1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 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 /* Set to true to allow experimental use of OpenFlow 1.4.
169 * This is false initially because OpenFlow 1.4 is not yet safe to use: it can
170 * abort due to unimplemented features. */
171 static bool allow_of14;
173 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
174 * expensive. This can cause bridge_reconfigure() to take a long time during
175 * which no other work can be done. To deal with this problem, we limit port
176 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
177 * call to bridge_reconfigure(). If there is more work to do after the limit
178 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
179 * This allows the rest of the code to catch up on important things like
180 * forwarding packets. */
181 #define OFP_PORT_ACTION_WINDOW 10
183 static void add_del_bridges(const struct ovsrec_open_vswitch *);
184 static void bridge_run__(void);
185 static void bridge_create(const struct ovsrec_bridge *);
186 static void bridge_destroy(struct bridge *);
187 static struct bridge *bridge_lookup(const char *name);
188 static unixctl_cb_func bridge_unixctl_dump_flows;
189 static unixctl_cb_func bridge_unixctl_reconnect;
190 static size_t bridge_get_controllers(const struct bridge *br,
191 struct ovsrec_controller ***controllersp);
192 static void bridge_collect_wanted_ports(struct bridge *,
193 const unsigned long *splinter_vlans,
194 struct shash *wanted_ports);
195 static void bridge_delete_ofprotos(void);
196 static void bridge_delete_or_reconfigure_ports(struct bridge *);
197 static void bridge_del_ports(struct bridge *,
198 const struct shash *wanted_ports);
199 static void bridge_add_ports(struct bridge *,
200 const struct shash *wanted_ports);
202 static void bridge_configure_datapath_id(struct bridge *);
203 static void bridge_configure_netflow(struct bridge *);
204 static void bridge_configure_forward_bpdu(struct bridge *);
205 static void bridge_configure_mac_table(struct bridge *);
206 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
207 static void bridge_configure_ipfix(struct bridge *);
208 static void bridge_configure_stp(struct bridge *);
209 static void bridge_configure_tables(struct bridge *);
210 static void bridge_configure_dp_desc(struct bridge *);
211 static void bridge_configure_remotes(struct bridge *,
212 const struct sockaddr_in *managers,
214 static void bridge_pick_local_hw_addr(struct bridge *,
215 uint8_t ea[ETH_ADDR_LEN],
216 struct iface **hw_addr_iface);
217 static uint64_t bridge_pick_datapath_id(struct bridge *,
218 const uint8_t bridge_ea[ETH_ADDR_LEN],
219 struct iface *hw_addr_iface);
220 static uint64_t dpid_from_hash(const void *, size_t nbytes);
221 static bool bridge_has_bond_fake_iface(const struct bridge *,
223 static bool port_is_bond_fake_iface(const struct port *);
225 static unixctl_cb_func qos_unixctl_show;
227 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
228 static void port_del_ifaces(struct port *);
229 static void port_destroy(struct port *);
230 static struct port *port_lookup(const struct bridge *, const char *name);
231 static void port_configure(struct port *);
232 static struct lacp_settings *port_configure_lacp(struct port *,
233 struct lacp_settings *);
234 static void port_configure_bond(struct port *, struct bond_settings *);
235 static bool port_is_synthetic(const struct port *);
237 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
238 static void run_system_stats(void);
240 static void bridge_configure_mirrors(struct bridge *);
241 static struct mirror *mirror_create(struct bridge *,
242 const struct ovsrec_mirror *);
243 static void mirror_destroy(struct mirror *);
244 static bool mirror_configure(struct mirror *);
245 static void mirror_refresh_stats(struct mirror *);
247 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
248 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
249 const struct ovsrec_port *);
250 static bool iface_is_internal(const struct ovsrec_interface *iface,
251 const struct ovsrec_bridge *br);
252 static const char *iface_get_type(const struct ovsrec_interface *,
253 const struct ovsrec_bridge *);
254 static void iface_destroy(struct iface *);
255 static struct iface *iface_lookup(const struct bridge *, const char *name);
256 static struct iface *iface_find(const char *name);
257 static struct iface *iface_from_ofp_port(const struct bridge *,
258 ofp_port_t ofp_port);
259 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
260 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
261 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
262 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
263 static void iface_configure_cfm(struct iface *);
264 static void iface_refresh_cfm_stats(struct iface *);
265 static void iface_refresh_stats(struct iface *);
266 static void iface_refresh_status(struct iface *);
267 static bool iface_is_synthetic(const struct iface *);
268 static ofp_port_t iface_get_requested_ofp_port(
269 const struct ovsrec_interface *);
270 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
272 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
274 * This is deprecated. It is only for compatibility with broken device drivers
275 * in old versions of Linux that do not properly support VLANs when VLAN
276 * devices are not used. When broken device drivers are no longer in
277 * widespread use, we will delete these interfaces. */
279 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
280 static bool vlan_splinters_enabled_anywhere;
282 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
283 static unsigned long int *collect_splinter_vlans(
284 const struct ovsrec_open_vswitch *);
285 static void configure_splinter_port(struct port *);
286 static void add_vlan_splinter_ports(struct bridge *,
287 const unsigned long int *splinter_vlans,
288 struct shash *ports);
291 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
293 struct shash iface_hints;
294 static bool initialized = false;
301 shash_init(&iface_hints);
304 for (i = 0; i < cfg->n_bridges; i++) {
305 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
308 for (j = 0; j < br_cfg->n_ports; j++) {
309 struct ovsrec_port *port_cfg = br_cfg->ports[j];
312 for (k = 0; k < port_cfg->n_interfaces; k++) {
313 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
314 struct iface_hint *iface_hint;
316 iface_hint = xmalloc(sizeof *iface_hint);
317 iface_hint->br_name = br_cfg->name;
318 iface_hint->br_type = br_cfg->datapath_type;
319 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
321 shash_add(&iface_hints, if_cfg->name, iface_hint);
327 ofproto_init(&iface_hints);
329 shash_destroy_free_data(&iface_hints);
333 /* Public functions. */
335 /* Initializes the bridge module, configuring it to obtain its configuration
336 * from an OVSDB server accessed over 'remote', which should be a string in a
337 * form acceptable to ovsdb_idl_create(). */
339 bridge_init(const char *remote)
341 /* Create connection to database. */
342 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
343 idl_seqno = ovsdb_idl_get_seqno(idl);
344 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
345 ovsdb_idl_verify_write_only(idl);
347 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
348 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
349 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
351 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
352 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
353 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
355 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
356 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
357 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
359 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
360 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
361 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
378 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
379 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
380 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
381 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
382 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
384 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
385 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
386 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
387 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
389 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
391 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
393 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
394 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
396 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
397 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
398 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
401 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
402 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
403 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
404 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
405 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
407 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
409 /* Register unixctl commands. */
410 unixctl_command_register("qos/show", "interface", 1, 1,
411 qos_unixctl_show, NULL);
412 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
413 bridge_unixctl_dump_flows, NULL);
414 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
415 bridge_unixctl_reconnect, NULL);
425 struct bridge *br, *next_br;
427 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
430 ovsdb_idl_destroy(idl);
433 /* Enables use of OpenFlow 1.4. This is off by default because OpenFlow 1.4 is
434 * not yet safe to use: it can abort due to unimplemented features. */
436 bridge_enable_of14(void)
441 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
442 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
443 * responsible for freeing '*managersp' (with free()).
445 * You may be asking yourself "why does ovs-vswitchd care?", because
446 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
447 * should not be and in fact is not directly involved in that. But
448 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
449 * it has to tell in-band control where the managers are to enable that.
450 * (Thus, only managers connected in-band are collected.)
453 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
454 struct sockaddr_in **managersp, size_t *n_managersp)
456 struct sockaddr_in *managers = NULL;
457 size_t n_managers = 0;
461 /* Collect all of the potential targets from the "targets" columns of the
462 * rows pointed to by "manager_options", excluding any that are
465 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
466 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
468 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
469 sset_find_and_delete(&targets, m->target);
471 sset_add(&targets, m->target);
475 /* Now extract the targets' IP addresses. */
476 if (!sset_is_empty(&targets)) {
479 managers = xmalloc(sset_count(&targets) * sizeof *managers);
480 SSET_FOR_EACH (target, &targets) {
481 struct sockaddr_storage ss;
483 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
485 && ss.ss_family == AF_INET) {
486 managers[n_managers++] = *(struct sockaddr_in *) &ss;
490 sset_destroy(&targets);
492 *managersp = managers;
493 *n_managersp = n_managers;
497 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
499 unsigned long int *splinter_vlans;
500 struct sockaddr_in *managers;
501 struct bridge *br, *next;
502 int sflow_bridge_number;
505 COVERAGE_INC(bridge_reconfigure);
507 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
508 OFPROTO_FLOW_LIMIT_DEFAULT));
509 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
510 OFPROTO_MAX_IDLE_DEFAULT));
513 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
514 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
516 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
517 * to 'ovs_cfg', with only very minimal configuration otherwise.
519 * This is mostly an update to bridge data structures. Nothing is pushed
520 * down to ofproto or lower layers. */
521 add_del_bridges(ovs_cfg);
522 splinter_vlans = collect_splinter_vlans(ovs_cfg);
523 HMAP_FOR_EACH (br, node, &all_bridges) {
524 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
525 bridge_del_ports(br, &br->wanted_ports);
527 free(splinter_vlans);
529 /* Start pushing configuration changes down to the ofproto layer:
531 * - Delete ofprotos that are no longer configured.
533 * - Delete ports that are no longer configured.
535 * - Reconfigure existing ports to their desired configurations, or
536 * delete them if not possible.
538 * We have to do all the deletions before we can do any additions, because
539 * the ports to be added might require resources that will be freed up by
540 * deletions (they might especially overlap in name). */
541 bridge_delete_ofprotos();
542 HMAP_FOR_EACH (br, node, &all_bridges) {
544 bridge_delete_or_reconfigure_ports(br);
548 /* Finish pushing configuration changes to the ofproto layer:
550 * - Create ofprotos that are missing.
552 * - Add ports that are missing. */
553 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
557 error = ofproto_create(br->name, br->type, &br->ofproto);
559 VLOG_ERR("failed to create bridge %s: %s", br->name,
560 ovs_strerror(error));
561 shash_destroy(&br->wanted_ports);
566 HMAP_FOR_EACH (br, node, &all_bridges) {
567 bridge_add_ports(br, &br->wanted_ports);
568 shash_destroy(&br->wanted_ports);
571 reconfigure_system_stats(ovs_cfg);
573 /* Complete the configuration. */
574 sflow_bridge_number = 0;
575 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
576 HMAP_FOR_EACH (br, node, &all_bridges) {
579 /* We need the datapath ID early to allow LACP ports to use it as the
580 * default system ID. */
581 bridge_configure_datapath_id(br);
583 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
586 port_configure(port);
588 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
589 iface_set_ofport(iface->cfg, iface->ofp_port);
590 iface_configure_cfm(iface);
591 iface_configure_qos(iface, port->cfg->qos);
592 iface_set_mac(br, port, iface);
593 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
597 bridge_configure_mirrors(br);
598 bridge_configure_forward_bpdu(br);
599 bridge_configure_mac_table(br);
600 bridge_configure_remotes(br, managers, n_managers);
601 bridge_configure_netflow(br);
602 bridge_configure_sflow(br, &sflow_bridge_number);
603 bridge_configure_ipfix(br);
604 bridge_configure_stp(br);
605 bridge_configure_tables(br);
606 bridge_configure_dp_desc(br);
608 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
609 /* XXX: Remove this warning message eventually. */
610 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
611 " moved to the Open_vSwitch table. Ignoring its"
612 " setting in the bridge table.");
617 /* The ofproto-dpif provider does some final reconfiguration in its
618 * ->type_run() function. We have to call it before notifying the database
619 * client that reconfiguration is complete, otherwise there is a very
620 * narrow race window in which e.g. ofproto/trace will not recognize the
621 * new configuration (sometimes this causes unit test failures). */
625 /* Delete ofprotos which aren't configured or have the wrong type. Create
626 * ofprotos which don't exist but need to. */
628 bridge_delete_ofprotos(void)
635 /* Delete ofprotos with no bridge or with the wrong type. */
638 ofproto_enumerate_types(&types);
639 SSET_FOR_EACH (type, &types) {
642 ofproto_enumerate_names(type, &names);
643 SSET_FOR_EACH (name, &names) {
644 br = bridge_lookup(name);
645 if (!br || strcmp(type, br->type)) {
646 ofproto_delete(name, type);
650 sset_destroy(&names);
651 sset_destroy(&types);
655 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
657 if (*n >= *allocated) {
658 ports = x2nrealloc(ports, allocated, sizeof *ports);
660 ports[(*n)++] = port;
665 bridge_delete_or_reconfigure_ports(struct bridge *br)
667 struct ofproto_port ofproto_port;
668 struct ofproto_port_dump dump;
670 /* List of "ofp_port"s to delete. We make a list instead of deleting them
671 * right away because ofproto implementations aren't necessarily able to
672 * iterate through a changing list of ports in an entirely robust way. */
680 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
681 ofp_port_t requested_ofp_port;
684 iface = iface_lookup(br, ofproto_port.name);
686 /* No such iface is configured, so we should delete this
689 * As a corner case exception, keep the port if it's a bond fake
691 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
692 && !strcmp(ofproto_port.type, "internal")) {
698 if (strcmp(ofproto_port.type, iface->type)
699 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
700 /* The interface is the wrong type or can't be configured.
705 /* If the requested OpenFlow port for 'iface' changed, and it's not
706 * already the correct port, then we might want to temporarily delete
707 * this interface, so we can add it back again with the new OpenFlow
709 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
710 if (iface->ofp_port != OFPP_LOCAL &&
711 requested_ofp_port != OFPP_NONE &&
712 requested_ofp_port != iface->ofp_port) {
713 ofp_port_t victim_request;
714 struct iface *victim;
716 /* Check for an existing OpenFlow port currently occupying
717 * 'iface''s requested port number. If there isn't one, then
718 * delete this port. Otherwise we need to consider further. */
719 victim = iface_from_ofp_port(br, requested_ofp_port);
724 /* 'victim' is a port currently using 'iface''s requested port
725 * number. Unless 'victim' specifically requested that port
726 * number, too, then we can delete both 'iface' and 'victim'
727 * temporarily. (We'll add both of them back again later with new
728 * OpenFlow port numbers.)
730 * If 'victim' did request port number 'requested_ofp_port', just
731 * like 'iface', then that's a configuration inconsistency that we
732 * can't resolve. We might as well let it keep its current port
734 victim_request = iface_get_requested_ofp_port(victim->cfg);
735 if (victim_request != requested_ofp_port) {
736 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
737 iface_destroy(victim);
746 iface_destroy(iface);
747 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
750 for (i = 0; i < n; i++) {
751 ofproto_port_del(br->ofproto, del[i]);
757 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
758 bool with_requested_port)
760 struct shash_node *port_node;
762 SHASH_FOR_EACH (port_node, wanted_ports) {
763 const struct ovsrec_port *port_cfg = port_node->data;
766 for (i = 0; i < port_cfg->n_interfaces; i++) {
767 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
768 ofp_port_t requested_ofp_port;
770 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
771 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
772 struct iface *iface = iface_lookup(br, iface_cfg->name);
775 iface_create(br, iface_cfg, port_cfg);
783 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
785 /* First add interfaces that request a particular port number. */
786 bridge_add_ports__(br, wanted_ports, true);
788 /* Then add interfaces that want automatic port number assignment.
789 * We add these afterward to avoid accidentally taking a specifically
790 * requested port number. */
791 bridge_add_ports__(br, wanted_ports, false);
795 port_configure(struct port *port)
797 const struct ovsrec_port *cfg = port->cfg;
798 struct bond_settings bond_settings;
799 struct lacp_settings lacp_settings;
800 struct ofproto_bundle_settings s;
803 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
804 configure_splinter_port(port);
813 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
814 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
815 s.slaves[s.n_slaves++] = iface->ofp_port;
820 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
824 /* Get VLAN trunks. */
827 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
831 if (cfg->vlan_mode) {
832 if (!strcmp(cfg->vlan_mode, "access")) {
833 s.vlan_mode = PORT_VLAN_ACCESS;
834 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
835 s.vlan_mode = PORT_VLAN_TRUNK;
836 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
837 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
838 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
839 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
841 /* This "can't happen" because ovsdb-server should prevent it. */
842 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
843 "back to trunk mode", port->name, cfg->vlan_mode);
844 s.vlan_mode = PORT_VLAN_TRUNK;
848 s.vlan_mode = PORT_VLAN_ACCESS;
850 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
854 s.vlan_mode = PORT_VLAN_TRUNK;
857 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
860 /* Get LACP settings. */
861 s.lacp = port_configure_lacp(port, &lacp_settings);
865 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
866 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
867 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
870 s.lacp_slaves = NULL;
873 /* Get bond settings. */
874 if (s.n_slaves > 1) {
875 s.bond = &bond_settings;
876 port_configure_bond(port, &bond_settings);
879 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
880 netdev_set_miimon_interval(iface->netdev, 0);
885 ofproto_bundle_register(port->bridge->ofproto, port, &s);
893 /* Pick local port hardware address and datapath ID for 'br'. */
895 bridge_configure_datapath_id(struct bridge *br)
897 uint8_t ea[ETH_ADDR_LEN];
899 struct iface *local_iface;
900 struct iface *hw_addr_iface;
903 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
904 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
906 int error = netdev_set_etheraddr(local_iface->netdev, ea);
908 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
909 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
910 "Ethernet address: %s",
911 br->name, ovs_strerror(error));
914 memcpy(br->ea, ea, ETH_ADDR_LEN);
916 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
917 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
918 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
919 ofproto_set_datapath_id(br->ofproto, dpid);
922 dpid_string = xasprintf("%016"PRIx64, dpid);
923 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
927 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
930 bridge_get_allowed_versions(struct bridge *br)
932 uint32_t allowed_versions;
934 if (!br->cfg->n_protocols)
937 allowed_versions = ofputil_versions_from_strings(br->cfg->protocols,
938 br->cfg->n_protocols);
940 allowed_versions &= ~(1u << OFP14_VERSION);
942 return allowed_versions;
945 /* Set NetFlow configuration on 'br'. */
947 bridge_configure_netflow(struct bridge *br)
949 struct ovsrec_netflow *cfg = br->cfg->netflow;
950 struct netflow_options opts;
953 ofproto_set_netflow(br->ofproto, NULL);
957 memset(&opts, 0, sizeof opts);
959 /* Get default NetFlow configuration from datapath.
960 * Apply overrides from 'cfg'. */
961 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
962 if (cfg->engine_type) {
963 opts.engine_type = *cfg->engine_type;
965 if (cfg->engine_id) {
966 opts.engine_id = *cfg->engine_id;
969 /* Configure active timeout interval. */
970 opts.active_timeout = cfg->active_timeout;
971 if (!opts.active_timeout) {
972 opts.active_timeout = -1;
973 } else if (opts.active_timeout < 0) {
974 VLOG_WARN("bridge %s: active timeout interval set to negative "
975 "value, using default instead (%d seconds)", br->name,
976 NF_ACTIVE_TIMEOUT_DEFAULT);
977 opts.active_timeout = -1;
980 /* Add engine ID to interface number to disambiguate bridgs? */
981 opts.add_id_to_iface = cfg->add_id_to_interface;
982 if (opts.add_id_to_iface) {
983 if (opts.engine_id > 0x7f) {
984 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
985 "another vswitch, choose an engine id less than 128",
988 if (hmap_count(&br->ports) > 508) {
989 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
990 "another port when more than 508 ports are used",
996 sset_init(&opts.collectors);
997 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1000 if (ofproto_set_netflow(br->ofproto, &opts)) {
1001 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1003 sset_destroy(&opts.collectors);
1006 /* Set sFlow configuration on 'br'. */
1008 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1010 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1011 struct ovsrec_controller **controllers;
1012 struct ofproto_sflow_options oso;
1013 size_t n_controllers;
1017 ofproto_set_sflow(br->ofproto, NULL);
1021 memset(&oso, 0, sizeof oso);
1023 sset_init(&oso.targets);
1024 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1026 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1027 if (cfg->sampling) {
1028 oso.sampling_rate = *cfg->sampling;
1031 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1033 oso.polling_interval = *cfg->polling;
1036 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1038 oso.header_len = *cfg->header;
1041 oso.sub_id = (*sflow_bridge_number)++;
1042 oso.agent_device = cfg->agent;
1044 oso.control_ip = NULL;
1045 n_controllers = bridge_get_controllers(br, &controllers);
1046 for (i = 0; i < n_controllers; i++) {
1047 if (controllers[i]->local_ip) {
1048 oso.control_ip = controllers[i]->local_ip;
1052 ofproto_set_sflow(br->ofproto, &oso);
1054 sset_destroy(&oso.targets);
1057 /* Returns whether a IPFIX row is valid. */
1059 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1061 return ipfix && ipfix->n_targets > 0;
1064 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1066 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1067 const struct bridge *br)
1069 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1072 /* Set IPFIX configuration on 'br'. */
1074 bridge_configure_ipfix(struct bridge *br)
1076 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1077 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1078 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1079 struct ofproto_ipfix_bridge_exporter_options be_opts;
1080 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1081 size_t n_fe_opts = 0;
1083 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1084 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1089 if (!valid_be_cfg && n_fe_opts == 0) {
1090 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1095 memset(&be_opts, 0, sizeof be_opts);
1097 sset_init(&be_opts.targets);
1098 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1100 if (be_cfg->sampling) {
1101 be_opts.sampling_rate = *be_cfg->sampling;
1103 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1105 if (be_cfg->obs_domain_id) {
1106 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1108 if (be_cfg->obs_point_id) {
1109 be_opts.obs_point_id = *be_cfg->obs_point_id;
1111 if (be_cfg->cache_active_timeout) {
1112 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1114 if (be_cfg->cache_max_flows) {
1115 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1119 if (n_fe_opts > 0) {
1120 struct ofproto_ipfix_flow_exporter_options *opts;
1121 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1123 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1124 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1125 opts->collector_set_id = fe_cfg->id;
1126 sset_init(&opts->targets);
1127 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1128 fe_cfg->ipfix->n_targets);
1129 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1130 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1131 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1132 ? *fe_cfg->ipfix->cache_max_flows : 0;
1138 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1142 sset_destroy(&be_opts.targets);
1145 if (n_fe_opts > 0) {
1146 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1148 for (i = 0; i < n_fe_opts; i++) {
1149 sset_destroy(&opts->targets);
1157 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1158 struct ofproto_port_stp_settings *port_s,
1159 int *port_num_counter, unsigned long *port_num_bitmap)
1161 const char *config_str;
1162 struct iface *iface;
1164 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1165 port_s->enable = false;
1168 port_s->enable = true;
1171 /* STP over bonds is not supported. */
1172 if (!list_is_singleton(&port->ifaces)) {
1173 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1175 port_s->enable = false;
1179 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1181 /* Internal ports shouldn't participate in spanning tree, so
1183 if (!strcmp(iface->type, "internal")) {
1184 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1185 port_s->enable = false;
1189 /* STP on mirror output ports is not supported. */
1190 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1191 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1192 port_s->enable = false;
1196 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1198 unsigned long int port_num = strtoul(config_str, NULL, 0);
1199 int port_idx = port_num - 1;
1201 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1202 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1203 port_s->enable = false;
1207 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1208 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1209 port->name, port_num);
1210 port_s->enable = false;
1213 bitmap_set1(port_num_bitmap, port_idx);
1214 port_s->port_num = port_idx;
1216 if (*port_num_counter >= STP_MAX_PORTS) {
1217 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1218 port_s->enable = false;
1222 port_s->port_num = (*port_num_counter)++;
1225 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1227 port_s->path_cost = strtoul(config_str, NULL, 10);
1229 enum netdev_features current;
1232 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1233 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1234 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1237 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1239 port_s->priority = strtoul(config_str, NULL, 0);
1241 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1245 /* Set spanning tree configuration on 'br'. */
1247 bridge_configure_stp(struct bridge *br)
1249 if (!br->cfg->stp_enable) {
1250 ofproto_set_stp(br->ofproto, NULL);
1252 struct ofproto_stp_settings br_s;
1253 const char *config_str;
1255 int port_num_counter;
1256 unsigned long *port_num_bitmap;
1258 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1260 uint8_t ea[ETH_ADDR_LEN];
1262 if (eth_addr_from_string(config_str, ea)) {
1263 br_s.system_id = eth_addr_to_uint64(ea);
1265 br_s.system_id = eth_addr_to_uint64(br->ea);
1266 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1267 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1270 br_s.system_id = eth_addr_to_uint64(br->ea);
1273 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1275 br_s.priority = strtoul(config_str, NULL, 0);
1277 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1280 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1282 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1284 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1287 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1289 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1291 br_s.max_age = STP_DEFAULT_MAX_AGE;
1294 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1296 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1298 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1301 /* Configure STP on the bridge. */
1302 if (ofproto_set_stp(br->ofproto, &br_s)) {
1303 VLOG_ERR("bridge %s: could not enable STP", br->name);
1307 /* Users must either set the port number with the "stp-port-num"
1308 * configuration on all ports or none. If manual configuration
1309 * is not done, then we allocate them sequentially. */
1310 port_num_counter = 0;
1311 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1312 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1313 struct ofproto_port_stp_settings port_s;
1314 struct iface *iface;
1316 port_configure_stp(br->ofproto, port, &port_s,
1317 &port_num_counter, port_num_bitmap);
1319 /* As bonds are not supported, just apply configuration to
1320 * all interfaces. */
1321 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1322 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1324 VLOG_ERR("port %s: could not enable STP", port->name);
1330 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1331 && port_num_counter) {
1332 VLOG_ERR("bridge %s: must manually configure all STP port "
1333 "IDs or none, disabling", br->name);
1334 ofproto_set_stp(br->ofproto, NULL);
1336 bitmap_free(port_num_bitmap);
1341 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1343 const struct port *port = port_lookup(br, name);
1344 return port && port_is_bond_fake_iface(port);
1348 port_is_bond_fake_iface(const struct port *port)
1350 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1354 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1356 struct bridge *br, *next;
1357 struct shash new_br;
1360 /* Collect new bridges' names and types. */
1361 shash_init(&new_br);
1362 for (i = 0; i < cfg->n_bridges; i++) {
1363 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1364 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1366 if (strchr(br_cfg->name, '/')) {
1367 /* Prevent remote ovsdb-server users from accessing arbitrary
1368 * directories, e.g. consider a bridge named "../../../etc/". */
1369 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1371 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1372 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1376 /* Get rid of deleted bridges or those whose types have changed.
1377 * Update 'cfg' of bridges that still exist. */
1378 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1379 br->cfg = shash_find_data(&new_br, br->name);
1380 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1381 br->cfg->datapath_type))) {
1386 /* Add new bridges. */
1387 for (i = 0; i < cfg->n_bridges; i++) {
1388 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1389 struct bridge *br = bridge_lookup(br_cfg->name);
1391 bridge_create(br_cfg);
1395 shash_destroy(&new_br);
1398 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1399 * Returns 0 if successful, otherwise a positive errno value. */
1401 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1402 struct netdev *netdev)
1404 return netdev_set_config(netdev, &iface_cfg->options);
1407 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1408 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1410 * If successful, returns 0 and stores the network device in '*netdevp'. On
1411 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1413 iface_do_create(const struct bridge *br,
1414 const struct ovsrec_interface *iface_cfg,
1415 const struct ovsrec_port *port_cfg,
1416 ofp_port_t *ofp_portp, struct netdev **netdevp)
1418 struct netdev *netdev = NULL;
1421 if (netdev_is_reserved_name(iface_cfg->name)) {
1422 VLOG_WARN("could not create interface %s, name is reserved",
1428 error = netdev_open(iface_cfg->name,
1429 iface_get_type(iface_cfg, br->cfg), &netdev);
1431 VLOG_WARN("could not open network device %s (%s)",
1432 iface_cfg->name, ovs_strerror(error));
1436 error = iface_set_netdev_config(iface_cfg, netdev);
1441 *ofp_portp = iface_pick_ofport(iface_cfg);
1442 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1447 VLOG_INFO("bridge %s: added interface %s on port %d",
1448 br->name, iface_cfg->name, *ofp_portp);
1450 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1451 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1459 netdev_close(netdev);
1463 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1464 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1465 * automatically allocated for the iface. Takes ownership of and
1466 * deallocates 'if_cfg'.
1468 * Return true if an iface is successfully created, false otherwise. */
1470 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1471 const struct ovsrec_port *port_cfg)
1473 struct netdev *netdev;
1474 struct iface *iface;
1475 ofp_port_t ofp_port;
1479 /* Do the bits that can fail up front. */
1480 ovs_assert(!iface_lookup(br, iface_cfg->name));
1481 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
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 /* Populate initial status in database. */
1508 iface_refresh_stats(iface);
1509 iface_refresh_status(iface);
1511 /* Add bond fake iface if necessary. */
1512 if (port_is_bond_fake_iface(port)) {
1513 struct ofproto_port ofproto_port;
1515 if (ofproto_port_query_by_name(br->ofproto, port->name,
1517 struct netdev *netdev;
1520 error = netdev_open(port->name, "internal", &netdev);
1522 ofp_port_t fake_ofp_port = OFPP_NONE;
1523 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1524 netdev_close(netdev);
1526 VLOG_WARN("could not open network device %s (%s)",
1527 port->name, ovs_strerror(error));
1530 /* Already exists, nothing to do. */
1531 ofproto_port_destroy(&ofproto_port);
1538 /* Set forward BPDU option. */
1540 bridge_configure_forward_bpdu(struct bridge *br)
1542 ofproto_set_forward_bpdu(br->ofproto,
1543 smap_get_bool(&br->cfg->other_config,
1548 /* Set MAC learning table configuration for 'br'. */
1550 bridge_configure_mac_table(struct bridge *br)
1552 const char *idle_time_str;
1555 const char *mac_table_size_str;
1558 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1559 idle_time = (idle_time_str && atoi(idle_time_str)
1560 ? atoi(idle_time_str)
1561 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1563 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1564 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1565 ? atoi(mac_table_size_str)
1568 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1572 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1573 const struct port *fake_br, struct iface **hw_addr_iface)
1575 struct hmapx mirror_output_ports;
1577 bool found_addr = false;
1581 /* Mirror output ports don't participate in picking the local hardware
1582 * address. ofproto can't help us find out whether a given port is a
1583 * mirror output because we haven't configured mirrors yet, so we need to
1584 * accumulate them ourselves. */
1585 hmapx_init(&mirror_output_ports);
1586 for (i = 0; i < br->cfg->n_mirrors; i++) {
1587 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1588 if (m->output_port) {
1589 hmapx_add(&mirror_output_ports, m->output_port);
1593 /* Otherwise choose the minimum non-local MAC address among all of the
1595 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1596 uint8_t iface_ea[ETH_ADDR_LEN];
1597 struct iface *candidate;
1598 struct iface *iface;
1600 /* Mirror output ports don't participate. */
1601 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1605 /* Choose the MAC address to represent the port. */
1607 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1608 /* Find the interface with this Ethernet address (if any) so that
1609 * we can provide the correct devname to the caller. */
1610 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1611 uint8_t candidate_ea[ETH_ADDR_LEN];
1612 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1613 && eth_addr_equals(iface_ea, candidate_ea)) {
1618 /* Choose the interface whose MAC address will represent the port.
1619 * The Linux kernel bonding code always chooses the MAC address of
1620 * the first slave added to a bond, and the Fedora networking
1621 * scripts always add slaves to a bond in alphabetical order, so
1622 * for compatibility we choose the interface with the name that is
1623 * first in alphabetical order. */
1624 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1625 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1630 /* The local port doesn't count (since we're trying to choose its
1631 * MAC address anyway). */
1632 if (iface->ofp_port == OFPP_LOCAL) {
1636 /* For fake bridges we only choose from ports with the same tag */
1637 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1638 if (!port->cfg->tag) {
1641 if (*port->cfg->tag != *fake_br->cfg->tag) {
1647 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1653 /* Compare against our current choice. */
1654 if (!eth_addr_is_multicast(iface_ea) &&
1655 !eth_addr_is_local(iface_ea) &&
1656 !eth_addr_is_reserved(iface_ea) &&
1657 !eth_addr_is_zero(iface_ea) &&
1658 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1660 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1661 *hw_addr_iface = iface;
1667 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1668 *hw_addr_iface = NULL;
1671 hmapx_destroy(&mirror_output_ports);
1675 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1676 struct iface **hw_addr_iface)
1679 *hw_addr_iface = NULL;
1681 /* Did the user request a particular MAC? */
1682 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1683 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1684 if (eth_addr_is_multicast(ea)) {
1685 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1686 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1687 } else if (eth_addr_is_zero(ea)) {
1688 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1694 /* Find a local hw address */
1695 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
1698 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1699 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1700 * an interface on 'br', then that interface must be passed in as
1701 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1702 * 'hw_addr_iface' must be passed in as a null pointer. */
1704 bridge_pick_datapath_id(struct bridge *br,
1705 const uint8_t bridge_ea[ETH_ADDR_LEN],
1706 struct iface *hw_addr_iface)
1709 * The procedure for choosing a bridge MAC address will, in the most
1710 * ordinary case, also choose a unique MAC that we can use as a datapath
1711 * ID. In some special cases, though, multiple bridges will end up with
1712 * the same MAC address. This is OK for the bridges, but it will confuse
1713 * the OpenFlow controller, because each datapath needs a unique datapath
1716 * Datapath IDs must be unique. It is also very desirable that they be
1717 * stable from one run to the next, so that policy set on a datapath
1720 const char *datapath_id;
1723 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1724 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1728 if (!hw_addr_iface) {
1730 * A purely internal bridge, that is, one that has no non-virtual
1731 * network devices on it at all, is difficult because it has no
1732 * natural unique identifier at all.
1734 * When the host is a XenServer, we handle this case by hashing the
1735 * host's UUID with the name of the bridge. Names of bridges are
1736 * persistent across XenServer reboots, although they can be reused if
1737 * an internal network is destroyed and then a new one is later
1738 * created, so this is fairly effective.
1740 * When the host is not a XenServer, we punt by using a random MAC
1741 * address on each run.
1743 const char *host_uuid = xenserver_get_host_uuid();
1745 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1746 dpid = dpid_from_hash(combined, strlen(combined));
1752 return eth_addr_to_uint64(bridge_ea);
1756 dpid_from_hash(const void *data, size_t n)
1758 uint8_t hash[SHA1_DIGEST_SIZE];
1760 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1761 sha1_bytes(data, n, hash);
1762 eth_addr_mark_random(hash);
1763 return eth_addr_to_uint64(hash);
1767 iface_refresh_status(struct iface *iface)
1771 enum netdev_features current;
1775 uint8_t mac[ETH_ADDR_LEN];
1779 if (iface_is_synthetic(iface)) {
1785 if (!netdev_get_status(iface->netdev, &smap)) {
1786 ovsrec_interface_set_status(iface->cfg, &smap);
1788 ovsrec_interface_set_status(iface->cfg, NULL);
1791 smap_destroy(&smap);
1793 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1794 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1796 ovsrec_interface_set_duplex(iface->cfg,
1797 netdev_features_is_full_duplex(current)
1799 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1801 ovsrec_interface_set_duplex(iface->cfg, NULL);
1802 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1805 error = netdev_get_mtu(iface->netdev, &mtu);
1808 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1810 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1813 error = netdev_get_etheraddr(iface->netdev, mac);
1815 char mac_string[32];
1817 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1818 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1820 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1823 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1824 * if there is no valid ifindex number. */
1825 ifindex64 = netdev_get_ifindex(iface->netdev);
1826 if (ifindex64 < 0) {
1829 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1832 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1835 iface_refresh_cfm_stats(struct iface *iface)
1837 const struct ovsrec_interface *cfg = iface->cfg;
1838 struct ofproto_cfm_status status;
1840 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1841 iface->ofp_port, &status)) {
1842 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1843 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1844 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1845 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1846 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1847 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1849 const char *reasons[CFM_FAULT_N_REASONS];
1850 int64_t cfm_health = status.health;
1851 int64_t cfm_flap_count = status.flap_count;
1852 bool faulted = status.faults != 0;
1855 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1858 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1859 int reason = 1 << i;
1860 if (status.faults & reason) {
1861 reasons[j++] = cfm_fault_reason_to_str(reason);
1864 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1866 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1868 if (status.remote_opstate >= 0) {
1869 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1870 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1872 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1875 ovsrec_interface_set_cfm_remote_mpids(cfg,
1876 (const int64_t *)status.rmps,
1878 if (cfm_health >= 0) {
1879 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1881 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1889 iface_refresh_stats(struct iface *iface)
1891 #define IFACE_STATS \
1892 IFACE_STAT(rx_packets, "rx_packets") \
1893 IFACE_STAT(tx_packets, "tx_packets") \
1894 IFACE_STAT(rx_bytes, "rx_bytes") \
1895 IFACE_STAT(tx_bytes, "tx_bytes") \
1896 IFACE_STAT(rx_dropped, "rx_dropped") \
1897 IFACE_STAT(tx_dropped, "tx_dropped") \
1898 IFACE_STAT(rx_errors, "rx_errors") \
1899 IFACE_STAT(tx_errors, "tx_errors") \
1900 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1901 IFACE_STAT(rx_over_errors, "rx_over_err") \
1902 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1903 IFACE_STAT(collisions, "collisions")
1905 #define IFACE_STAT(MEMBER, NAME) + 1
1906 enum { N_IFACE_STATS = IFACE_STATS };
1908 int64_t values[N_IFACE_STATS];
1909 char *keys[N_IFACE_STATS];
1912 struct netdev_stats stats;
1914 if (iface_is_synthetic(iface)) {
1918 /* Intentionally ignore return value, since errors will set 'stats' to
1919 * all-1s, and we will deal with that correctly below. */
1920 netdev_get_stats(iface->netdev, &stats);
1922 /* Copy statistics into keys[] and values[]. */
1924 #define IFACE_STAT(MEMBER, NAME) \
1925 if (stats.MEMBER != UINT64_MAX) { \
1927 values[n] = stats.MEMBER; \
1932 ovs_assert(n <= N_IFACE_STATS);
1934 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1939 br_refresh_stp_status(struct bridge *br)
1941 struct smap smap = SMAP_INITIALIZER(&smap);
1942 struct ofproto *ofproto = br->ofproto;
1943 struct ofproto_stp_status status;
1945 if (ofproto_get_stp_status(ofproto, &status)) {
1949 if (!status.enabled) {
1950 ovsrec_bridge_set_status(br->cfg, NULL);
1954 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1955 STP_ID_ARGS(status.bridge_id));
1956 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1957 STP_ID_ARGS(status.designated_root));
1958 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1960 ovsrec_bridge_set_status(br->cfg, &smap);
1961 smap_destroy(&smap);
1965 port_refresh_stp_status(struct port *port)
1967 struct ofproto *ofproto = port->bridge->ofproto;
1968 struct iface *iface;
1969 struct ofproto_port_stp_status status;
1972 if (port_is_synthetic(port)) {
1976 /* STP doesn't currently support bonds. */
1977 if (!list_is_singleton(&port->ifaces)) {
1978 ovsrec_port_set_status(port->cfg, NULL);
1982 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1983 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1987 if (!status.enabled) {
1988 ovsrec_port_set_status(port->cfg, NULL);
1992 /* Set Status column. */
1994 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1995 smap_add(&smap, "stp_state", stp_state_name(status.state));
1996 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1997 smap_add(&smap, "stp_role", stp_role_name(status.role));
1998 ovsrec_port_set_status(port->cfg, &smap);
1999 smap_destroy(&smap);
2003 port_refresh_stp_stats(struct port *port)
2005 struct ofproto *ofproto = port->bridge->ofproto;
2006 struct iface *iface;
2007 struct ofproto_port_stp_stats stats;
2009 int64_t int_values[3];
2011 if (port_is_synthetic(port)) {
2015 /* STP doesn't currently support bonds. */
2016 if (!list_is_singleton(&port->ifaces)) {
2020 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2021 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2025 if (!stats.enabled) {
2026 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2030 /* Set Statistics column. */
2031 keys[0] = "stp_tx_count";
2032 int_values[0] = stats.tx_count;
2033 keys[1] = "stp_rx_count";
2034 int_values[1] = stats.rx_count;
2035 keys[2] = "stp_error_count";
2036 int_values[2] = stats.error_count;
2038 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2039 ARRAY_SIZE(int_values));
2043 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2045 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2049 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2051 bool enable = enable_system_stats(cfg);
2053 system_stats_enable(enable);
2055 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2060 run_system_stats(void)
2062 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2065 stats = system_stats_run();
2067 struct ovsdb_idl_txn *txn;
2068 struct ovsdb_datum datum;
2070 txn = ovsdb_idl_txn_create(idl);
2071 ovsdb_datum_from_smap(&datum, stats);
2072 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2074 ovsdb_idl_txn_commit(txn);
2075 ovsdb_idl_txn_destroy(txn);
2082 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2085 case OFPCR12_ROLE_EQUAL:
2087 case OFPCR12_ROLE_MASTER:
2089 case OFPCR12_ROLE_SLAVE:
2091 case OFPCR12_ROLE_NOCHANGE:
2093 return "*** INVALID ROLE ***";
2098 refresh_controller_status(void)
2102 const struct ovsrec_controller *cfg;
2106 /* Accumulate status for controllers on all bridges. */
2107 HMAP_FOR_EACH (br, node, &all_bridges) {
2108 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2111 /* Update each controller in the database with current status. */
2112 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2113 struct ofproto_controller_info *cinfo =
2114 shash_find_data(&info, cfg->target);
2117 struct smap smap = SMAP_INITIALIZER(&smap);
2118 const char **values = cinfo->pairs.values;
2119 const char **keys = cinfo->pairs.keys;
2122 for (i = 0; i < cinfo->pairs.n; i++) {
2123 smap_add(&smap, keys[i], values[i]);
2126 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2127 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2129 ovsrec_controller_set_status(cfg, &smap);
2130 smap_destroy(&smap);
2132 ovsrec_controller_set_is_connected(cfg, false);
2133 ovsrec_controller_set_role(cfg, NULL);
2134 ovsrec_controller_set_status(cfg, NULL);
2138 ofproto_free_ofproto_controller_info(&info);
2143 * Some information in the database must be kept as up-to-date as possible to
2144 * allow controllers to respond rapidly to network outages. We call these
2145 * statistics "instant" stats.
2147 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2148 * assuming that they've changed. The only means we have to determine whether
2149 * they have changed are:
2151 * - Try to commit changes to the database. If nothing changed, then
2152 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2155 * - instant_stats_run() is called late in the run loop, after anything that
2156 * might change any of the instant stats.
2158 * We use these two facts together to avoid waking the process up every
2159 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2162 /* Minimum interval between writing updates to the instant stats to the
2164 #define INSTANT_INTERVAL_MSEC 100
2166 /* Current instant stats database transaction, NULL if there is no ongoing
2168 static struct ovsdb_idl_txn *instant_txn;
2170 /* Next time (in msec on monotonic clock) at which we will update the instant
2172 static long long int instant_next_txn = LLONG_MIN;
2174 /* True if the run loop has run since we last saw that the instant stats were
2175 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2176 * to refresh the instant stats. */
2177 static bool instant_stats_could_have_changed;
2180 instant_stats_run(void)
2182 enum ovsdb_idl_txn_status status;
2184 instant_stats_could_have_changed = true;
2190 if (time_msec() < instant_next_txn) {
2193 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2195 seq = seq_read(connectivity_seq_get());
2196 if (seq == connectivity_seqno) {
2199 connectivity_seqno = seq;
2201 instant_txn = ovsdb_idl_txn_create(idl);
2202 HMAP_FOR_EACH (br, node, &all_bridges) {
2203 struct iface *iface;
2206 br_refresh_stp_status(br);
2208 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2209 port_refresh_stp_status(port);
2212 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2213 enum netdev_flags flags;
2215 const char *link_state;
2216 int64_t link_resets;
2219 if (iface_is_synthetic(iface)) {
2223 current = ofproto_port_is_lacp_current(br->ofproto,
2227 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2229 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2232 error = netdev_get_flags(iface->netdev, &flags);
2234 const char *state = flags & NETDEV_UP ? "up" : "down";
2235 ovsrec_interface_set_admin_state(iface->cfg, state);
2237 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2240 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2241 ovsrec_interface_set_link_state(iface->cfg, link_state);
2243 link_resets = netdev_get_carrier_resets(iface->netdev);
2244 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2246 iface_refresh_cfm_stats(iface);
2249 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2251 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2252 smap_destroy(&smap);
2257 status = ovsdb_idl_txn_commit(instant_txn);
2258 if (status != TXN_INCOMPLETE) {
2259 ovsdb_idl_txn_destroy(instant_txn);
2262 if (status == TXN_UNCHANGED) {
2263 instant_stats_could_have_changed = false;
2268 instant_stats_wait(void)
2271 ovsdb_idl_txn_wait(instant_txn);
2272 } else if (instant_stats_could_have_changed) {
2273 poll_timer_wait_until(instant_next_txn);
2284 /* Let each datapath type do the work that it needs to do. */
2286 ofproto_enumerate_types(&types);
2287 SSET_FOR_EACH (type, &types) {
2288 ofproto_type_run(type);
2290 sset_destroy(&types);
2292 /* Let each bridge do the work that it needs to do. */
2293 HMAP_FOR_EACH (br, node, &all_bridges) {
2294 ofproto_run(br->ofproto);
2301 static struct ovsrec_open_vswitch null_cfg;
2302 const struct ovsrec_open_vswitch *cfg;
2304 bool vlan_splinters_changed;
2307 ovsrec_open_vswitch_init(&null_cfg);
2311 if (ovsdb_idl_is_lock_contended(idl)) {
2312 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2313 struct bridge *br, *next_br;
2315 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2316 "disabling this process (pid %ld) until it goes away",
2317 (long int) getpid());
2319 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2322 /* Since we will not be running system_stats_run() in this process
2323 * with the current situation of multiple ovs-vswitchd daemons,
2324 * disable system stats collection. */
2325 system_stats_enable(false);
2327 } else if (!ovsdb_idl_has_lock(idl)) {
2330 cfg = ovsrec_open_vswitch_first(idl);
2332 /* Initialize the ofproto library. This only needs to run once, but
2333 * it must be done after the configuration is set. If the
2334 * initialization has already occurred, bridge_init_ofproto()
2335 * returns immediately. */
2336 bridge_init_ofproto(cfg);
2338 /* Once the value of flow-restore-wait is false, we no longer should
2339 * check its value from the database. */
2340 if (cfg && ofproto_get_flow_restore_wait()) {
2341 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2342 "flow-restore-wait", false));
2347 /* Re-configure SSL. We do this on every trip through the main loop,
2348 * instead of just when the database changes, because the contents of the
2349 * key and certificate files can change without the database changing.
2351 * We do this before bridge_reconfigure() because that function might
2352 * initiate SSL connections and thus requires SSL to be configured. */
2353 if (cfg && cfg->ssl) {
2354 const struct ovsrec_ssl *ssl = cfg->ssl;
2356 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2357 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2360 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2361 * usage has changed. */
2362 vlan_splinters_changed = false;
2363 if (vlan_splinters_enabled_anywhere) {
2364 HMAP_FOR_EACH (br, node, &all_bridges) {
2365 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2366 vlan_splinters_changed = true;
2372 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2373 struct ovsdb_idl_txn *txn;
2375 idl_seqno = ovsdb_idl_get_seqno(idl);
2376 txn = ovsdb_idl_txn_create(idl);
2377 bridge_reconfigure(cfg ? cfg : &null_cfg);
2380 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2383 /* If we are completing our initial configuration for this run
2384 * of ovs-vswitchd, then keep the transaction around to monitor
2385 * it for completion. */
2386 if (initial_config_done) {
2387 ovsdb_idl_txn_commit(txn);
2388 ovsdb_idl_txn_destroy(txn);
2390 initial_config_done = true;
2391 daemonize_txn = txn;
2395 if (daemonize_txn) {
2396 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2397 if (status != TXN_INCOMPLETE) {
2398 ovsdb_idl_txn_destroy(daemonize_txn);
2399 daemonize_txn = NULL;
2401 /* ovs-vswitchd has completed initialization, so allow the
2402 * process that forked us to exit successfully. */
2403 daemonize_complete();
2405 vlog_enable_async();
2407 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2411 /* Refresh interface and mirror stats if necessary. */
2412 if (time_msec() >= iface_stats_timer) {
2414 struct ovsdb_idl_txn *txn;
2416 txn = ovsdb_idl_txn_create(idl);
2417 HMAP_FOR_EACH (br, node, &all_bridges) {
2421 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2422 struct iface *iface;
2424 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2425 iface_refresh_stats(iface);
2426 iface_refresh_status(iface);
2429 port_refresh_stp_stats(port);
2432 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2433 mirror_refresh_stats(m);
2437 refresh_controller_status();
2438 ovsdb_idl_txn_commit(txn);
2439 ovsdb_idl_txn_destroy(txn); /* XXX */
2442 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2446 instant_stats_run();
2455 ovsdb_idl_wait(idl);
2456 if (daemonize_txn) {
2457 ovsdb_idl_txn_wait(daemonize_txn);
2461 ofproto_enumerate_types(&types);
2462 SSET_FOR_EACH (type, &types) {
2463 ofproto_type_wait(type);
2465 sset_destroy(&types);
2467 if (!hmap_is_empty(&all_bridges)) {
2470 HMAP_FOR_EACH (br, node, &all_bridges) {
2471 ofproto_wait(br->ofproto);
2473 poll_timer_wait_until(iface_stats_timer);
2476 system_stats_wait();
2477 instant_stats_wait();
2480 /* Adds some memory usage statistics for bridges into 'usage', for use with
2481 * memory_report(). */
2483 bridge_get_memory_usage(struct simap *usage)
2490 ofproto_enumerate_types(&types);
2491 SSET_FOR_EACH (type, &types) {
2492 ofproto_type_get_memory_usage(type, usage);
2494 sset_destroy(&types);
2496 HMAP_FOR_EACH (br, node, &all_bridges) {
2497 ofproto_get_memory_usage(br->ofproto, usage);
2501 /* QoS unixctl user interface functions. */
2503 struct qos_unixctl_show_cbdata {
2505 struct iface *iface;
2509 qos_unixctl_show_queue(unsigned int queue_id,
2510 const struct smap *details,
2511 struct iface *iface,
2514 struct netdev_queue_stats stats;
2515 struct smap_node *node;
2518 ds_put_cstr(ds, "\n");
2520 ds_put_format(ds, "Queue %u:\n", queue_id);
2522 ds_put_cstr(ds, "Default:\n");
2525 SMAP_FOR_EACH (node, details) {
2526 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2529 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2531 if (stats.tx_packets != UINT64_MAX) {
2532 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2535 if (stats.tx_bytes != UINT64_MAX) {
2536 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2539 if (stats.tx_errors != UINT64_MAX) {
2540 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2543 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2544 queue_id, ovs_strerror(error));
2549 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2550 const char *argv[], void *aux OVS_UNUSED)
2552 struct ds ds = DS_EMPTY_INITIALIZER;
2553 struct smap smap = SMAP_INITIALIZER(&smap);
2554 struct iface *iface;
2556 struct smap_node *node;
2558 iface = iface_find(argv[1]);
2560 unixctl_command_reply_error(conn, "no such interface");
2564 netdev_get_qos(iface->netdev, &type, &smap);
2566 if (*type != '\0') {
2567 struct netdev_queue_dump dump;
2568 struct smap details;
2569 unsigned int queue_id;
2571 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2573 SMAP_FOR_EACH (node, &smap) {
2574 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2577 smap_init(&details);
2578 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2579 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2581 smap_destroy(&details);
2583 unixctl_command_reply(conn, ds_cstr(&ds));
2585 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2586 unixctl_command_reply_error(conn, ds_cstr(&ds));
2589 smap_destroy(&smap);
2593 /* Bridge reconfiguration functions. */
2595 bridge_create(const struct ovsrec_bridge *br_cfg)
2599 ovs_assert(!bridge_lookup(br_cfg->name));
2600 br = xzalloc(sizeof *br);
2602 br->name = xstrdup(br_cfg->name);
2603 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2606 /* Derive the default Ethernet address from the bridge's UUID. This should
2607 * be unique and it will be stable between ovs-vswitchd runs. */
2608 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2609 eth_addr_mark_random(br->default_ea);
2611 hmap_init(&br->ports);
2612 hmap_init(&br->ifaces);
2613 hmap_init(&br->iface_by_name);
2614 hmap_init(&br->mirrors);
2616 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2620 bridge_destroy(struct bridge *br)
2623 struct mirror *mirror, *next_mirror;
2624 struct port *port, *next_port;
2626 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2629 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2630 mirror_destroy(mirror);
2633 hmap_remove(&all_bridges, &br->node);
2634 ofproto_destroy(br->ofproto);
2635 hmap_destroy(&br->ifaces);
2636 hmap_destroy(&br->ports);
2637 hmap_destroy(&br->iface_by_name);
2638 hmap_destroy(&br->mirrors);
2645 static struct bridge *
2646 bridge_lookup(const char *name)
2650 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2651 if (!strcmp(br->name, name)) {
2658 /* Handle requests for a listing of all flows known by the OpenFlow
2659 * stack, including those normally hidden. */
2661 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2662 const char *argv[], void *aux OVS_UNUSED)
2667 br = bridge_lookup(argv[1]);
2669 unixctl_command_reply_error(conn, "Unknown bridge");
2674 ofproto_get_all_flows(br->ofproto, &results);
2676 unixctl_command_reply(conn, ds_cstr(&results));
2677 ds_destroy(&results);
2680 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2681 * connections and reconnect. If BRIDGE is not specified, then all bridges
2682 * drop their controller connections and reconnect. */
2684 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2685 const char *argv[], void *aux OVS_UNUSED)
2689 br = bridge_lookup(argv[1]);
2691 unixctl_command_reply_error(conn, "Unknown bridge");
2694 ofproto_reconnect_controllers(br->ofproto);
2696 HMAP_FOR_EACH (br, node, &all_bridges) {
2697 ofproto_reconnect_controllers(br->ofproto);
2700 unixctl_command_reply(conn, NULL);
2704 bridge_get_controllers(const struct bridge *br,
2705 struct ovsrec_controller ***controllersp)
2707 struct ovsrec_controller **controllers;
2708 size_t n_controllers;
2710 controllers = br->cfg->controller;
2711 n_controllers = br->cfg->n_controller;
2713 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2719 *controllersp = controllers;
2721 return n_controllers;
2725 bridge_collect_wanted_ports(struct bridge *br,
2726 const unsigned long int *splinter_vlans,
2727 struct shash *wanted_ports)
2731 shash_init(wanted_ports);
2733 for (i = 0; i < br->cfg->n_ports; i++) {
2734 const char *name = br->cfg->ports[i]->name;
2735 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2736 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2740 if (bridge_get_controllers(br, NULL)
2741 && !shash_find(wanted_ports, br->name)) {
2742 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2743 br->name, br->name);
2745 ovsrec_interface_init(&br->synth_local_iface);
2746 ovsrec_port_init(&br->synth_local_port);
2748 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2749 br->synth_local_port.n_interfaces = 1;
2750 br->synth_local_port.name = br->name;
2752 br->synth_local_iface.name = br->name;
2753 br->synth_local_iface.type = "internal";
2755 br->synth_local_ifacep = &br->synth_local_iface;
2757 shash_add(wanted_ports, br->name, &br->synth_local_port);
2760 if (splinter_vlans) {
2761 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2765 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2766 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2767 * 'br' needs to complete its configuration. */
2769 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2771 struct shash_node *port_node;
2772 struct port *port, *next;
2774 /* Get rid of deleted ports.
2775 * Get rid of deleted interfaces on ports that still exist. */
2776 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2777 port->cfg = shash_find_data(wanted_ports, port->name);
2781 port_del_ifaces(port);
2785 /* Update iface->cfg and iface->type in interfaces that still exist. */
2786 SHASH_FOR_EACH (port_node, wanted_ports) {
2787 const struct ovsrec_port *port = port_node->data;
2790 for (i = 0; i < port->n_interfaces; i++) {
2791 const struct ovsrec_interface *cfg = port->interfaces[i];
2792 struct iface *iface = iface_lookup(br, cfg->name);
2793 const char *type = iface_get_type(cfg, br->cfg);
2798 } else if (!strcmp(type, "null")) {
2799 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2800 " may be removed in February 2013. Please email"
2801 " dev@openvswitch.org with concerns.",
2804 /* We will add new interfaces later. */
2810 /* Initializes 'oc' appropriately as a management service controller for
2813 * The caller must free oc->target when it is no longer needed. */
2815 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2816 struct ofproto_controller *oc)
2818 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2819 oc->max_backoff = 0;
2820 oc->probe_interval = 60;
2821 oc->band = OFPROTO_OUT_OF_BAND;
2823 oc->burst_limit = 0;
2824 oc->enable_async_msgs = true;
2827 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2829 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2830 struct ofproto_controller *oc)
2834 oc->target = c->target;
2835 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2836 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2837 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2838 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2839 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2840 oc->burst_limit = (c->controller_burst_limit
2841 ? *c->controller_burst_limit : 0);
2842 oc->enable_async_msgs = (!c->enable_async_messages
2843 || *c->enable_async_messages);
2844 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2845 if (dscp < 0 || dscp > 63) {
2846 dscp = DSCP_DEFAULT;
2851 /* Configures the IP stack for 'br''s local interface properly according to the
2852 * configuration in 'c'. */
2854 bridge_configure_local_iface_netdev(struct bridge *br,
2855 struct ovsrec_controller *c)
2857 struct netdev *netdev;
2858 struct in_addr mask, gateway;
2860 struct iface *local_iface;
2863 /* If there's no local interface or no IP address, give up. */
2864 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2865 if (!local_iface || !c->local_ip
2866 || !inet_pton(AF_INET, c->local_ip, &ip)) {
2870 /* Bring up the local interface. */
2871 netdev = local_iface->netdev;
2872 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2874 /* Configure the IP address and netmask. */
2875 if (!c->local_netmask
2876 || !inet_pton(AF_INET, c->local_netmask, &mask)
2878 mask.s_addr = guess_netmask(ip.s_addr);
2880 if (!netdev_set_in4(netdev, ip, mask)) {
2881 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2882 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2885 /* Configure the default gateway. */
2886 if (c->local_gateway
2887 && inet_pton(AF_INET, c->local_gateway, &gateway)
2888 && gateway.s_addr) {
2889 if (!netdev_add_router(netdev, gateway)) {
2890 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2891 br->name, IP_ARGS(gateway.s_addr));
2896 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2897 * in either string are treated as equal to any number of slashes in the other,
2898 * e.g. "x///y" is equal to "x/y".
2900 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2901 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2902 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2903 * 'b' against a prefix of 'a'.
2906 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2908 const char *b_start = b;
2910 if (b - b_start >= b_stoplen) {
2912 } else if (*a != *b) {
2914 } else if (*a == '/') {
2915 a += strspn(a, "/");
2916 b += strspn(b, "/");
2917 } else if (*a == '\0') {
2927 bridge_configure_remotes(struct bridge *br,
2928 const struct sockaddr_in *managers, size_t n_managers)
2930 bool disable_in_band;
2932 struct ovsrec_controller **controllers;
2933 size_t n_controllers;
2935 enum ofproto_fail_mode fail_mode;
2937 struct ofproto_controller *ocs;
2941 /* Check if we should disable in-band control on this bridge. */
2942 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2945 /* Set OpenFlow queue ID for in-band control. */
2946 ofproto_set_in_band_queue(br->ofproto,
2947 smap_get_int(&br->cfg->other_config,
2948 "in-band-queue", -1));
2950 if (disable_in_band) {
2951 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2953 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2956 n_controllers = bridge_get_controllers(br, &controllers);
2958 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2961 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2962 for (i = 0; i < n_controllers; i++) {
2963 struct ovsrec_controller *c = controllers[i];
2965 if (!strncmp(c->target, "punix:", 6)
2966 || !strncmp(c->target, "unix:", 5)) {
2967 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2970 if (!strncmp(c->target, "unix:", 5)) {
2971 /* Connect to a listening socket */
2972 whitelist = xasprintf("unix:%s/", ovs_rundir());
2973 if (strchr(c->target, '/') &&
2974 !equal_pathnames(c->target, whitelist,
2975 strlen(whitelist))) {
2976 /* Absolute path specified, but not in ovs_rundir */
2977 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2978 "controller \"%s\" due to possibility for "
2979 "remote exploit. Instead, specify socket "
2980 "in whitelisted \"%s\" or connect to "
2981 "\"unix:%s/%s.mgmt\" (which is always "
2982 "available without special configuration).",
2983 br->name, c->target, whitelist,
2984 ovs_rundir(), br->name);
2989 whitelist = xasprintf("punix:%s/%s.controller",
2990 ovs_rundir(), br->name);
2991 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2992 /* Prevent remote ovsdb-server users from accessing
2993 * arbitrary Unix domain sockets and overwriting arbitrary
2995 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2996 "controller \"%s\" due to possibility of "
2997 "overwriting local files. Instead, specify "
2998 "whitelisted \"%s\" or connect to "
2999 "\"unix:%s/%s.mgmt\" (which is always "
3000 "available without special configuration).",
3001 br->name, c->target, whitelist,
3002 ovs_rundir(), br->name);
3011 bridge_configure_local_iface_netdev(br, c);
3012 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3013 if (disable_in_band) {
3014 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3019 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3020 bridge_get_allowed_versions(br));
3021 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3024 /* Set the fail-mode. */
3025 fail_mode = !br->cfg->fail_mode
3026 || !strcmp(br->cfg->fail_mode, "standalone")
3027 ? OFPROTO_FAIL_STANDALONE
3028 : OFPROTO_FAIL_SECURE;
3029 ofproto_set_fail_mode(br->ofproto, fail_mode);
3031 /* Configure OpenFlow controller connection snooping. */
3032 if (!ofproto_has_snoops(br->ofproto)) {
3036 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3037 ovs_rundir(), br->name));
3038 ofproto_set_snoops(br->ofproto, &snoops);
3039 sset_destroy(&snoops);
3044 bridge_configure_tables(struct bridge *br)
3046 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3050 n_tables = ofproto_get_n_tables(br->ofproto);
3052 for (i = 0; i < n_tables; i++) {
3053 struct ofproto_table_settings s;
3056 s.max_flows = UINT_MAX;
3059 s.n_prefix_fields = 0;
3060 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3062 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3063 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3066 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3067 s.max_flows = *cfg->flow_limit;
3069 if (cfg->overflow_policy
3070 && !strcmp(cfg->overflow_policy, "evict")) {
3072 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3073 for (k = 0; k < cfg->n_groups; k++) {
3074 const char *string = cfg->groups[k];
3077 msg = mf_parse_subfield__(&s.groups[k], &string);
3079 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3080 "'groups' (%s)", br->name, i, msg);
3082 } else if (*string) {
3083 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3084 "element '%s' contains trailing garbage",
3085 br->name, i, cfg->groups[k]);
3091 /* Prefix lookup fields. */
3092 s.n_prefix_fields = 0;
3093 for (k = 0; k < cfg->n_prefixes; k++) {
3094 const char *name = cfg->prefixes[k];
3095 const struct mf_field *mf = mf_from_name(name);
3097 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3101 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3102 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3103 "%s", br->name, name);
3106 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3107 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3108 "field not used: %s", br->name, name);
3111 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3113 if (s.n_prefix_fields > 0) {
3115 struct ds ds = DS_EMPTY_INITIALIZER;
3116 for (k = 0; k < s.n_prefix_fields; k++) {
3118 ds_put_char(&ds, ',');
3120 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3122 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3123 br->name, i, ds_cstr(&ds));
3128 ofproto_configure_table(br->ofproto, i, &s);
3132 for (; j < br->cfg->n_flow_tables; j++) {
3133 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3134 "%"PRId64" not supported by this datapath", br->name,
3135 br->cfg->key_flow_tables[j]);
3140 bridge_configure_dp_desc(struct bridge *br)
3142 ofproto_set_dp_desc(br->ofproto,
3143 smap_get(&br->cfg->other_config, "dp-desc"));
3146 /* Port functions. */
3148 static void iface_destroy__(struct iface *);
3150 static struct port *
3151 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3155 port = xzalloc(sizeof *port);
3157 port->name = xstrdup(cfg->name);
3159 list_init(&port->ifaces);
3161 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3165 /* Deletes interfaces from 'port' that are no longer configured for it. */
3167 port_del_ifaces(struct port *port)
3169 struct iface *iface, *next;
3170 struct sset new_ifaces;
3173 /* Collect list of new interfaces. */
3174 sset_init(&new_ifaces);
3175 for (i = 0; i < port->cfg->n_interfaces; i++) {
3176 const char *name = port->cfg->interfaces[i]->name;
3177 const char *type = port->cfg->interfaces[i]->type;
3178 if (strcmp(type, "null")) {
3179 sset_add(&new_ifaces, name);
3183 /* Get rid of deleted interfaces. */
3184 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3185 if (!sset_contains(&new_ifaces, iface->name)) {
3186 iface_destroy(iface);
3190 sset_destroy(&new_ifaces);
3194 port_destroy(struct port *port)
3197 struct bridge *br = port->bridge;
3198 struct iface *iface, *next;
3201 ofproto_bundle_unregister(br->ofproto, port);
3204 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3205 iface_destroy__(iface);
3208 hmap_remove(&br->ports, &port->hmap_node);
3214 static struct port *
3215 port_lookup(const struct bridge *br, const char *name)
3219 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3221 if (!strcmp(port->name, name)) {
3229 enable_lacp(struct port *port, bool *activep)
3231 if (!port->cfg->lacp) {
3232 /* XXX when LACP implementation has been sufficiently tested, enable by
3233 * default and make active on bonded ports. */
3235 } else if (!strcmp(port->cfg->lacp, "off")) {
3237 } else if (!strcmp(port->cfg->lacp, "active")) {
3240 } else if (!strcmp(port->cfg->lacp, "passive")) {
3244 VLOG_WARN("port %s: unknown LACP mode %s",
3245 port->name, port->cfg->lacp);
3250 static struct lacp_settings *
3251 port_configure_lacp(struct port *port, struct lacp_settings *s)
3253 const char *lacp_time, *system_id;
3256 if (!enable_lacp(port, &s->active)) {
3260 s->name = port->name;
3262 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3264 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3265 ETH_ADDR_SCAN_ARGS(s->id))) {
3266 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3267 " address.", port->name, system_id);
3271 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3274 if (eth_addr_is_zero(s->id)) {
3275 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3279 /* Prefer bondable links if unspecified. */
3280 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3282 s->priority = (priority > 0 && priority <= UINT16_MAX
3284 : UINT16_MAX - !list_is_short(&port->ifaces));
3286 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3287 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3289 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3290 "lacp-fallback-ab", false);
3296 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3298 int priority, portid, key;
3300 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3301 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3303 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3305 if (portid <= 0 || portid > UINT16_MAX) {
3306 portid = ofp_to_u16(iface->ofp_port);
3309 if (priority <= 0 || priority > UINT16_MAX) {
3310 priority = UINT16_MAX;
3313 if (key < 0 || key > UINT16_MAX) {
3317 s->name = iface->name;
3319 s->priority = priority;
3324 port_configure_bond(struct port *port, struct bond_settings *s)
3326 const char *detect_s;
3327 struct iface *iface;
3328 int miimon_interval;
3330 s->name = port->name;
3332 if (port->cfg->bond_mode) {
3333 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3334 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3335 port->name, port->cfg->bond_mode,
3336 bond_mode_to_string(s->balance));
3339 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3341 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3342 * active-backup. At some point we should remove this warning. */
3343 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3344 " in previous versions, the default bond_mode was"
3345 " balance-slb", port->name,
3346 bond_mode_to_string(s->balance));
3348 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3349 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3350 "please use another bond type or disable flood_vlans",
3354 miimon_interval = smap_get_int(&port->cfg->other_config,
3355 "bond-miimon-interval", 0);
3356 if (miimon_interval <= 0) {
3357 miimon_interval = 200;
3360 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3361 if (!detect_s || !strcmp(detect_s, "carrier")) {
3362 miimon_interval = 0;
3363 } else if (strcmp(detect_s, "miimon")) {
3364 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3365 "defaulting to carrier", port->name, detect_s);
3366 miimon_interval = 0;
3369 s->up_delay = MAX(0, port->cfg->bond_updelay);
3370 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3371 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3372 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3373 "bond-rebalance-interval", 10000);
3374 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3375 s->rebalance_interval = 1000;
3378 s->fake_iface = port->cfg->bond_fake_iface;
3380 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3381 "lacp-fallback-ab", false);
3383 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3384 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3388 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3389 * instead of obtaining it from the database. */
3391 port_is_synthetic(const struct port *port)
3393 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3396 /* Interface functions. */
3399 iface_is_internal(const struct ovsrec_interface *iface,
3400 const struct ovsrec_bridge *br)
3402 /* The local port and "internal" ports are always "internal". */
3403 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3406 /* Returns the correct network device type for interface 'iface' in bridge
3409 iface_get_type(const struct ovsrec_interface *iface,
3410 const struct ovsrec_bridge *br)
3414 /* The local port always has type "internal". Other ports take
3415 * their type from the database and default to "system" if none is
3417 if (iface_is_internal(iface, br)) {
3420 type = iface->type[0] ? iface->type : "system";
3423 return ofproto_port_open_type(br->datapath_type, type);
3427 iface_destroy__(struct iface *iface)
3430 struct port *port = iface->port;
3431 struct bridge *br = port->bridge;
3433 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3434 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3437 if (iface->ofp_port != OFPP_NONE) {
3438 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3441 list_remove(&iface->port_elem);
3442 hmap_remove(&br->iface_by_name, &iface->name_node);
3444 netdev_close(iface->netdev);
3452 iface_destroy(struct iface *iface)
3455 struct port *port = iface->port;
3457 iface_destroy__(iface);
3458 if (list_is_empty(&port->ifaces)) {
3464 static struct iface *
3465 iface_lookup(const struct bridge *br, const char *name)
3467 struct iface *iface;
3469 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3470 &br->iface_by_name) {
3471 if (!strcmp(iface->name, name)) {
3479 static struct iface *
3480 iface_find(const char *name)
3482 const struct bridge *br;
3484 HMAP_FOR_EACH (br, node, &all_bridges) {
3485 struct iface *iface = iface_lookup(br, name);
3494 static struct iface *
3495 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3497 struct iface *iface;
3499 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3501 if (iface->ofp_port == ofp_port) {
3508 /* Set Ethernet address of 'iface', if one is specified in the configuration
3511 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3513 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3514 struct iface *hw_addr_iface;
3516 if (strcmp(iface->type, "internal")) {
3520 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3522 } else if (port->cfg->fake_bridge) {
3523 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3524 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3529 if (iface->ofp_port == OFPP_LOCAL) {
3530 VLOG_ERR("interface %s: ignoring mac in Interface record "
3531 "(use Bridge record to set local port's mac)",
3533 } else if (eth_addr_is_multicast(mac)) {
3534 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3537 int error = netdev_set_etheraddr(iface->netdev, mac);
3539 VLOG_ERR("interface %s: setting MAC failed (%s)",
3540 iface->name, ovs_strerror(error));
3546 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3548 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3550 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3551 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3552 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3556 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3557 * sets the "ofport" field to -1.
3559 * This is appropriate when 'if_cfg''s interface cannot be created or is
3560 * otherwise invalid. */
3562 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3564 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3565 iface_set_ofport(if_cfg, OFPP_NONE);
3566 ovsrec_interface_set_status(if_cfg, NULL);
3567 ovsrec_interface_set_admin_state(if_cfg, NULL);
3568 ovsrec_interface_set_duplex(if_cfg, NULL);
3569 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3570 ovsrec_interface_set_link_state(if_cfg, NULL);
3571 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3572 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3573 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3574 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3575 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3576 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3577 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3578 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3583 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3585 union ovsdb_atom atom;
3587 atom.integer = target;
3588 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3592 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3594 struct ofpbuf queues_buf;
3596 ofpbuf_init(&queues_buf, 0);
3598 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3599 netdev_set_qos(iface->netdev, NULL, NULL);
3601 const struct ovsdb_datum *queues;
3602 struct netdev_queue_dump dump;
3603 unsigned int queue_id;
3604 struct smap details;
3608 /* Configure top-level Qos for 'iface'. */
3609 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3611 /* Deconfigure queues that were deleted. */
3612 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3614 smap_init(&details);
3615 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3616 if (!queue_ids_include(queues, queue_id)) {
3617 netdev_delete_queue(iface->netdev, queue_id);
3620 smap_destroy(&details);
3622 /* Configure queues for 'iface'. */
3624 for (i = 0; i < qos->n_queues; i++) {
3625 const struct ovsrec_queue *queue = qos->value_queues[i];
3626 unsigned int queue_id = qos->key_queues[i];
3628 if (queue_id == 0) {
3632 if (queue->n_dscp == 1) {
3633 struct ofproto_port_queue *port_queue;
3635 port_queue = ofpbuf_put_uninit(&queues_buf,
3636 sizeof *port_queue);
3637 port_queue->queue = queue_id;
3638 port_queue->dscp = queue->dscp[0];
3641 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3644 struct smap details;
3646 smap_init(&details);
3647 netdev_set_queue(iface->netdev, 0, &details);
3648 smap_destroy(&details);
3652 if (iface->ofp_port != OFPP_NONE) {
3653 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
3654 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
3656 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3657 port_queues, n_queues);
3660 netdev_set_policing(iface->netdev,
3661 iface->cfg->ingress_policing_rate,
3662 iface->cfg->ingress_policing_burst);
3664 ofpbuf_uninit(&queues_buf);
3668 iface_configure_cfm(struct iface *iface)
3670 const struct ovsrec_interface *cfg = iface->cfg;
3671 const char *opstate_str;
3672 const char *cfm_ccm_vlan;
3673 struct cfm_settings s;
3674 struct smap netdev_args;
3676 if (!cfg->n_cfm_mpid) {
3677 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3681 s.check_tnl_key = false;
3682 smap_init(&netdev_args);
3683 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3684 const char *key = smap_get(&netdev_args, "key");
3685 const char *in_key = smap_get(&netdev_args, "in_key");
3687 s.check_tnl_key = (key && !strcmp(key, "flow"))
3688 || (in_key && !strcmp(in_key, "flow"));
3690 smap_destroy(&netdev_args);
3692 s.mpid = *cfg->cfm_mpid;
3693 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3694 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3695 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3697 if (s.interval <= 0) {
3701 if (!cfm_ccm_vlan) {
3703 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3704 s.ccm_vlan = CFM_RANDOM_VLAN;
3706 s.ccm_vlan = atoi(cfm_ccm_vlan);
3707 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3712 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3714 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3716 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3717 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3719 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3722 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3723 * instead of obtaining it from the database. */
3725 iface_is_synthetic(const struct iface *iface)
3727 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3731 iface_validate_ofport__(size_t n, int64_t *ofport)
3733 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3734 ? u16_to_ofp(*ofport)
3739 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3741 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3745 iface_pick_ofport(const struct ovsrec_interface *cfg)
3747 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3748 return (requested_ofport != OFPP_NONE
3750 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3753 /* Port mirroring. */
3755 static struct mirror *
3756 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3760 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3761 if (uuid_equals(uuid, &m->uuid)) {
3769 bridge_configure_mirrors(struct bridge *br)
3771 const struct ovsdb_datum *mc;
3772 unsigned long *flood_vlans;
3773 struct mirror *m, *next;
3776 /* Get rid of deleted mirrors. */
3777 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3778 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3779 union ovsdb_atom atom;
3781 atom.uuid = m->uuid;
3782 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3787 /* Add new mirrors and reconfigure existing ones. */
3788 for (i = 0; i < br->cfg->n_mirrors; i++) {
3789 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3790 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3792 m = mirror_create(br, cfg);
3795 if (!mirror_configure(m)) {
3800 /* Update flooded vlans (for RSPAN). */
3801 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3802 br->cfg->n_flood_vlans);
3803 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3804 bitmap_free(flood_vlans);
3807 static struct mirror *
3808 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3812 m = xzalloc(sizeof *m);
3813 m->uuid = cfg->header_.uuid;
3814 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3816 m->name = xstrdup(cfg->name);
3822 mirror_destroy(struct mirror *m)
3825 struct bridge *br = m->bridge;
3828 ofproto_mirror_unregister(br->ofproto, m);
3831 hmap_remove(&br->mirrors, &m->hmap_node);
3838 mirror_collect_ports(struct mirror *m,
3839 struct ovsrec_port **in_ports, int n_in_ports,
3840 void ***out_portsp, size_t *n_out_portsp)
3842 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3843 size_t n_out_ports = 0;
3846 for (i = 0; i < n_in_ports; i++) {
3847 const char *name = in_ports[i]->name;
3848 struct port *port = port_lookup(m->bridge, name);
3850 out_ports[n_out_ports++] = port;
3852 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3853 "port %s", m->bridge->name, m->name, name);
3856 *out_portsp = out_ports;
3857 *n_out_portsp = n_out_ports;
3861 mirror_configure(struct mirror *m)
3863 const struct ovsrec_mirror *cfg = m->cfg;
3864 struct ofproto_mirror_settings s;
3867 if (strcmp(cfg->name, m->name)) {
3869 m->name = xstrdup(cfg->name);
3873 /* Get output port or VLAN. */
3874 if (cfg->output_port) {
3875 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3876 if (!s.out_bundle) {
3877 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3878 m->bridge->name, m->name);
3881 s.out_vlan = UINT16_MAX;
3883 if (cfg->output_vlan) {
3884 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3885 "output vlan; ignoring output vlan",
3886 m->bridge->name, m->name);
3888 } else if (cfg->output_vlan) {
3889 /* The database should prevent invalid VLAN values. */
3890 s.out_bundle = NULL;
3891 s.out_vlan = *cfg->output_vlan;
3893 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3894 m->bridge->name, m->name);
3898 /* Get port selection. */
3899 if (cfg->select_all) {
3900 size_t n_ports = hmap_count(&m->bridge->ports);
3901 void **ports = xmalloc(n_ports * sizeof *ports);
3906 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3916 /* Get ports, dropping ports that don't exist.
3917 * The IDL ensures that there are no duplicates. */
3918 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3919 &s.srcs, &s.n_srcs);
3920 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3921 &s.dsts, &s.n_dsts);
3924 /* Get VLAN selection. */
3925 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3928 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3931 if (s.srcs != s.dsts) {
3940 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3942 * This is deprecated. It is only for compatibility with broken device drivers
3943 * in old versions of Linux that do not properly support VLANs when VLAN
3944 * devices are not used. When broken device drivers are no longer in
3945 * widespread use, we will delete these interfaces. */
3947 static struct ovsrec_port **recs;
3948 static size_t n_recs, allocated_recs;
3950 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3951 * splinters are reconfigured. */
3953 register_rec(struct ovsrec_port *rec)
3955 if (n_recs >= allocated_recs) {
3956 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3958 recs[n_recs++] = rec;
3961 /* Frees all of the ports registered with register_reg(). */
3963 free_registered_recs(void)
3967 for (i = 0; i < n_recs; i++) {
3968 struct ovsrec_port *port = recs[i];
3971 for (j = 0; j < port->n_interfaces; j++) {
3972 struct ovsrec_interface *iface = port->interfaces[j];
3977 smap_destroy(&port->other_config);
3978 free(port->interfaces);
3986 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3989 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3991 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3995 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3998 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3999 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4000 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4003 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4004 * use, returns NULL.
4006 * Updates 'vlan_splinters_enabled_anywhere'. */
4007 static unsigned long int *
4008 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4010 unsigned long int *splinter_vlans;
4011 struct sset splinter_ifaces;
4012 const char *real_dev_name;
4013 struct shash *real_devs;
4014 struct shash_node *node;
4018 /* Free space allocated for synthesized ports and interfaces, since we're
4019 * in the process of reconstructing all of them. */
4020 free_registered_recs();
4022 splinter_vlans = bitmap_allocate(4096);
4023 sset_init(&splinter_ifaces);
4024 vlan_splinters_enabled_anywhere = false;
4025 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4026 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4029 for (j = 0; j < br_cfg->n_ports; j++) {
4030 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4033 for (k = 0; k < port_cfg->n_interfaces; k++) {
4034 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4036 if (vlan_splinters_is_enabled(iface_cfg)) {
4037 vlan_splinters_enabled_anywhere = true;
4038 sset_add(&splinter_ifaces, iface_cfg->name);
4039 vlan_bitmap_from_array__(port_cfg->trunks,
4045 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4046 bitmap_set1(splinter_vlans, *port_cfg->tag);
4051 if (!vlan_splinters_enabled_anywhere) {
4052 free(splinter_vlans);
4053 sset_destroy(&splinter_ifaces);
4057 HMAP_FOR_EACH (br, node, &all_bridges) {
4059 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4063 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4064 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4065 * device to be created for it. */
4066 bitmap_set0(splinter_vlans, 0);
4067 bitmap_set0(splinter_vlans, 4095);
4069 /* Delete all VLAN devices that we don't need. */
4071 real_devs = vlandev_get_real_devs();
4072 SHASH_FOR_EACH (node, real_devs) {
4073 const struct vlan_real_dev *real_dev = node->data;
4074 const struct vlan_dev *vlan_dev;
4075 bool real_dev_has_splinters;
4077 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4079 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4080 if (!real_dev_has_splinters
4081 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4082 struct netdev *netdev;
4084 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4085 if (!netdev_get_in4(netdev, NULL, NULL) ||
4086 !netdev_get_in6(netdev, NULL)) {
4087 /* It has an IP address configured, so we don't own
4088 * it. Don't delete it. */
4090 vlandev_del(vlan_dev->name);
4092 netdev_close(netdev);
4099 /* Add all VLAN devices that we need. */
4100 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4103 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4104 if (!vlandev_get_name(real_dev_name, vid)) {
4105 vlandev_add(real_dev_name, vid);
4112 sset_destroy(&splinter_ifaces);
4114 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4115 free(splinter_vlans);
4118 return splinter_vlans;
4121 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4124 configure_splinter_port(struct port *port)
4126 struct ofproto *ofproto = port->bridge->ofproto;
4127 ofp_port_t realdev_ofp_port;
4128 const char *realdev_name;
4129 struct iface *vlandev, *realdev;
4131 ofproto_bundle_unregister(port->bridge->ofproto, port);
4133 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4136 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4137 realdev = iface_lookup(port->bridge, realdev_name);
4138 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4140 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4144 static struct ovsrec_port *
4145 synthesize_splinter_port(const char *real_dev_name,
4146 const char *vlan_dev_name, int vid)
4148 struct ovsrec_interface *iface;
4149 struct ovsrec_port *port;
4151 iface = xmalloc(sizeof *iface);
4152 ovsrec_interface_init(iface);
4153 iface->name = xstrdup(vlan_dev_name);
4154 iface->type = "system";
4156 port = xmalloc(sizeof *port);
4157 ovsrec_port_init(port);
4158 port->interfaces = xmemdup(&iface, sizeof iface);
4159 port->n_interfaces = 1;
4160 port->name = xstrdup(vlan_dev_name);
4161 port->vlan_mode = "splinter";
4162 port->tag = xmalloc(sizeof *port->tag);
4165 smap_add(&port->other_config, "realdev", real_dev_name);
4171 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4172 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4173 * 1-bit in the 'splinter_vlans' bitmap. */
4175 add_vlan_splinter_ports(struct bridge *br,
4176 const unsigned long int *splinter_vlans,
4177 struct shash *ports)
4181 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4182 * we're modifying 'ports'. */
4183 for (i = 0; i < br->cfg->n_ports; i++) {
4184 const char *name = br->cfg->ports[i]->name;
4185 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4188 for (j = 0; j < port_cfg->n_interfaces; j++) {
4189 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4191 if (vlan_splinters_is_enabled(iface_cfg)) {
4192 const char *real_dev_name;
4195 real_dev_name = iface_cfg->name;
4196 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4197 const char *vlan_dev_name;
4199 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4201 && !shash_find(ports, vlan_dev_name)) {
4202 shash_add(ports, vlan_dev_name,
4203 synthesize_splinter_port(
4204 real_dev_name, vlan_dev_name, vid));
4213 mirror_refresh_stats(struct mirror *m)
4215 struct ofproto *ofproto = m->bridge->ofproto;
4216 uint64_t tx_packets, tx_bytes;
4219 size_t stat_cnt = 0;
4221 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4222 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4226 if (tx_packets != UINT64_MAX) {
4227 keys[stat_cnt] = "tx_packets";
4228 values[stat_cnt] = tx_packets;
4231 if (tx_bytes != UINT64_MAX) {
4232 keys[stat_cnt] = "tx_bytes";
4233 values[stat_cnt] = tx_bytes;
4237 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);