1 /* Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
28 #include "dynamic-string.h"
35 #include "ofp-print.h"
37 #include "ofproto/ofproto.h"
38 #include "poll-loop.h"
41 #include "socket-util.h"
42 #include "stream-ssl.h"
44 #include "system-stats.h"
48 #include "vswitchd/vswitch-idl.h"
49 #include "xenserver.h"
51 #include "sflow_api.h"
52 #include "vlan-bitmap.h"
54 VLOG_DEFINE_THIS_MODULE(bridge);
56 COVERAGE_DEFINE(bridge_reconfigure);
59 /* These members are always valid. */
60 struct list port_elem; /* Element in struct port's "ifaces" list. */
61 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
62 struct port *port; /* Containing port. */
63 char *name; /* Host network device name. */
64 tag_type tag; /* Tag associated with this interface. */
66 /* These members are valid only after bridge_reconfigure() causes them to
68 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
69 int ofp_port; /* OpenFlow port number, -1 if unknown. */
70 struct netdev *netdev; /* Network device. */
71 const char *type; /* Usually same as cfg->type. */
72 const struct ovsrec_interface *cfg;
76 struct uuid uuid; /* UUID of this "mirror" record in database. */
77 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
78 struct bridge *bridge;
83 struct bridge *bridge;
84 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
87 const struct ovsrec_port *cfg;
89 /* An ordinary bridge port has 1 interface.
90 * A bridge port for bonding has at least 2 interfaces. */
91 struct list ifaces; /* List of "struct iface"s. */
95 struct hmap_node node; /* In 'all_bridges'. */
96 char *name; /* User-specified arbitrary name. */
97 char *type; /* Datapath type. */
98 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
99 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
100 const struct ovsrec_bridge *cfg;
102 /* OpenFlow switch processing. */
103 struct ofproto *ofproto; /* OpenFlow switch. */
106 struct hmap ports; /* "struct port"s indexed by name. */
107 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
108 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
110 /* Port mirroring. */
111 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
113 /* Synthetic local port if necessary. */
114 struct ovsrec_port synth_local_port;
115 struct ovsrec_interface synth_local_iface;
116 struct ovsrec_interface *synth_local_ifacep;
119 /* All bridges, indexed by name. */
120 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
122 /* OVSDB IDL used to obtain configuration. */
123 static struct ovsdb_idl *idl;
125 /* Each time this timer expires, the bridge fetches systems and interface
126 * statistics and pushes them into the database. */
127 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
128 static long long int stats_timer = LLONG_MIN;
130 /* Stores the time after which rate limited statistics may be written to the
131 * database. Only updated when changes to the database require rate limiting.
133 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
134 static long long int db_limiter = LLONG_MIN;
136 static void add_del_bridges(const struct ovsrec_open_vswitch *);
137 static void bridge_del_ofprotos(void);
138 static bool bridge_add_ofprotos(struct bridge *);
139 static void bridge_create(const struct ovsrec_bridge *);
140 static void bridge_destroy(struct bridge *);
141 static struct bridge *bridge_lookup(const char *name);
142 static unixctl_cb_func bridge_unixctl_dump_flows;
143 static unixctl_cb_func bridge_unixctl_reconnect;
144 static size_t bridge_get_controllers(const struct bridge *br,
145 struct ovsrec_controller ***controllersp);
146 static void bridge_add_del_ports(struct bridge *);
147 static void bridge_add_ofproto_ports(struct bridge *);
148 static void bridge_del_ofproto_ports(struct bridge *);
149 static void bridge_refresh_ofp_port(struct bridge *);
150 static void bridge_configure_datapath_id(struct bridge *);
151 static void bridge_configure_flow_eviction_threshold(struct bridge *);
152 static void bridge_configure_netflow(struct bridge *);
153 static void bridge_configure_forward_bpdu(struct bridge *);
154 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
155 static void bridge_configure_remotes(struct bridge *,
156 const struct sockaddr_in *managers,
158 static void bridge_pick_local_hw_addr(struct bridge *,
159 uint8_t ea[ETH_ADDR_LEN],
160 struct iface **hw_addr_iface);
161 static uint64_t bridge_pick_datapath_id(struct bridge *,
162 const uint8_t bridge_ea[ETH_ADDR_LEN],
163 struct iface *hw_addr_iface);
164 static uint64_t dpid_from_hash(const void *, size_t nbytes);
165 static bool bridge_has_bond_fake_iface(const struct bridge *,
167 static bool port_is_bond_fake_iface(const struct port *);
169 static unixctl_cb_func qos_unixctl_show;
171 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
172 static void port_add_ifaces(struct port *);
173 static void port_del_ifaces(struct port *);
174 static void port_destroy(struct port *);
175 static struct port *port_lookup(const struct bridge *, const char *name);
176 static void port_configure(struct port *);
177 static struct lacp_settings *port_configure_lacp(struct port *,
178 struct lacp_settings *);
179 static void port_configure_bond(struct port *, struct bond_settings *,
180 uint32_t *bond_stable_ids);
182 static void bridge_configure_mirrors(struct bridge *);
183 static struct mirror *mirror_create(struct bridge *,
184 const struct ovsrec_mirror *);
185 static void mirror_destroy(struct mirror *);
186 static bool mirror_configure(struct mirror *, const struct ovsrec_mirror *);
188 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
189 static struct iface *iface_create(struct port *port,
190 const struct ovsrec_interface *if_cfg);
191 static void iface_destroy(struct iface *);
192 static struct iface *iface_lookup(const struct bridge *, const char *name);
193 static struct iface *iface_find(const char *name);
194 static struct iface *iface_from_ofp_port(const struct bridge *,
196 static void iface_set_mac(struct iface *);
197 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
198 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
199 static void iface_configure_cfm(struct iface *);
200 static bool iface_refresh_cfm_stats(struct iface *);
201 static void iface_refresh_stats(struct iface *);
202 static void iface_refresh_status(struct iface *);
203 static bool iface_get_carrier(const struct iface *);
204 static bool iface_is_synthetic(const struct iface *);
206 static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
208 static void shash_to_ovs_idl_map(struct shash *,
209 char ***keys, char ***values, size_t *n);
211 /* Public functions. */
213 /* Initializes the bridge module, configuring it to obtain its configuration
214 * from an OVSDB server accessed over 'remote', which should be a string in a
215 * form acceptable to ovsdb_idl_create(). */
217 bridge_init(const char *remote)
219 /* Create connection to database. */
220 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
221 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
223 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
224 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
225 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
226 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
227 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
228 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
229 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
231 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
232 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
234 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
235 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
237 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
238 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
239 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
240 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
241 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
242 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
243 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
244 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
245 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
247 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
248 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
249 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
250 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
252 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
254 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
256 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
258 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
260 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
262 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
263 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
264 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
265 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
266 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
268 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
270 /* Register unixctl commands. */
271 unixctl_command_register("qos/show", qos_unixctl_show, NULL);
272 unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
274 unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
284 struct bridge *br, *next_br;
286 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
289 ovsdb_idl_destroy(idl);
292 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
293 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
294 * responsible for freeing '*managersp' (with free()).
296 * You may be asking yourself "why does ovs-vswitchd care?", because
297 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
298 * should not be and in fact is not directly involved in that. But
299 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
300 * it has to tell in-band control where the managers are to enable that.
301 * (Thus, only managers connected in-band are collected.)
304 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
305 struct sockaddr_in **managersp, size_t *n_managersp)
307 struct sockaddr_in *managers = NULL;
308 size_t n_managers = 0;
312 /* Collect all of the potential targets from the "targets" columns of the
313 * rows pointed to by "manager_options", excluding any that are
316 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
317 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
319 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
320 sset_find_and_delete(&targets, m->target);
322 sset_add(&targets, m->target);
326 /* Now extract the targets' IP addresses. */
327 if (!sset_is_empty(&targets)) {
330 managers = xmalloc(sset_count(&targets) * sizeof *managers);
331 SSET_FOR_EACH (target, &targets) {
332 struct sockaddr_in *sin = &managers[n_managers];
334 if ((!strncmp(target, "tcp:", 4)
335 && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) ||
336 (!strncmp(target, "ssl:", 4)
337 && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) {
342 sset_destroy(&targets);
344 *managersp = managers;
345 *n_managersp = n_managers;
349 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
351 struct sockaddr_in *managers;
352 struct bridge *br, *next;
353 int sflow_bridge_number;
356 COVERAGE_INC(bridge_reconfigure);
358 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
359 * iface"s according to 'ovs_cfg', with only very minimal configuration
362 * This is purely an update to bridge data structures. Nothing is pushed
363 * down to ofproto or lower layers. */
364 add_del_bridges(ovs_cfg);
365 HMAP_FOR_EACH (br, node, &all_bridges) {
366 bridge_add_del_ports(br);
369 /* Delete all datapaths and datapath ports that are no longer configured.
371 * The kernel will reject any attempt to add a given port to a datapath if
372 * that port already belongs to a different datapath, so we must do all
373 * port deletions before any port additions. A datapath always has a
374 * "local port" so we must delete not-configured datapaths too. */
375 bridge_del_ofprotos();
376 HMAP_FOR_EACH (br, node, &all_bridges) {
378 bridge_del_ofproto_ports(br);
382 /* Create datapaths and datapath ports that are missing.
384 * After this is done, we have our final set of bridges, ports, and
385 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
386 * has at least one iface, every "struct iface" has a valid ofp_port and
388 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
390 if (bridge_add_ofprotos(br)) {
391 bridge_del_ofproto_ports(br);
397 HMAP_FOR_EACH (br, node, &all_bridges) {
398 bridge_refresh_ofp_port(br);
399 bridge_add_ofproto_ports(br);
402 /* Complete the configuration. */
403 sflow_bridge_number = 0;
404 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
405 HMAP_FOR_EACH (br, node, &all_bridges) {
408 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
411 port_configure(port);
413 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
414 iface_configure_cfm(iface);
415 iface_configure_qos(iface, port->cfg->qos);
416 iface_set_mac(iface);
419 bridge_configure_mirrors(br);
420 bridge_configure_datapath_id(br);
421 bridge_configure_flow_eviction_threshold(br);
422 bridge_configure_forward_bpdu(br);
423 bridge_configure_remotes(br, managers, n_managers);
424 bridge_configure_netflow(br);
425 bridge_configure_sflow(br, &sflow_bridge_number);
429 /* ovs-vswitchd has completed initialization, so allow the process that
430 * forked us to exit successfully. */
431 daemonize_complete();
434 /* Iterate over all ofprotos and delete any of them that do not have a
435 * configured bridge or that are the wrong type. */
437 bridge_del_ofprotos(void)
445 ofproto_enumerate_types(&types);
446 SSET_FOR_EACH (type, &types) {
449 ofproto_enumerate_names(type, &names);
450 SSET_FOR_EACH (name, &names) {
451 struct bridge *br = bridge_lookup(name);
452 if (!br || strcmp(type, br->type)) {
453 ofproto_delete(name, type);
457 sset_destroy(&names);
458 sset_destroy(&types);
462 bridge_add_ofprotos(struct bridge *br)
464 int error = ofproto_create(br->name, br->type, &br->ofproto);
466 VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
473 port_configure(struct port *port)
475 const struct ovsrec_port *cfg = port->cfg;
476 struct bond_settings bond_settings;
477 struct lacp_settings lacp_settings;
478 struct ofproto_bundle_settings s;
486 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
487 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
488 s.slaves[s.n_slaves++] = iface->ofp_port;
494 if (list_is_short(&port->ifaces)) {
495 if (*cfg->tag >= 0 && *cfg->tag <= 4095) {
497 VLOG_DBG("port %s: assigning VLAN tag %d", port->name, s.vlan);
500 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
501 * they even work as-is. But they have not been tested. */
502 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
507 /* Get VLAN trunks. */
509 if (s.vlan < 0 && cfg->n_trunks) {
510 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
511 } else if (s.vlan >= 0 && cfg->n_trunks) {
512 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
516 /* Get LACP settings. */
517 s.lacp = port_configure_lacp(port, &lacp_settings);
521 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
522 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
523 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
526 s.lacp_slaves = NULL;
529 /* Get bond settings. */
530 if (s.n_slaves > 1) {
531 s.bond = &bond_settings;
532 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
533 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
536 s.bond_stable_ids = NULL;
538 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
539 netdev_set_miimon_interval(iface->netdev, 0);
544 ofproto_bundle_register(port->bridge->ofproto, port, &s);
550 free(s.bond_stable_ids);
553 /* Pick local port hardware address and datapath ID for 'br'. */
555 bridge_configure_datapath_id(struct bridge *br)
557 uint8_t ea[ETH_ADDR_LEN];
559 struct iface *local_iface;
560 struct iface *hw_addr_iface;
563 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
564 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
566 int error = netdev_set_etheraddr(local_iface->netdev, ea);
568 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
569 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
570 "Ethernet address: %s",
571 br->name, strerror(error));
574 memcpy(br->ea, ea, ETH_ADDR_LEN);
576 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
577 ofproto_set_datapath_id(br->ofproto, dpid);
579 dpid_string = xasprintf("%016"PRIx64, dpid);
580 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
584 /* Set NetFlow configuration on 'br'. */
586 bridge_configure_netflow(struct bridge *br)
588 struct ovsrec_netflow *cfg = br->cfg->netflow;
589 struct netflow_options opts;
592 ofproto_set_netflow(br->ofproto, NULL);
596 memset(&opts, 0, sizeof opts);
598 /* Get default NetFlow configuration from datapath.
599 * Apply overrides from 'cfg'. */
600 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
601 if (cfg->engine_type) {
602 opts.engine_type = *cfg->engine_type;
604 if (cfg->engine_id) {
605 opts.engine_id = *cfg->engine_id;
608 /* Configure active timeout interval. */
609 opts.active_timeout = cfg->active_timeout;
610 if (!opts.active_timeout) {
611 opts.active_timeout = -1;
612 } else if (opts.active_timeout < 0) {
613 VLOG_WARN("bridge %s: active timeout interval set to negative "
614 "value, using default instead (%d seconds)", br->name,
615 NF_ACTIVE_TIMEOUT_DEFAULT);
616 opts.active_timeout = -1;
619 /* Add engine ID to interface number to disambiguate bridgs? */
620 opts.add_id_to_iface = cfg->add_id_to_interface;
621 if (opts.add_id_to_iface) {
622 if (opts.engine_id > 0x7f) {
623 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
624 "another vswitch, choose an engine id less than 128",
627 if (hmap_count(&br->ports) > 508) {
628 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
629 "another port when more than 508 ports are used",
635 sset_init(&opts.collectors);
636 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
639 if (ofproto_set_netflow(br->ofproto, &opts)) {
640 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
642 sset_destroy(&opts.collectors);
645 /* Set sFlow configuration on 'br'. */
647 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
649 const struct ovsrec_sflow *cfg = br->cfg->sflow;
650 struct ovsrec_controller **controllers;
651 struct ofproto_sflow_options oso;
652 size_t n_controllers;
656 ofproto_set_sflow(br->ofproto, NULL);
660 memset(&oso, 0, sizeof oso);
662 sset_init(&oso.targets);
663 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
665 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
667 oso.sampling_rate = *cfg->sampling;
670 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
672 oso.polling_interval = *cfg->polling;
675 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
677 oso.header_len = *cfg->header;
680 oso.sub_id = (*sflow_bridge_number)++;
681 oso.agent_device = cfg->agent;
683 oso.control_ip = NULL;
684 n_controllers = bridge_get_controllers(br, &controllers);
685 for (i = 0; i < n_controllers; i++) {
686 if (controllers[i]->local_ip) {
687 oso.control_ip = controllers[i]->local_ip;
691 ofproto_set_sflow(br->ofproto, &oso);
693 sset_destroy(&oso.targets);
697 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
699 const struct port *port = port_lookup(br, name);
700 return port && port_is_bond_fake_iface(port);
704 port_is_bond_fake_iface(const struct port *port)
706 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
710 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
712 struct bridge *br, *next;
716 /* Collect new bridges' names and types. */
718 for (i = 0; i < cfg->n_bridges; i++) {
719 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
720 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
722 if (strchr(br_cfg->name, '/')) {
723 /* Prevent remote ovsdb-server users from accessing arbitrary
724 * directories, e.g. consider a bridge named "../../../etc/". */
725 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
727 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
728 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
732 /* Get rid of deleted bridges or those whose types have changed.
733 * Update 'cfg' of bridges that still exist. */
734 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
735 br->cfg = shash_find_data(&new_br, br->name);
736 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
737 br->cfg->datapath_type))) {
742 /* Add new bridges. */
743 for (i = 0; i < cfg->n_bridges; i++) {
744 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
745 struct bridge *br = bridge_lookup(br_cfg->name);
747 bridge_create(br_cfg);
751 shash_destroy(&new_br);
754 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
757 * The kernel will reject any attempt to add a given port to a datapath if that
758 * port already belongs to a different datapath, so we must do all port
759 * deletions before any port additions. */
761 bridge_del_ofproto_ports(struct bridge *br)
763 struct ofproto_port_dump dump;
764 struct ofproto_port ofproto_port;
766 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
767 const char *name = ofproto_port.name;
772 /* Ignore the local port. We can't change it anyhow. */
773 if (!strcmp(name, br->name)) {
777 /* Get the type that 'ofproto_port' should have (ordinarily the
778 * type of its corresponding iface) or NULL if it should be
780 iface = iface_lookup(br, name);
781 type = (iface ? iface->type
782 : bridge_has_bond_fake_iface(br, name) ? "internal"
785 /* If it's the wrong type then delete the ofproto port. */
787 && !strcmp(ofproto_port.type, type)
788 && (!iface || !iface->netdev
789 || !strcmp(netdev_get_type(iface->netdev), type))) {
792 error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
794 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
795 br->name, name, strerror(error));
798 netdev_close(iface->netdev);
799 iface->netdev = NULL;
805 iface_set_ofp_port(struct iface *iface, int ofp_port)
807 struct bridge *br = iface->port->bridge;
809 assert(iface->ofp_port < 0 && ofp_port >= 0);
810 iface->ofp_port = ofp_port;
811 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
812 iface_set_ofport(iface->cfg, ofp_port);
816 bridge_refresh_ofp_port(struct bridge *br)
818 struct ofproto_port_dump dump;
819 struct ofproto_port ofproto_port;
822 /* Clear all the "ofp_port"es. */
823 hmap_clear(&br->ifaces);
824 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
827 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
828 iface->ofp_port = -1;
832 /* Obtain the correct "ofp_port"s from ofproto. */
833 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
834 struct iface *iface = iface_lookup(br, ofproto_port.name);
836 if (iface->ofp_port >= 0) {
837 VLOG_WARN("bridge %s: interface %s reported twice",
838 br->name, ofproto_port.name);
839 } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
840 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
841 br->name, ofproto_port.ofp_port);
843 iface_set_ofp_port(iface, ofproto_port.ofp_port);
849 /* Add an ofproto port for any "struct iface" that doesn't have one.
850 * Delete any "struct iface" for which this fails.
851 * Delete any "struct port" that thereby ends up with no ifaces. */
853 bridge_add_ofproto_ports(struct bridge *br)
855 struct port *port, *next_port;
857 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
858 struct iface *iface, *next_iface;
859 struct ofproto_port ofproto_port;
861 LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
865 /* Open the netdev or reconfigure it. */
867 shash_from_ovs_idl_map(iface->cfg->key_options,
868 iface->cfg->value_options,
869 iface->cfg->n_options, &args);
870 if (!iface->netdev) {
871 struct netdev_options options;
872 options.name = iface->name;
873 options.type = iface->type;
874 options.args = &args;
875 options.ethertype = NETDEV_ETH_TYPE_NONE;
876 error = netdev_open(&options, &iface->netdev);
878 error = netdev_set_config(iface->netdev, &args);
880 shash_destroy(&args);
882 VLOG_WARN("could not %s network device %s (%s)",
883 iface->netdev ? "reconfigure" : "open",
884 iface->name, strerror(error));
887 /* Add the port, if necessary. */
888 if (iface->netdev && iface->ofp_port < 0) {
892 error = ofproto_port_add(br->ofproto, iface->netdev,
895 iface_set_ofp_port(iface, ofp_port);
897 netdev_close(iface->netdev);
898 iface->netdev = NULL;
902 /* Populate stats columns in new Interface rows. */
903 if (iface->netdev && !iface->cfg->mtu) {
904 iface_refresh_stats(iface);
905 iface_refresh_status(iface);
908 /* Delete the iface if */
909 if (iface->netdev && iface->ofp_port >= 0) {
910 VLOG_DBG("bridge %s: interface %s is on port %d",
911 br->name, iface->name, iface->ofp_port);
914 VLOG_ERR("bridge %s: missing %s interface, dropping",
915 br->name, iface->name);
917 /* We already reported a related error, don't bother
920 iface_set_ofport(iface->cfg, -1);
921 iface_destroy(iface);
924 if (list_is_empty(&port->ifaces)) {
925 VLOG_WARN("%s port has no interfaces, dropping", port->name);
930 /* Add bond fake iface if necessary. */
931 if (port_is_bond_fake_iface(port)) {
932 if (ofproto_port_query_by_name(br->ofproto, port->name,
934 struct netdev_options options;
935 struct netdev *netdev;
938 options.name = port->name;
939 options.type = "internal";
941 options.ethertype = NETDEV_ETH_TYPE_NONE;
942 error = netdev_open(&options, &netdev);
944 ofproto_port_add(br->ofproto, netdev, NULL);
945 netdev_close(netdev);
947 VLOG_WARN("could not open network device %s (%s)",
948 port->name, strerror(error));
951 /* Already exists, nothing to do. */
952 ofproto_port_destroy(&ofproto_port);
959 get_ovsrec_key_value(const struct ovsdb_idl_row *row,
960 const struct ovsdb_idl_column *column,
963 const struct ovsdb_datum *datum;
964 union ovsdb_atom atom;
967 datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
968 atom.string = (char *) key;
969 idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
970 return idx == UINT_MAX ? NULL : datum->values[idx].string;
974 bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
976 return get_ovsrec_key_value(&br_cfg->header_,
977 &ovsrec_bridge_col_other_config, key);
980 /* Set Flow eviction threshold */
982 bridge_configure_flow_eviction_threshold(struct bridge *br)
984 const char *threshold_str;
987 threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold");
989 threshold = strtoul(threshold_str, NULL, 10);
991 threshold = OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT;
993 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
996 /* Set forward BPDU option. */
998 bridge_configure_forward_bpdu(struct bridge *br)
1000 const char *forward_bpdu_str;
1001 bool forward_bpdu = false;
1003 forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu");
1004 if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) {
1005 forward_bpdu = true;
1007 ofproto_set_forward_bpdu(br->ofproto, forward_bpdu);
1011 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1012 struct iface **hw_addr_iface)
1018 *hw_addr_iface = NULL;
1020 /* Did the user request a particular MAC? */
1021 hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
1022 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1023 if (eth_addr_is_multicast(ea)) {
1024 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1025 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1026 } else if (eth_addr_is_zero(ea)) {
1027 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1033 /* Otherwise choose the minimum non-local MAC address among all of the
1035 memset(ea, 0xff, ETH_ADDR_LEN);
1036 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1037 uint8_t iface_ea[ETH_ADDR_LEN];
1038 struct iface *candidate;
1039 struct iface *iface;
1041 /* Mirror output ports don't participate. */
1042 if (ofproto_is_mirror_output_bundle(br->ofproto, port)) {
1046 /* Choose the MAC address to represent the port. */
1048 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1049 /* Find the interface with this Ethernet address (if any) so that
1050 * we can provide the correct devname to the caller. */
1051 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1052 uint8_t candidate_ea[ETH_ADDR_LEN];
1053 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1054 && eth_addr_equals(iface_ea, candidate_ea)) {
1059 /* Choose the interface whose MAC address will represent the port.
1060 * The Linux kernel bonding code always chooses the MAC address of
1061 * the first slave added to a bond, and the Fedora networking
1062 * scripts always add slaves to a bond in alphabetical order, so
1063 * for compatibility we choose the interface with the name that is
1064 * first in alphabetical order. */
1065 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1066 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1071 /* The local port doesn't count (since we're trying to choose its
1072 * MAC address anyway). */
1073 if (iface->ofp_port == OFPP_LOCAL) {
1078 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1084 /* Compare against our current choice. */
1085 if (!eth_addr_is_multicast(iface_ea) &&
1086 !eth_addr_is_local(iface_ea) &&
1087 !eth_addr_is_reserved(iface_ea) &&
1088 !eth_addr_is_zero(iface_ea) &&
1089 eth_addr_compare_3way(iface_ea, ea) < 0)
1091 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1092 *hw_addr_iface = iface;
1095 if (eth_addr_is_multicast(ea)) {
1096 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1097 *hw_addr_iface = NULL;
1098 VLOG_WARN("bridge %s: using default bridge Ethernet "
1099 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1101 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
1102 br->name, ETH_ADDR_ARGS(ea));
1106 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1107 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1108 * an interface on 'br', then that interface must be passed in as
1109 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1110 * 'hw_addr_iface' must be passed in as a null pointer. */
1112 bridge_pick_datapath_id(struct bridge *br,
1113 const uint8_t bridge_ea[ETH_ADDR_LEN],
1114 struct iface *hw_addr_iface)
1117 * The procedure for choosing a bridge MAC address will, in the most
1118 * ordinary case, also choose a unique MAC that we can use as a datapath
1119 * ID. In some special cases, though, multiple bridges will end up with
1120 * the same MAC address. This is OK for the bridges, but it will confuse
1121 * the OpenFlow controller, because each datapath needs a unique datapath
1124 * Datapath IDs must be unique. It is also very desirable that they be
1125 * stable from one run to the next, so that policy set on a datapath
1128 const char *datapath_id;
1131 datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
1132 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1136 if (hw_addr_iface) {
1138 if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
1140 * A bridge whose MAC address is taken from a VLAN network device
1141 * (that is, a network device created with vconfig(8) or similar
1142 * tool) will have the same MAC address as a bridge on the VLAN
1143 * device's physical network device.
1145 * Handle this case by hashing the physical network device MAC
1146 * along with the VLAN identifier.
1148 uint8_t buf[ETH_ADDR_LEN + 2];
1149 memcpy(buf, bridge_ea, ETH_ADDR_LEN);
1150 buf[ETH_ADDR_LEN] = vlan >> 8;
1151 buf[ETH_ADDR_LEN + 1] = vlan;
1152 return dpid_from_hash(buf, sizeof buf);
1155 * Assume that this bridge's MAC address is unique, since it
1156 * doesn't fit any of the cases we handle specially.
1161 * A purely internal bridge, that is, one that has no non-virtual
1162 * network devices on it at all, is more difficult because it has no
1163 * natural unique identifier at all.
1165 * When the host is a XenServer, we handle this case by hashing the
1166 * host's UUID with the name of the bridge. Names of bridges are
1167 * persistent across XenServer reboots, although they can be reused if
1168 * an internal network is destroyed and then a new one is later
1169 * created, so this is fairly effective.
1171 * When the host is not a XenServer, we punt by using a random MAC
1172 * address on each run.
1174 const char *host_uuid = xenserver_get_host_uuid();
1176 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1177 dpid = dpid_from_hash(combined, strlen(combined));
1183 return eth_addr_to_uint64(bridge_ea);
1187 dpid_from_hash(const void *data, size_t n)
1189 uint8_t hash[SHA1_DIGEST_SIZE];
1191 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1192 sha1_bytes(data, n, hash);
1193 eth_addr_mark_random(hash);
1194 return eth_addr_to_uint64(hash);
1198 iface_refresh_status(struct iface *iface)
1202 enum netdev_flags flags;
1209 if (iface_is_synthetic(iface)) {
1215 if (!netdev_get_status(iface->netdev, &sh)) {
1217 char **keys, **values;
1219 shash_to_ovs_idl_map(&sh, &keys, &values, &n);
1220 ovsrec_interface_set_status(iface->cfg, keys, values, n);
1225 ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
1228 shash_destroy_free_data(&sh);
1230 error = netdev_get_flags(iface->netdev, &flags);
1232 ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
1235 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1238 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1240 ovsrec_interface_set_duplex(iface->cfg,
1241 netdev_features_is_full_duplex(current)
1243 /* warning: uint64_t -> int64_t conversion */
1244 bps = netdev_features_to_bps(current);
1245 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1248 ovsrec_interface_set_duplex(iface->cfg, NULL);
1249 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1252 ovsrec_interface_set_link_state(iface->cfg,
1253 iface_get_carrier(iface) ? "up" : "down");
1255 error = netdev_get_mtu(iface->netdev, &mtu);
1256 if (!error && mtu != INT_MAX) {
1258 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1261 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1265 /* Writes 'iface''s CFM statistics to the database. Returns true if anything
1266 * changed, false otherwise. */
1268 iface_refresh_cfm_stats(struct iface *iface)
1270 const struct ovsrec_interface *cfg = iface->cfg;
1271 bool changed = false;
1274 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1281 if (cfg->n_cfm_fault != 1 || cfg->cfm_fault[0] != fault) {
1282 bool fault_bool = fault;
1283 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1291 iface_refresh_lacp_stats(struct iface *iface)
1293 struct ofproto *ofproto = iface->port->bridge->ofproto;
1294 int old = iface->cfg->lacp_current ? *iface->cfg->lacp_current : -1;
1295 int new = ofproto_port_is_lacp_current(ofproto, iface->ofp_port);
1299 ovsrec_interface_set_lacp_current(iface->cfg, ¤t, new >= 0);
1305 iface_refresh_stats(struct iface *iface)
1307 #define IFACE_STATS \
1308 IFACE_STAT(rx_packets, "rx_packets") \
1309 IFACE_STAT(tx_packets, "tx_packets") \
1310 IFACE_STAT(rx_bytes, "rx_bytes") \
1311 IFACE_STAT(tx_bytes, "tx_bytes") \
1312 IFACE_STAT(rx_dropped, "rx_dropped") \
1313 IFACE_STAT(tx_dropped, "tx_dropped") \
1314 IFACE_STAT(rx_errors, "rx_errors") \
1315 IFACE_STAT(tx_errors, "tx_errors") \
1316 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1317 IFACE_STAT(rx_over_errors, "rx_over_err") \
1318 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1319 IFACE_STAT(collisions, "collisions")
1321 #define IFACE_STAT(MEMBER, NAME) NAME,
1322 static char *keys[] = { IFACE_STATS };
1324 int64_t values[ARRAY_SIZE(keys)];
1327 struct netdev_stats stats;
1329 if (iface_is_synthetic(iface)) {
1333 /* Intentionally ignore return value, since errors will set 'stats' to
1334 * all-1s, and we will deal with that correctly below. */
1335 netdev_get_stats(iface->netdev, &stats);
1337 /* Copy statistics into values[] array. */
1339 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1342 assert(i == ARRAY_SIZE(keys));
1344 ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
1349 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1353 /* Use other-config:enable-system-stats by preference. */
1354 enable = get_ovsrec_key_value(&cfg->header_,
1355 &ovsrec_open_vswitch_col_other_config,
1356 "enable-statistics");
1358 return !strcmp(enable, "true");
1361 /* Disable by default. */
1366 refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
1368 struct ovsdb_datum datum;
1372 if (enable_system_stats(cfg)) {
1373 get_system_stats(&stats);
1376 ovsdb_datum_from_shash(&datum, &stats);
1377 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1381 static inline const char *
1382 nx_role_to_str(enum nx_role role)
1387 case NX_ROLE_MASTER:
1392 return "*** INVALID ROLE ***";
1397 refresh_controller_status(void)
1401 const struct ovsrec_controller *cfg;
1405 /* Accumulate status for controllers on all bridges. */
1406 HMAP_FOR_EACH (br, node, &all_bridges) {
1407 ofproto_get_ofproto_controller_info(br->ofproto, &info);
1410 /* Update each controller in the database with current status. */
1411 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
1412 struct ofproto_controller_info *cinfo =
1413 shash_find_data(&info, cfg->target);
1416 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
1417 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
1418 ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
1419 (char **) cinfo->pairs.values,
1422 ovsrec_controller_set_is_connected(cfg, false);
1423 ovsrec_controller_set_role(cfg, NULL);
1424 ovsrec_controller_set_status(cfg, NULL, NULL, 0);
1428 ofproto_free_ofproto_controller_info(&info);
1434 const struct ovsrec_open_vswitch *cfg;
1436 bool datapath_destroyed;
1437 bool database_changed;
1440 /* (Re)configure if necessary. */
1441 database_changed = ovsdb_idl_run(idl);
1442 if (ovsdb_idl_is_lock_contended(idl)) {
1443 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
1444 struct bridge *br, *next_br;
1446 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
1447 "disabling this process until it goes away");
1449 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
1453 } else if (!ovsdb_idl_has_lock(idl)) {
1456 cfg = ovsrec_open_vswitch_first(idl);
1458 /* Let each bridge do the work that it needs to do. */
1459 datapath_destroyed = false;
1460 HMAP_FOR_EACH (br, node, &all_bridges) {
1461 int error = ofproto_run(br->ofproto);
1463 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1464 VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
1465 "forcing reconfiguration", br->name);
1466 datapath_destroyed = true;
1470 /* Re-configure SSL. We do this on every trip through the main loop,
1471 * instead of just when the database changes, because the contents of the
1472 * key and certificate files can change without the database changing.
1474 * We do this before bridge_reconfigure() because that function might
1475 * initiate SSL connections and thus requires SSL to be configured. */
1476 if (cfg && cfg->ssl) {
1477 const struct ovsrec_ssl *ssl = cfg->ssl;
1479 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
1480 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
1483 if (database_changed || datapath_destroyed) {
1485 struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
1487 bridge_reconfigure(cfg);
1489 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
1490 ovsdb_idl_txn_commit(txn);
1491 ovsdb_idl_txn_destroy(txn); /* XXX */
1493 /* We still need to reconfigure to avoid dangling pointers to
1494 * now-destroyed ovsrec structures inside bridge data. */
1495 static const struct ovsrec_open_vswitch null_cfg;
1497 bridge_reconfigure(&null_cfg);
1501 /* Refresh system and interface stats if necessary. */
1502 if (time_msec() >= stats_timer) {
1504 struct ovsdb_idl_txn *txn;
1506 txn = ovsdb_idl_txn_create(idl);
1507 HMAP_FOR_EACH (br, node, &all_bridges) {
1510 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1511 struct iface *iface;
1513 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1514 iface_refresh_stats(iface);
1515 iface_refresh_status(iface);
1519 refresh_system_stats(cfg);
1520 refresh_controller_status();
1521 ovsdb_idl_txn_commit(txn);
1522 ovsdb_idl_txn_destroy(txn); /* XXX */
1525 stats_timer = time_msec() + STATS_INTERVAL;
1528 if (time_msec() >= db_limiter) {
1529 struct ovsdb_idl_txn *txn;
1530 bool changed = false;
1532 txn = ovsdb_idl_txn_create(idl);
1533 HMAP_FOR_EACH (br, node, &all_bridges) {
1536 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1537 struct iface *iface;
1539 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1540 changed = iface_refresh_cfm_stats(iface) || changed;
1541 changed = iface_refresh_lacp_stats(iface) || changed;
1547 db_limiter = time_msec() + DB_LIMIT_INTERVAL;
1550 ovsdb_idl_txn_commit(txn);
1551 ovsdb_idl_txn_destroy(txn);
1558 ovsdb_idl_wait(idl);
1559 if (!hmap_is_empty(&all_bridges)) {
1562 HMAP_FOR_EACH (br, node, &all_bridges) {
1563 ofproto_wait(br->ofproto);
1565 poll_timer_wait_until(stats_timer);
1567 if (db_limiter > time_msec()) {
1568 poll_timer_wait_until(db_limiter);
1573 /* QoS unixctl user interface functions. */
1575 struct qos_unixctl_show_cbdata {
1577 struct iface *iface;
1581 qos_unixctl_show_cb(unsigned int queue_id,
1582 const struct shash *details,
1585 struct qos_unixctl_show_cbdata *data = aux;
1586 struct ds *ds = data->ds;
1587 struct iface *iface = data->iface;
1588 struct netdev_queue_stats stats;
1589 struct shash_node *node;
1592 ds_put_cstr(ds, "\n");
1594 ds_put_format(ds, "Queue %u:\n", queue_id);
1596 ds_put_cstr(ds, "Default:\n");
1599 SHASH_FOR_EACH (node, details) {
1600 ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
1603 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
1605 if (stats.tx_packets != UINT64_MAX) {
1606 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
1609 if (stats.tx_bytes != UINT64_MAX) {
1610 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
1613 if (stats.tx_errors != UINT64_MAX) {
1614 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
1617 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
1618 queue_id, strerror(error));
1623 qos_unixctl_show(struct unixctl_conn *conn,
1624 const char *args, void *aux OVS_UNUSED)
1626 struct ds ds = DS_EMPTY_INITIALIZER;
1627 struct shash sh = SHASH_INITIALIZER(&sh);
1628 struct iface *iface;
1630 struct shash_node *node;
1631 struct qos_unixctl_show_cbdata data;
1634 iface = iface_find(args);
1636 unixctl_command_reply(conn, 501, "no such interface");
1640 netdev_get_qos(iface->netdev, &type, &sh);
1642 if (*type != '\0') {
1643 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
1645 SHASH_FOR_EACH (node, &sh) {
1646 ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
1651 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
1654 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
1656 unixctl_command_reply(conn, 200, ds_cstr(&ds));
1658 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
1659 unixctl_command_reply(conn, 501, ds_cstr(&ds));
1662 shash_destroy_free_data(&sh);
1666 /* Bridge reconfiguration functions. */
1668 bridge_create(const struct ovsrec_bridge *br_cfg)
1672 assert(!bridge_lookup(br_cfg->name));
1673 br = xzalloc(sizeof *br);
1675 br->name = xstrdup(br_cfg->name);
1676 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
1679 /* Derive the default Ethernet address from the bridge's UUID. This should
1680 * be unique and it will be stable between ovs-vswitchd runs. */
1681 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
1682 eth_addr_mark_random(br->default_ea);
1684 hmap_init(&br->ports);
1685 hmap_init(&br->ifaces);
1686 hmap_init(&br->iface_by_name);
1687 hmap_init(&br->mirrors);
1689 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
1693 bridge_destroy(struct bridge *br)
1696 struct mirror *mirror, *next_mirror;
1697 struct port *port, *next_port;
1699 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
1702 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
1703 mirror_destroy(mirror);
1705 hmap_remove(&all_bridges, &br->node);
1706 ofproto_destroy(br->ofproto);
1707 hmap_destroy(&br->ifaces);
1708 hmap_destroy(&br->ports);
1709 hmap_destroy(&br->iface_by_name);
1710 hmap_destroy(&br->mirrors);
1717 static struct bridge *
1718 bridge_lookup(const char *name)
1722 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
1723 if (!strcmp(br->name, name)) {
1730 /* Handle requests for a listing of all flows known by the OpenFlow
1731 * stack, including those normally hidden. */
1733 bridge_unixctl_dump_flows(struct unixctl_conn *conn,
1734 const char *args, void *aux OVS_UNUSED)
1739 br = bridge_lookup(args);
1741 unixctl_command_reply(conn, 501, "Unknown bridge");
1746 ofproto_get_all_flows(br->ofproto, &results);
1748 unixctl_command_reply(conn, 200, ds_cstr(&results));
1749 ds_destroy(&results);
1752 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1753 * connections and reconnect. If BRIDGE is not specified, then all bridges
1754 * drop their controller connections and reconnect. */
1756 bridge_unixctl_reconnect(struct unixctl_conn *conn,
1757 const char *args, void *aux OVS_UNUSED)
1760 if (args[0] != '\0') {
1761 br = bridge_lookup(args);
1763 unixctl_command_reply(conn, 501, "Unknown bridge");
1766 ofproto_reconnect_controllers(br->ofproto);
1768 HMAP_FOR_EACH (br, node, &all_bridges) {
1769 ofproto_reconnect_controllers(br->ofproto);
1772 unixctl_command_reply(conn, 200, NULL);
1776 bridge_get_controllers(const struct bridge *br,
1777 struct ovsrec_controller ***controllersp)
1779 struct ovsrec_controller **controllers;
1780 size_t n_controllers;
1782 controllers = br->cfg->controller;
1783 n_controllers = br->cfg->n_controller;
1785 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
1791 *controllersp = controllers;
1793 return n_controllers;
1796 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
1797 * those configured in 'br->cfg'. */
1799 bridge_add_del_ports(struct bridge *br)
1801 struct port *port, *next;
1802 struct shash_node *node;
1803 struct shash new_ports;
1806 /* Collect new ports. */
1807 shash_init(&new_ports);
1808 for (i = 0; i < br->cfg->n_ports; i++) {
1809 const char *name = br->cfg->ports[i]->name;
1810 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
1811 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1815 if (bridge_get_controllers(br, NULL)
1816 && !shash_find(&new_ports, br->name)) {
1817 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1818 br->name, br->name);
1820 br->synth_local_port.interfaces = &br->synth_local_ifacep;
1821 br->synth_local_port.n_interfaces = 1;
1822 br->synth_local_port.name = br->name;
1824 br->synth_local_iface.name = br->name;
1825 br->synth_local_iface.type = "internal";
1827 br->synth_local_ifacep = &br->synth_local_iface;
1829 shash_add(&new_ports, br->name, &br->synth_local_port);
1832 /* Get rid of deleted ports.
1833 * Get rid of deleted interfaces on ports that still exist.
1834 * Update 'cfg' of ports that still exist. */
1835 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
1836 port->cfg = shash_find_data(&new_ports, port->name);
1840 port_del_ifaces(port);
1844 /* Create new ports.
1845 * Add new interfaces to existing ports. */
1846 SHASH_FOR_EACH (node, &new_ports) {
1847 struct port *port = port_lookup(br, node->name);
1849 struct ovsrec_port *cfg = node->data;
1850 port = port_create(br, cfg);
1852 port_add_ifaces(port);
1853 if (list_is_empty(&port->ifaces)) {
1854 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1855 br->name, port->name);
1859 shash_destroy(&new_ports);
1862 /* Initializes 'oc' appropriately as a management service controller for
1865 * The caller must free oc->target when it is no longer needed. */
1867 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
1868 struct ofproto_controller *oc)
1870 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
1871 oc->max_backoff = 0;
1872 oc->probe_interval = 60;
1873 oc->band = OFPROTO_OUT_OF_BAND;
1875 oc->burst_limit = 0;
1878 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1880 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
1881 struct ofproto_controller *oc)
1883 oc->target = c->target;
1884 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
1885 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
1886 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
1887 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
1888 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
1889 oc->burst_limit = (c->controller_burst_limit
1890 ? *c->controller_burst_limit : 0);
1893 /* Configures the IP stack for 'br''s local interface properly according to the
1894 * configuration in 'c'. */
1896 bridge_configure_local_iface_netdev(struct bridge *br,
1897 struct ovsrec_controller *c)
1899 struct netdev *netdev;
1900 struct in_addr mask, gateway;
1902 struct iface *local_iface;
1905 /* If there's no local interface or no IP address, give up. */
1906 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1907 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
1911 /* Bring up the local interface. */
1912 netdev = local_iface->netdev;
1913 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1915 /* Configure the IP address and netmask. */
1916 if (!c->local_netmask
1917 || !inet_aton(c->local_netmask, &mask)
1919 mask.s_addr = guess_netmask(ip.s_addr);
1921 if (!netdev_set_in4(netdev, ip, mask)) {
1922 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
1923 br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
1926 /* Configure the default gateway. */
1927 if (c->local_gateway
1928 && inet_aton(c->local_gateway, &gateway)
1929 && gateway.s_addr) {
1930 if (!netdev_add_router(netdev, gateway)) {
1931 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
1932 br->name, IP_ARGS(&gateway.s_addr));
1938 bridge_configure_remotes(struct bridge *br,
1939 const struct sockaddr_in *managers, size_t n_managers)
1941 const char *disable_ib_str, *queue_id_str;
1942 bool disable_in_band = false;
1945 struct ovsrec_controller **controllers;
1946 size_t n_controllers;
1948 enum ofproto_fail_mode fail_mode;
1950 struct ofproto_controller *ocs;
1954 /* Check if we should disable in-band control on this bridge. */
1955 disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band");
1956 if (disable_ib_str && !strcmp(disable_ib_str, "true")) {
1957 disable_in_band = true;
1960 /* Set OpenFlow queue ID for in-band control. */
1961 queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue");
1962 queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
1963 ofproto_set_in_band_queue(br->ofproto, queue_id);
1965 if (disable_in_band) {
1966 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
1968 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
1971 n_controllers = bridge_get_controllers(br, &controllers);
1973 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
1976 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
1977 for (i = 0; i < n_controllers; i++) {
1978 struct ovsrec_controller *c = controllers[i];
1980 if (!strncmp(c->target, "punix:", 6)
1981 || !strncmp(c->target, "unix:", 5)) {
1982 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1984 /* Prevent remote ovsdb-server users from accessing arbitrary Unix
1985 * domain sockets and overwriting arbitrary local files. */
1986 VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket "
1987 "controller \"%s\" due to possibility for remote "
1988 "exploit", br->name, c->target);
1992 bridge_configure_local_iface_netdev(br, c);
1993 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
1994 if (disable_in_band) {
1995 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2000 ofproto_set_controllers(br->ofproto, ocs, n_ocs);
2001 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2004 /* Set the fail-mode. */
2005 fail_mode = !br->cfg->fail_mode
2006 || !strcmp(br->cfg->fail_mode, "standalone")
2007 ? OFPROTO_FAIL_STANDALONE
2008 : OFPROTO_FAIL_SECURE;
2009 ofproto_set_fail_mode(br->ofproto, fail_mode);
2011 /* Configure OpenFlow controller connection snooping. */
2012 if (!ofproto_has_snoops(br->ofproto)) {
2016 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2017 ovs_rundir(), br->name));
2018 ofproto_set_snoops(br->ofproto, &snoops);
2019 sset_destroy(&snoops);
2023 /* Port functions. */
2025 static struct port *
2026 port_create(struct bridge *br, const struct ovsrec_port *cfg)
2030 port = xzalloc(sizeof *port);
2032 port->name = xstrdup(cfg->name);
2034 list_init(&port->ifaces);
2036 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
2038 VLOG_INFO("created port %s on bridge %s", port->name, br->name);
2044 get_port_other_config(const struct ovsrec_port *port, const char *key,
2045 const char *default_value)
2049 value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config,
2051 return value ? value : default_value;
2055 get_interface_other_config(const struct ovsrec_interface *iface,
2056 const char *key, const char *default_value)
2060 value = get_ovsrec_key_value(&iface->header_,
2061 &ovsrec_interface_col_other_config, key);
2062 return value ? value : default_value;
2065 /* Deletes interfaces from 'port' that are no longer configured for it. */
2067 port_del_ifaces(struct port *port)
2069 struct iface *iface, *next;
2070 struct sset new_ifaces;
2073 /* Collect list of new interfaces. */
2074 sset_init(&new_ifaces);
2075 for (i = 0; i < port->cfg->n_interfaces; i++) {
2076 const char *name = port->cfg->interfaces[i]->name;
2077 const char *type = port->cfg->interfaces[i]->name;
2078 if (strcmp(type, "null")) {
2079 sset_add(&new_ifaces, name);
2083 /* Get rid of deleted interfaces. */
2084 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2085 if (!sset_contains(&new_ifaces, iface->name)) {
2086 iface_destroy(iface);
2090 sset_destroy(&new_ifaces);
2093 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2096 port_add_ifaces(struct port *port)
2098 struct shash new_ifaces;
2099 struct shash_node *node;
2102 /* Collect new ifaces. */
2103 shash_init(&new_ifaces);
2104 for (i = 0; i < port->cfg->n_interfaces; i++) {
2105 const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
2106 if (strcmp(cfg->type, "null")
2107 && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
2108 VLOG_WARN("port %s: %s specified twice as port interface",
2109 port->name, cfg->name);
2110 iface_set_ofport(cfg, -1);
2114 /* Create new interfaces.
2115 * Update interface types and 'cfg' members. */
2116 SHASH_FOR_EACH (node, &new_ifaces) {
2117 const struct ovsrec_interface *cfg = node->data;
2118 const char *iface_name = node->name;
2119 struct iface *iface;
2121 iface = iface_lookup(port->bridge, iface_name);
2123 iface = iface_create(port, cfg);
2128 /* Determine interface type. The local port always has type
2129 * "internal". Other ports take their type from the database and
2130 * default to "system" if none is specified. */
2131 iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
2132 : cfg->type[0] ? cfg->type
2135 shash_destroy(&new_ifaces);
2139 port_destroy(struct port *port)
2142 struct bridge *br = port->bridge;
2143 struct iface *iface, *next;
2146 ofproto_bundle_unregister(br->ofproto, port);
2149 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
2150 iface_destroy(iface);
2153 hmap_remove(&br->ports, &port->hmap_node);
2155 VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
2162 static struct port *
2163 port_lookup(const struct bridge *br, const char *name)
2167 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
2169 if (!strcmp(port->name, name)) {
2177 enable_lacp(struct port *port, bool *activep)
2179 if (!port->cfg->lacp) {
2180 /* XXX when LACP implementation has been sufficiently tested, enable by
2181 * default and make active on bonded ports. */
2183 } else if (!strcmp(port->cfg->lacp, "off")) {
2185 } else if (!strcmp(port->cfg->lacp, "active")) {
2188 } else if (!strcmp(port->cfg->lacp, "passive")) {
2192 VLOG_WARN("port %s: unknown LACP mode %s",
2193 port->name, port->cfg->lacp);
2198 static struct lacp_settings *
2199 port_configure_lacp(struct port *port, struct lacp_settings *s)
2201 const char *lacp_time;
2202 long long int custom_time;
2205 if (!enable_lacp(port, &s->active)) {
2209 s->name = port->name;
2210 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
2212 /* Prefer bondable links if unspecified. */
2213 priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
2215 s->priority = (priority > 0 && priority <= UINT16_MAX
2217 : UINT16_MAX - !list_is_short(&port->ifaces));
2219 s->heartbeat = !strcmp(get_port_other_config(port->cfg,
2224 lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
2225 custom_time = atoi(lacp_time);
2226 if (!strcmp(lacp_time, "fast")) {
2227 s->lacp_time = LACP_TIME_FAST;
2228 } else if (!strcmp(lacp_time, "slow")) {
2229 s->lacp_time = LACP_TIME_SLOW;
2230 } else if (custom_time > 0) {
2231 s->lacp_time = LACP_TIME_CUSTOM;
2232 s->custom_time = custom_time;
2234 s->lacp_time = LACP_TIME_SLOW;
2241 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
2243 int priority, portid, key;
2245 portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
2246 priority = atoi(get_interface_other_config(iface->cfg,
2247 "lacp-port-priority", "0"));
2248 key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
2251 if (portid <= 0 || portid > UINT16_MAX) {
2252 portid = iface->ofp_port;
2255 if (priority <= 0 || priority > UINT16_MAX) {
2256 priority = UINT16_MAX;
2259 if (key < 0 || key > UINT16_MAX) {
2263 s->name = iface->name;
2265 s->priority = priority;
2270 port_configure_bond(struct port *port, struct bond_settings *s,
2271 uint32_t *bond_stable_ids)
2273 const char *detect_s;
2274 struct iface *iface;
2275 int miimon_interval;
2278 s->name = port->name;
2279 s->balance = BM_SLB;
2280 if (port->cfg->bond_mode
2281 && !bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
2282 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2283 port->name, port->cfg->bond_mode,
2284 bond_mode_to_string(s->balance));
2286 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
2287 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2288 "please use another bond type or disable flood_vlans",
2292 miimon_interval = atoi(get_port_other_config(port->cfg,
2293 "bond-miimon-interval", "0"));
2294 if (miimon_interval <= 0) {
2295 miimon_interval = 200;
2298 detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
2299 if (!strcmp(detect_s, "carrier")) {
2300 miimon_interval = 0;
2301 } else if (strcmp(detect_s, "miimon")) {
2302 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2303 "defaulting to carrier", port->name, detect_s);
2304 miimon_interval = 0;
2307 s->up_delay = MAX(0, port->cfg->bond_updelay);
2308 s->down_delay = MAX(0, port->cfg->bond_downdelay);
2309 s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
2310 s->rebalance_interval = atoi(
2311 get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
2312 if (s->rebalance_interval < 1000) {
2313 s->rebalance_interval = 1000;
2316 s->fake_iface = port->cfg->bond_fake_iface;
2319 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2320 long long stable_id;
2322 stable_id = atoll(get_interface_other_config(iface->cfg,
2323 "bond-stable-id", "0"));
2324 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
2325 stable_id = iface->ofp_port;
2327 bond_stable_ids[i++] = stable_id;
2329 netdev_set_miimon_interval(iface->netdev, miimon_interval);
2333 /* Interface functions. */
2335 static struct iface *
2336 iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
2338 struct bridge *br = port->bridge;
2339 struct iface *iface;
2340 char *name = if_cfg->name;
2342 iface = xzalloc(sizeof *iface);
2344 iface->name = xstrdup(name);
2345 iface->ofp_port = -1;
2346 iface->tag = tag_create_random();
2347 iface->netdev = NULL;
2348 iface->cfg = if_cfg;
2350 hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
2352 list_push_back(&port->ifaces, &iface->port_elem);
2354 VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
2360 iface_destroy(struct iface *iface)
2363 struct port *port = iface->port;
2364 struct bridge *br = port->bridge;
2366 if (br->ofproto && iface->ofp_port >= 0) {
2367 ofproto_port_unregister(br->ofproto, iface->ofp_port);
2370 if (iface->ofp_port >= 0) {
2371 hmap_remove(&br->ifaces, &iface->ofp_port_node);
2374 list_remove(&iface->port_elem);
2375 hmap_remove(&br->iface_by_name, &iface->name_node);
2377 netdev_close(iface->netdev);
2384 static struct iface *
2385 iface_lookup(const struct bridge *br, const char *name)
2387 struct iface *iface;
2389 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
2390 &br->iface_by_name) {
2391 if (!strcmp(iface->name, name)) {
2399 static struct iface *
2400 iface_find(const char *name)
2402 const struct bridge *br;
2404 HMAP_FOR_EACH (br, node, &all_bridges) {
2405 struct iface *iface = iface_lookup(br, name);
2414 static struct iface *
2415 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
2417 struct iface *iface;
2419 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
2420 hash_int(ofp_port, 0), &br->ifaces) {
2421 if (iface->ofp_port == ofp_port) {
2428 /* Set Ethernet address of 'iface', if one is specified in the configuration
2431 iface_set_mac(struct iface *iface)
2433 uint8_t ea[ETH_ADDR_LEN];
2435 if (!strcmp(iface->type, "internal")
2436 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
2437 if (iface->ofp_port == OFPP_LOCAL) {
2438 VLOG_ERR("interface %s: ignoring mac in Interface record "
2439 "(use Bridge record to set local port's mac)",
2441 } else if (eth_addr_is_multicast(ea)) {
2442 VLOG_ERR("interface %s: cannot set MAC to multicast address",
2445 int error = netdev_set_etheraddr(iface->netdev, ea);
2447 VLOG_ERR("interface %s: setting MAC failed (%s)",
2448 iface->name, strerror(error));
2454 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
2456 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
2458 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
2459 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
2463 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
2465 * The value strings in '*shash' are taken directly from values[], not copied,
2466 * so the caller should not modify or free them. */
2468 shash_from_ovs_idl_map(char **keys, char **values, size_t n,
2469 struct shash *shash)
2474 for (i = 0; i < n; i++) {
2475 shash_add(shash, keys[i], values[i]);
2479 /* Creates 'keys' and 'values' arrays from 'shash'.
2481 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
2482 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
2483 * are populated with with strings taken directly from 'shash' and thus have
2484 * the same ownership of the key-value pairs in shash.
2487 shash_to_ovs_idl_map(struct shash *shash,
2488 char ***keys, char ***values, size_t *n)
2492 struct shash_node *sn;
2494 count = shash_count(shash);
2496 k = xmalloc(count * sizeof *k);
2497 v = xmalloc(count * sizeof *v);
2500 SHASH_FOR_EACH(sn, shash) {
2511 struct iface_delete_queues_cbdata {
2512 struct netdev *netdev;
2513 const struct ovsdb_datum *queues;
2517 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
2519 union ovsdb_atom atom;
2521 atom.integer = target;
2522 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
2526 iface_delete_queues(unsigned int queue_id,
2527 const struct shash *details OVS_UNUSED, void *cbdata_)
2529 struct iface_delete_queues_cbdata *cbdata = cbdata_;
2531 if (!queue_ids_include(cbdata->queues, queue_id)) {
2532 netdev_delete_queue(cbdata->netdev, queue_id);
2537 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
2539 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
2540 netdev_set_qos(iface->netdev, NULL, NULL);
2542 struct iface_delete_queues_cbdata cbdata;
2543 struct shash details;
2547 /* Configure top-level Qos for 'iface'. */
2548 shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
2549 qos->n_other_config, &details);
2550 netdev_set_qos(iface->netdev, qos->type, &details);
2551 shash_destroy(&details);
2553 /* Deconfigure queues that were deleted. */
2554 cbdata.netdev = iface->netdev;
2555 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
2557 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
2559 /* Configure queues for 'iface'. */
2561 for (i = 0; i < qos->n_queues; i++) {
2562 const struct ovsrec_queue *queue = qos->value_queues[i];
2563 unsigned int queue_id = qos->key_queues[i];
2565 if (queue_id == 0) {
2569 shash_from_ovs_idl_map(queue->key_other_config,
2570 queue->value_other_config,
2571 queue->n_other_config, &details);
2572 netdev_set_queue(iface->netdev, queue_id, &details);
2573 shash_destroy(&details);
2576 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2577 VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default "
2578 "queue (queue 0). Using default configuration.",
2581 shash_init(&details);
2582 netdev_set_queue(iface->netdev, 0, &details);
2583 shash_destroy(&details);
2587 netdev_set_policing(iface->netdev,
2588 iface->cfg->ingress_policing_rate,
2589 iface->cfg->ingress_policing_burst);
2593 iface_configure_cfm(struct iface *iface)
2595 const struct ovsrec_interface *cfg = iface->cfg;
2596 struct cfm_settings s;
2597 uint16_t remote_mpid;
2599 if (!cfg->n_cfm_mpid || !cfg->n_cfm_remote_mpid) {
2600 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
2604 s.mpid = *cfg->cfm_mpid;
2605 remote_mpid = *cfg->cfm_remote_mpid;
2606 s.remote_mpids = &remote_mpid;
2607 s.n_remote_mpids = 1;
2609 s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
2611 if (s.interval <= 0) {
2615 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
2618 /* Read carrier or miimon status directly from 'iface''s netdev, according to
2619 * how 'iface''s port is configured.
2621 * Returns true if 'iface' is up, false otherwise. */
2623 iface_get_carrier(const struct iface *iface)
2626 return netdev_get_carrier(iface->netdev);
2629 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
2630 * instead of obtaining it from the database. */
2632 iface_is_synthetic(const struct iface *iface)
2634 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
2637 /* Port mirroring. */
2639 static struct mirror *
2640 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
2644 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
2645 if (uuid_equals(uuid, &m->uuid)) {
2653 bridge_configure_mirrors(struct bridge *br)
2655 const struct ovsdb_datum *mc;
2656 unsigned long *flood_vlans;
2657 struct mirror *m, *next;
2660 /* Get rid of deleted mirrors. */
2661 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
2662 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
2663 union ovsdb_atom atom;
2665 atom.uuid = m->uuid;
2666 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
2671 /* Add new mirrors and reconfigure existing ones. */
2672 for (i = 0; i < br->cfg->n_mirrors; i++) {
2673 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
2674 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
2676 m = mirror_create(br, cfg);
2678 if (!mirror_configure(m, cfg)) {
2683 /* Update flooded vlans (for RSPAN). */
2684 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
2685 br->cfg->n_flood_vlans);
2686 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
2687 bitmap_free(flood_vlans);
2690 static struct mirror *
2691 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
2695 m = xzalloc(sizeof *m);
2696 m->uuid = cfg->header_.uuid;
2697 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
2699 m->name = xstrdup(cfg->name);
2705 mirror_destroy(struct mirror *m)
2708 struct bridge *br = m->bridge;
2711 ofproto_mirror_unregister(br->ofproto, m);
2714 hmap_remove(&br->mirrors, &m->hmap_node);
2721 mirror_collect_ports(struct mirror *m,
2722 struct ovsrec_port **in_ports, int n_in_ports,
2723 void ***out_portsp, size_t *n_out_portsp)
2725 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
2726 size_t n_out_ports = 0;
2729 for (i = 0; i < n_in_ports; i++) {
2730 const char *name = in_ports[i]->name;
2731 struct port *port = port_lookup(m->bridge, name);
2733 out_ports[n_out_ports++] = port;
2735 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
2736 "port %s", m->bridge->name, m->name, name);
2739 *out_portsp = out_ports;
2740 *n_out_portsp = n_out_ports;
2744 mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg)
2746 struct ofproto_mirror_settings s;
2749 if (strcmp(cfg->name, m->name)) {
2751 m->name = xstrdup(cfg->name);
2755 /* Get output port or VLAN. */
2756 if (cfg->output_port) {
2757 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
2758 if (!s.out_bundle) {
2759 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
2760 m->bridge->name, m->name);
2763 s.out_vlan = UINT16_MAX;
2765 if (cfg->output_vlan) {
2766 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
2767 "output vlan; ignoring output vlan",
2768 m->bridge->name, m->name);
2770 } else if (cfg->output_vlan) {
2771 /* The database should prevent invalid VLAN values. */
2772 s.out_bundle = NULL;
2773 s.out_vlan = *cfg->output_vlan;
2775 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
2776 m->bridge->name, m->name);
2780 /* Get port selection. */
2781 if (cfg->select_all) {
2782 size_t n_ports = hmap_count(&m->bridge->ports);
2783 void **ports = xmalloc(n_ports * sizeof *ports);
2788 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
2798 /* Get ports, dropping ports that don't exist.
2799 * The IDL ensures that there are no duplicates. */
2800 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
2801 &s.srcs, &s.n_srcs);
2802 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
2803 &s.dsts, &s.n_dsts);
2806 /* Get VLAN selection. */
2807 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
2810 ofproto_mirror_register(m->bridge->ofproto, m, &s);
2813 if (s.srcs != s.dsts) {