X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=ofproto%2Fofproto-dpif.c;h=be7f807d5af7d013223ff87c6a55612082a4c342;hb=04f01c24b522f893b663a34126beb6775176e536;hp=80e97e04c7439e8a9e3facdf3c10bdbad5767c68;hpb=0a37839c03307cc4cc726d27a7348c05d469c5eb;p=sliver-openvswitch.git diff --git a/ofproto/ofproto-dpif.c b/ofproto/ofproto-dpif.c index 80e97e04c..be7f807d5 100644 --- a/ofproto/ofproto-dpif.c +++ b/ofproto/ofproto-dpif.c @@ -52,6 +52,7 @@ #include "ofproto-dpif-governor.h" #include "ofproto-dpif-ipfix.h" #include "ofproto-dpif-mirror.h" +#include "ofproto-dpif-monitor.h" #include "ofproto-dpif-sflow.h" #include "ofproto-dpif-upcall.h" #include "ofproto-dpif-xlate.h" @@ -361,8 +362,6 @@ ofport_dpif_cast(const struct ofport *ofport) } static void port_run(struct ofport_dpif *); -static void port_run_fast(struct ofport_dpif *); -static void port_wait(struct ofport_dpif *); static int set_bfd(struct ofport *, const struct smap *); static int set_cfm(struct ofport *, const struct cfm_settings *); static void ofport_update_peer(struct ofport_dpif *); @@ -470,8 +469,9 @@ struct ofproto_dpif { struct classifier facets; /* Contains 'struct facet's. */ long long int consistency_rl; - struct netdev_stats stats; /* To account packets generated and consumed in - * userspace. */ + struct ovs_mutex stats_mutex; + struct netdev_stats stats OVS_GUARDED; /* To account packets generated and + * consumed in userspace. */ /* Spanning tree. */ struct stp *stp; @@ -526,9 +526,6 @@ static int expire(struct dpif_backer *); /* NetFlow. */ static void send_netflow_active_timeouts(struct ofproto_dpif *); -/* Utilities. */ -static int send_packet(const struct ofport_dpif *, struct ofpbuf *packet); - /* Global variables. */ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); @@ -548,11 +545,11 @@ ofproto_dpif_flow_mod(struct ofproto_dpif *ofproto, * Takes ownership of 'pin' and pin->packet. */ void ofproto_dpif_send_packet_in(struct ofproto_dpif *ofproto, - struct ofputil_packet_in *pin) + struct ofproto_packet_in *pin) { if (!guarded_list_push_back(&ofproto->pins, &pin->list_node, 1024)) { COVERAGE_INC(packet_in_overflow); - free(CONST_CAST(void *, pin->packet)); + free(CONST_CAST(void *, pin->up.packet)); free(pin); } } @@ -1233,6 +1230,7 @@ construct(struct ofproto *ofproto_) ofproto->ml = mac_learning_create(MAC_ENTRY_DEFAULT_IDLE_TIME); ofproto->mbridge = mbridge_create(); ofproto->has_bonded_bundles = false; + ovs_mutex_init(&ofproto->stats_mutex); ovs_mutex_init(&ofproto->vsp_mutex); classifier_init(&ofproto->facets); @@ -1361,7 +1359,7 @@ destruct(struct ofproto *ofproto_) { struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_); struct rule_dpif *rule, *next_rule; - struct ofputil_packet_in *pin, *next_pin; + struct ofproto_packet_in *pin, *next_pin; struct facet *facet, *next_facet; struct cls_cursor cursor; struct oftable *table; @@ -1399,7 +1397,7 @@ destruct(struct ofproto *ofproto_) guarded_list_pop_all(&ofproto->pins, &pins); LIST_FOR_EACH_SAFE (pin, next_pin, list_node, &pins) { list_remove(&pin->list_node); - free(CONST_CAST(void *, pin->packet)); + free(CONST_CAST(void *, pin->up.packet)); free(pin); } guarded_list_destroy(&ofproto->pins); @@ -1420,6 +1418,7 @@ destruct(struct ofproto *ofproto_) sset_destroy(&ofproto->ghost_ports); sset_destroy(&ofproto->port_poll_set); + ovs_mutex_destroy(&ofproto->stats_mutex); ovs_mutex_destroy(&ofproto->vsp_mutex); close_dpif_backer(ofproto->backer); @@ -1429,8 +1428,7 @@ static int run_fast(struct ofproto *ofproto_) { struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_); - struct ofputil_packet_in *pin, *next_pin; - struct ofport_dpif *ofport; + struct ofproto_packet_in *pin, *next_pin; struct list pins; /* Do not perform any periodic activity required by 'ofproto' while @@ -1443,14 +1441,10 @@ run_fast(struct ofproto *ofproto_) LIST_FOR_EACH_SAFE (pin, next_pin, list_node, &pins) { connmgr_send_packet_in(ofproto->up.connmgr, pin); list_remove(&pin->list_node); - free(CONST_CAST(void *, pin->packet)); + free(CONST_CAST(void *, pin->up.packet)); free(pin); } - HMAP_FOR_EACH (ofport, up.hmap_node, &ofproto->up.ports) { - port_run_fast(ofport); - } - return 0; } @@ -1536,7 +1530,6 @@ static void wait(struct ofproto *ofproto_) { struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_); - struct ofport_dpif *ofport; struct ofbundle *bundle; if (ofproto_get_flow_restore_wait()) { @@ -1549,9 +1542,6 @@ wait(struct ofproto *ofproto_) if (ofproto->ipfix) { dpif_ipfix_wait(ofproto->ipfix); } - HMAP_FOR_EACH (ofport, up.hmap_node, &ofproto->up.ports) { - port_wait(ofport); - } HMAP_FOR_EACH (bundle, hmap_node, &ofproto->bundles) { bundle_wait(bundle); } @@ -1814,6 +1804,9 @@ port_modified(struct ofport *port_) bfd_set_netdev(port->bfd, port->up.netdev); } + ofproto_dpif_monitor_port_update(port, port->bfd, port->cfm, + port->up.pp.hw_addr); + if (port->is_tunnel && tnl_port_reconfigure(port, port->up.netdev, port->odp_port)) { ofproto_dpif_cast(port->up.ofproto)->backer->need_revalidate = @@ -1904,11 +1897,9 @@ static int set_cfm(struct ofport *ofport_, const struct cfm_settings *s) { struct ofport_dpif *ofport = ofport_dpif_cast(ofport_); - int error; + int error = 0; - if (!s) { - error = 0; - } else { + if (s) { if (!ofport->cfm) { struct ofproto_dpif *ofproto; @@ -1918,13 +1909,17 @@ set_cfm(struct ofport *ofport_, const struct cfm_settings *s) } if (cfm_configure(ofport->cfm, s)) { - return 0; + error = 0; + goto out; } error = EINVAL; } cfm_unref(ofport->cfm); ofport->cfm = NULL; +out: + ofproto_dpif_monitor_port_update(ofport, ofport->bfd, ofport->cfm, + ofport->up.pp.hw_addr); return error; } @@ -1958,7 +1953,8 @@ set_bfd(struct ofport *ofport_, const struct smap *cfg) if (ofport->bfd != old) { ofproto->backer->need_revalidate = REV_RECONFIGURE; } - + ofproto_dpif_monitor_port_update(ofport, ofport->bfd, ofport->cfm, + ofport->up.pp.hw_addr); return 0; } @@ -1996,7 +1992,7 @@ send_bpdu_cb(struct ofpbuf *pkt, int port_num, void *ofproto_) VLOG_WARN_RL(&rl, "%s: cannot send BPDU on port %d " "with unknown MAC", ofproto->up.name, port_num); } else { - send_packet(ofport, pkt); + ofproto_dpif_send_packet(ofport, pkt); } } ofpbuf_delete(pkt); @@ -2592,7 +2588,7 @@ send_pdu_cb(void *port_, const void *pdu, size_t pdu_size) pdu_size); memcpy(packet_pdu, pdu, pdu_size); - send_packet(port, &packet); + ofproto_dpif_send_packet(port, &packet); ofpbuf_uninit(&packet); } else { VLOG_ERR_RL(&rl, "port %s: cannot obtain Ethernet address of iface " @@ -2629,7 +2625,7 @@ bundle_send_learning_packets(struct ofbundle *bundle) LIST_FOR_EACH (learning_packet, list_node, &packets) { int ret; - ret = send_packet(learning_packet->private_p, learning_packet); + ret = ofproto_dpif_send_packet(learning_packet->private_p, learning_packet); if (ret) { error = ret; n_errors++; @@ -2843,28 +2839,6 @@ ofport_update_peer(struct ofport_dpif *ofport) free(peer_name); } -static void -port_run_fast(struct ofport_dpif *ofport) -{ - if (ofport->cfm && cfm_should_send_ccm(ofport->cfm)) { - struct ofpbuf packet; - - ofpbuf_init(&packet, 0); - cfm_compose_ccm(ofport->cfm, &packet, ofport->up.pp.hw_addr); - send_packet(ofport, &packet); - ofpbuf_uninit(&packet); - } - - if (ofport->bfd && bfd_should_send_packet(ofport->bfd)) { - struct ofpbuf packet; - - ofpbuf_init(&packet, 0); - bfd_put_packet(ofport->bfd, &packet, ofport->up.pp.hw_addr); - send_packet(ofport, &packet); - ofpbuf_uninit(&packet); - } -} - static void port_run(struct ofport_dpif *ofport) { @@ -2876,12 +2850,9 @@ port_run(struct ofport_dpif *ofport) ofport->carrier_seq = carrier_seq; - port_run_fast(ofport); - if (ofport->cfm) { int cfm_opup = cfm_get_opup(ofport->cfm); - cfm_run(ofport->cfm); cfm_enable = !cfm_get_fault(ofport->cfm); if (cfm_opup >= 0) { @@ -2890,7 +2861,6 @@ port_run(struct ofport_dpif *ofport) } if (ofport->bfd) { - bfd_run(ofport->bfd); bfd_enable = bfd_forwarding(ofport->bfd); } @@ -2913,18 +2883,6 @@ port_run(struct ofport_dpif *ofport) ofport->may_enable = enable; } -static void -port_wait(struct ofport_dpif *ofport) -{ - if (ofport->cfm) { - cfm_wait(ofport->cfm); - } - - if (ofport->bfd) { - bfd_wait(ofport->bfd); - } -} - static int port_query_by_name(const struct ofproto *ofproto_, const char *devname, struct ofproto_port *ofproto_port) @@ -3038,10 +2996,11 @@ port_get_stats(const struct ofport *ofport_, struct netdev_stats *stats) if (!error && ofport_->ofp_port == OFPP_LOCAL) { struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto); + ovs_mutex_lock(&ofproto->stats_mutex); /* ofproto->stats.tx_packets represents packets that we created * internally and sent to some port (e.g. packets sent with - * send_packet()). Account for them as if they had come from - * OFPP_LOCAL and got forwarded. */ + * ofproto_dpif_send_packet()). Account for them as if they had + * come from OFPP_LOCAL and got forwarded. */ if (stats->rx_packets != UINT64_MAX) { stats->rx_packets += ofproto->stats.tx_packets; @@ -3062,6 +3021,7 @@ port_get_stats(const struct ofport *ofport_, struct netdev_stats *stats) if (stats->tx_bytes != UINT64_MAX) { stats->tx_bytes += ofproto->stats.rx_bytes; } + ovs_mutex_unlock(&ofproto->stats_mutex); } return error; @@ -3836,24 +3796,49 @@ facet_free(struct facet *facet) } } -/* Executes, within 'ofproto', the 'n_actions' actions in 'actions' on - * 'packet', which arrived on 'in_port'. */ -static bool -execute_odp_actions(struct ofproto_dpif *ofproto, const struct flow *flow, - const struct nlattr *odp_actions, size_t actions_len, - struct ofpbuf *packet) +/* Executes, within 'ofproto', the actions in 'rule' or 'ofpacts' on 'packet'. + * 'flow' must reflect the data in 'packet'. */ +int +ofproto_dpif_execute_actions(struct ofproto_dpif *ofproto, + const struct flow *flow, + struct rule_dpif *rule, + const struct ofpact *ofpacts, size_t ofpacts_len, + struct ofpbuf *packet) { struct odputil_keybuf keybuf; + struct dpif_flow_stats stats; + struct xlate_out xout; + struct xlate_in xin; + ofp_port_t in_port; struct ofpbuf key; int error; + ovs_assert((rule != NULL) != (ofpacts != NULL)); + + dpif_flow_stats_extract(flow, packet, time_msec(), &stats); + if (rule) { + rule_dpif_credit_stats(rule, &stats); + } + + xlate_in_init(&xin, ofproto, flow, rule, stats.tcp_flags, packet); + xin.ofpacts = ofpacts; + xin.ofpacts_len = ofpacts_len; + xin.resubmit_stats = &stats; + xlate_actions(&xin, &xout); + ofpbuf_use_stack(&key, &keybuf, sizeof keybuf); - odp_flow_key_from_flow(&key, flow, - ofp_port_to_odp_port(ofproto, flow->in_port.ofp_port)); + in_port = flow->in_port.ofp_port; + if (in_port == OFPP_NONE) { + in_port = OFPP_LOCAL; + } + odp_flow_key_from_flow(&key, flow, ofp_port_to_odp_port(ofproto, in_port)); error = dpif_execute(ofproto->backer->dpif, key.data, key.size, - odp_actions, actions_len, packet); - return !error; + xout.odp_actions.data, xout.odp_actions.size, packet, + (xout.slow & SLOW_ACTION) != 0); + xlate_out_uninit(&xout); + + return error; } /* Remove 'facet' from its ofproto and free up the associated memory: @@ -4076,15 +4061,11 @@ facet_check_consistency(struct facet *facet) struct xlate_out xout; struct xlate_in xin; - - struct rule_dpif *rule; bool ok; /* Check the datapath actions for consistency. */ - rule_dpif_lookup(facet->ofproto, &facet->flow, NULL, &rule); - xlate_in_init(&xin, facet->ofproto, &facet->flow, rule, 0, NULL); + xlate_in_init(&xin, facet->ofproto, &facet->flow, NULL, 0, NULL); xlate_actions(&xin, &xout); - rule_dpif_unref(rule); ok = ofpbuf_equal(&facet->xout.odp_actions, &xout.odp_actions) && facet->xout.slow == xout.slow; @@ -4236,7 +4217,6 @@ flow_push_stats(struct ofproto_dpif *ofproto, struct flow *flow, struct dpif_flow_stats *stats, bool may_learn) { struct ofport_dpif *in_port; - struct rule_dpif *rule; struct xlate_in xin; in_port = get_ofp_port(ofproto, flow->in_port.ofp_port); @@ -4244,13 +4224,10 @@ flow_push_stats(struct ofproto_dpif *ofproto, struct flow *flow, netdev_vport_inc_rx(in_port->up.netdev, stats); } - rule_dpif_lookup(ofproto, flow, NULL, &rule); - rule_dpif_credit_stats(rule, stats); - xlate_in_init(&xin, ofproto, flow, rule, stats->tcp_flags, NULL); + xlate_in_init(&xin, ofproto, flow, NULL, stats->tcp_flags, NULL); xin.resubmit_stats = stats; xin.may_learn = may_learn; xlate_actions_for_side_effects(&xin); - rule_dpif_unref(rule); } static void @@ -4327,9 +4304,15 @@ rule_dpif_credit_stats(struct rule_dpif *rule, } bool -rule_dpif_fail_open(const struct rule_dpif *rule) +rule_dpif_is_fail_open(const struct rule_dpif *rule) { - return rule->up.cr.priority == FAIL_OPEN_PRIORITY; + return is_fail_open_rule(&rule->up); +} + +bool +rule_dpif_is_table_miss(const struct rule_dpif *rule) +{ + return rule_is_table_miss(&rule->up); } ovs_be64 @@ -4651,9 +4634,7 @@ rule_dpif_lookup_in_table(struct ofproto_dpif *ofproto, cls_rule = classifier_lookup(cls, &ofpc_normal_flow, wc); } else if (frag && ofproto->up.frag_handling == OFPC_FRAG_DROP) { cls_rule = &ofproto->drop_frags_rule->up.cr; - if (wc) { - flow_wildcards_init_exact(wc); - } + /* Frag mask in wc already set above. */ } else { cls_rule = classifier_lookup(cls, flow, wc); } @@ -4780,21 +4761,8 @@ rule_dpif_execute(struct rule_dpif *rule, const struct flow *flow, struct ofpbuf *packet) { struct ofproto_dpif *ofproto = ofproto_dpif_cast(rule->up.ofproto); - struct dpif_flow_stats stats; - struct xlate_out xout; - struct xlate_in xin; - - dpif_flow_stats_extract(flow, packet, time_msec(), &stats); - rule_dpif_credit_stats(rule, &stats); - - xlate_in_init(&xin, ofproto, flow, rule, stats.tcp_flags, packet); - xin.resubmit_stats = &stats; - xlate_actions(&xin, &xout); - - execute_odp_actions(ofproto, flow, xout.odp_actions.data, - xout.odp_actions.size, packet); - xlate_out_uninit(&xout); + ofproto_dpif_execute_actions(ofproto, flow, rule, NULL, 0, packet); } static enum ofperr @@ -4825,55 +4793,18 @@ rule_modify_actions(struct rule *rule_, bool reset_counters) /* Sends 'packet' out 'ofport'. * May modify 'packet'. * Returns 0 if successful, otherwise a positive errno value. */ -static int -send_packet(const struct ofport_dpif *ofport, struct ofpbuf *packet) +int +ofproto_dpif_send_packet(const struct ofport_dpif *ofport, struct ofpbuf *packet) { struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto); - uint64_t odp_actions_stub[1024 / 8]; - struct ofpbuf key, odp_actions; - struct dpif_flow_stats stats; - struct odputil_keybuf keybuf; - struct ofpact_output output; - struct xlate_out xout; - struct xlate_in xin; - struct flow flow; - union flow_in_port in_port_; int error; - ofpbuf_use_stub(&odp_actions, odp_actions_stub, sizeof odp_actions_stub); - ofpbuf_use_stack(&key, &keybuf, sizeof keybuf); - - /* Use OFPP_NONE as the in_port to avoid special packet processing. */ - in_port_.ofp_port = OFPP_NONE; - flow_extract(packet, 0, 0, NULL, &in_port_, &flow); - odp_flow_key_from_flow(&key, &flow, ofp_port_to_odp_port(ofproto, - OFPP_LOCAL)); - dpif_flow_stats_extract(&flow, packet, time_msec(), &stats); - - ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); - output.port = ofport->up.ofp_port; - output.max_len = 0; - - xlate_in_init(&xin, ofproto, &flow, NULL, 0, packet); - xin.ofpacts_len = sizeof output; - xin.ofpacts = &output.ofpact; - xin.resubmit_stats = &stats; - xlate_actions(&xin, &xout); - - error = dpif_execute(ofproto->backer->dpif, - key.data, key.size, - xout.odp_actions.data, xout.odp_actions.size, - packet); - xlate_out_uninit(&xout); - - if (error) { - VLOG_WARN_RL(&rl, "%s: failed to send packet on port %s (%s)", - ofproto->up.name, netdev_get_name(ofport->up.netdev), - ovs_strerror(error)); - } + error = xlate_send_packet(ofport, packet); + ovs_mutex_lock(&ofproto->stats_mutex); ofproto->stats.tx_packets++; ofproto->stats.tx_bytes += packet->size; + ovs_mutex_unlock(&ofproto->stats_mutex); return error; } @@ -4936,30 +4867,9 @@ packet_out(struct ofproto *ofproto_, struct ofpbuf *packet, const struct ofpact *ofpacts, size_t ofpacts_len) { struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_); - struct odputil_keybuf keybuf; - struct dpif_flow_stats stats; - struct xlate_out xout; - struct xlate_in xin; - struct ofpbuf key; - - - ofpbuf_use_stack(&key, &keybuf, sizeof keybuf); - odp_flow_key_from_flow(&key, flow, - ofp_port_to_odp_port(ofproto, - flow->in_port.ofp_port)); - - dpif_flow_stats_extract(flow, packet, time_msec(), &stats); - - xlate_in_init(&xin, ofproto, flow, NULL, stats.tcp_flags, packet); - xin.resubmit_stats = &stats; - xin.ofpacts_len = ofpacts_len; - xin.ofpacts = ofpacts; - - xlate_actions(&xin, &xout); - dpif_execute(ofproto->backer->dpif, key.data, key.size, - xout.odp_actions.data, xout.odp_actions.size, packet); - xlate_out_uninit(&xout); + ofproto_dpif_execute_actions(ofproto, flow, NULL, ofpacts, + ofpacts_len, packet); return 0; } @@ -5144,7 +5054,7 @@ trace_format_rule(struct ds *result, int level, const struct rule_dpif *rule) actions = rule_dpif_get_actions(rule); ds_put_char_multiple(result, '\t', level); - ds_put_cstr(result, "OpenFlow "); + ds_put_cstr(result, "OpenFlow actions="); ofpacts_format(actions->ofpacts, actions->ofpacts_len, result); ds_put_char(result, '\n'); @@ -5295,7 +5205,7 @@ ofproto_unixctl_trace(struct unixctl_conn *conn, int argc, const char *argv[], goto exit; } ds_put_format(&result, "Bridge: %s\n", ofproto->up.name); - } else if (!parse_ofp_exact_flow(&flow, argv[argc - 1])) { + } else if (!parse_ofp_exact_flow(&flow, NULL, argv[argc - 1], NULL)) { if (argc != 3) { unixctl_command_reply_error(conn, "Must specify bridge name"); goto exit; @@ -5399,27 +5309,19 @@ ofproto_trace(struct ofproto_dpif *ofproto, const struct flow *flow, trace.xout.odp_actions.size); if (trace.xout.slow) { + enum slow_path_reason slow; + ds_put_cstr(ds, "\nThis flow is handled by the userspace " "slow path because it:"); - switch (trace.xout.slow) { - case SLOW_CFM: - ds_put_cstr(ds, "\n\t- Consists of CFM packets."); - break; - case SLOW_LACP: - ds_put_cstr(ds, "\n\t- Consists of LACP packets."); - break; - case SLOW_STP: - ds_put_cstr(ds, "\n\t- Consists of STP packets."); - break; - case SLOW_BFD: - ds_put_cstr(ds, "\n\t- Consists of BFD packets."); - break; - case SLOW_CONTROLLER: - ds_put_cstr(ds, "\n\t- Sends \"packet-in\" messages " - "to the OpenFlow controller."); - break; - case __SLOW_MAX: - NOT_REACHED(); + + slow = trace.xout.slow; + while (slow) { + enum slow_path_reason bit = rightmost_1bit(slow); + + ds_put_format(ds, "\n\t- %s.", + slow_path_reason_to_explanation(bit)); + + slow &= ~bit; } }