1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
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.
21 #include "async-append.h"
25 #include "connectivity.h"
29 #include "dynamic-string.h"
36 #include "mac-learning.h"
37 #include "meta-flow.h"
39 #include "ofp-print.h"
42 #include "ofproto/bond.h"
43 #include "ofproto/ofproto.h"
44 #include "poll-loop.h"
49 #include "socket-util.h"
51 #include "stream-ssl.h"
53 #include "system-stats.h"
58 #include "lib/vswitch-idl.h"
59 #include "xenserver.h"
61 #include "sflow_api.h"
62 #include "vlan-bitmap.h"
64 VLOG_DEFINE_THIS_MODULE(bridge);
66 COVERAGE_DEFINE(bridge_reconfigure);
69 /* These members are always valid.
71 * They are immutable: they never change between iface_create() and
73 struct list port_elem; /* Element in struct port's "ifaces" list. */
74 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
75 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
76 struct port *port; /* Containing port. */
77 char *name; /* Host network device name. */
78 struct netdev *netdev; /* Network device. */
79 ofp_port_t ofp_port; /* OpenFlow port number. */
81 /* These members are valid only within bridge_reconfigure(). */
82 const char *type; /* Usually same as cfg->type. */
83 const struct ovsrec_interface *cfg;
87 struct uuid uuid; /* UUID of this "mirror" record in database. */
88 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
89 struct bridge *bridge;
91 const struct ovsrec_mirror *cfg;
95 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
96 struct bridge *bridge;
99 const struct ovsrec_port *cfg;
101 /* An ordinary bridge port has 1 interface.
102 * A bridge port for bonding has at least 2 interfaces. */
103 struct list ifaces; /* List of "struct iface"s. */
107 struct hmap_node node; /* In 'all_bridges'. */
108 char *name; /* User-specified arbitrary name. */
109 char *type; /* Datapath type. */
110 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
111 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
112 const struct ovsrec_bridge *cfg;
114 /* OpenFlow switch processing. */
115 struct ofproto *ofproto; /* OpenFlow switch. */
118 struct hmap ports; /* "struct port"s indexed by name. */
119 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
120 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
122 /* Port mirroring. */
123 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
125 /* Used during reconfiguration. */
126 struct shash wanted_ports;
128 /* Synthetic local port if necessary. */
129 struct ovsrec_port synth_local_port;
130 struct ovsrec_interface synth_local_iface;
131 struct ovsrec_interface *synth_local_ifacep;
134 /* All bridges, indexed by name. */
135 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
137 /* OVSDB IDL used to obtain configuration. */
138 static struct ovsdb_idl *idl;
140 /* We want to complete daemonization, fully detaching from our parent process,
141 * only after we have completed our initial configuration, committed our state
142 * to the database, and received confirmation back from the database server
143 * that it applied the commit. This allows our parent process to know that,
144 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
145 * to have already been filled in. (It is only "very likely" rather than
146 * certain because there is always a slim possibility that the transaction will
147 * fail or that some other client has added new bridges, ports, etc. while
148 * ovs-vswitchd was configuring using an old configuration.)
150 * We only need to do this once for our initial configuration at startup, so
151 * 'initial_config_done' tracks whether we've already done it. While we are
152 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
153 * itself and is otherwise NULL. */
154 static bool initial_config_done;
155 static struct ovsdb_idl_txn *daemonize_txn;
157 /* Most recently processed IDL sequence number. */
158 static unsigned int idl_seqno;
160 /* Track changes to port connectivity. */
161 static uint64_t connectivity_seqno = LLONG_MIN;
163 /* Each time this timer expires, the bridge fetches interface and mirror
164 * statistics and pushes them into the database. */
165 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
166 static long long int iface_stats_timer = LLONG_MIN;
168 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
169 * expensive. This can cause bridge_reconfigure() to take a long time during
170 * which no other work can be done. To deal with this problem, we limit port
171 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
172 * call to bridge_reconfigure(). If there is more work to do after the limit
173 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
174 * This allows the rest of the code to catch up on important things like
175 * forwarding packets. */
176 #define OFP_PORT_ACTION_WINDOW 10
178 static void add_del_bridges(const struct ovsrec_open_vswitch *);
179 static void bridge_create(const struct ovsrec_bridge *);
180 static void bridge_destroy(struct bridge *);
181 static struct bridge *bridge_lookup(const char *name);
182 static unixctl_cb_func bridge_unixctl_dump_flows;
183 static unixctl_cb_func bridge_unixctl_reconnect;
184 static size_t bridge_get_controllers(const struct bridge *br,
185 struct ovsrec_controller ***controllersp);
186 static void bridge_collect_wanted_ports(struct bridge *,
187 const unsigned long *splinter_vlans,
188 struct shash *wanted_ports);
189 static void bridge_delete_ofprotos(void);
190 static void bridge_delete_or_reconfigure_ports(struct bridge *);
191 static void bridge_del_ports(struct bridge *,
192 const struct shash *wanted_ports);
193 static void bridge_add_ports(struct bridge *,
194 const struct shash *wanted_ports);
196 static void bridge_configure_flow_miss_model(const char *opt);
197 static void bridge_configure_datapath_id(struct bridge *);
198 static void bridge_configure_netflow(struct bridge *);
199 static void bridge_configure_forward_bpdu(struct bridge *);
200 static void bridge_configure_mac_table(struct bridge *);
201 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
202 static void bridge_configure_ipfix(struct bridge *);
203 static void bridge_configure_stp(struct bridge *);
204 static void bridge_configure_tables(struct bridge *);
205 static void bridge_configure_dp_desc(struct bridge *);
206 static void bridge_configure_remotes(struct bridge *,
207 const struct sockaddr_in *managers,
209 static void bridge_pick_local_hw_addr(struct bridge *,
210 uint8_t ea[ETH_ADDR_LEN],
211 struct iface **hw_addr_iface);
212 static uint64_t bridge_pick_datapath_id(struct bridge *,
213 const uint8_t bridge_ea[ETH_ADDR_LEN],
214 struct iface *hw_addr_iface);
215 static uint64_t dpid_from_hash(const void *, size_t nbytes);
216 static bool bridge_has_bond_fake_iface(const struct bridge *,
218 static bool port_is_bond_fake_iface(const struct port *);
220 static unixctl_cb_func qos_unixctl_show;
222 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
223 static void port_del_ifaces(struct port *);
224 static void port_destroy(struct port *);
225 static struct port *port_lookup(const struct bridge *, const char *name);
226 static void port_configure(struct port *);
227 static struct lacp_settings *port_configure_lacp(struct port *,
228 struct lacp_settings *);
229 static void port_configure_bond(struct port *, struct bond_settings *);
230 static bool port_is_synthetic(const struct port *);
232 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
233 static void run_system_stats(void);
235 static void bridge_configure_mirrors(struct bridge *);
236 static struct mirror *mirror_create(struct bridge *,
237 const struct ovsrec_mirror *);
238 static void mirror_destroy(struct mirror *);
239 static bool mirror_configure(struct mirror *);
240 static void mirror_refresh_stats(struct mirror *);
242 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
243 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
244 const struct ovsrec_port *);
245 static bool iface_is_internal(const struct ovsrec_interface *iface,
246 const struct ovsrec_bridge *br);
247 static const char *iface_get_type(const struct ovsrec_interface *,
248 const struct ovsrec_bridge *);
249 static void iface_destroy(struct iface *);
250 static struct iface *iface_lookup(const struct bridge *, const char *name);
251 static struct iface *iface_find(const char *name);
252 static struct iface *iface_from_ofp_port(const struct bridge *,
253 ofp_port_t ofp_port);
254 static void iface_set_mac(struct iface *);
255 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
256 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
257 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
258 static void iface_configure_cfm(struct iface *);
259 static void iface_refresh_cfm_stats(struct iface *);
260 static void iface_refresh_stats(struct iface *);
261 static void iface_refresh_status(struct iface *);
262 static bool iface_is_synthetic(const struct iface *);
263 static ofp_port_t iface_get_requested_ofp_port(
264 const struct ovsrec_interface *);
265 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
267 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
269 * This is deprecated. It is only for compatibility with broken device drivers
270 * in old versions of Linux that do not properly support VLANs when VLAN
271 * devices are not used. When broken device drivers are no longer in
272 * widespread use, we will delete these interfaces. */
274 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
275 static bool vlan_splinters_enabled_anywhere;
277 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
278 static unsigned long int *collect_splinter_vlans(
279 const struct ovsrec_open_vswitch *);
280 static void configure_splinter_port(struct port *);
281 static void add_vlan_splinter_ports(struct bridge *,
282 const unsigned long int *splinter_vlans,
283 struct shash *ports);
286 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
288 struct shash iface_hints;
289 static bool initialized = false;
296 shash_init(&iface_hints);
299 for (i = 0; i < cfg->n_bridges; i++) {
300 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
303 for (j = 0; j < br_cfg->n_ports; j++) {
304 struct ovsrec_port *port_cfg = br_cfg->ports[j];
307 for (k = 0; k < port_cfg->n_interfaces; k++) {
308 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
309 struct iface_hint *iface_hint;
311 iface_hint = xmalloc(sizeof *iface_hint);
312 iface_hint->br_name = br_cfg->name;
313 iface_hint->br_type = br_cfg->datapath_type;
314 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
316 shash_add(&iface_hints, if_cfg->name, iface_hint);
322 ofproto_init(&iface_hints);
324 shash_destroy_free_data(&iface_hints);
328 /* Public functions. */
330 /* Initializes the bridge module, configuring it to obtain its configuration
331 * from an OVSDB server accessed over 'remote', which should be a string in a
332 * form acceptable to ovsdb_idl_create(). */
334 bridge_init(const char *remote)
336 /* Create connection to database. */
337 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
338 idl_seqno = ovsdb_idl_get_seqno(idl);
339 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
340 ovsdb_idl_verify_write_only(idl);
342 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
343 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
344 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
345 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
346 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
347 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
348 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
350 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
351 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
352 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
354 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
355 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
356 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
357 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
359 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
360 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
361 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
362 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
378 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
380 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
381 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
382 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
383 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
385 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
387 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
389 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
390 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
392 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
393 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
394 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
395 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
397 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
398 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
399 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
400 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
401 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
403 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
405 /* Register unixctl commands. */
406 unixctl_command_register("qos/show", "interface", 1, 1,
407 qos_unixctl_show, NULL);
408 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
409 bridge_unixctl_dump_flows, NULL);
410 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
411 bridge_unixctl_reconnect, NULL);
421 struct bridge *br, *next_br;
423 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
426 ovsdb_idl_destroy(idl);
429 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
430 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
431 * responsible for freeing '*managersp' (with free()).
433 * You may be asking yourself "why does ovs-vswitchd care?", because
434 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
435 * should not be and in fact is not directly involved in that. But
436 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
437 * it has to tell in-band control where the managers are to enable that.
438 * (Thus, only managers connected in-band are collected.)
441 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
442 struct sockaddr_in **managersp, size_t *n_managersp)
444 struct sockaddr_in *managers = NULL;
445 size_t n_managers = 0;
449 /* Collect all of the potential targets from the "targets" columns of the
450 * rows pointed to by "manager_options", excluding any that are
453 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
454 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
456 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
457 sset_find_and_delete(&targets, m->target);
459 sset_add(&targets, m->target);
463 /* Now extract the targets' IP addresses. */
464 if (!sset_is_empty(&targets)) {
467 managers = xmalloc(sset_count(&targets) * sizeof *managers);
468 SSET_FOR_EACH (target, &targets) {
469 struct sockaddr_in *sin = &managers[n_managers];
471 if (stream_parse_target_with_default_port(target,
478 sset_destroy(&targets);
480 *managersp = managers;
481 *n_managersp = n_managers;
485 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
487 unsigned long int *splinter_vlans;
488 struct sockaddr_in *managers;
489 struct bridge *br, *next;
490 int sflow_bridge_number;
493 COVERAGE_INC(bridge_reconfigure);
495 ofproto_set_flow_eviction_threshold(
496 smap_get_int(&ovs_cfg->other_config, "flow-eviction-threshold",
497 OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT));
499 ofproto_set_n_handler_threads(
500 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0));
502 bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config,
503 "force-miss-model"));
505 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
506 * to 'ovs_cfg', with only very minimal configuration otherwise.
508 * This is mostly an update to bridge data structures. Nothing is pushed
509 * down to ofproto or lower layers. */
510 add_del_bridges(ovs_cfg);
511 splinter_vlans = collect_splinter_vlans(ovs_cfg);
512 HMAP_FOR_EACH (br, node, &all_bridges) {
513 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
514 bridge_del_ports(br, &br->wanted_ports);
516 free(splinter_vlans);
518 /* Start pushing configuration changes down to the ofproto layer:
520 * - Delete ofprotos that are no longer configured.
522 * - Delete ports that are no longer configured.
524 * - Reconfigure existing ports to their desired configurations, or
525 * delete them if not possible.
527 * We have to do all the deletions before we can do any additions, because
528 * the ports to be added might require resources that will be freed up by
529 * deletions (they might especially overlap in name). */
530 bridge_delete_ofprotos();
531 HMAP_FOR_EACH (br, node, &all_bridges) {
533 bridge_delete_or_reconfigure_ports(br);
537 /* Finish pushing configuration changes to the ofproto layer:
539 * - Create ofprotos that are missing.
541 * - Add ports that are missing. */
542 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
546 error = ofproto_create(br->name, br->type, &br->ofproto);
548 VLOG_ERR("failed to create bridge %s: %s", br->name,
549 ovs_strerror(error));
550 shash_destroy(&br->wanted_ports);
555 HMAP_FOR_EACH (br, node, &all_bridges) {
556 bridge_add_ports(br, &br->wanted_ports);
557 shash_destroy(&br->wanted_ports);
560 reconfigure_system_stats(ovs_cfg);
562 /* Complete the configuration. */
563 sflow_bridge_number = 0;
564 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
565 HMAP_FOR_EACH (br, node, &all_bridges) {
568 /* We need the datapath ID early to allow LACP ports to use it as the
569 * default system ID. */
570 bridge_configure_datapath_id(br);
572 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
575 port_configure(port);
577 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
578 iface_configure_cfm(iface);
579 iface_configure_qos(iface, port->cfg->qos);
580 iface_set_mac(iface);
581 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
585 bridge_configure_mirrors(br);
586 bridge_configure_forward_bpdu(br);
587 bridge_configure_mac_table(br);
588 bridge_configure_remotes(br, managers, n_managers);
589 bridge_configure_netflow(br);
590 bridge_configure_sflow(br, &sflow_bridge_number);
591 bridge_configure_ipfix(br);
592 bridge_configure_stp(br);
593 bridge_configure_tables(br);
594 bridge_configure_dp_desc(br);
596 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
597 /* XXX: Remove this warning message eventually. */
598 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
599 " moved to the Open_vSwitch table. Ignoring its"
600 " setting in the bridge table.");
606 /* Delete ofprotos which aren't configured or have the wrong type. Create
607 * ofprotos which don't exist but need to. */
609 bridge_delete_ofprotos(void)
616 /* Delete ofprotos with no bridge or with the wrong type. */
619 ofproto_enumerate_types(&types);
620 SSET_FOR_EACH (type, &types) {
623 ofproto_enumerate_names(type, &names);
624 SSET_FOR_EACH (name, &names) {
625 br = bridge_lookup(name);
626 if (!br || strcmp(type, br->type)) {
627 ofproto_delete(name, type);
631 sset_destroy(&names);
632 sset_destroy(&types);
636 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
638 if (*n >= *allocated) {
639 ports = x2nrealloc(ports, allocated, sizeof *ports);
641 ports[(*n)++] = port;
646 bridge_delete_or_reconfigure_ports(struct bridge *br)
648 struct ofproto_port ofproto_port;
649 struct ofproto_port_dump dump;
651 /* List of "ofp_port"s to delete. We make a list instead of deleting them
652 * right away because ofproto implementations aren't necessarily able to
653 * iterate through a changing list of ports in an entirely robust way. */
661 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
662 ofp_port_t requested_ofp_port;
665 iface = iface_lookup(br, ofproto_port.name);
667 /* No such iface is configured, so we should delete this
670 * As a corner case exception, keep the port if it's a bond fake
672 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
673 && !strcmp(ofproto_port.type, "internal")) {
679 if (strcmp(ofproto_port.type, iface->type)
680 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
681 /* The interface is the wrong type or can't be configured.
686 /* If the requested OpenFlow port for 'iface' changed, and it's not
687 * already the correct port, then we might want to temporarily delete
688 * this interface, so we can add it back again with the new OpenFlow
690 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
691 if (iface->ofp_port != OFPP_LOCAL &&
692 requested_ofp_port != OFPP_NONE &&
693 requested_ofp_port != iface->ofp_port) {
694 ofp_port_t victim_request;
695 struct iface *victim;
697 /* Check for an existing OpenFlow port currently occupying
698 * 'iface''s requested port number. If there isn't one, then
699 * delete this port. Otherwise we need to consider further. */
700 victim = iface_from_ofp_port(br, requested_ofp_port);
705 /* 'victim' is a port currently using 'iface''s requested port
706 * number. Unless 'victim' specifically requested that port
707 * number, too, then we can delete both 'iface' and 'victim'
708 * temporarily. (We'll add both of them back again later with new
709 * OpenFlow port numbers.)
711 * If 'victim' did request port number 'requested_ofp_port', just
712 * like 'iface', then that's a configuration inconsistency that we
713 * can't resolve. We might as well let it keep its current port
715 victim_request = iface_get_requested_ofp_port(victim->cfg);
716 if (victim_request != requested_ofp_port) {
717 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
718 iface_destroy(victim);
727 iface_destroy(iface);
728 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
731 for (i = 0; i < n; i++) {
732 ofproto_port_del(br->ofproto, del[i]);
738 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
739 bool with_requested_port)
741 struct shash_node *port_node;
743 SHASH_FOR_EACH (port_node, wanted_ports) {
744 const struct ovsrec_port *port_cfg = port_node->data;
747 for (i = 0; i < port_cfg->n_interfaces; i++) {
748 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
749 ofp_port_t requested_ofp_port;
751 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
752 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
753 struct iface *iface = iface_lookup(br, iface_cfg->name);
756 iface_create(br, iface_cfg, port_cfg);
764 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
766 /* First add interfaces that request a particular port number. */
767 bridge_add_ports__(br, wanted_ports, true);
769 /* Then add interfaces that want automatic port number assignment.
770 * We add these afterward to avoid accidentally taking a specifically
771 * requested port number. */
772 bridge_add_ports__(br, wanted_ports, false);
776 port_configure(struct port *port)
778 const struct ovsrec_port *cfg = port->cfg;
779 struct bond_settings bond_settings;
780 struct lacp_settings lacp_settings;
781 struct ofproto_bundle_settings s;
784 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
785 configure_splinter_port(port);
794 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
795 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
796 s.slaves[s.n_slaves++] = iface->ofp_port;
801 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
805 /* Get VLAN trunks. */
808 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
812 if (cfg->vlan_mode) {
813 if (!strcmp(cfg->vlan_mode, "access")) {
814 s.vlan_mode = PORT_VLAN_ACCESS;
815 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
816 s.vlan_mode = PORT_VLAN_TRUNK;
817 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
818 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
819 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
820 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
822 /* This "can't happen" because ovsdb-server should prevent it. */
823 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
824 s.vlan_mode = PORT_VLAN_TRUNK;
828 s.vlan_mode = PORT_VLAN_ACCESS;
830 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
834 s.vlan_mode = PORT_VLAN_TRUNK;
837 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
840 /* Get LACP settings. */
841 s.lacp = port_configure_lacp(port, &lacp_settings);
845 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
846 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
847 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
850 s.lacp_slaves = NULL;
853 /* Get bond settings. */
854 if (s.n_slaves > 1) {
855 s.bond = &bond_settings;
856 port_configure_bond(port, &bond_settings);
859 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
860 netdev_set_miimon_interval(iface->netdev, 0);
865 ofproto_bundle_register(port->bridge->ofproto, port, &s);
874 bridge_configure_flow_miss_model(const char *opt)
876 enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO;
879 if (strcmp(opt, "with-facets")) {
880 model = OFPROTO_HANDLE_MISS_WITH_FACETS;
881 } else if (strcmp(opt, "without-facets")) {
882 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
886 ofproto_set_flow_miss_model(model);
889 /* Pick local port hardware address and datapath ID for 'br'. */
891 bridge_configure_datapath_id(struct bridge *br)
893 uint8_t ea[ETH_ADDR_LEN];
895 struct iface *local_iface;
896 struct iface *hw_addr_iface;
899 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
900 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
902 int error = netdev_set_etheraddr(local_iface->netdev, ea);
904 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
905 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
906 "Ethernet address: %s",
907 br->name, ovs_strerror(error));
910 memcpy(br->ea, ea, ETH_ADDR_LEN);
912 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
913 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
914 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
915 ofproto_set_datapath_id(br->ofproto, dpid);
918 dpid_string = xasprintf("%016"PRIx64, dpid);
919 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
923 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
926 bridge_get_allowed_versions(struct bridge *br)
928 if (!br->cfg->n_protocols)
931 return ofputil_versions_from_strings(br->cfg->protocols,
932 br->cfg->n_protocols);
935 /* Set NetFlow configuration on 'br'. */
937 bridge_configure_netflow(struct bridge *br)
939 struct ovsrec_netflow *cfg = br->cfg->netflow;
940 struct netflow_options opts;
943 ofproto_set_netflow(br->ofproto, NULL);
947 memset(&opts, 0, sizeof opts);
949 /* Get default NetFlow configuration from datapath.
950 * Apply overrides from 'cfg'. */
951 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
952 if (cfg->engine_type) {
953 opts.engine_type = *cfg->engine_type;
955 if (cfg->engine_id) {
956 opts.engine_id = *cfg->engine_id;
959 /* Configure active timeout interval. */
960 opts.active_timeout = cfg->active_timeout;
961 if (!opts.active_timeout) {
962 opts.active_timeout = -1;
963 } else if (opts.active_timeout < 0) {
964 VLOG_WARN("bridge %s: active timeout interval set to negative "
965 "value, using default instead (%d seconds)", br->name,
966 NF_ACTIVE_TIMEOUT_DEFAULT);
967 opts.active_timeout = -1;
970 /* Add engine ID to interface number to disambiguate bridgs? */
971 opts.add_id_to_iface = cfg->add_id_to_interface;
972 if (opts.add_id_to_iface) {
973 if (opts.engine_id > 0x7f) {
974 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
975 "another vswitch, choose an engine id less than 128",
978 if (hmap_count(&br->ports) > 508) {
979 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
980 "another port when more than 508 ports are used",
986 sset_init(&opts.collectors);
987 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
990 if (ofproto_set_netflow(br->ofproto, &opts)) {
991 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
993 sset_destroy(&opts.collectors);
996 /* Set sFlow configuration on 'br'. */
998 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1000 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1001 struct ovsrec_controller **controllers;
1002 struct ofproto_sflow_options oso;
1003 size_t n_controllers;
1007 ofproto_set_sflow(br->ofproto, NULL);
1011 memset(&oso, 0, sizeof oso);
1013 sset_init(&oso.targets);
1014 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1016 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1017 if (cfg->sampling) {
1018 oso.sampling_rate = *cfg->sampling;
1021 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1023 oso.polling_interval = *cfg->polling;
1026 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1028 oso.header_len = *cfg->header;
1031 oso.sub_id = (*sflow_bridge_number)++;
1032 oso.agent_device = cfg->agent;
1034 oso.control_ip = NULL;
1035 n_controllers = bridge_get_controllers(br, &controllers);
1036 for (i = 0; i < n_controllers; i++) {
1037 if (controllers[i]->local_ip) {
1038 oso.control_ip = controllers[i]->local_ip;
1042 ofproto_set_sflow(br->ofproto, &oso);
1044 sset_destroy(&oso.targets);
1047 /* Returns whether a IPFIX row is valid. */
1049 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1051 return ipfix && ipfix->n_targets > 0;
1054 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1056 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1057 const struct bridge *br)
1059 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1062 /* Set IPFIX configuration on 'br'. */
1064 bridge_configure_ipfix(struct bridge *br)
1066 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1067 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1068 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1069 struct ofproto_ipfix_bridge_exporter_options be_opts;
1070 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1071 size_t n_fe_opts = 0;
1073 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1074 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1079 if (!valid_be_cfg && n_fe_opts == 0) {
1080 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1085 memset(&be_opts, 0, sizeof be_opts);
1087 sset_init(&be_opts.targets);
1088 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1090 if (be_cfg->sampling) {
1091 be_opts.sampling_rate = *be_cfg->sampling;
1093 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1095 if (be_cfg->obs_domain_id) {
1096 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1098 if (be_cfg->obs_point_id) {
1099 be_opts.obs_point_id = *be_cfg->obs_point_id;
1101 if (be_cfg->cache_active_timeout) {
1102 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1104 if (be_cfg->cache_max_flows) {
1105 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1109 if (n_fe_opts > 0) {
1110 struct ofproto_ipfix_flow_exporter_options *opts;
1111 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1113 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1114 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1115 opts->collector_set_id = fe_cfg->id;
1116 sset_init(&opts->targets);
1117 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1118 fe_cfg->ipfix->n_targets);
1119 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1120 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1121 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1122 ? *fe_cfg->ipfix->cache_max_flows : 0;
1128 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1132 sset_destroy(&be_opts.targets);
1135 if (n_fe_opts > 0) {
1136 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1138 for (i = 0; i < n_fe_opts; i++) {
1139 sset_destroy(&opts->targets);
1147 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1148 struct ofproto_port_stp_settings *port_s,
1149 int *port_num_counter, unsigned long *port_num_bitmap)
1151 const char *config_str;
1152 struct iface *iface;
1154 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1155 port_s->enable = false;
1158 port_s->enable = true;
1161 /* STP over bonds is not supported. */
1162 if (!list_is_singleton(&port->ifaces)) {
1163 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1165 port_s->enable = false;
1169 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1171 /* Internal ports shouldn't participate in spanning tree, so
1173 if (!strcmp(iface->type, "internal")) {
1174 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1175 port_s->enable = false;
1179 /* STP on mirror output ports is not supported. */
1180 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1181 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1182 port_s->enable = false;
1186 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1188 unsigned long int port_num = strtoul(config_str, NULL, 0);
1189 int port_idx = port_num - 1;
1191 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1192 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1193 port_s->enable = false;
1197 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1198 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1199 port->name, port_num);
1200 port_s->enable = false;
1203 bitmap_set1(port_num_bitmap, port_idx);
1204 port_s->port_num = port_idx;
1206 if (*port_num_counter >= STP_MAX_PORTS) {
1207 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1208 port_s->enable = false;
1212 port_s->port_num = (*port_num_counter)++;
1215 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1217 port_s->path_cost = strtoul(config_str, NULL, 10);
1219 enum netdev_features current;
1222 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1223 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1224 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1227 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1229 port_s->priority = strtoul(config_str, NULL, 0);
1231 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1235 /* Set spanning tree configuration on 'br'. */
1237 bridge_configure_stp(struct bridge *br)
1239 if (!br->cfg->stp_enable) {
1240 ofproto_set_stp(br->ofproto, NULL);
1242 struct ofproto_stp_settings br_s;
1243 const char *config_str;
1245 int port_num_counter;
1246 unsigned long *port_num_bitmap;
1248 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1250 uint8_t ea[ETH_ADDR_LEN];
1252 if (eth_addr_from_string(config_str, ea)) {
1253 br_s.system_id = eth_addr_to_uint64(ea);
1255 br_s.system_id = eth_addr_to_uint64(br->ea);
1256 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1257 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1260 br_s.system_id = eth_addr_to_uint64(br->ea);
1263 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1265 br_s.priority = strtoul(config_str, NULL, 0);
1267 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1270 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1272 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1274 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1277 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1279 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1281 br_s.max_age = STP_DEFAULT_MAX_AGE;
1284 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1286 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1288 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1291 /* Configure STP on the bridge. */
1292 if (ofproto_set_stp(br->ofproto, &br_s)) {
1293 VLOG_ERR("bridge %s: could not enable STP", br->name);
1297 /* Users must either set the port number with the "stp-port-num"
1298 * configuration on all ports or none. If manual configuration
1299 * is not done, then we allocate them sequentially. */
1300 port_num_counter = 0;
1301 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1302 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1303 struct ofproto_port_stp_settings port_s;
1304 struct iface *iface;
1306 port_configure_stp(br->ofproto, port, &port_s,
1307 &port_num_counter, port_num_bitmap);
1309 /* As bonds are not supported, just apply configuration to
1310 * all interfaces. */
1311 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1312 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1314 VLOG_ERR("port %s: could not enable STP", port->name);
1320 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1321 && port_num_counter) {
1322 VLOG_ERR("bridge %s: must manually configure all STP port "
1323 "IDs or none, disabling", br->name);
1324 ofproto_set_stp(br->ofproto, NULL);
1326 bitmap_free(port_num_bitmap);
1331 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1333 const struct port *port = port_lookup(br, name);
1334 return port && port_is_bond_fake_iface(port);
1338 port_is_bond_fake_iface(const struct port *port)
1340 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1344 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1346 struct bridge *br, *next;
1347 struct shash new_br;
1350 /* Collect new bridges' names and types. */
1351 shash_init(&new_br);
1352 for (i = 0; i < cfg->n_bridges; i++) {
1353 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1354 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1356 if (strchr(br_cfg->name, '/')) {
1357 /* Prevent remote ovsdb-server users from accessing arbitrary
1358 * directories, e.g. consider a bridge named "../../../etc/". */
1359 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1361 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1362 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1366 /* Get rid of deleted bridges or those whose types have changed.
1367 * Update 'cfg' of bridges that still exist. */
1368 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1369 br->cfg = shash_find_data(&new_br, br->name);
1370 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1371 br->cfg->datapath_type))) {
1376 /* Add new bridges. */
1377 for (i = 0; i < cfg->n_bridges; i++) {
1378 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1379 struct bridge *br = bridge_lookup(br_cfg->name);
1381 bridge_create(br_cfg);
1385 shash_destroy(&new_br);
1388 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1389 * Returns 0 if successful, otherwise a positive errno value. */
1391 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1392 struct netdev *netdev)
1394 return netdev_set_config(netdev, &iface_cfg->options);
1397 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1398 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1399 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1402 * If successful, returns 0 and stores the network device in '*netdevp'. On
1403 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1405 iface_do_create(const struct bridge *br,
1406 const struct ovsrec_interface *iface_cfg,
1407 const struct ovsrec_port *port_cfg,
1408 ofp_port_t *ofp_portp, struct netdev **netdevp)
1410 struct netdev *netdev = NULL;
1413 if (netdev_is_reserved_name(iface_cfg->name)) {
1414 VLOG_WARN("could not create interface %s, name is reserved",
1420 error = netdev_open(iface_cfg->name,
1421 iface_get_type(iface_cfg, br->cfg), &netdev);
1423 VLOG_WARN("could not open network device %s (%s)",
1424 iface_cfg->name, ovs_strerror(error));
1428 error = iface_set_netdev_config(iface_cfg, netdev);
1433 *ofp_portp = iface_pick_ofport(iface_cfg);
1434 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1439 VLOG_INFO("bridge %s: added interface %s on port %d",
1440 br->name, iface_cfg->name, *ofp_portp);
1442 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1443 || iface_is_internal(iface_cfg, br->cfg)) {
1444 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1452 netdev_close(netdev);
1456 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1457 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1458 * automatically allocated for the iface. Takes ownership of and
1459 * deallocates 'if_cfg'.
1461 * Return true if an iface is successfully created, false otherwise. */
1463 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1464 const struct ovsrec_port *port_cfg)
1466 struct netdev *netdev;
1467 struct iface *iface;
1468 ofp_port_t ofp_port;
1472 /* Do the bits that can fail up front.
1474 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1475 * internal datastructures may not be consistent. Eventually, when port
1476 * additions and deletions are cheaper, these calls should be removed. */
1478 ovs_assert(!iface_lookup(br, iface_cfg->name));
1479 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1482 iface_set_ofport(iface_cfg, OFPP_NONE);
1483 iface_clear_db_record(iface_cfg);
1487 /* Get or create the port structure. */
1488 port = port_lookup(br, port_cfg->name);
1490 port = port_create(br, port_cfg);
1493 /* Create the iface structure. */
1494 iface = xzalloc(sizeof *iface);
1495 list_push_back(&port->ifaces, &iface->port_elem);
1496 hmap_insert(&br->iface_by_name, &iface->name_node,
1497 hash_string(iface_cfg->name, 0));
1499 iface->name = xstrdup(iface_cfg->name);
1500 iface->ofp_port = ofp_port;
1501 iface->netdev = netdev;
1502 iface->type = iface_get_type(iface_cfg, br->cfg);
1503 iface->cfg = iface_cfg;
1504 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1505 hash_ofp_port(ofp_port));
1507 iface_set_ofport(iface->cfg, ofp_port);
1509 /* Populate initial status in database. */
1510 iface_refresh_stats(iface);
1511 iface_refresh_status(iface);
1513 /* Add bond fake iface if necessary. */
1514 if (port_is_bond_fake_iface(port)) {
1515 struct ofproto_port ofproto_port;
1517 if (ofproto_port_query_by_name(br->ofproto, port->name,
1519 struct netdev *netdev;
1522 error = netdev_open(port->name, "internal", &netdev);
1524 ofp_port_t fake_ofp_port = OFPP_NONE;
1525 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1526 netdev_close(netdev);
1528 VLOG_WARN("could not open network device %s (%s)",
1529 port->name, ovs_strerror(error));
1532 /* Already exists, nothing to do. */
1533 ofproto_port_destroy(&ofproto_port);
1540 /* Set forward BPDU option. */
1542 bridge_configure_forward_bpdu(struct bridge *br)
1544 ofproto_set_forward_bpdu(br->ofproto,
1545 smap_get_bool(&br->cfg->other_config,
1550 /* Set MAC learning table configuration for 'br'. */
1552 bridge_configure_mac_table(struct bridge *br)
1554 const char *idle_time_str;
1557 const char *mac_table_size_str;
1560 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1561 idle_time = (idle_time_str && atoi(idle_time_str)
1562 ? atoi(idle_time_str)
1563 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1565 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1566 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1567 ? atoi(mac_table_size_str)
1570 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1574 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1575 struct iface **hw_addr_iface)
1577 struct hmapx mirror_output_ports;
1580 bool found_addr = false;
1584 *hw_addr_iface = NULL;
1586 /* Did the user request a particular MAC? */
1587 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1588 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1589 if (eth_addr_is_multicast(ea)) {
1590 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1591 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1592 } else if (eth_addr_is_zero(ea)) {
1593 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1599 /* Mirror output ports don't participate in picking the local hardware
1600 * address. ofproto can't help us find out whether a given port is a
1601 * mirror output because we haven't configured mirrors yet, so we need to
1602 * accumulate them ourselves. */
1603 hmapx_init(&mirror_output_ports);
1604 for (i = 0; i < br->cfg->n_mirrors; i++) {
1605 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1606 if (m->output_port) {
1607 hmapx_add(&mirror_output_ports, m->output_port);
1611 /* Otherwise choose the minimum non-local MAC address among all of the
1613 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1614 uint8_t iface_ea[ETH_ADDR_LEN];
1615 struct iface *candidate;
1616 struct iface *iface;
1618 /* Mirror output ports don't participate. */
1619 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1623 /* Choose the MAC address to represent the port. */
1625 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1626 /* Find the interface with this Ethernet address (if any) so that
1627 * we can provide the correct devname to the caller. */
1628 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1629 uint8_t candidate_ea[ETH_ADDR_LEN];
1630 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1631 && eth_addr_equals(iface_ea, candidate_ea)) {
1636 /* Choose the interface whose MAC address will represent the port.
1637 * The Linux kernel bonding code always chooses the MAC address of
1638 * the first slave added to a bond, and the Fedora networking
1639 * scripts always add slaves to a bond in alphabetical order, so
1640 * for compatibility we choose the interface with the name that is
1641 * first in alphabetical order. */
1642 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1643 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1648 /* The local port doesn't count (since we're trying to choose its
1649 * MAC address anyway). */
1650 if (iface->ofp_port == OFPP_LOCAL) {
1655 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1661 /* Compare against our current choice. */
1662 if (!eth_addr_is_multicast(iface_ea) &&
1663 !eth_addr_is_local(iface_ea) &&
1664 !eth_addr_is_reserved(iface_ea) &&
1665 !eth_addr_is_zero(iface_ea) &&
1666 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1668 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1669 *hw_addr_iface = iface;
1675 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1676 *hw_addr_iface = NULL;
1679 hmapx_destroy(&mirror_output_ports);
1682 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1683 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1684 * an interface on 'br', then that interface must be passed in as
1685 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1686 * 'hw_addr_iface' must be passed in as a null pointer. */
1688 bridge_pick_datapath_id(struct bridge *br,
1689 const uint8_t bridge_ea[ETH_ADDR_LEN],
1690 struct iface *hw_addr_iface)
1693 * The procedure for choosing a bridge MAC address will, in the most
1694 * ordinary case, also choose a unique MAC that we can use as a datapath
1695 * ID. In some special cases, though, multiple bridges will end up with
1696 * the same MAC address. This is OK for the bridges, but it will confuse
1697 * the OpenFlow controller, because each datapath needs a unique datapath
1700 * Datapath IDs must be unique. It is also very desirable that they be
1701 * stable from one run to the next, so that policy set on a datapath
1704 const char *datapath_id;
1707 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1708 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1712 if (!hw_addr_iface) {
1714 * A purely internal bridge, that is, one that has no non-virtual
1715 * network devices on it at all, is difficult because it has no
1716 * natural unique identifier at all.
1718 * When the host is a XenServer, we handle this case by hashing the
1719 * host's UUID with the name of the bridge. Names of bridges are
1720 * persistent across XenServer reboots, although they can be reused if
1721 * an internal network is destroyed and then a new one is later
1722 * created, so this is fairly effective.
1724 * When the host is not a XenServer, we punt by using a random MAC
1725 * address on each run.
1727 const char *host_uuid = xenserver_get_host_uuid();
1729 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1730 dpid = dpid_from_hash(combined, strlen(combined));
1736 return eth_addr_to_uint64(bridge_ea);
1740 dpid_from_hash(const void *data, size_t n)
1742 uint8_t hash[SHA1_DIGEST_SIZE];
1744 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1745 sha1_bytes(data, n, hash);
1746 eth_addr_mark_random(hash);
1747 return eth_addr_to_uint64(hash);
1751 iface_refresh_status(struct iface *iface)
1755 enum netdev_features current;
1759 uint8_t mac[ETH_ADDR_LEN];
1763 if (iface_is_synthetic(iface)) {
1769 if (!netdev_get_status(iface->netdev, &smap)) {
1770 ovsrec_interface_set_status(iface->cfg, &smap);
1772 ovsrec_interface_set_status(iface->cfg, NULL);
1775 smap_destroy(&smap);
1777 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1778 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1780 ovsrec_interface_set_duplex(iface->cfg,
1781 netdev_features_is_full_duplex(current)
1783 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1785 ovsrec_interface_set_duplex(iface->cfg, NULL);
1786 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1789 error = netdev_get_mtu(iface->netdev, &mtu);
1792 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1794 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1797 error = netdev_get_etheraddr(iface->netdev, mac);
1799 char mac_string[32];
1801 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1802 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1804 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1807 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1808 * if there is no valid ifindex number. */
1809 ifindex64 = netdev_get_ifindex(iface->netdev);
1810 if (ifindex64 < 0) {
1813 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1816 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1819 iface_refresh_cfm_stats(struct iface *iface)
1821 const struct ovsrec_interface *cfg = iface->cfg;
1822 struct ofproto_cfm_status status;
1824 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1825 iface->ofp_port, &status)) {
1826 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1827 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1828 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1829 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1830 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1831 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1833 const char *reasons[CFM_FAULT_N_REASONS];
1834 int64_t cfm_health = status.health;
1835 int64_t cfm_flap_count = status.flap_count;
1836 bool faulted = status.faults != 0;
1839 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1842 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1843 int reason = 1 << i;
1844 if (status.faults & reason) {
1845 reasons[j++] = cfm_fault_reason_to_str(reason);
1848 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1850 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1852 if (status.remote_opstate >= 0) {
1853 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1854 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1856 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1859 ovsrec_interface_set_cfm_remote_mpids(cfg,
1860 (const int64_t *)status.rmps,
1862 if (cfm_health >= 0) {
1863 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1865 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1873 iface_refresh_stats(struct iface *iface)
1875 #define IFACE_STATS \
1876 IFACE_STAT(rx_packets, "rx_packets") \
1877 IFACE_STAT(tx_packets, "tx_packets") \
1878 IFACE_STAT(rx_bytes, "rx_bytes") \
1879 IFACE_STAT(tx_bytes, "tx_bytes") \
1880 IFACE_STAT(rx_dropped, "rx_dropped") \
1881 IFACE_STAT(tx_dropped, "tx_dropped") \
1882 IFACE_STAT(rx_errors, "rx_errors") \
1883 IFACE_STAT(tx_errors, "tx_errors") \
1884 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1885 IFACE_STAT(rx_over_errors, "rx_over_err") \
1886 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1887 IFACE_STAT(collisions, "collisions")
1889 #define IFACE_STAT(MEMBER, NAME) + 1
1890 enum { N_IFACE_STATS = IFACE_STATS };
1892 int64_t values[N_IFACE_STATS];
1893 char *keys[N_IFACE_STATS];
1896 struct netdev_stats stats;
1898 if (iface_is_synthetic(iface)) {
1902 /* Intentionally ignore return value, since errors will set 'stats' to
1903 * all-1s, and we will deal with that correctly below. */
1904 netdev_get_stats(iface->netdev, &stats);
1906 /* Copy statistics into keys[] and values[]. */
1908 #define IFACE_STAT(MEMBER, NAME) \
1909 if (stats.MEMBER != UINT64_MAX) { \
1911 values[n] = stats.MEMBER; \
1916 ovs_assert(n <= N_IFACE_STATS);
1918 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1923 br_refresh_stp_status(struct bridge *br)
1925 struct smap smap = SMAP_INITIALIZER(&smap);
1926 struct ofproto *ofproto = br->ofproto;
1927 struct ofproto_stp_status status;
1929 if (ofproto_get_stp_status(ofproto, &status)) {
1933 if (!status.enabled) {
1934 ovsrec_bridge_set_status(br->cfg, NULL);
1938 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1939 STP_ID_ARGS(status.bridge_id));
1940 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1941 STP_ID_ARGS(status.designated_root));
1942 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1944 ovsrec_bridge_set_status(br->cfg, &smap);
1945 smap_destroy(&smap);
1949 port_refresh_stp_status(struct port *port)
1951 struct ofproto *ofproto = port->bridge->ofproto;
1952 struct iface *iface;
1953 struct ofproto_port_stp_status status;
1956 if (port_is_synthetic(port)) {
1960 /* STP doesn't currently support bonds. */
1961 if (!list_is_singleton(&port->ifaces)) {
1962 ovsrec_port_set_status(port->cfg, NULL);
1966 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1967 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1971 if (!status.enabled) {
1972 ovsrec_port_set_status(port->cfg, NULL);
1976 /* Set Status column. */
1978 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1979 smap_add(&smap, "stp_state", stp_state_name(status.state));
1980 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1981 smap_add(&smap, "stp_role", stp_role_name(status.role));
1982 ovsrec_port_set_status(port->cfg, &smap);
1983 smap_destroy(&smap);
1987 port_refresh_stp_stats(struct port *port)
1989 struct ofproto *ofproto = port->bridge->ofproto;
1990 struct iface *iface;
1991 struct ofproto_port_stp_stats stats;
1993 int64_t int_values[3];
1995 if (port_is_synthetic(port)) {
1999 /* STP doesn't currently support bonds. */
2000 if (!list_is_singleton(&port->ifaces)) {
2004 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2005 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2009 if (!stats.enabled) {
2010 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2014 /* Set Statistics column. */
2015 keys[0] = "stp_tx_count";
2016 int_values[0] = stats.tx_count;
2017 keys[1] = "stp_rx_count";
2018 int_values[1] = stats.rx_count;
2019 keys[2] = "stp_error_count";
2020 int_values[2] = stats.error_count;
2022 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2023 ARRAY_SIZE(int_values));
2027 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2029 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2033 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2035 bool enable = enable_system_stats(cfg);
2037 system_stats_enable(enable);
2039 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2044 run_system_stats(void)
2046 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2049 stats = system_stats_run();
2051 struct ovsdb_idl_txn *txn;
2052 struct ovsdb_datum datum;
2054 txn = ovsdb_idl_txn_create(idl);
2055 ovsdb_datum_from_smap(&datum, stats);
2056 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2058 ovsdb_idl_txn_commit(txn);
2059 ovsdb_idl_txn_destroy(txn);
2065 static inline const char *
2066 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2069 case OFPCR12_ROLE_EQUAL:
2071 case OFPCR12_ROLE_MASTER:
2073 case OFPCR12_ROLE_SLAVE:
2075 case OFPCR12_ROLE_NOCHANGE:
2077 return "*** INVALID ROLE ***";
2082 refresh_controller_status(void)
2086 const struct ovsrec_controller *cfg;
2090 /* Accumulate status for controllers on all bridges. */
2091 HMAP_FOR_EACH (br, node, &all_bridges) {
2092 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2095 /* Update each controller in the database with current status. */
2096 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2097 struct ofproto_controller_info *cinfo =
2098 shash_find_data(&info, cfg->target);
2101 struct smap smap = SMAP_INITIALIZER(&smap);
2102 const char **values = cinfo->pairs.values;
2103 const char **keys = cinfo->pairs.keys;
2106 for (i = 0; i < cinfo->pairs.n; i++) {
2107 smap_add(&smap, keys[i], values[i]);
2110 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2111 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2113 ovsrec_controller_set_status(cfg, &smap);
2114 smap_destroy(&smap);
2116 ovsrec_controller_set_is_connected(cfg, false);
2117 ovsrec_controller_set_role(cfg, NULL);
2118 ovsrec_controller_set_status(cfg, NULL);
2122 ofproto_free_ofproto_controller_info(&info);
2127 * Some information in the database must be kept as up-to-date as possible to
2128 * allow controllers to respond rapidly to network outages. We call these
2129 * statistics "instant" stats.
2131 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2132 * assuming that they've changed. The only means we have to determine whether
2133 * they have changed are:
2135 * - Try to commit changes to the database. If nothing changed, then
2136 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2139 * - instant_stats_run() is called late in the run loop, after anything that
2140 * might change any of the instant stats.
2142 * We use these two facts together to avoid waking the process up every
2143 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2146 /* Minimum interval between writing updates to the instant stats to the
2148 #define INSTANT_INTERVAL_MSEC 100
2150 /* Current instant stats database transaction, NULL if there is no ongoing
2152 static struct ovsdb_idl_txn *instant_txn;
2154 /* Next time (in msec on monotonic clock) at which we will update the instant
2156 static long long int instant_next_txn = LLONG_MIN;
2158 /* True if the run loop has run since we last saw that the instant stats were
2159 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2160 * to refresh the instant stats. */
2161 static bool instant_stats_could_have_changed;
2164 instant_stats_run(void)
2166 enum ovsdb_idl_txn_status status;
2168 instant_stats_could_have_changed = true;
2174 if (time_msec() < instant_next_txn) {
2177 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2179 seq = seq_read(connectivity_seq_get());
2180 if (seq == connectivity_seqno) {
2183 connectivity_seqno = seq;
2185 instant_txn = ovsdb_idl_txn_create(idl);
2186 HMAP_FOR_EACH (br, node, &all_bridges) {
2187 struct iface *iface;
2190 br_refresh_stp_status(br);
2192 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2193 port_refresh_stp_status(port);
2196 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2197 enum netdev_flags flags;
2199 const char *link_state;
2200 int64_t link_resets;
2203 if (iface_is_synthetic(iface)) {
2207 current = ofproto_port_is_lacp_current(br->ofproto,
2211 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2213 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2216 error = netdev_get_flags(iface->netdev, &flags);
2218 const char *state = flags & NETDEV_UP ? "up" : "down";
2219 ovsrec_interface_set_admin_state(iface->cfg, state);
2221 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2224 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2225 ovsrec_interface_set_link_state(iface->cfg, link_state);
2227 link_resets = netdev_get_carrier_resets(iface->netdev);
2228 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2230 iface_refresh_cfm_stats(iface);
2233 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2235 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2236 smap_destroy(&smap);
2241 status = ovsdb_idl_txn_commit(instant_txn);
2242 if (status != TXN_INCOMPLETE) {
2243 ovsdb_idl_txn_destroy(instant_txn);
2246 if (status == TXN_UNCHANGED) {
2247 instant_stats_could_have_changed = false;
2252 instant_stats_wait(void)
2255 ovsdb_idl_txn_wait(instant_txn);
2256 } else if (instant_stats_could_have_changed) {
2257 poll_timer_wait_until(instant_next_txn);
2261 /* Performs periodic activity required by bridges that needs to be done with
2262 * the least possible latency.
2264 * It makes sense to call this function a couple of times per poll loop, to
2265 * provide a significant performance boost on some benchmarks with ofprotos
2266 * that use the ofproto-dpif implementation. */
2268 bridge_run_fast(void)
2275 ofproto_enumerate_types(&types);
2276 SSET_FOR_EACH (type, &types) {
2277 ofproto_type_run_fast(type);
2279 sset_destroy(&types);
2281 HMAP_FOR_EACH (br, node, &all_bridges) {
2282 ofproto_run_fast(br->ofproto);
2289 static struct ovsrec_open_vswitch null_cfg;
2290 const struct ovsrec_open_vswitch *cfg;
2294 bool vlan_splinters_changed;
2297 ovsrec_open_vswitch_init(&null_cfg);
2301 if (ovsdb_idl_is_lock_contended(idl)) {
2302 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2303 struct bridge *br, *next_br;
2305 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2306 "disabling this process (pid %ld) until it goes away",
2307 (long int) getpid());
2309 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2312 /* Since we will not be running system_stats_run() in this process
2313 * with the current situation of multiple ovs-vswitchd daemons,
2314 * disable system stats collection. */
2315 system_stats_enable(false);
2317 } else if (!ovsdb_idl_has_lock(idl)) {
2320 cfg = ovsrec_open_vswitch_first(idl);
2322 /* Initialize the ofproto library. This only needs to run once, but
2323 * it must be done after the configuration is set. If the
2324 * initialization has already occurred, bridge_init_ofproto()
2325 * returns immediately. */
2326 bridge_init_ofproto(cfg);
2328 /* Once the value of flow-restore-wait is false, we no longer should
2329 * check its value from the database. */
2330 if (cfg && ofproto_get_flow_restore_wait()) {
2331 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2332 "flow-restore-wait", false));
2335 /* Let each datapath type do the work that it needs to do. */
2337 ofproto_enumerate_types(&types);
2338 SSET_FOR_EACH (type, &types) {
2339 ofproto_type_run(type);
2341 sset_destroy(&types);
2343 /* Let each bridge do the work that it needs to do. */
2344 HMAP_FOR_EACH (br, node, &all_bridges) {
2345 ofproto_run(br->ofproto);
2348 /* Re-configure SSL. We do this on every trip through the main loop,
2349 * instead of just when the database changes, because the contents of the
2350 * key and certificate files can change without the database changing.
2352 * We do this before bridge_reconfigure() because that function might
2353 * initiate SSL connections and thus requires SSL to be configured. */
2354 if (cfg && cfg->ssl) {
2355 const struct ovsrec_ssl *ssl = cfg->ssl;
2357 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2358 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2361 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2362 * usage has changed. */
2363 vlan_splinters_changed = false;
2364 if (vlan_splinters_enabled_anywhere) {
2365 HMAP_FOR_EACH (br, node, &all_bridges) {
2366 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2367 vlan_splinters_changed = true;
2373 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2374 struct ovsdb_idl_txn *txn;
2376 idl_seqno = ovsdb_idl_get_seqno(idl);
2377 txn = ovsdb_idl_txn_create(idl);
2378 bridge_reconfigure(cfg ? cfg : &null_cfg);
2381 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2384 /* If we are completing our initial configuration for this run
2385 * of ovs-vswitchd, then keep the transaction around to monitor
2386 * it for completion. */
2387 if (initial_config_done) {
2388 ovsdb_idl_txn_commit(txn);
2389 ovsdb_idl_txn_destroy(txn);
2391 initial_config_done = true;
2392 daemonize_txn = txn;
2396 if (daemonize_txn) {
2397 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2398 if (status != TXN_INCOMPLETE) {
2399 ovsdb_idl_txn_destroy(daemonize_txn);
2400 daemonize_txn = NULL;
2402 /* ovs-vswitchd has completed initialization, so allow the
2403 * process that forked us to exit successfully. */
2404 daemonize_complete();
2406 vlog_enable_async();
2408 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2412 /* Refresh interface and mirror stats if necessary. */
2413 if (time_msec() >= iface_stats_timer) {
2415 struct ovsdb_idl_txn *txn;
2417 txn = ovsdb_idl_txn_create(idl);
2418 HMAP_FOR_EACH (br, node, &all_bridges) {
2422 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2423 struct iface *iface;
2425 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2426 iface_refresh_stats(iface);
2427 iface_refresh_status(iface);
2430 port_refresh_stp_stats(port);
2433 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2434 mirror_refresh_stats(m);
2438 refresh_controller_status();
2439 ovsdb_idl_txn_commit(txn);
2440 ovsdb_idl_txn_destroy(txn); /* XXX */
2443 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2447 instant_stats_run();
2456 ovsdb_idl_wait(idl);
2457 if (daemonize_txn) {
2458 ovsdb_idl_txn_wait(daemonize_txn);
2462 ofproto_enumerate_types(&types);
2463 SSET_FOR_EACH (type, &types) {
2464 ofproto_type_wait(type);
2466 sset_destroy(&types);
2468 if (!hmap_is_empty(&all_bridges)) {
2471 HMAP_FOR_EACH (br, node, &all_bridges) {
2472 ofproto_wait(br->ofproto);
2474 poll_timer_wait_until(iface_stats_timer);
2477 system_stats_wait();
2478 instant_stats_wait();
2481 /* Adds some memory usage statistics for bridges into 'usage', for use with
2482 * memory_report(). */
2484 bridge_get_memory_usage(struct simap *usage)
2488 HMAP_FOR_EACH (br, node, &all_bridges) {
2489 ofproto_get_memory_usage(br->ofproto, usage);
2493 /* QoS unixctl user interface functions. */
2495 struct qos_unixctl_show_cbdata {
2497 struct iface *iface;
2501 qos_unixctl_show_queue(unsigned int queue_id,
2502 const struct smap *details,
2503 struct iface *iface,
2506 struct netdev_queue_stats stats;
2507 struct smap_node *node;
2510 ds_put_cstr(ds, "\n");
2512 ds_put_format(ds, "Queue %u:\n", queue_id);
2514 ds_put_cstr(ds, "Default:\n");
2517 SMAP_FOR_EACH (node, details) {
2518 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2521 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2523 if (stats.tx_packets != UINT64_MAX) {
2524 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2527 if (stats.tx_bytes != UINT64_MAX) {
2528 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2531 if (stats.tx_errors != UINT64_MAX) {
2532 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2535 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2536 queue_id, ovs_strerror(error));
2541 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2542 const char *argv[], void *aux OVS_UNUSED)
2544 struct ds ds = DS_EMPTY_INITIALIZER;
2545 struct smap smap = SMAP_INITIALIZER(&smap);
2546 struct iface *iface;
2548 struct smap_node *node;
2550 iface = iface_find(argv[1]);
2552 unixctl_command_reply_error(conn, "no such interface");
2556 netdev_get_qos(iface->netdev, &type, &smap);
2558 if (*type != '\0') {
2559 struct netdev_queue_dump dump;
2560 struct smap details;
2561 unsigned int queue_id;
2563 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2565 SMAP_FOR_EACH (node, &smap) {
2566 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2569 smap_init(&details);
2570 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2571 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2573 smap_destroy(&details);
2575 unixctl_command_reply(conn, ds_cstr(&ds));
2577 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2578 unixctl_command_reply_error(conn, ds_cstr(&ds));
2581 smap_destroy(&smap);
2585 /* Bridge reconfiguration functions. */
2587 bridge_create(const struct ovsrec_bridge *br_cfg)
2591 ovs_assert(!bridge_lookup(br_cfg->name));
2592 br = xzalloc(sizeof *br);
2594 br->name = xstrdup(br_cfg->name);
2595 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2598 /* Derive the default Ethernet address from the bridge's UUID. This should
2599 * be unique and it will be stable between ovs-vswitchd runs. */
2600 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2601 eth_addr_mark_random(br->default_ea);
2603 hmap_init(&br->ports);
2604 hmap_init(&br->ifaces);
2605 hmap_init(&br->iface_by_name);
2606 hmap_init(&br->mirrors);
2608 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2612 bridge_destroy(struct bridge *br)
2615 struct mirror *mirror, *next_mirror;
2616 struct port *port, *next_port;
2618 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2621 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2622 mirror_destroy(mirror);
2625 hmap_remove(&all_bridges, &br->node);
2626 ofproto_destroy(br->ofproto);
2627 hmap_destroy(&br->ifaces);
2628 hmap_destroy(&br->ports);
2629 hmap_destroy(&br->iface_by_name);
2630 hmap_destroy(&br->mirrors);
2637 static struct bridge *
2638 bridge_lookup(const char *name)
2642 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2643 if (!strcmp(br->name, name)) {
2650 /* Handle requests for a listing of all flows known by the OpenFlow
2651 * stack, including those normally hidden. */
2653 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2654 const char *argv[], void *aux OVS_UNUSED)
2659 br = bridge_lookup(argv[1]);
2661 unixctl_command_reply_error(conn, "Unknown bridge");
2666 ofproto_get_all_flows(br->ofproto, &results);
2668 unixctl_command_reply(conn, ds_cstr(&results));
2669 ds_destroy(&results);
2672 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2673 * connections and reconnect. If BRIDGE is not specified, then all bridges
2674 * drop their controller connections and reconnect. */
2676 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2677 const char *argv[], void *aux OVS_UNUSED)
2681 br = bridge_lookup(argv[1]);
2683 unixctl_command_reply_error(conn, "Unknown bridge");
2686 ofproto_reconnect_controllers(br->ofproto);
2688 HMAP_FOR_EACH (br, node, &all_bridges) {
2689 ofproto_reconnect_controllers(br->ofproto);
2692 unixctl_command_reply(conn, NULL);
2696 bridge_get_controllers(const struct bridge *br,
2697 struct ovsrec_controller ***controllersp)
2699 struct ovsrec_controller **controllers;
2700 size_t n_controllers;
2702 controllers = br->cfg->controller;
2703 n_controllers = br->cfg->n_controller;
2705 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2711 *controllersp = controllers;
2713 return n_controllers;
2717 bridge_collect_wanted_ports(struct bridge *br,
2718 const unsigned long int *splinter_vlans,
2719 struct shash *wanted_ports)
2723 shash_init(wanted_ports);
2725 for (i = 0; i < br->cfg->n_ports; i++) {
2726 const char *name = br->cfg->ports[i]->name;
2727 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2728 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2732 if (bridge_get_controllers(br, NULL)
2733 && !shash_find(wanted_ports, br->name)) {
2734 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2735 br->name, br->name);
2737 ovsrec_interface_init(&br->synth_local_iface);
2738 ovsrec_port_init(&br->synth_local_port);
2740 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2741 br->synth_local_port.n_interfaces = 1;
2742 br->synth_local_port.name = br->name;
2744 br->synth_local_iface.name = br->name;
2745 br->synth_local_iface.type = "internal";
2747 br->synth_local_ifacep = &br->synth_local_iface;
2749 shash_add(wanted_ports, br->name, &br->synth_local_port);
2752 if (splinter_vlans) {
2753 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2757 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2758 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2759 * 'br' needs to complete its configuration. */
2761 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2763 struct shash_node *port_node;
2764 struct port *port, *next;
2766 /* Get rid of deleted ports.
2767 * Get rid of deleted interfaces on ports that still exist. */
2768 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2769 port->cfg = shash_find_data(wanted_ports, port->name);
2773 port_del_ifaces(port);
2777 /* Update iface->cfg and iface->type in interfaces that still exist. */
2778 SHASH_FOR_EACH (port_node, wanted_ports) {
2779 const struct ovsrec_port *port = port_node->data;
2782 for (i = 0; i < port->n_interfaces; i++) {
2783 const struct ovsrec_interface *cfg = port->interfaces[i];
2784 struct iface *iface = iface_lookup(br, cfg->name);
2785 const char *type = iface_get_type(cfg, br->cfg);
2790 } else if (!strcmp(type, "null")) {
2791 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2792 " may be removed in February 2013. Please email"
2793 " dev@openvswitch.org with concerns.",
2796 /* We will add new interfaces later. */
2802 /* Initializes 'oc' appropriately as a management service controller for
2805 * The caller must free oc->target when it is no longer needed. */
2807 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2808 struct ofproto_controller *oc)
2810 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2811 oc->max_backoff = 0;
2812 oc->probe_interval = 60;
2813 oc->band = OFPROTO_OUT_OF_BAND;
2815 oc->burst_limit = 0;
2816 oc->enable_async_msgs = true;
2819 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2821 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2822 struct ofproto_controller *oc)
2826 oc->target = c->target;
2827 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2828 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2829 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2830 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2831 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2832 oc->burst_limit = (c->controller_burst_limit
2833 ? *c->controller_burst_limit : 0);
2834 oc->enable_async_msgs = (!c->enable_async_messages
2835 || *c->enable_async_messages);
2836 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2837 if (dscp < 0 || dscp > 63) {
2838 dscp = DSCP_DEFAULT;
2843 /* Configures the IP stack for 'br''s local interface properly according to the
2844 * configuration in 'c'. */
2846 bridge_configure_local_iface_netdev(struct bridge *br,
2847 struct ovsrec_controller *c)
2849 struct netdev *netdev;
2850 struct in_addr mask, gateway;
2852 struct iface *local_iface;
2855 /* If there's no local interface or no IP address, give up. */
2856 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2857 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2861 /* Bring up the local interface. */
2862 netdev = local_iface->netdev;
2863 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2865 /* Configure the IP address and netmask. */
2866 if (!c->local_netmask
2867 || !inet_aton(c->local_netmask, &mask)
2869 mask.s_addr = guess_netmask(ip.s_addr);
2871 if (!netdev_set_in4(netdev, ip, mask)) {
2872 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2873 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2876 /* Configure the default gateway. */
2877 if (c->local_gateway
2878 && inet_aton(c->local_gateway, &gateway)
2879 && gateway.s_addr) {
2880 if (!netdev_add_router(netdev, gateway)) {
2881 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2882 br->name, IP_ARGS(gateway.s_addr));
2887 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2888 * in either string are treated as equal to any number of slashes in the other,
2889 * e.g. "x///y" is equal to "x/y".
2891 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2892 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2893 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2894 * 'b' against a prefix of 'a'.
2897 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2899 const char *b_start = b;
2901 if (b - b_start >= b_stoplen) {
2903 } else if (*a != *b) {
2905 } else if (*a == '/') {
2906 a += strspn(a, "/");
2907 b += strspn(b, "/");
2908 } else if (*a == '\0') {
2918 bridge_configure_remotes(struct bridge *br,
2919 const struct sockaddr_in *managers, size_t n_managers)
2921 bool disable_in_band;
2923 struct ovsrec_controller **controllers;
2924 size_t n_controllers;
2926 enum ofproto_fail_mode fail_mode;
2928 struct ofproto_controller *ocs;
2932 /* Check if we should disable in-band control on this bridge. */
2933 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2936 /* Set OpenFlow queue ID for in-band control. */
2937 ofproto_set_in_band_queue(br->ofproto,
2938 smap_get_int(&br->cfg->other_config,
2939 "in-band-queue", -1));
2941 if (disable_in_band) {
2942 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2944 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2947 n_controllers = bridge_get_controllers(br, &controllers);
2949 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2952 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2953 for (i = 0; i < n_controllers; i++) {
2954 struct ovsrec_controller *c = controllers[i];
2956 if (!strncmp(c->target, "punix:", 6)
2957 || !strncmp(c->target, "unix:", 5)) {
2958 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2961 if (!strncmp(c->target, "unix:", 5)) {
2962 /* Connect to a listening socket */
2963 whitelist = xasprintf("unix:%s/", ovs_rundir());
2964 if (strchr(c->target, '/') &&
2965 !equal_pathnames(c->target, whitelist,
2966 strlen(whitelist))) {
2967 /* Absolute path specified, but not in ovs_rundir */
2968 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2969 "controller \"%s\" due to possibility for "
2970 "remote exploit. Instead, specify socket "
2971 "in whitelisted \"%s\" or connect to "
2972 "\"unix:%s/%s.mgmt\" (which is always "
2973 "available without special configuration).",
2974 br->name, c->target, whitelist,
2975 ovs_rundir(), br->name);
2980 whitelist = xasprintf("punix:%s/%s.controller",
2981 ovs_rundir(), br->name);
2982 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2983 /* Prevent remote ovsdb-server users from accessing
2984 * arbitrary Unix domain sockets and overwriting arbitrary
2986 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2987 "controller \"%s\" due to possibility of "
2988 "overwriting local files. Instead, specify "
2989 "whitelisted \"%s\" or connect to "
2990 "\"unix:%s/%s.mgmt\" (which is always "
2991 "available without special configuration).",
2992 br->name, c->target, whitelist,
2993 ovs_rundir(), br->name);
3002 bridge_configure_local_iface_netdev(br, c);
3003 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3004 if (disable_in_band) {
3005 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3010 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3011 bridge_get_allowed_versions(br));
3012 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3015 /* Set the fail-mode. */
3016 fail_mode = !br->cfg->fail_mode
3017 || !strcmp(br->cfg->fail_mode, "standalone")
3018 ? OFPROTO_FAIL_STANDALONE
3019 : OFPROTO_FAIL_SECURE;
3020 ofproto_set_fail_mode(br->ofproto, fail_mode);
3022 /* Configure OpenFlow controller connection snooping. */
3023 if (!ofproto_has_snoops(br->ofproto)) {
3027 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3028 ovs_rundir(), br->name));
3029 ofproto_set_snoops(br->ofproto, &snoops);
3030 sset_destroy(&snoops);
3035 bridge_configure_tables(struct bridge *br)
3037 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3041 n_tables = ofproto_get_n_tables(br->ofproto);
3043 for (i = 0; i < n_tables; i++) {
3044 struct ofproto_table_settings s;
3047 s.max_flows = UINT_MAX;
3050 s.n_prefix_fields = 0;
3051 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3053 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3054 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3057 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3058 s.max_flows = *cfg->flow_limit;
3060 if (cfg->overflow_policy
3061 && !strcmp(cfg->overflow_policy, "evict")) {
3063 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3064 for (k = 0; k < cfg->n_groups; k++) {
3065 const char *string = cfg->groups[k];
3068 msg = mf_parse_subfield__(&s.groups[k], &string);
3070 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3071 "'groups' (%s)", br->name, i, msg);
3073 } else if (*string) {
3074 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3075 "element '%s' contains trailing garbage",
3076 br->name, i, cfg->groups[k]);
3082 /* Prefix lookup fields. */
3083 s.n_prefix_fields = 0;
3084 for (k = 0; k < cfg->n_prefixes; k++) {
3085 const char *name = cfg->prefixes[k];
3086 const struct mf_field *mf = mf_from_name(name);
3088 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3092 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3093 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3094 "%s", br->name, name);
3097 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3098 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3099 "field not used: %s", br->name, name);
3102 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3104 if (s.n_prefix_fields > 0) {
3106 struct ds ds = DS_EMPTY_INITIALIZER;
3107 for (k = 0; k < s.n_prefix_fields; k++) {
3109 ds_put_char(&ds, ',');
3111 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3113 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3114 br->name, i, ds_cstr(&ds));
3119 ofproto_configure_table(br->ofproto, i, &s);
3123 for (; j < br->cfg->n_flow_tables; j++) {
3124 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3125 "%"PRId64" not supported by this datapath", br->name,
3126 br->cfg->key_flow_tables[j]);
3131 bridge_configure_dp_desc(struct bridge *br)
3133 ofproto_set_dp_desc(br->ofproto,
3134 smap_get(&br->cfg->other_config, "dp-desc"));
3137 /* Port functions. */
3139 static void iface_destroy__(struct iface *);
3141 static struct port *
3142 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3146 port = xzalloc(sizeof *port);
3148 port->name = xstrdup(cfg->name);
3150 list_init(&port->ifaces);
3152 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3156 /* Deletes interfaces from 'port' that are no longer configured for it. */
3158 port_del_ifaces(struct port *port)
3160 struct iface *iface, *next;
3161 struct sset new_ifaces;
3164 /* Collect list of new interfaces. */
3165 sset_init(&new_ifaces);
3166 for (i = 0; i < port->cfg->n_interfaces; i++) {
3167 const char *name = port->cfg->interfaces[i]->name;
3168 const char *type = port->cfg->interfaces[i]->type;
3169 if (strcmp(type, "null")) {
3170 sset_add(&new_ifaces, name);
3174 /* Get rid of deleted interfaces. */
3175 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3176 if (!sset_contains(&new_ifaces, iface->name)) {
3177 iface_destroy(iface);
3181 sset_destroy(&new_ifaces);
3185 port_destroy(struct port *port)
3188 struct bridge *br = port->bridge;
3189 struct iface *iface, *next;
3192 ofproto_bundle_unregister(br->ofproto, port);
3195 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3196 iface_destroy__(iface);
3199 hmap_remove(&br->ports, &port->hmap_node);
3205 static struct port *
3206 port_lookup(const struct bridge *br, const char *name)
3210 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3212 if (!strcmp(port->name, name)) {
3220 enable_lacp(struct port *port, bool *activep)
3222 if (!port->cfg->lacp) {
3223 /* XXX when LACP implementation has been sufficiently tested, enable by
3224 * default and make active on bonded ports. */
3226 } else if (!strcmp(port->cfg->lacp, "off")) {
3228 } else if (!strcmp(port->cfg->lacp, "active")) {
3231 } else if (!strcmp(port->cfg->lacp, "passive")) {
3235 VLOG_WARN("port %s: unknown LACP mode %s",
3236 port->name, port->cfg->lacp);
3241 static struct lacp_settings *
3242 port_configure_lacp(struct port *port, struct lacp_settings *s)
3244 const char *lacp_time, *system_id;
3247 if (!enable_lacp(port, &s->active)) {
3251 s->name = port->name;
3253 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3255 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3256 ETH_ADDR_SCAN_ARGS(s->id))) {
3257 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3258 " address.", port->name, system_id);
3262 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3265 if (eth_addr_is_zero(s->id)) {
3266 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3270 /* Prefer bondable links if unspecified. */
3271 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3273 s->priority = (priority > 0 && priority <= UINT16_MAX
3275 : UINT16_MAX - !list_is_short(&port->ifaces));
3277 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3278 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3280 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3281 "lacp-fallback-ab", false);
3287 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3289 int priority, portid, key;
3291 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3292 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3294 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3296 if (portid <= 0 || portid > UINT16_MAX) {
3297 portid = ofp_to_u16(iface->ofp_port);
3300 if (priority <= 0 || priority > UINT16_MAX) {
3301 priority = UINT16_MAX;
3304 if (key < 0 || key > UINT16_MAX) {
3308 s->name = iface->name;
3310 s->priority = priority;
3315 port_configure_bond(struct port *port, struct bond_settings *s)
3317 const char *detect_s;
3318 struct iface *iface;
3319 int miimon_interval;
3321 s->name = port->name;
3323 if (port->cfg->bond_mode) {
3324 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3325 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3326 port->name, port->cfg->bond_mode,
3327 bond_mode_to_string(s->balance));
3330 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3332 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3333 * active-backup. At some point we should remove this warning. */
3334 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3335 " in previous versions, the default bond_mode was"
3336 " balance-slb", port->name,
3337 bond_mode_to_string(s->balance));
3339 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3340 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3341 "please use another bond type or disable flood_vlans",
3345 miimon_interval = smap_get_int(&port->cfg->other_config,
3346 "bond-miimon-interval", 0);
3347 if (miimon_interval <= 0) {
3348 miimon_interval = 200;
3351 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3352 if (!detect_s || !strcmp(detect_s, "carrier")) {
3353 miimon_interval = 0;
3354 } else if (strcmp(detect_s, "miimon")) {
3355 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3356 "defaulting to carrier", port->name, detect_s);
3357 miimon_interval = 0;
3360 s->up_delay = MAX(0, port->cfg->bond_updelay);
3361 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3362 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3363 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3364 "bond-rebalance-interval", 10000);
3365 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3366 s->rebalance_interval = 1000;
3369 s->fake_iface = port->cfg->bond_fake_iface;
3371 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3372 "lacp-fallback-ab", false);
3374 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3375 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3379 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3380 * instead of obtaining it from the database. */
3382 port_is_synthetic(const struct port *port)
3384 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3387 /* Interface functions. */
3390 iface_is_internal(const struct ovsrec_interface *iface,
3391 const struct ovsrec_bridge *br)
3393 /* The local port and "internal" ports are always "internal". */
3394 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3397 /* Returns the correct network device type for interface 'iface' in bridge
3400 iface_get_type(const struct ovsrec_interface *iface,
3401 const struct ovsrec_bridge *br)
3405 /* The local port always has type "internal". Other ports take
3406 * their type from the database and default to "system" if none is
3408 if (iface_is_internal(iface, br)) {
3411 type = iface->type[0] ? iface->type : "system";
3414 return ofproto_port_open_type(br->datapath_type, type);
3418 iface_destroy__(struct iface *iface)
3421 struct port *port = iface->port;
3422 struct bridge *br = port->bridge;
3424 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3425 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3428 if (iface->ofp_port != OFPP_NONE) {
3429 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3432 list_remove(&iface->port_elem);
3433 hmap_remove(&br->iface_by_name, &iface->name_node);
3435 netdev_close(iface->netdev);
3443 iface_destroy(struct iface *iface)
3446 struct port *port = iface->port;
3448 iface_destroy__(iface);
3449 if (list_is_empty(&port->ifaces)) {
3455 static struct iface *
3456 iface_lookup(const struct bridge *br, const char *name)
3458 struct iface *iface;
3460 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3461 &br->iface_by_name) {
3462 if (!strcmp(iface->name, name)) {
3470 static struct iface *
3471 iface_find(const char *name)
3473 const struct bridge *br;
3475 HMAP_FOR_EACH (br, node, &all_bridges) {
3476 struct iface *iface = iface_lookup(br, name);
3485 static struct iface *
3486 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3488 struct iface *iface;
3490 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3492 if (iface->ofp_port == ofp_port) {
3499 /* Set Ethernet address of 'iface', if one is specified in the configuration
3502 iface_set_mac(struct iface *iface)
3504 uint8_t ea[ETH_ADDR_LEN];
3506 if (!strcmp(iface->type, "internal")
3507 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3508 if (iface->ofp_port == OFPP_LOCAL) {
3509 VLOG_ERR("interface %s: ignoring mac in Interface record "
3510 "(use Bridge record to set local port's mac)",
3512 } else if (eth_addr_is_multicast(ea)) {
3513 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3516 int error = netdev_set_etheraddr(iface->netdev, ea);
3518 VLOG_ERR("interface %s: setting MAC failed (%s)",
3519 iface->name, ovs_strerror(error));
3525 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3527 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3529 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3530 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3531 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3535 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3536 * sets the "ofport" field to -1.
3538 * This is appropriate when 'if_cfg''s interface cannot be created or is
3539 * otherwise invalid. */
3541 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3543 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3544 ovsrec_interface_set_status(if_cfg, NULL);
3545 ovsrec_interface_set_admin_state(if_cfg, NULL);
3546 ovsrec_interface_set_duplex(if_cfg, NULL);
3547 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3548 ovsrec_interface_set_link_state(if_cfg, NULL);
3549 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3550 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3551 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3552 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3553 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3554 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3555 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3556 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3561 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3563 union ovsdb_atom atom;
3565 atom.integer = target;
3566 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3570 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3572 struct ofpbuf queues_buf;
3574 ofpbuf_init(&queues_buf, 0);
3576 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3577 netdev_set_qos(iface->netdev, NULL, NULL);
3579 const struct ovsdb_datum *queues;
3580 struct netdev_queue_dump dump;
3581 unsigned int queue_id;
3582 struct smap details;
3586 /* Configure top-level Qos for 'iface'. */
3587 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3589 /* Deconfigure queues that were deleted. */
3590 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3592 smap_init(&details);
3593 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3594 if (!queue_ids_include(queues, queue_id)) {
3595 netdev_delete_queue(iface->netdev, queue_id);
3598 smap_destroy(&details);
3600 /* Configure queues for 'iface'. */
3602 for (i = 0; i < qos->n_queues; i++) {
3603 const struct ovsrec_queue *queue = qos->value_queues[i];
3604 unsigned int queue_id = qos->key_queues[i];
3606 if (queue_id == 0) {
3610 if (queue->n_dscp == 1) {
3611 struct ofproto_port_queue *port_queue;
3613 port_queue = ofpbuf_put_uninit(&queues_buf,
3614 sizeof *port_queue);
3615 port_queue->queue = queue_id;
3616 port_queue->dscp = queue->dscp[0];
3619 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3622 struct smap details;
3624 smap_init(&details);
3625 netdev_set_queue(iface->netdev, 0, &details);
3626 smap_destroy(&details);
3630 if (iface->ofp_port != OFPP_NONE) {
3631 const struct ofproto_port_queue *port_queues = queues_buf.data;
3632 size_t n_queues = queues_buf.size / sizeof *port_queues;
3634 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3635 port_queues, n_queues);
3638 netdev_set_policing(iface->netdev,
3639 iface->cfg->ingress_policing_rate,
3640 iface->cfg->ingress_policing_burst);
3642 ofpbuf_uninit(&queues_buf);
3646 iface_configure_cfm(struct iface *iface)
3648 const struct ovsrec_interface *cfg = iface->cfg;
3649 const char *opstate_str;
3650 const char *cfm_ccm_vlan;
3651 struct cfm_settings s;
3652 struct smap netdev_args;
3654 if (!cfg->n_cfm_mpid) {
3655 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3659 s.check_tnl_key = false;
3660 smap_init(&netdev_args);
3661 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3662 const char *key = smap_get(&netdev_args, "key");
3663 const char *in_key = smap_get(&netdev_args, "in_key");
3665 s.check_tnl_key = (key && !strcmp(key, "flow"))
3666 || (in_key && !strcmp(in_key, "flow"));
3668 smap_destroy(&netdev_args);
3670 s.mpid = *cfg->cfm_mpid;
3671 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3672 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3673 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3675 if (s.interval <= 0) {
3679 if (!cfm_ccm_vlan) {
3681 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3682 s.ccm_vlan = CFM_RANDOM_VLAN;
3684 s.ccm_vlan = atoi(cfm_ccm_vlan);
3685 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3690 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3692 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3694 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3695 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3697 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3700 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3701 * instead of obtaining it from the database. */
3703 iface_is_synthetic(const struct iface *iface)
3705 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3709 iface_validate_ofport__(size_t n, int64_t *ofport)
3711 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3712 ? u16_to_ofp(*ofport)
3717 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3719 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3723 iface_pick_ofport(const struct ovsrec_interface *cfg)
3725 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3726 return (requested_ofport != OFPP_NONE
3728 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3731 /* Port mirroring. */
3733 static struct mirror *
3734 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3738 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3739 if (uuid_equals(uuid, &m->uuid)) {
3747 bridge_configure_mirrors(struct bridge *br)
3749 const struct ovsdb_datum *mc;
3750 unsigned long *flood_vlans;
3751 struct mirror *m, *next;
3754 /* Get rid of deleted mirrors. */
3755 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3756 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3757 union ovsdb_atom atom;
3759 atom.uuid = m->uuid;
3760 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3765 /* Add new mirrors and reconfigure existing ones. */
3766 for (i = 0; i < br->cfg->n_mirrors; i++) {
3767 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3768 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3770 m = mirror_create(br, cfg);
3773 if (!mirror_configure(m)) {
3778 /* Update flooded vlans (for RSPAN). */
3779 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3780 br->cfg->n_flood_vlans);
3781 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3782 bitmap_free(flood_vlans);
3785 static struct mirror *
3786 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3790 m = xzalloc(sizeof *m);
3791 m->uuid = cfg->header_.uuid;
3792 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3794 m->name = xstrdup(cfg->name);
3800 mirror_destroy(struct mirror *m)
3803 struct bridge *br = m->bridge;
3806 ofproto_mirror_unregister(br->ofproto, m);
3809 hmap_remove(&br->mirrors, &m->hmap_node);
3816 mirror_collect_ports(struct mirror *m,
3817 struct ovsrec_port **in_ports, int n_in_ports,
3818 void ***out_portsp, size_t *n_out_portsp)
3820 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3821 size_t n_out_ports = 0;
3824 for (i = 0; i < n_in_ports; i++) {
3825 const char *name = in_ports[i]->name;
3826 struct port *port = port_lookup(m->bridge, name);
3828 out_ports[n_out_ports++] = port;
3830 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3831 "port %s", m->bridge->name, m->name, name);
3834 *out_portsp = out_ports;
3835 *n_out_portsp = n_out_ports;
3839 mirror_configure(struct mirror *m)
3841 const struct ovsrec_mirror *cfg = m->cfg;
3842 struct ofproto_mirror_settings s;
3845 if (strcmp(cfg->name, m->name)) {
3847 m->name = xstrdup(cfg->name);
3851 /* Get output port or VLAN. */
3852 if (cfg->output_port) {
3853 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3854 if (!s.out_bundle) {
3855 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3856 m->bridge->name, m->name);
3859 s.out_vlan = UINT16_MAX;
3861 if (cfg->output_vlan) {
3862 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3863 "output vlan; ignoring output vlan",
3864 m->bridge->name, m->name);
3866 } else if (cfg->output_vlan) {
3867 /* The database should prevent invalid VLAN values. */
3868 s.out_bundle = NULL;
3869 s.out_vlan = *cfg->output_vlan;
3871 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3872 m->bridge->name, m->name);
3876 /* Get port selection. */
3877 if (cfg->select_all) {
3878 size_t n_ports = hmap_count(&m->bridge->ports);
3879 void **ports = xmalloc(n_ports * sizeof *ports);
3884 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3894 /* Get ports, dropping ports that don't exist.
3895 * The IDL ensures that there are no duplicates. */
3896 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3897 &s.srcs, &s.n_srcs);
3898 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3899 &s.dsts, &s.n_dsts);
3902 /* Get VLAN selection. */
3903 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3906 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3909 if (s.srcs != s.dsts) {
3918 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3920 * This is deprecated. It is only for compatibility with broken device drivers
3921 * in old versions of Linux that do not properly support VLANs when VLAN
3922 * devices are not used. When broken device drivers are no longer in
3923 * widespread use, we will delete these interfaces. */
3925 static struct ovsrec_port **recs;
3926 static size_t n_recs, allocated_recs;
3928 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3929 * splinters are reconfigured. */
3931 register_rec(struct ovsrec_port *rec)
3933 if (n_recs >= allocated_recs) {
3934 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3936 recs[n_recs++] = rec;
3939 /* Frees all of the ports registered with register_reg(). */
3941 free_registered_recs(void)
3945 for (i = 0; i < n_recs; i++) {
3946 struct ovsrec_port *port = recs[i];
3949 for (j = 0; j < port->n_interfaces; j++) {
3950 struct ovsrec_interface *iface = port->interfaces[j];
3955 smap_destroy(&port->other_config);
3956 free(port->interfaces);
3964 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3967 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3969 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3973 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3976 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3977 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3978 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3981 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3982 * use, returns NULL.
3984 * Updates 'vlan_splinters_enabled_anywhere'. */
3985 static unsigned long int *
3986 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3988 unsigned long int *splinter_vlans;
3989 struct sset splinter_ifaces;
3990 const char *real_dev_name;
3991 struct shash *real_devs;
3992 struct shash_node *node;
3996 /* Free space allocated for synthesized ports and interfaces, since we're
3997 * in the process of reconstructing all of them. */
3998 free_registered_recs();
4000 splinter_vlans = bitmap_allocate(4096);
4001 sset_init(&splinter_ifaces);
4002 vlan_splinters_enabled_anywhere = false;
4003 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4004 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4007 for (j = 0; j < br_cfg->n_ports; j++) {
4008 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4011 for (k = 0; k < port_cfg->n_interfaces; k++) {
4012 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4014 if (vlan_splinters_is_enabled(iface_cfg)) {
4015 vlan_splinters_enabled_anywhere = true;
4016 sset_add(&splinter_ifaces, iface_cfg->name);
4017 vlan_bitmap_from_array__(port_cfg->trunks,
4023 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4024 bitmap_set1(splinter_vlans, *port_cfg->tag);
4029 if (!vlan_splinters_enabled_anywhere) {
4030 free(splinter_vlans);
4031 sset_destroy(&splinter_ifaces);
4035 HMAP_FOR_EACH (br, node, &all_bridges) {
4037 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4041 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4042 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4043 * device to be created for it. */
4044 bitmap_set0(splinter_vlans, 0);
4045 bitmap_set0(splinter_vlans, 4095);
4047 /* Delete all VLAN devices that we don't need. */
4049 real_devs = vlandev_get_real_devs();
4050 SHASH_FOR_EACH (node, real_devs) {
4051 const struct vlan_real_dev *real_dev = node->data;
4052 const struct vlan_dev *vlan_dev;
4053 bool real_dev_has_splinters;
4055 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4057 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4058 if (!real_dev_has_splinters
4059 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4060 struct netdev *netdev;
4062 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4063 if (!netdev_get_in4(netdev, NULL, NULL) ||
4064 !netdev_get_in6(netdev, NULL)) {
4065 /* It has an IP address configured, so we don't own
4066 * it. Don't delete it. */
4068 vlandev_del(vlan_dev->name);
4070 netdev_close(netdev);
4077 /* Add all VLAN devices that we need. */
4078 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4081 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4082 if (!vlandev_get_name(real_dev_name, vid)) {
4083 vlandev_add(real_dev_name, vid);
4090 sset_destroy(&splinter_ifaces);
4092 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4093 free(splinter_vlans);
4096 return splinter_vlans;
4099 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4102 configure_splinter_port(struct port *port)
4104 struct ofproto *ofproto = port->bridge->ofproto;
4105 ofp_port_t realdev_ofp_port;
4106 const char *realdev_name;
4107 struct iface *vlandev, *realdev;
4109 ofproto_bundle_unregister(port->bridge->ofproto, port);
4111 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4114 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4115 realdev = iface_lookup(port->bridge, realdev_name);
4116 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4118 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4122 static struct ovsrec_port *
4123 synthesize_splinter_port(const char *real_dev_name,
4124 const char *vlan_dev_name, int vid)
4126 struct ovsrec_interface *iface;
4127 struct ovsrec_port *port;
4129 iface = xmalloc(sizeof *iface);
4130 ovsrec_interface_init(iface);
4131 iface->name = xstrdup(vlan_dev_name);
4132 iface->type = "system";
4134 port = xmalloc(sizeof *port);
4135 ovsrec_port_init(port);
4136 port->interfaces = xmemdup(&iface, sizeof iface);
4137 port->n_interfaces = 1;
4138 port->name = xstrdup(vlan_dev_name);
4139 port->vlan_mode = "splinter";
4140 port->tag = xmalloc(sizeof *port->tag);
4143 smap_add(&port->other_config, "realdev", real_dev_name);
4149 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4150 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4151 * 1-bit in the 'splinter_vlans' bitmap. */
4153 add_vlan_splinter_ports(struct bridge *br,
4154 const unsigned long int *splinter_vlans,
4155 struct shash *ports)
4159 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4160 * we're modifying 'ports'. */
4161 for (i = 0; i < br->cfg->n_ports; i++) {
4162 const char *name = br->cfg->ports[i]->name;
4163 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4166 for (j = 0; j < port_cfg->n_interfaces; j++) {
4167 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4169 if (vlan_splinters_is_enabled(iface_cfg)) {
4170 const char *real_dev_name;
4173 real_dev_name = iface_cfg->name;
4174 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4175 const char *vlan_dev_name;
4177 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4179 && !shash_find(ports, vlan_dev_name)) {
4180 shash_add(ports, vlan_dev_name,
4181 synthesize_splinter_port(
4182 real_dev_name, vlan_dev_name, vid));
4191 mirror_refresh_stats(struct mirror *m)
4193 struct ofproto *ofproto = m->bridge->ofproto;
4194 uint64_t tx_packets, tx_bytes;
4197 size_t stat_cnt = 0;
4199 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4200 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4204 if (tx_packets != UINT64_MAX) {
4205 keys[stat_cnt] = "tx_packets";
4206 values[stat_cnt] = tx_packets;
4209 if (tx_bytes != UINT64_MAX) {
4210 keys[stat_cnt] = "tx_bytes";
4211 values[stat_cnt] = tx_bytes;
4215 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);