, port->cfg->bond_mode, - bond_mode_to_string(s->balance)); + s->balance = BM_AB; + if (port->cfg->bond_mode) { + if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) { + VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s", + port->name, port->cfg->bond_mode, + bond_mode_to_string(s->balance)); + } + } else { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); + + /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to + * active-backup. At some point we should remove this warning. */ + VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that" + " in previous versions, the default bond_mode was" + " balance-slb", port->name, + bond_mode_to_string(s->balance)); } if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) { VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, " @@ -2734,7 +2870,7 @@ port_configure_bond(struct port *port, struct bond_settings *s, s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0")); s->rebalance_interval = atoi( get_port_other_config(port->cfg, "bond-rebalance-interval", "10000")); - if (s->rebalance_interval < 1000) { + if (s->rebalance_interval && s->rebalance_interval < 1000) { s->rebalance_interval = 1000; } @@ -3082,6 +3218,8 @@ iface_configure_cfm(struct iface *iface) "0")); s.ccm_vlan = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_vlan", "0")); + s.ccm_pcp = atoi(get_interface_other_config(iface->cfg, "cfm_ccm_pcp", + "0")); if (s.interval <= 0) { s.interval = 1000; } @@ -3145,7 +3283,8 @@ bridge_configure_mirrors(struct bridge *br) if (!m) { m = mirror_create(br, cfg); } - if (!mirror_configure(m, cfg)) { + m->cfg = cfg; + if (!mirror_configure(m)) { mirror_destroy(m); } } @@ -3211,8 +3350,9 @@ mirror_collect_ports(struct mirror *m, } static bool -mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg) +mirror_configure(struct mirror *m) { + const struct ovsrec_mirror *cfg = m->cfg; struct ofproto_mirror_settings s; /* Set name. */ @@ -3356,8 +3496,13 @@ collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg) struct bridge *br; size_t i; - splinter_vlans = NULL; + /* Free space allocated for synthesized ports and interfaces, since we're + * in the process of reconstructing all of them. */ + free_registered_blocks(); + + splinter_vlans = bitmap_allocate(4096); sset_init(&splinter_ifaces); + vlan_splinters_enabled_anywhere = false; for (i = 0; i < ovs_cfg->n_bridges; i++) { struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i]; size_t j; @@ -3370,21 +3515,22 @@ collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg) struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k]; if (vlan_splinters_is_enabled(iface_cfg)) { + vlan_splinters_enabled_anywhere = true; sset_add(&splinter_ifaces, iface_cfg->name); - - if (!splinter_vlans) { - splinter_vlans = bitmap_allocate(4096); - } vlan_bitmap_from_array__(port_cfg->trunks, port_cfg->n_trunks, splinter_vlans); } } + + if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) { + bitmap_set1(splinter_vlans, *port_cfg->tag); + } } } - vlan_splinters_enabled_anywhere = splinter_vlans != NULL; - if (!splinter_vlans) { + if (!vlan_splinters_enabled_anywhere) { + free(splinter_vlans); sset_destroy(&splinter_ifaces); return NULL; } @@ -3523,8 +3669,6 @@ add_vlan_splinter_ports(struct bridge *br, { size_t i; - free_registered_blocks(); - /* We iterate through 'br->cfg->ports' instead of 'ports' here because * we're modifying 'ports'. */ for (i = 0; i < br->cfg->n_ports; i++) { @@ -3555,3 +3699,31 @@ add_vlan_splinter_ports(struct bridge *br, } } } + +static void +mirror_refresh_stats(struct mirror *m) +{ + struct ofproto *ofproto = m->bridge->ofproto; + uint64_t tx_packets, tx_bytes; + char *keys[2]; + int64_t values[2]; + size_t stat_cnt = 0; + + if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) { + ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0); + return; + } + + if (tx_packets != UINT64_MAX) { + keys[stat_cnt] = "tx_packets"; + values[stat_cnt] = tx_packets; + stat_cnt++; + } + if (tx_bytes != UINT64_MAX) { + keys[stat_cnt] = "tx_bytes"; + values[stat_cnt] = tx_bytes; + stat_cnt++; + } + + ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt); +}