X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=60a0d6e1c7d1de2ac3a4b6023efbfe8a221d6520;hb=3dffcf0701f7e7753710348d34eb5be007525877;hp=b7f068bf0cd181407bc7cf74b8cca5b52c774bab;hpb=efdd908807544de7bee8644fafc1400e6a32d936;p=sliver-openvswitch.git diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index b7f068bf0..60a0d6e1c 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -37,6 +37,7 @@ #include "dynamic-string.h" #include "flow.h" #include "hash.h" +#include "hmap.h" #include "jsonrpc.h" #include "list.h" #include "mac-learning.h" @@ -46,9 +47,9 @@ #include "ofpbuf.h" #include "ofproto/netflow.h" #include "ofproto/ofproto.h" +#include "ovsdb-data.h" #include "packets.h" #include "poll-loop.h" -#include "port-array.h" #include "proc-net-compat.h" #include "process.h" #include "sha1.h" @@ -56,17 +57,18 @@ #include "socket-util.h" #include "stream-ssl.h" #include "svec.h" +#include "system-stats.h" #include "timeval.h" #include "util.h" #include "unixctl.h" #include "vconn.h" #include "vswitchd/vswitch-idl.h" #include "xenserver.h" +#include "vlog.h" #include "xtoxll.h" #include "sflow_api.h" -#define THIS_MODULE VLM_bridge -#include "vlog.h" +VLOG_DEFINE_THIS_MODULE(bridge) struct dst { uint16_t vlan; @@ -83,6 +85,7 @@ struct iface { /* These members are valid only after bridge_reconfigure() causes them to * be initialized. */ + struct hmap_node dp_ifidx_node; /* In struct bridge's "ifaces" hmap. */ int dp_ifidx; /* Index within kernel datapath. */ struct netdev *netdev; /* Network device. */ bool enabled; /* May be chosen for flows? */ @@ -104,6 +107,7 @@ struct mirror { struct bridge *bridge; size_t idx; char *name; + struct uuid uuid; /* UUID of this "mirror" record in database. */ /* Selection criteria. */ struct shash src_ports; /* Name is port name; data is always NULL. */ @@ -139,7 +143,7 @@ struct port { 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? */ - long bond_next_fake_iface_update; /* Next update to fake bond stats. */ + long long int bond_next_fake_iface_update; /* Time of next update. */ int bond_rebalance_interval; /* Interval between rebalances, in ms. */ long long int bond_next_rebalance; /* Next rebalancing time. */ @@ -162,7 +166,7 @@ struct bridge { /* Kernel datapath information. */ struct dpif *dpif; /* Datapath. */ - struct port_array ifaces; /* Indexed by kernel datapath port number. */ + struct hmap ifaces; /* Contains "struct iface"s. */ /* Bridge ports. */ struct port **ports; @@ -186,10 +190,10 @@ static struct list all_bridges = LIST_INITIALIZER(&all_bridges); /* OVSDB IDL used to obtain configuration. */ static struct ovsdb_idl *idl; -/* Each time this timer expires, the bridge fetches statistics for every - * interface and pushes them into the database. */ -#define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */ -static long long int iface_stats_timer = LLONG_MIN; +/* Each time this timer expires, the bridge fetches systems and interface + * statistics and pushes them into the database. */ +#define STATS_INTERVAL (5 * 1000) /* In milliseconds. */ +static long long int stats_timer = LLONG_MIN; static struct bridge *bridge_create(const struct ovsrec_bridge *br_cfg); static void bridge_destroy(struct bridge *); @@ -197,13 +201,10 @@ static struct bridge *bridge_lookup(const char *name); static unixctl_cb_func bridge_unixctl_dump_flows; static unixctl_cb_func bridge_unixctl_reconnect; static int bridge_run_one(struct bridge *); -static size_t bridge_get_controllers(const struct ovsrec_open_vswitch *ovs_cfg, - const struct bridge *br, +static size_t bridge_get_controllers(const struct bridge *br, struct ovsrec_controller ***controllersp); -static void bridge_reconfigure_one(const struct ovsrec_open_vswitch *, - struct bridge *); -static void bridge_reconfigure_remotes(const struct ovsrec_open_vswitch *, - struct bridge *, +static void bridge_reconfigure_one(struct bridge *); +static void bridge_reconfigure_remotes(struct bridge *, const struct sockaddr_in *managers, size_t n_managers); static void bridge_get_all_ifaces(const struct bridge *, struct shash *ifaces); @@ -239,13 +240,13 @@ static void port_update_bond_compat(struct port *); static void port_update_vlan_compat(struct port *); static void port_update_bonding(struct port *); -static struct mirror *mirror_create(struct bridge *, const char *name); +static void mirror_create(struct bridge *, struct ovsrec_mirror *); static void mirror_destroy(struct mirror *); static void mirror_reconfigure(struct bridge *); static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *); static bool vlan_is_mirrored(const struct mirror *, int vlan); -static struct iface *iface_create(struct port *port, +static struct iface *iface_create(struct port *port, const struct ovsrec_interface *if_cfg); static void iface_destroy(struct iface *); static struct iface *iface_lookup(const struct bridge *, const char *name); @@ -269,6 +270,19 @@ bridge_init(const char *remote) /* Create connection to database. */ idl = ovsdb_idl_create(remote, &ovsrec_idl_class); + ovsdb_idl_set_write_only(idl, &ovsrec_open_vswitch_col_cur_cfg); + ovsdb_idl_set_write_only(idl, &ovsrec_open_vswitch_col_statistics); + ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids); + + ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids); + + ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids); + ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge); + + ovsdb_idl_set_write_only(idl, &ovsrec_interface_col_ofport); + ovsdb_idl_set_write_only(idl, &ovsrec_interface_col_statistics); + ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids); + /* Register unixctl commands. */ unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL); unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows, @@ -294,7 +308,7 @@ bridge_configure_once(const struct ovsrec_open_vswitch *cfg) } already_configured_once = true; - iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL; + stats_timer = time_msec() + STATS_INTERVAL; /* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */ svec_init(&bridge_names); @@ -345,19 +359,6 @@ bridge_configure_once(const struct ovsrec_open_vswitch *cfg) svec_destroy(&dpif_types); } -#ifdef HAVE_OPENSSL -static void -bridge_configure_ssl(const struct ovsrec_ssl *ssl) -{ - /* XXX SSL should be configurable on a per-bridge basis. */ - if (ssl) { - stream_ssl_set_private_key_file(ssl->private_key); - stream_ssl_set_certificate_file(ssl->certificate); - stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert); - } -} -#endif - /* Attempt to create the network device 'iface_name' through the netdev * library. */ static int @@ -374,6 +375,20 @@ set_up_iface(const struct ovsrec_interface *iface_cfg, struct iface *iface, xstrdup(iface_cfg->value_options[i])); } + /* Include 'other_config' keys in hash of netdev options. The + * namespace of 'other_config' and 'options' must be disjoint. + * Prefer 'options' keys over 'other_config' keys. */ + for (i = 0; i < iface_cfg->n_other_config; i++) { + char *value = xstrdup(iface_cfg->value_other_config[i]); + if (!shash_add_once(&options, iface_cfg->key_other_config[i], + value)) { + VLOG_WARN("%s: \"other_config\" key %s conflicts with existing " + "\"other_config\" or \"options\" entry...ignoring", + iface_cfg->name, iface_cfg->key_other_config[i]); + free(value); + } + } + if (create) { struct netdev_options netdev_options; @@ -560,7 +575,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) /* Collect old and new bridges. */ shash_init(&old_br); shash_init(&new_br); - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { shash_add(&old_br, br->name, br); } for (i = 0; i < ovs_cfg->n_bridges; i++) { @@ -571,7 +586,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) } /* Get rid of deleted bridges and add new bridges. */ - LIST_FOR_EACH_SAFE (br, next, struct bridge, node, &all_bridges) { + LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) { struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name); if (br_cfg) { br->cfg = br_cfg; @@ -597,14 +612,9 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) shash_destroy(&old_br); shash_destroy(&new_br); -#ifdef HAVE_OPENSSL - /* Configure SSL. */ - bridge_configure_ssl(ovs_cfg->ssl); -#endif - /* Reconfigure all bridges. */ - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { - bridge_reconfigure_one(ovs_cfg, br); + LIST_FOR_EACH (br, node, &all_bridges) { + bridge_reconfigure_one(br); } /* Add and delete ports on all datapaths. @@ -612,7 +622,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) * The kernel will reject any attempt to add a given port to a datapath if * that port already belongs to a different datapath, so we must do all * port deletions before any port additions. */ - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { struct odp_port *dpif_ports; size_t n_dpif_ports; struct shash want_ifaces; @@ -634,11 +644,10 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) shash_destroy(&want_ifaces); free(dpif_ports); } - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { struct odp_port *dpif_ports; size_t n_dpif_ports; struct shash cur_ifaces, want_ifaces; - struct shash_node *node; /* Get the set of interfaces currently in this datapath. */ dpif_port_list(br->dpif, &dpif_ports, &n_dpif_ports); @@ -684,7 +693,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) shash_destroy(&want_ifaces); } sflow_bridge_number = 0; - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { uint8_t ea[8]; uint64_t dpid; struct iface *local_iface; @@ -758,7 +767,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) opts.collectors.n = nf_cfg->n_targets; opts.collectors.names = nf_cfg->targets; if (ofproto_set_netflow(br->ofproto, &opts)) { - VLOG_ERR("bridge %s: problem setting netflow collectors", + VLOG_ERR("bridge %s: problem setting netflow collectors", br->name); } } else { @@ -771,7 +780,6 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) struct ovsrec_controller **controllers; struct ofproto_sflow_options oso; size_t n_controllers; - size_t i; memset(&oso, 0, sizeof oso); @@ -797,7 +805,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) oso.agent_device = sflow_cfg->agent; oso.control_ip = NULL; - n_controllers = bridge_get_controllers(ovs_cfg, br, &controllers); + n_controllers = bridge_get_controllers(br, &controllers); for (i = 0; i < n_controllers; i++) { if (controllers[i]->local_ip) { oso.control_ip = controllers[i]->local_ip; @@ -820,9 +828,9 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) * yet; when a controller is configured, resetting the datapath ID will * immediately disconnect from the controller, so it's better to set * the datapath ID before the controller. */ - bridge_reconfigure_remotes(ovs_cfg, br, managers, n_managers); + bridge_reconfigure_remotes(br, managers, n_managers); } - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { for (i = 0; i < br->n_ports; i++) { struct port *port = br->ports[i]; int j; @@ -835,7 +843,7 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) } } } - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { iterate_and_prune_ifaces(br, set_iface_properties, NULL); } @@ -843,25 +851,25 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) } static const char * -get_ovsrec_key_value(const char *key, char **keys, char **values, size_t n) +get_ovsrec_key_value(const struct ovsdb_idl_row *row, + const struct ovsdb_idl_column *column, + const char *key) { - size_t i; - - for (i = 0; i < n; i++) { - if (!strcmp(keys[i], key)) { - return values[i]; - } - } - return NULL; + const struct ovsdb_datum *datum; + union ovsdb_atom atom; + unsigned int idx; + + datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING); + atom.string = (char *) key; + idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING); + return idx == UINT_MAX ? NULL : datum->values[idx].string; } static const char * bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key) { - return get_ovsrec_key_value(key, - br_cfg->key_other_config, - br_cfg->value_other_config, - br_cfg->n_other_config); + return get_ovsrec_key_value(&br_cfg->header_, + &ovsrec_bridge_col_other_config, key); } static void @@ -1104,15 +1112,32 @@ iface_refresh_stats(struct iface *iface) ovsrec_interface_set_statistics(iface->cfg, keys, values, n); } +static void +refresh_system_stats(const struct ovsrec_open_vswitch *cfg) +{ + struct ovsdb_datum datum; + struct shash stats; + + shash_init(&stats); + get_system_stats(&stats); + + ovsdb_datum_from_shash(&datum, &stats); + ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics, + &datum); +} + void bridge_run(void) { + const struct ovsrec_open_vswitch *cfg; + bool datapath_destroyed; + bool database_changed; struct bridge *br; /* Let each bridge do the work that it needs to do. */ datapath_destroyed = false; - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { int error = bridge_run_one(br); if (error) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); @@ -1123,8 +1148,9 @@ bridge_run(void) } /* (Re)configure if necessary. */ - if (ovsdb_idl_run(idl) || datapath_destroyed) { - const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl); + database_changed = ovsdb_idl_run(idl); + cfg = ovsrec_open_vswitch_first(idl); + if (database_changed || datapath_destroyed) { if (cfg) { struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl); @@ -1143,28 +1169,43 @@ bridge_run(void) } } - /* Refresh interface stats if necessary. */ - if (time_msec() >= iface_stats_timer) { - struct ovsdb_idl_txn *txn; +#ifdef HAVE_OPENSSL + /* Re-configure SSL. We do this on every trip through the main loop, + * instead of just when the database changes, because the contents of the + * key and certificate files can change without the database changing. */ + if (cfg && cfg->ssl) { + const struct ovsrec_ssl *ssl = cfg->ssl; - txn = ovsdb_idl_txn_create(idl); - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { - size_t i; + stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate); + stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert); + } +#endif - for (i = 0; i < br->n_ports; i++) { - struct port *port = br->ports[i]; - size_t j; + /* Refresh system and interface stats if necessary. */ + if (time_msec() >= stats_timer) { + if (cfg) { + struct ovsdb_idl_txn *txn; - for (j = 0; j < port->n_ifaces; j++) { - struct iface *iface = port->ifaces[j]; - iface_refresh_stats(iface); + txn = ovsdb_idl_txn_create(idl); + LIST_FOR_EACH (br, node, &all_bridges) { + size_t i; + + for (i = 0; i < br->n_ports; i++) { + struct port *port = br->ports[i]; + size_t j; + + for (j = 0; j < port->n_ifaces; j++) { + struct iface *iface = port->ifaces[j]; + iface_refresh_stats(iface); + } } } + refresh_system_stats(cfg); + ovsdb_idl_txn_commit(txn); + ovsdb_idl_txn_destroy(txn); /* XXX */ } - ovsdb_idl_txn_commit(txn); - ovsdb_idl_txn_destroy(txn); /* XXX */ - iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL; + stats_timer = time_msec() + STATS_INTERVAL; } } @@ -1173,9 +1214,9 @@ bridge_wait(void) { struct bridge *br; - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { ofproto_wait(br->ofproto); - if (ofproto_has_controller(br->ofproto)) { + if (ofproto_has_primary_controller(br->ofproto)) { continue; } @@ -1183,7 +1224,7 @@ bridge_wait(void) bond_wait(br); } ovsdb_idl_wait(idl); - poll_timer_wait_until(iface_stats_timer); + poll_timer_wait_until(stats_timer); } /* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s @@ -1232,7 +1273,7 @@ bridge_unixctl_fdb_show(struct unixctl_conn *conn, } ds_put_cstr(&ds, " port VLAN MAC Age\n"); - LIST_FOR_EACH (e, struct mac_entry, lru_node, &br->ml->lrus) { + LIST_FOR_EACH (e, lru_node, &br->ml->lrus) { if (e->port < 0 || e->port >= br->n_ports) { continue; } @@ -1278,7 +1319,7 @@ bridge_create(const struct ovsrec_bridge *br_cfg) br->ml = mac_learning_create(); eth_addr_nicira_random(br->default_ea); - port_array_init(&br->ifaces); + hmap_init(&br->ifaces); shash_init(&br->port_by_name); shash_init(&br->iface_by_name); @@ -1310,7 +1351,7 @@ bridge_destroy(struct bridge *br) dpif_close(br->dpif); ofproto_destroy(br->ofproto); mac_learning_destroy(br->ml); - port_array_destroy(&br->ifaces); + hmap_destroy(&br->ifaces); shash_destroy(&br->port_by_name); shash_destroy(&br->iface_by_name); free(br->ports); @@ -1324,7 +1365,7 @@ bridge_lookup(const char *name) { struct bridge *br; - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { if (!strcmp(br->name, name)) { return br; } @@ -1340,7 +1381,7 @@ bridge_unixctl_dump_flows(struct unixctl_conn *conn, { struct bridge *br; struct ds results; - + br = bridge_lookup(args); if (!br) { unixctl_command_reply(conn, 501, "Unknown bridge"); @@ -1370,7 +1411,7 @@ bridge_unixctl_reconnect(struct unixctl_conn *conn, } ofproto_reconnect_controllers(br->ofproto); } else { - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { ofproto_reconnect_controllers(br->ofproto); } } @@ -1397,20 +1438,14 @@ bridge_run_one(struct bridge *br) } static size_t -bridge_get_controllers(const struct ovsrec_open_vswitch *ovs_cfg, - const struct bridge *br, +bridge_get_controllers(const struct bridge *br, struct ovsrec_controller ***controllersp) { struct ovsrec_controller **controllers; size_t n_controllers; - if (br->cfg->n_controller) { - controllers = br->cfg->controller; - n_controllers = br->cfg->n_controller; - } else { - controllers = ovs_cfg->controller; - n_controllers = ovs_cfg->n_controller; - } + controllers = br->cfg->controller; + n_controllers = br->cfg->n_controller; if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) { controllers = NULL; @@ -1424,13 +1459,12 @@ bridge_get_controllers(const struct ovsrec_open_vswitch *ovs_cfg, } static void -bridge_reconfigure_one(const struct ovsrec_open_vswitch *ovs_cfg, - struct bridge *br) +bridge_reconfigure_one(struct bridge *br) { struct shash old_ports, new_ports; - struct svec listeners, old_listeners; struct svec snoops, old_snoops; struct shash_node *node; + enum ofproto_fail_mode fail_mode; size_t i; /* Collect old ports. */ @@ -1453,7 +1487,7 @@ bridge_reconfigure_one(const struct ovsrec_open_vswitch *ovs_cfg, * user didn't specify one. * * XXX perhaps we should synthesize a port ourselves in this case. */ - if (bridge_get_controllers(ovs_cfg, br, NULL)) { + if (bridge_get_controllers(br, NULL)) { char local_name[IF_NAMESIZE]; int error; @@ -1499,22 +1533,21 @@ bridge_reconfigure_one(const struct ovsrec_open_vswitch *ovs_cfg, shash_destroy(&old_ports); shash_destroy(&new_ports); + /* Set the fail-mode */ + fail_mode = !br->cfg->fail_mode + || !strcmp(br->cfg->fail_mode, "standalone") + ? OFPROTO_FAIL_STANDALONE + : OFPROTO_FAIL_SECURE; + if (ofproto_get_fail_mode(br->ofproto) != fail_mode + && !ofproto_has_primary_controller(br->ofproto)) { + ofproto_flush_flows(br->ofproto); + } + ofproto_set_fail_mode(br->ofproto, fail_mode); + /* Delete all flows if we're switching from connected to standalone or vice * versa. (XXX Should we delete all flows if we are switching from one * controller to another?) */ - /* Configure OpenFlow management listener. */ - svec_init(&listeners); - svec_add_nocopy(&listeners, xasprintf("punix:%s/%s.mgmt", - ovs_rundir, br->name)); - svec_init(&old_listeners); - ofproto_get_listeners(br->ofproto, &old_listeners); - if (!svec_equal(&listeners, &old_listeners)) { - ofproto_set_listeners(br->ofproto, &listeners); - } - svec_destroy(&listeners); - svec_destroy(&old_listeners); - /* Configure OpenFlow controller connection snooping. */ svec_init(&snoops); svec_add_nocopy(&snoops, xasprintf("punix:%s/%s.snoop", @@ -1530,107 +1563,155 @@ bridge_reconfigure_one(const struct ovsrec_open_vswitch *ovs_cfg, mirror_reconfigure(br); } +/* Initializes 'oc' appropriately as a management service controller for + * 'br'. + * + * The caller must free oc->target when it is no longer needed. */ +static void +bridge_ofproto_controller_for_mgmt(const struct bridge *br, + struct ofproto_controller *oc) +{ + oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir, br->name); + oc->max_backoff = 0; + oc->probe_interval = 60; + oc->band = OFPROTO_OUT_OF_BAND; + oc->accept_re = NULL; + oc->update_resolv_conf = false; + oc->rate_limit = 0; + oc->burst_limit = 0; +} + +/* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */ static void -bridge_reconfigure_remotes(const struct ovsrec_open_vswitch *ovs_cfg, - struct bridge *br, +bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c, + struct ofproto_controller *oc) +{ + oc->target = c->target; + oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8; + oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5; + oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band") + ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND); + oc->accept_re = c->discover_accept_regex; + oc->update_resolv_conf = c->discover_update_resolv_conf; + oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0; + oc->burst_limit = (c->controller_burst_limit + ? *c->controller_burst_limit : 0); +} + +/* Configures the IP stack for 'br''s local interface properly according to the + * configuration in 'c'. */ +static void +bridge_configure_local_iface_netdev(struct bridge *br, + struct ovsrec_controller *c) +{ + struct netdev *netdev; + struct in_addr mask, gateway; + + struct iface *local_iface; + struct in_addr ip; + + /* Controller discovery does its own TCP/IP configuration later. */ + if (strcmp(c->target, "discover")) { + return; + } + + /* If there's no local interface or no IP address, give up. */ + local_iface = bridge_get_local_iface(br); + if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) { + return; + } + + /* Bring up the local interface. */ + netdev = local_iface->netdev; + netdev_turn_flags_on(netdev, NETDEV_UP, true); + + /* Configure the IP address and netmask. */ + if (!c->local_netmask + || !inet_aton(c->local_netmask, &mask) + || !mask.s_addr) { + mask.s_addr = guess_netmask(ip.s_addr); + } + if (!netdev_set_in4(netdev, ip, mask)) { + VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT, + br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr)); + } + + /* Configure the default gateway. */ + if (c->local_gateway + && inet_aton(c->local_gateway, &gateway) + && gateway.s_addr) { + if (!netdev_add_router(netdev, gateway)) { + VLOG_INFO("bridge %s: configured gateway "IP_FMT, + br->name, IP_ARGS(&gateway.s_addr)); + } + } +} + +static void +bridge_reconfigure_remotes(struct bridge *br, const struct sockaddr_in *managers, size_t n_managers) { struct ovsrec_controller **controllers; size_t n_controllers; + bool had_primary; + + struct ofproto_controller *ocs; + size_t n_ocs; + size_t i; ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers); + had_primary = ofproto_has_primary_controller(br->ofproto); + + n_controllers = bridge_get_controllers(br, &controllers); - n_controllers = bridge_get_controllers(ovs_cfg, br, &controllers); - if (ofproto_has_controller(br->ofproto) != (n_controllers != 0)) { + ocs = xmalloc((n_controllers + 1) * sizeof *ocs); + n_ocs = 0; + + bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]); + for (i = 0; i < n_controllers; i++) { + struct ovsrec_controller *c = controllers[i]; + + if (!strncmp(c->target, "punix:", 6) + || !strncmp(c->target, "unix:", 5)) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + + /* Prevent remote ovsdb-server users from accessing arbitrary Unix + * domain sockets and overwriting arbitrary local files. */ + VLOG_ERR_RL(&rl, "%s: not adding Unix domain socket controller " + "\"%s\" due to possibility for remote exploit", + dpif_name(br->dpif), c->target); + continue; + } + + bridge_configure_local_iface_netdev(br, c); + bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs++]); + } + + ofproto_set_controllers(br->ofproto, ocs, n_ocs); + free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */ + free(ocs); + + if (had_primary != ofproto_has_primary_controller(br->ofproto)) { ofproto_flush_flows(br->ofproto); } - if (!n_controllers) { + /* If there are no controllers and the bridge is in standalone + * mode, set up a flow that matches every packet and directs + * them to OFPP_NORMAL (which goes to us). Otherwise, the + * switch is in secure mode and we won't pass any traffic until + * a controller has been defined and it tells us to do so. */ + if (!n_controllers + && ofproto_get_fail_mode(br->ofproto) == OFPROTO_FAIL_STANDALONE) { union ofp_action action; flow_t flow; - /* Clear out controllers. */ - ofproto_set_controllers(br->ofproto, NULL, 0); - - /* Set up a flow that matches every packet and directs them to - * OFPP_NORMAL (which goes to us). */ memset(&action, 0, sizeof action); action.type = htons(OFPAT_OUTPUT); action.output.len = htons(sizeof action); action.output.port = htons(OFPP_NORMAL); memset(&flow, 0, sizeof flow); ofproto_add_flow(br->ofproto, &flow, OVSFW_ALL, 0, &action, 1, 0); - } else { - struct ofproto_controller *ocs; - size_t i; - - ocs = xmalloc(n_controllers * sizeof *ocs); - for (i = 0; i < n_controllers; i++) { - struct ovsrec_controller *c = controllers[i]; - struct ofproto_controller *oc = &ocs[i]; - - if (strcmp(c->target, "discover")) { - struct iface *local_iface; - struct in_addr ip; - - local_iface = bridge_get_local_iface(br); - if (local_iface && c->local_ip - && inet_aton(c->local_ip, &ip)) { - struct netdev *netdev = local_iface->netdev; - struct in_addr mask, gateway; - - if (!c->local_netmask - || !inet_aton(c->local_netmask, &mask)) { - mask.s_addr = 0; - } - if (!c->local_gateway - || !inet_aton(c->local_gateway, &gateway)) { - gateway.s_addr = 0; - } - - netdev_turn_flags_on(netdev, NETDEV_UP, true); - if (!mask.s_addr) { - mask.s_addr = guess_netmask(ip.s_addr); - } - if (!netdev_set_in4(netdev, ip, mask)) { - VLOG_INFO("bridge %s: configured IP address "IP_FMT", " - "netmask "IP_FMT, - br->name, IP_ARGS(&ip.s_addr), - IP_ARGS(&mask.s_addr)); - } - - if (gateway.s_addr) { - if (!netdev_add_router(netdev, gateway)) { - VLOG_INFO("bridge %s: configured gateway "IP_FMT, - br->name, IP_ARGS(&gateway.s_addr)); - } - } - } - } - - oc->target = c->target; - oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8; - oc->probe_interval = (c->inactivity_probe - ? *c->inactivity_probe / 1000 : 5); - oc->fail = (!c->fail_mode - || !strcmp(c->fail_mode, "standalone") - || !strcmp(c->fail_mode, "open") - ? OFPROTO_FAIL_STANDALONE - : OFPROTO_FAIL_SECURE); - oc->band = (!c->connection_mode - || !strcmp(c->connection_mode, "in-band") - ? OFPROTO_IN_BAND - : OFPROTO_OUT_OF_BAND); - oc->accept_re = c->discover_accept_regex; - oc->update_resolv_conf = c->discover_update_resolv_conf; - oc->rate_limit = (c->controller_rate_limit - ? *c->controller_rate_limit : 0); - oc->burst_limit = (c->controller_burst_limit - ? *c->controller_burst_limit : 0); - } - ofproto_set_controllers(br->ofproto, ocs, n_controllers); - free(ocs); } } @@ -1675,7 +1756,7 @@ bridge_fetch_dp_ifaces(struct bridge *br) iface->dp_ifidx = -1; } } - port_array_clear(&br->ifaces); + hmap_clear(&br->ifaces); dpif_port_list(br->dpif, &dpif_ports, &n_dpif_ports); for (i = 0; i < n_dpif_ports; i++) { @@ -1689,8 +1770,9 @@ bridge_fetch_dp_ifaces(struct bridge *br) VLOG_WARN("%s reported interface %"PRIu16" twice", dpif_name(br->dpif), p->port); } else { - port_array_set(&br->ifaces, p->port, iface); iface->dp_ifidx = p->port; + hmap_insert(&br->ifaces, &iface->dp_ifidx_node, + hash_int(iface->dp_ifidx, 0)); } if (iface->cfg) { @@ -2311,7 +2393,7 @@ is_admissible(struct bridge *br, const flow_t *flow, bool have_packet, static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown " - "interface %"PRIu16, br->name, flow->in_port); + "interface %"PRIu16, br->name, flow->in_port); } *in_portp = NULL; @@ -2469,11 +2551,12 @@ bridge_normal_ofhook_cb(const flow_t *flow, const struct ofpbuf *packet, struct bridge *br = br_; COVERAGE_INC(bridge_process_flow); + return process_flow(br, flow, packet, actions, tags, nf_output_iface); } static void -bridge_account_flow_ofhook_cb(const flow_t *flow, +bridge_account_flow_ofhook_cb(const flow_t *flow, tag_type tags, const union odp_action *actions, size_t n_actions, unsigned long long int n_bytes, void *br_) @@ -2481,20 +2564,24 @@ bridge_account_flow_ofhook_cb(const flow_t *flow, struct bridge *br = br_; const union odp_action *a; struct port *in_port; - tag_type tags = 0; + tag_type dummy = 0; int vlan; - /* Feed information from the active flows back into the learning table - * to ensure that table is always in sync with what is actually flowing - * through the datapath. */ - if (is_admissible(br, flow, false, &tags, &vlan, &in_port)) { + /* Feed information from the active flows back into the learning table to + * ensure that table is always in sync with what is actually flowing + * through the datapath. + * + * We test that 'tags' is nonzero to ensure that only flows that include an + * OFPP_NORMAL action are used for learning. This works because + * bridge_normal_ofhook_cb() always sets a nonzero tag value. */ + if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) { update_learning_table(br, flow, vlan, in_port); } + /* Account for bond slave utilization. */ if (!br->has_bonded_ports) { return; } - for (a = actions; a < &actions[n_actions]; a++) { if (a->type == ODPAT_OUTPUT) { struct port *out_port = port_from_dp_ifidx(br, a->output.port); @@ -2767,7 +2854,6 @@ bond_rebalance_port(struct port *port) * smallest hashes instead of the biggest ones. There is little * reason behind this decision; we could use the opposite sort * order to shift away big hashes ahead of small ones. */ - size_t i; bool order_swapped; for (i = 0; i < from->n_hashes; i++) { @@ -2848,7 +2934,7 @@ bond_send_learning_packets(struct port *port) ofpbuf_init(&packet, 128); error = n_packets = n_errors = 0; - LIST_FOR_EACH (e, struct mac_entry, lru_node, &br->ml->lrus) { + LIST_FOR_EACH (e, lru_node, &br->ml->lrus) { union ofp_action actions[2], *a; uint16_t dp_ifidx; tag_type tags = 0; @@ -2910,7 +2996,7 @@ bond_unixctl_list(struct unixctl_conn *conn, ds_put_cstr(&ds, "bridge\tbond\tslaves\n"); - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { size_t i; for (i = 0; i < br->n_ports; i++) { @@ -2939,7 +3025,7 @@ bond_find(const char *name) { const struct bridge *br; - LIST_FOR_EACH (br, struct bridge, node, &all_bridges) { + LIST_FOR_EACH (br, node, &all_bridges) { size_t i; for (i = 0; i < br->n_ports; i++) { @@ -2999,8 +3085,7 @@ bond_unixctl_show(struct unixctl_conn *conn, hash, be->tx_bytes / 1024); /* MACs. */ - LIST_FOR_EACH (me, struct mac_entry, lru_node, - &port->bridge->ml->lrus) { + LIST_FOR_EACH (me, lru_node, &port->bridge->ml->lrus) { uint16_t dp_ifidx; tag_type tags = 0; if (bond_hash(me->mac) == hash @@ -3237,10 +3322,10 @@ static const char * get_port_other_config(const struct ovsrec_port *port, const char *key, const char *default_value) { - const char *value = get_ovsrec_key_value(key, - port->key_other_config, - port->value_other_config, - port->n_other_config); + const char *value; + + value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config, + key); return value ? value : default_value; } @@ -3285,7 +3370,7 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) if (port->updelay < 0) { port->updelay = 0; } - port->updelay = cfg->bond_downdelay; + port->downdelay = cfg->bond_downdelay; if (port->downdelay < 0) { port->downdelay = 0; } @@ -3352,7 +3437,6 @@ port_reconfigure(struct port *port, const struct ovsrec_port *cfg) trunks = NULL; if (vlan < 0 && cfg->n_trunks) { size_t n_errors; - size_t i; trunks = bitmap_allocate(4096); n_errors = 0; @@ -3515,15 +3599,15 @@ port_update_bond_compat(struct port *port) /* 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 + * 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 + * 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; @@ -3651,7 +3735,7 @@ iface_destroy(struct iface *iface) shash_find_and_delete_assert(&br->iface_by_name, iface->name); if (iface->dp_ifidx >= 0) { - port_array_set(&br->ifaces, iface->dp_ifidx, NULL); + hmap_remove(&br->ifaces, &iface->dp_ifidx_node); } del = port->ifaces[iface->port_ifidx] = port->ifaces[--port->n_ifaces]; @@ -3681,7 +3765,15 @@ iface_lookup(const struct bridge *br, const char *name) static struct iface * iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx) { - return port_array_get(&br->ifaces, dp_ifidx); + struct iface *iface; + + HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node, + hash_int(dp_ifidx, 0), &br->ifaces) { + if (iface->dp_ifidx == dp_ifidx) { + return iface; + } + } + return NULL; } /* Returns true if 'iface' is the name of an "internal" interface on bridge @@ -3754,27 +3846,16 @@ shash_from_ovs_idl_map(char **keys, char **values, size_t n, struct iface_delete_queues_cbdata { struct netdev *netdev; - const int64_t *queue_ids; - size_t n_queue_ids; + const struct ovsdb_datum *queues; }; static bool -queue_ids_include(const int64_t *ids, size_t n, int64_t target) +queue_ids_include(const struct ovsdb_datum *queues, int64_t target) { - size_t low = 0; - size_t high = n; - - while (low < high) { - size_t mid = low + (high - low) / 2; - if (target > ids[mid]) { - high = mid; - } else if (target < ids[mid]) { - low = mid + 1; - } else { - return true; - } - } - return false; + union ovsdb_atom atom; + + atom.integer = target; + return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX; } static void @@ -3783,7 +3864,7 @@ iface_delete_queues(unsigned int queue_id, { struct iface_delete_queues_cbdata *cbdata = cbdata_; - if (!queue_ids_include(cbdata->queue_ids, cbdata->n_queue_ids, queue_id)) { + if (!queue_ids_include(cbdata->queues, queue_id)) { netdev_delete_queue(cbdata->netdev, queue_id); } } @@ -3806,8 +3887,8 @@ iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos) /* Deconfigure queues that were deleted. */ cbdata.netdev = iface->netdev; - cbdata.queue_ids = qos->key_queues; - cbdata.n_queue_ids = qos->n_queues; + cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER, + OVSDB_TYPE_UUID); netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata); /* Configure queues for 'iface'. */ @@ -3826,50 +3907,51 @@ iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos) /* Port mirroring. */ +static struct mirror * +mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid) +{ + int i; + + for (i = 0; i < MAX_MIRRORS; i++) { + struct mirror *m = br->mirrors[i]; + if (m && uuid_equals(uuid, &m->uuid)) { + return m; + } + } + return NULL; +} + static void mirror_reconfigure(struct bridge *br) { - struct shash old_mirrors, new_mirrors; - struct shash_node *node; unsigned long *rspan_vlans; int i; - /* Collect old mirrors. */ - shash_init(&old_mirrors); + /* Get rid of deleted mirrors. */ for (i = 0; i < MAX_MIRRORS; i++) { - if (br->mirrors[i]) { - shash_add(&old_mirrors, br->mirrors[i]->name, br->mirrors[i]); + struct mirror *m = br->mirrors[i]; + if (m) { + const struct ovsdb_datum *mc; + union ovsdb_atom atom; + + mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID); + atom.uuid = br->mirrors[i]->uuid; + if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) { + mirror_destroy(m); + } } } - /* Collect new mirrors. */ - shash_init(&new_mirrors); + /* Add new mirrors and reconfigure existing ones. */ for (i = 0; i < br->cfg->n_mirrors; i++) { struct ovsrec_mirror *cfg = br->cfg->mirrors[i]; - if (!shash_add_once(&new_mirrors, cfg->name, cfg)) { - VLOG_WARN("bridge %s: %s specified twice as mirror", - br->name, cfg->name); - } - } - - /* Get rid of deleted mirrors and add new mirrors. */ - SHASH_FOR_EACH (node, &old_mirrors) { - if (!shash_find(&new_mirrors, node->name)) { - mirror_destroy(node->data); - } - } - SHASH_FOR_EACH (node, &new_mirrors) { - struct mirror *mirror = shash_find_data(&old_mirrors, node->name); - if (!mirror) { - mirror = mirror_create(br, node->name); - if (!mirror) { - break; - } + struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid); + if (m) { + mirror_reconfigure_one(m, cfg); + } else { + mirror_create(br, cfg); } - mirror_reconfigure_one(mirror, node->data); } - shash_destroy(&old_mirrors); - shash_destroy(&new_mirrors); /* Update port reserved status. */ for (i = 0; i < br->n_ports; i++) { @@ -3904,8 +3986,8 @@ mirror_reconfigure(struct bridge *br) } } -static struct mirror * -mirror_create(struct bridge *br, const char *name) +static void +mirror_create(struct bridge *br, struct ovsrec_mirror *cfg) { struct mirror *m; size_t i; @@ -3913,21 +3995,21 @@ mirror_create(struct bridge *br, const char *name) for (i = 0; ; i++) { if (i >= MAX_MIRRORS) { VLOG_WARN("bridge %s: maximum of %d port mirrors reached, " - "cannot create %s", br->name, MAX_MIRRORS, name); - return NULL; + "cannot create %s", br->name, MAX_MIRRORS, cfg->name); + return; } if (!br->mirrors[i]) { break; } } - VLOG_INFO("created port mirror %s on bridge %s", name, br->name); + VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name); bridge_flush(br); br->mirrors[i] = m = xzalloc(sizeof *m); m->bridge = br; m->idx = i; - m->name = xstrdup(name); + m->name = xstrdup(cfg->name); shash_init(&m->src_ports); shash_init(&m->dst_ports); m->vlans = NULL; @@ -3935,7 +4017,7 @@ mirror_create(struct bridge *br, const char *name) m->out_vlan = -1; m->out_port = NULL; - return m; + mirror_reconfigure_one(m, cfg); } static void @@ -3955,6 +4037,7 @@ mirror_destroy(struct mirror *m) free(m->vlans); m->bridge->mirrors[m->idx] = NULL; + free(m->name); free(m); bridge_flush(br); @@ -4036,6 +4119,12 @@ mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg) int *vlans; size_t i; + /* Set name. */ + if (strcmp(cfg->name, m->name)) { + free(m->name); + m->name = xstrdup(cfg->name); + } + /* Get output port. */ if (cfg->output_port) { out_port = port_lookup(m->bridge, cfg->output_port->name);