X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=6abb0f6cab0e67a767bcfd977af65396d7662501;hb=2658ae89b2231df0abc95f43afc2b40e1db11298;hp=ce3a78871e0c0d70d454d2a231e257bab66d8e36;hpb=a70e4b2a0a8e0dead971c37872fb3c454908f2b6;p=sliver-openvswitch.git diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index ce3a78871..6abb0f6ca 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" @@ -39,12 +42,14 @@ #include "sha1.h" #include "shash.h" #include "socket-util.h" +#include "stream.h" #include "stream-ssl.h" #include "sset.h" #include "system-stats.h" #include "timeval.h" #include "util.h" #include "unixctl.h" +#include "vlandev.h" #include "vswitchd/vswitch-idl.h" #include "xenserver.h" #include "vlog.h" @@ -70,6 +75,7 @@ struct iface { struct netdev *netdev; /* Network device. */ const char *type; /* Usually same as cfg->type. */ const struct ovsrec_interface *cfg; + bool need_refresh; /* Refresh iface after create. */ }; struct mirror { @@ -77,6 +83,7 @@ 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 { @@ -143,7 +150,8 @@ static unixctl_cb_func bridge_unixctl_dump_flows; static unixctl_cb_func bridge_unixctl_reconnect; static size_t bridge_get_controllers(const struct bridge *br, struct ovsrec_controller ***controllersp); -static void bridge_add_del_ports(struct bridge *); +static void bridge_add_del_ports(struct bridge *, + const unsigned long int *splinter_vlans); static void bridge_add_ofproto_ports(struct bridge *); static void bridge_del_ofproto_ports(struct bridge *); static void bridge_refresh_ofp_port(struct bridge *); @@ -151,8 +159,10 @@ 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); @@ -190,7 +200,8 @@ 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, @@ -209,11 +220,32 @@ static void iface_refresh_cfm_stats(struct iface *); static void iface_refresh_stats(struct iface *); static void iface_refresh_status(struct iface *); static bool iface_is_synthetic(const struct iface *); +static const char *get_interface_other_config(const struct ovsrec_interface *, + const char *key, + const char *default_value); 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); + +/* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.) + * + * This is deprecated. It is only for compatibility with broken device drivers + * in old versions of Linux that do not properly support VLANs when VLAN + * devices are not used. When broken device drivers are no longer in + * widespread use, we will delete these interfaces. */ + +/* True if VLAN splinters are enabled on any interface, false otherwise.*/ +static bool vlan_splinters_enabled_anywhere; + +static bool vlan_splinters_is_enabled(const struct ovsrec_interface *); +static unsigned long int *collect_splinter_vlans( + const struct ovsrec_open_vswitch *); +static void configure_splinter_port(struct port *); +static void add_vlan_splinter_ports(struct bridge *, + const unsigned long int *splinter_vlans, + struct shash *ports); /* Public functions. */ @@ -254,6 +286,7 @@ bridge_init(const char *remote) ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault); + ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current); ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids); @@ -268,6 +301,7 @@ 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); @@ -282,14 +316,16 @@ 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 @@ -345,10 +381,10 @@ 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++; } } @@ -362,6 +398,7 @@ collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg, static void bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) { + unsigned long int *splinter_vlans; struct sockaddr_in *managers; struct bridge *br, *next; int sflow_bridge_number; @@ -376,9 +413,11 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) * This is mostly an update to bridge data structures. Very little is * pushed down to ofproto or lower layers. */ add_del_bridges(ovs_cfg); + splinter_vlans = collect_splinter_vlans(ovs_cfg); HMAP_FOR_EACH (br, node, &all_bridges) { - bridge_add_del_ports(br); + bridge_add_del_ports(br, splinter_vlans); } + free(splinter_vlans); /* Delete all datapaths and datapath ports that are no longer configured. * @@ -400,8 +439,12 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) * has at least one iface, every "struct iface" has a valid ofp_port and * netdev. */ HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) { - if (!br->ofproto && !bridge_add_ofprotos(br)) { - bridge_destroy(br); + if (!br->ofproto) { + if (bridge_add_ofprotos(br)) { + bridge_del_ofproto_ports(br); + } else { + bridge_destroy(br); + } } } HMAP_FOR_EACH (br, node, &all_bridges) { @@ -415,6 +458,10 @@ 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; @@ -427,13 +474,14 @@ 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); @@ -489,6 +537,11 @@ port_configure(struct port *port) struct ofproto_bundle_settings s; struct iface *iface; + if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) { + configure_splinter_port(port); + return; + } + /* Get name. */ s.name = port->name; @@ -501,17 +554,8 @@ 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. */ @@ -1094,6 +1138,12 @@ bridge_add_ofproto_ports(struct bridge *br) VLOG_WARN("could not open network device %s (%s)", iface->name, strerror(error)); } + + if (iface->netdev + && port->cfg->vlan_mode + && !strcmp(port->cfg->vlan_mode, "splinter")) { + netdev_turn_flags_on(iface->netdev, NETDEV_UP, true); + } } else { error = 0; } @@ -1133,9 +1183,10 @@ bridge_add_ofproto_ports(struct bridge *br) } /* Populate stats columns in new Interface rows. */ - if (iface->netdev && !iface->cfg->mtu) { + if (iface->netdev && iface->need_refresh) { iface_refresh_stats(iface); iface_refresh_status(iface); + iface->need_refresh = false; } /* Delete the iface if we failed. */ @@ -1234,14 +1285,30 @@ 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; @@ -1258,6 +1325,18 @@ 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) { @@ -1266,7 +1345,7 @@ 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; } @@ -1329,6 +1408,8 @@ 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 @@ -1361,33 +1442,10 @@ bridge_pick_datapath_id(struct bridge *br, return dpid; } - if (hw_addr_iface) { - int vlan; - if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) { - /* - * A bridge whose MAC address is taken from a VLAN network device - * (that is, a network device created with vconfig(8) or similar - * tool) will have the same MAC address as a bridge on the VLAN - * device's physical network device. - * - * Handle this case by hashing the physical network device MAC - * along with the VLAN identifier. - */ - uint8_t buf[ETH_ADDR_LEN + 2]; - memcpy(buf, bridge_ea, ETH_ADDR_LEN); - buf[ETH_ADDR_LEN] = vlan >> 8; - buf[ETH_ADDR_LEN + 1] = vlan; - return dpid_from_hash(buf, sizeof buf); - } else { - /* - * Assume that this bridge's MAC address is unique, since it - * doesn't fit any of the cases we handle specially. - */ - } - } else { + if (!hw_addr_iface) { /* * A purely internal bridge, that is, one that has no non-virtual - * network devices on it at all, is more difficult because it has no + * network devices on it at all, is difficult because it has no * natural unique identifier at all. * * When the host is a XenServer, we handle this case by hashing the @@ -1503,10 +1561,23 @@ iface_refresh_cfm_stats(struct iface *iface) fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto, iface->ofp_port); if (fault >= 0) { + const char *reasons[CFM_FAULT_N_REASONS]; bool fault_bool = fault; + size_t i, j; + + j = 0; + for (i = 0; i < CFM_FAULT_N_REASONS; i++) { + int reason = 1 << i; + if (fault & reason) { + reasons[j++] = cfm_fault_reason_to_str(reason); + } + } + ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1); + ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j); } else { ovsrec_interface_set_cfm_fault(cfg, NULL, 0); + ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0); } error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto, @@ -1765,12 +1836,28 @@ 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 datapath_destroyed; + bool vlan_splinters_changed; bool database_changed; struct bridge *br; @@ -1793,15 +1880,8 @@ 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, @@ -1817,7 +1897,19 @@ bridge_run(void) stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert); } - if (database_changed || datapath_destroyed) { + /* If VLAN splinters are in use, then we need to reconfigure if VLAN usage + * has changed. */ + vlan_splinters_changed = false; + if (vlan_splinters_enabled_anywhere) { + HMAP_FOR_EACH (br, node, &all_bridges) { + if (ofproto_has_vlan_usage_changed(br->ofproto)) { + vlan_splinters_changed = true; + break; + } + } + } + + if (database_changed || vlan_splinters_changed) { if (cfg) { struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl); @@ -1843,6 +1935,7 @@ 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; @@ -1852,6 +1945,11 @@ 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(); @@ -1979,8 +2077,8 @@ 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); @@ -1990,9 +2088,9 @@ 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"); + unixctl_command_reply_error(conn, "no such interface"); return; } @@ -2012,10 +2110,10 @@ qos_unixctl_show(struct unixctl_conn *conn, if (error) { ds_put_format(&ds, "failed to dump queues: %s", strerror(error)); } - unixctl_command_reply(conn, 200, ds_cstr(&ds)); + unixctl_command_reply(conn, ds_cstr(&ds)); } else { ds_put_format(&ds, "QoS not configured on %s\n", iface->name); - unixctl_command_reply(conn, 501, ds_cstr(&ds)); + unixctl_command_reply_error(conn, ds_cstr(&ds)); } shash_destroy_free_data(&sh); @@ -2089,22 +2187,22 @@ 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"); + unixctl_command_reply_error(conn, "Unknown bridge"); return; } ds_init(&results); ofproto_get_all_flows(br->ofproto, &results); - unixctl_command_reply(conn, 200, ds_cstr(&results)); + unixctl_command_reply(conn, ds_cstr(&results)); ds_destroy(&results); } @@ -2112,14 +2210,14 @@ 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"); + unixctl_command_reply_error(conn, "Unknown bridge"); return; } ofproto_reconnect_controllers(br->ofproto); @@ -2128,7 +2226,7 @@ bridge_unixctl_reconnect(struct unixctl_conn *conn, ofproto_reconnect_controllers(br->ofproto); } } - unixctl_command_reply(conn, 200, NULL); + unixctl_command_reply(conn, NULL); } static size_t @@ -2155,7 +2253,8 @@ bridge_get_controllers(const struct bridge *br, /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match * those configured in 'br->cfg'. */ static void -bridge_add_del_ports(struct bridge *br) +bridge_add_del_ports(struct bridge *br, + const unsigned long int *splinter_vlans) { struct port *port, *next; struct shash_node *node; @@ -2188,6 +2287,10 @@ bridge_add_del_ports(struct bridge *br) shash_add(&new_ports, br->name, &br->synth_local_port); } + if (splinter_vlans) { + add_vlan_splinter_ports(br, splinter_vlans, &new_ports); + } + /* Get rid of deleted ports. * Get rid of deleted interfaces on ports that still exist. */ HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) { @@ -2231,6 +2334,7 @@ bridge_ofproto_controller_for_mgmt(const struct bridge *br, oc->band = OFPROTO_OUT_OF_BAND; oc->rate_limit = 0; oc->burst_limit = 0; + oc->enable_async_msgs = true; } /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */ @@ -2246,6 +2350,8 @@ bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c, oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0; oc->burst_limit = (c->controller_burst_limit ? *c->controller_burst_limit : 0); + oc->enable_async_msgs = (!c->enable_async_messages + || *c->enable_async_messages); } /* Configures the IP stack for 'br''s local interface properly according to the @@ -2410,6 +2516,66 @@ 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. */ @@ -2591,7 +2757,7 @@ 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; @@ -2600,7 +2766,23 @@ 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", @@ -2613,7 +2795,6 @@ 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")) { @@ -2669,12 +2850,22 @@ 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, " @@ -2702,7 +2893,7 @@ 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; } @@ -2747,6 +2938,7 @@ iface_create(struct port *port, const struct ovsrec_interface *if_cfg) iface->tag = tag_create_random(); iface->netdev = NULL; iface->cfg = if_cfg; + iface->need_refresh = true; hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0)); @@ -2878,6 +3070,7 @@ iface_clear_db_record(const struct ovsrec_interface *if_cfg) ovsrec_interface_set_link_state(if_cfg, NULL); ovsrec_interface_set_mtu(if_cfg, NULL, 0); ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0); + ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0); ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0); ovsrec_interface_set_lacp_current(if_cfg, NULL, 0); ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0); @@ -3010,11 +3203,9 @@ iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) shash_destroy(&details); } if (!queue_zero) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); - VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default " - "queue (queue 0). Packets not directed to a " - "correctly configured queue may be dropped.", - iface->name); + shash_init(&details); + netdev_set_queue(iface->netdev, 0, &details); + shash_destroy(&details); } } @@ -3050,6 +3241,8 @@ 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; } @@ -3113,7 +3306,8 @@ 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); } } @@ -3179,8 +3373,9 @@ 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. */ @@ -3256,3 +3451,302 @@ mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg) return true; } + +/* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.) + * + * This is deprecated. It is only for compatibility with broken device drivers + * in old versions of Linux that do not properly support VLANs when VLAN + * devices are not used. When broken device drivers are no longer in + * widespread use, we will delete these interfaces. */ + +static void **blocks; +static size_t n_blocks, allocated_blocks; + +/* Adds 'block' to a list of blocks that have to be freed with free() when the + * VLAN splinters are reconfigured. */ +static void +register_block(void *block) +{ + if (n_blocks >= allocated_blocks) { + blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks); + } + blocks[n_blocks++] = block; +} + +/* Frees all of the blocks registered with register_block(). */ +static void +free_registered_blocks(void) +{ + size_t i; + + for (i = 0; i < n_blocks; i++) { + free(blocks[i]); + } + n_blocks = 0; +} + +/* Returns true if VLAN splinters are enabled on 'iface_cfg', false + * otherwise. */ +static bool +vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg) +{ + const char *value; + + value = get_interface_other_config(iface_cfg, "enable-vlan-splinters", ""); + return !strcmp(value, "true"); +} + +/* Figures out the set of VLANs that are in use for the purpose of VLAN + * splinters. + * + * If VLAN splinters are enabled on at least one interface and any VLANs are in + * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and + * 4095 will not be set). The caller is responsible for freeing the bitmap, + * with free(). + * + * If VLANs splinters are not enabled on any interface or if no VLANs are in + * use, returns NULL. + * + * Updates 'vlan_splinters_enabled_anywhere'. */ +static unsigned long int * +collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg) +{ + unsigned long int *splinter_vlans; + struct sset splinter_ifaces; + const char *real_dev_name; + struct shash *real_devs; + struct shash_node *node; + struct bridge *br; + size_t i; + + /* 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; + + for (j = 0; j < br_cfg->n_ports; j++) { + struct ovsrec_port *port_cfg = br_cfg->ports[j]; + int k; + + for (k = 0; k < port_cfg->n_interfaces; k++) { + 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); + 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); + } + } + } + + if (!vlan_splinters_enabled_anywhere) { + free(splinter_vlans); + sset_destroy(&splinter_ifaces); + return NULL; + } + + HMAP_FOR_EACH (br, node, &all_bridges) { + if (br->ofproto) { + ofproto_get_vlan_usage(br->ofproto, splinter_vlans); + } + } + + /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on + * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN + * device to be created for it. */ + bitmap_set0(splinter_vlans, 0); + bitmap_set0(splinter_vlans, 4095); + + /* Delete all VLAN devices that we don't need. */ + vlandev_refresh(); + real_devs = vlandev_get_real_devs(); + SHASH_FOR_EACH (node, real_devs) { + const struct vlan_real_dev *real_dev = node->data; + const struct vlan_dev *vlan_dev; + bool real_dev_has_splinters; + + real_dev_has_splinters = sset_contains(&splinter_ifaces, + real_dev->name); + HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) { + if (!real_dev_has_splinters + || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) { + struct netdev *netdev; + + if (!netdev_open(vlan_dev->name, "system", &netdev)) { + if (!netdev_get_in4(netdev, NULL, NULL) || + !netdev_get_in6(netdev, NULL)) { + vlandev_del(vlan_dev->name); + } else { + /* It has an IP address configured, so we don't own + * it. Don't delete it. */ + } + netdev_close(netdev); + } + } + + } + } + + /* Add all VLAN devices that we need. */ + SSET_FOR_EACH (real_dev_name, &splinter_ifaces) { + int vid; + + BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) { + if (!vlandev_get_name(real_dev_name, vid)) { + vlandev_add(real_dev_name, vid); + } + } + } + + vlandev_refresh(); + + sset_destroy(&splinter_ifaces); + + if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) { + free(splinter_vlans); + return NULL; + } + return splinter_vlans; +} + +/* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to + * ofproto. */ +static void +configure_splinter_port(struct port *port) +{ + struct ofproto *ofproto = port->bridge->ofproto; + uint16_t realdev_ofp_port; + const char *realdev_name; + struct iface *vlandev, *realdev; + + ofproto_bundle_unregister(port->bridge->ofproto, port); + + vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface, + port_elem); + + realdev_name = get_port_other_config(port->cfg, "realdev", NULL); + realdev = iface_lookup(port->bridge, realdev_name); + realdev_ofp_port = realdev ? realdev->ofp_port : 0; + + ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port, + *port->cfg->tag); +} + +static struct ovsrec_port * +synthesize_splinter_port(const char *real_dev_name, + const char *vlan_dev_name, int vid) +{ + struct ovsrec_interface *iface; + struct ovsrec_port *port; + + iface = xzalloc(sizeof *iface); + iface->name = xstrdup(vlan_dev_name); + iface->type = "system"; + + port = xzalloc(sizeof *port); + port->interfaces = xmemdup(&iface, sizeof iface); + port->n_interfaces = 1; + port->name = xstrdup(vlan_dev_name); + port->vlan_mode = "splinter"; + port->tag = xmalloc(sizeof *port->tag); + *port->tag = vid; + port->key_other_config = xmalloc(sizeof *port->key_other_config); + port->key_other_config[0] = "realdev"; + port->value_other_config = xmalloc(sizeof *port->value_other_config); + port->value_other_config[0] = xstrdup(real_dev_name); + port->n_other_config = 1; + + register_block(iface); + register_block(iface->name); + register_block(port); + register_block(port->interfaces); + register_block(port->name); + register_block(port->tag); + register_block(port->key_other_config); + register_block(port->value_other_config); + register_block(port->value_other_config[0]); + + return port; +} + +/* For each interface with 'br' that has VLAN splinters enabled, adds a + * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a + * 1-bit in the 'splinter_vlans' bitmap. */ +static void +add_vlan_splinter_ports(struct bridge *br, + const unsigned long int *splinter_vlans, + struct shash *ports) +{ + size_t i; + + /* We iterate through 'br->cfg->ports' instead of 'ports' here because + * we're modifying 'ports'. */ + for (i = 0; i < br->cfg->n_ports; i++) { + const char *name = br->cfg->ports[i]->name; + struct ovsrec_port *port_cfg = shash_find_data(ports, name); + size_t j; + + for (j = 0; j < port_cfg->n_interfaces; j++) { + struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j]; + + if (vlan_splinters_is_enabled(iface_cfg)) { + const char *real_dev_name; + uint16_t vid; + + real_dev_name = iface_cfg->name; + BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) { + const char *vlan_dev_name; + + vlan_dev_name = vlandev_get_name(real_dev_name, vid); + if (vlan_dev_name + && !shash_find(ports, vlan_dev_name)) { + shash_add(ports, vlan_dev_name, + synthesize_splinter_port( + real_dev_name, vlan_dev_name, vid)); + } + } + } + } + } +} + +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); +}