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_ERR("unknown VLAN mode %s", cfg->vlan_mode);
843 s.vlan_mode = PORT_VLAN_TRUNK;
847 s.vlan_mode = PORT_VLAN_ACCESS;
849 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
853 s.vlan_mode = PORT_VLAN_TRUNK;
856 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
859 /* Get LACP settings. */
860 s.lacp = port_configure_lacp(port, &lacp_settings);
864 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
865 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
866 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
869 s.lacp_slaves = NULL;
872 /* Get bond settings. */
873 if (s.n_slaves > 1) {
874 s.bond = &bond_settings;
875 port_configure_bond(port, &bond_settings);
878 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
879 netdev_set_miimon_interval(iface->netdev, 0);
884 ofproto_bundle_register(port->bridge->ofproto, port, &s);
892 /* Pick local port hardware address and datapath ID for 'br'. */
894 bridge_configure_datapath_id(struct bridge *br)
896 uint8_t ea[ETH_ADDR_LEN];
898 struct iface *local_iface;
899 struct iface *hw_addr_iface;
902 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
903 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
905 int error = netdev_set_etheraddr(local_iface->netdev, ea);
907 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
908 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
909 "Ethernet address: %s",
910 br->name, ovs_strerror(error));
913 memcpy(br->ea, ea, ETH_ADDR_LEN);
915 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
916 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
917 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
918 ofproto_set_datapath_id(br->ofproto, dpid);
921 dpid_string = xasprintf("%016"PRIx64, dpid);
922 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
926 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
929 bridge_get_allowed_versions(struct bridge *br)
931 uint32_t allowed_versions;
933 if (!br->cfg->n_protocols)
936 allowed_versions = ofputil_versions_from_strings(br->cfg->protocols,
937 br->cfg->n_protocols);
939 allowed_versions &= ~(1u << OFP14_VERSION);
941 return allowed_versions;
944 /* Set NetFlow configuration on 'br'. */
946 bridge_configure_netflow(struct bridge *br)
948 struct ovsrec_netflow *cfg = br->cfg->netflow;
949 struct netflow_options opts;
952 ofproto_set_netflow(br->ofproto, NULL);
956 memset(&opts, 0, sizeof opts);
958 /* Get default NetFlow configuration from datapath.
959 * Apply overrides from 'cfg'. */
960 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
961 if (cfg->engine_type) {
962 opts.engine_type = *cfg->engine_type;
964 if (cfg->engine_id) {
965 opts.engine_id = *cfg->engine_id;
968 /* Configure active timeout interval. */
969 opts.active_timeout = cfg->active_timeout;
970 if (!opts.active_timeout) {
971 opts.active_timeout = -1;
972 } else if (opts.active_timeout < 0) {
973 VLOG_WARN("bridge %s: active timeout interval set to negative "
974 "value, using default instead (%d seconds)", br->name,
975 NF_ACTIVE_TIMEOUT_DEFAULT);
976 opts.active_timeout = -1;
979 /* Add engine ID to interface number to disambiguate bridgs? */
980 opts.add_id_to_iface = cfg->add_id_to_interface;
981 if (opts.add_id_to_iface) {
982 if (opts.engine_id > 0x7f) {
983 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
984 "another vswitch, choose an engine id less than 128",
987 if (hmap_count(&br->ports) > 508) {
988 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
989 "another port when more than 508 ports are used",
995 sset_init(&opts.collectors);
996 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
999 if (ofproto_set_netflow(br->ofproto, &opts)) {
1000 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1002 sset_destroy(&opts.collectors);
1005 /* Set sFlow configuration on 'br'. */
1007 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1009 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1010 struct ovsrec_controller **controllers;
1011 struct ofproto_sflow_options oso;
1012 size_t n_controllers;
1016 ofproto_set_sflow(br->ofproto, NULL);
1020 memset(&oso, 0, sizeof oso);
1022 sset_init(&oso.targets);
1023 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1025 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1026 if (cfg->sampling) {
1027 oso.sampling_rate = *cfg->sampling;
1030 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1032 oso.polling_interval = *cfg->polling;
1035 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1037 oso.header_len = *cfg->header;
1040 oso.sub_id = (*sflow_bridge_number)++;
1041 oso.agent_device = cfg->agent;
1043 oso.control_ip = NULL;
1044 n_controllers = bridge_get_controllers(br, &controllers);
1045 for (i = 0; i < n_controllers; i++) {
1046 if (controllers[i]->local_ip) {
1047 oso.control_ip = controllers[i]->local_ip;
1051 ofproto_set_sflow(br->ofproto, &oso);
1053 sset_destroy(&oso.targets);
1056 /* Returns whether a IPFIX row is valid. */
1058 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1060 return ipfix && ipfix->n_targets > 0;
1063 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1065 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1066 const struct bridge *br)
1068 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1071 /* Set IPFIX configuration on 'br'. */
1073 bridge_configure_ipfix(struct bridge *br)
1075 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1076 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1077 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1078 struct ofproto_ipfix_bridge_exporter_options be_opts;
1079 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1080 size_t n_fe_opts = 0;
1082 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1083 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1088 if (!valid_be_cfg && n_fe_opts == 0) {
1089 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1094 memset(&be_opts, 0, sizeof be_opts);
1096 sset_init(&be_opts.targets);
1097 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1099 if (be_cfg->sampling) {
1100 be_opts.sampling_rate = *be_cfg->sampling;
1102 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1104 if (be_cfg->obs_domain_id) {
1105 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1107 if (be_cfg->obs_point_id) {
1108 be_opts.obs_point_id = *be_cfg->obs_point_id;
1110 if (be_cfg->cache_active_timeout) {
1111 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1113 if (be_cfg->cache_max_flows) {
1114 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1118 if (n_fe_opts > 0) {
1119 struct ofproto_ipfix_flow_exporter_options *opts;
1120 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1122 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1123 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1124 opts->collector_set_id = fe_cfg->id;
1125 sset_init(&opts->targets);
1126 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1127 fe_cfg->ipfix->n_targets);
1128 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1129 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1130 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1131 ? *fe_cfg->ipfix->cache_max_flows : 0;
1137 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1141 sset_destroy(&be_opts.targets);
1144 if (n_fe_opts > 0) {
1145 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1147 for (i = 0; i < n_fe_opts; i++) {
1148 sset_destroy(&opts->targets);
1156 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1157 struct ofproto_port_stp_settings *port_s,
1158 int *port_num_counter, unsigned long *port_num_bitmap)
1160 const char *config_str;
1161 struct iface *iface;
1163 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1164 port_s->enable = false;
1167 port_s->enable = true;
1170 /* STP over bonds is not supported. */
1171 if (!list_is_singleton(&port->ifaces)) {
1172 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1174 port_s->enable = false;
1178 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1180 /* Internal ports shouldn't participate in spanning tree, so
1182 if (!strcmp(iface->type, "internal")) {
1183 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1184 port_s->enable = false;
1188 /* STP on mirror output ports is not supported. */
1189 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1190 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1191 port_s->enable = false;
1195 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1197 unsigned long int port_num = strtoul(config_str, NULL, 0);
1198 int port_idx = port_num - 1;
1200 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1201 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1202 port_s->enable = false;
1206 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1207 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1208 port->name, port_num);
1209 port_s->enable = false;
1212 bitmap_set1(port_num_bitmap, port_idx);
1213 port_s->port_num = port_idx;
1215 if (*port_num_counter >= STP_MAX_PORTS) {
1216 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1217 port_s->enable = false;
1221 port_s->port_num = (*port_num_counter)++;
1224 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1226 port_s->path_cost = strtoul(config_str, NULL, 10);
1228 enum netdev_features current;
1231 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1232 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1233 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1236 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1238 port_s->priority = strtoul(config_str, NULL, 0);
1240 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1244 /* Set spanning tree configuration on 'br'. */
1246 bridge_configure_stp(struct bridge *br)
1248 if (!br->cfg->stp_enable) {
1249 ofproto_set_stp(br->ofproto, NULL);
1251 struct ofproto_stp_settings br_s;
1252 const char *config_str;
1254 int port_num_counter;
1255 unsigned long *port_num_bitmap;
1257 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1259 uint8_t ea[ETH_ADDR_LEN];
1261 if (eth_addr_from_string(config_str, ea)) {
1262 br_s.system_id = eth_addr_to_uint64(ea);
1264 br_s.system_id = eth_addr_to_uint64(br->ea);
1265 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1266 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1269 br_s.system_id = eth_addr_to_uint64(br->ea);
1272 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1274 br_s.priority = strtoul(config_str, NULL, 0);
1276 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1279 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1281 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1283 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1286 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1288 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1290 br_s.max_age = STP_DEFAULT_MAX_AGE;
1293 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1295 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1297 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1300 /* Configure STP on the bridge. */
1301 if (ofproto_set_stp(br->ofproto, &br_s)) {
1302 VLOG_ERR("bridge %s: could not enable STP", br->name);
1306 /* Users must either set the port number with the "stp-port-num"
1307 * configuration on all ports or none. If manual configuration
1308 * is not done, then we allocate them sequentially. */
1309 port_num_counter = 0;
1310 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1311 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1312 struct ofproto_port_stp_settings port_s;
1313 struct iface *iface;
1315 port_configure_stp(br->ofproto, port, &port_s,
1316 &port_num_counter, port_num_bitmap);
1318 /* As bonds are not supported, just apply configuration to
1319 * all interfaces. */
1320 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1321 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1323 VLOG_ERR("port %s: could not enable STP", port->name);
1329 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1330 && port_num_counter) {
1331 VLOG_ERR("bridge %s: must manually configure all STP port "
1332 "IDs or none, disabling", br->name);
1333 ofproto_set_stp(br->ofproto, NULL);
1335 bitmap_free(port_num_bitmap);
1340 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1342 const struct port *port = port_lookup(br, name);
1343 return port && port_is_bond_fake_iface(port);
1347 port_is_bond_fake_iface(const struct port *port)
1349 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1353 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1355 struct bridge *br, *next;
1356 struct shash new_br;
1359 /* Collect new bridges' names and types. */
1360 shash_init(&new_br);
1361 for (i = 0; i < cfg->n_bridges; i++) {
1362 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1363 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1365 if (strchr(br_cfg->name, '/')) {
1366 /* Prevent remote ovsdb-server users from accessing arbitrary
1367 * directories, e.g. consider a bridge named "../../../etc/". */
1368 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1370 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1371 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1375 /* Get rid of deleted bridges or those whose types have changed.
1376 * Update 'cfg' of bridges that still exist. */
1377 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1378 br->cfg = shash_find_data(&new_br, br->name);
1379 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1380 br->cfg->datapath_type))) {
1385 /* Add new bridges. */
1386 for (i = 0; i < cfg->n_bridges; i++) {
1387 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1388 struct bridge *br = bridge_lookup(br_cfg->name);
1390 bridge_create(br_cfg);
1394 shash_destroy(&new_br);
1397 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1398 * Returns 0 if successful, otherwise a positive errno value. */
1400 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1401 struct netdev *netdev)
1403 return netdev_set_config(netdev, &iface_cfg->options);
1406 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1407 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1409 * If successful, returns 0 and stores the network device in '*netdevp'. On
1410 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1412 iface_do_create(const struct bridge *br,
1413 const struct ovsrec_interface *iface_cfg,
1414 const struct ovsrec_port *port_cfg,
1415 ofp_port_t *ofp_portp, struct netdev **netdevp)
1417 struct netdev *netdev = NULL;
1420 if (netdev_is_reserved_name(iface_cfg->name)) {
1421 VLOG_WARN("could not create interface %s, name is reserved",
1427 error = netdev_open(iface_cfg->name,
1428 iface_get_type(iface_cfg, br->cfg), &netdev);
1430 VLOG_WARN("could not open network device %s (%s)",
1431 iface_cfg->name, ovs_strerror(error));
1435 error = iface_set_netdev_config(iface_cfg, netdev);
1440 *ofp_portp = iface_pick_ofport(iface_cfg);
1441 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1446 VLOG_INFO("bridge %s: added interface %s on port %d",
1447 br->name, iface_cfg->name, *ofp_portp);
1449 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1450 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1458 netdev_close(netdev);
1462 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1463 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1464 * automatically allocated for the iface. Takes ownership of and
1465 * deallocates 'if_cfg'.
1467 * Return true if an iface is successfully created, false otherwise. */
1469 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1470 const struct ovsrec_port *port_cfg)
1472 struct netdev *netdev;
1473 struct iface *iface;
1474 ofp_port_t ofp_port;
1478 /* Do the bits that can fail up front. */
1479 ovs_assert(!iface_lookup(br, iface_cfg->name));
1480 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1482 iface_clear_db_record(iface_cfg);
1486 /* Get or create the port structure. */
1487 port = port_lookup(br, port_cfg->name);
1489 port = port_create(br, port_cfg);
1492 /* Create the iface structure. */
1493 iface = xzalloc(sizeof *iface);
1494 list_push_back(&port->ifaces, &iface->port_elem);
1495 hmap_insert(&br->iface_by_name, &iface->name_node,
1496 hash_string(iface_cfg->name, 0));
1498 iface->name = xstrdup(iface_cfg->name);
1499 iface->ofp_port = ofp_port;
1500 iface->netdev = netdev;
1501 iface->type = iface_get_type(iface_cfg, br->cfg);
1502 iface->cfg = iface_cfg;
1503 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1504 hash_ofp_port(ofp_port));
1506 /* Populate initial status in database. */
1507 iface_refresh_stats(iface);
1508 iface_refresh_status(iface);
1510 /* Add bond fake iface if necessary. */
1511 if (port_is_bond_fake_iface(port)) {
1512 struct ofproto_port ofproto_port;
1514 if (ofproto_port_query_by_name(br->ofproto, port->name,
1516 struct netdev *netdev;
1519 error = netdev_open(port->name, "internal", &netdev);
1521 ofp_port_t fake_ofp_port = OFPP_NONE;
1522 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1523 netdev_close(netdev);
1525 VLOG_WARN("could not open network device %s (%s)",
1526 port->name, ovs_strerror(error));
1529 /* Already exists, nothing to do. */
1530 ofproto_port_destroy(&ofproto_port);
1537 /* Set forward BPDU option. */
1539 bridge_configure_forward_bpdu(struct bridge *br)
1541 ofproto_set_forward_bpdu(br->ofproto,
1542 smap_get_bool(&br->cfg->other_config,
1547 /* Set MAC learning table configuration for 'br'. */
1549 bridge_configure_mac_table(struct bridge *br)
1551 const char *idle_time_str;
1554 const char *mac_table_size_str;
1557 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1558 idle_time = (idle_time_str && atoi(idle_time_str)
1559 ? atoi(idle_time_str)
1560 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1562 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1563 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1564 ? atoi(mac_table_size_str)
1567 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1571 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1572 const struct port *fake_br, struct iface **hw_addr_iface)
1574 struct hmapx mirror_output_ports;
1576 bool found_addr = false;
1580 /* Mirror output ports don't participate in picking the local hardware
1581 * address. ofproto can't help us find out whether a given port is a
1582 * mirror output because we haven't configured mirrors yet, so we need to
1583 * accumulate them ourselves. */
1584 hmapx_init(&mirror_output_ports);
1585 for (i = 0; i < br->cfg->n_mirrors; i++) {
1586 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1587 if (m->output_port) {
1588 hmapx_add(&mirror_output_ports, m->output_port);
1592 /* Otherwise choose the minimum non-local MAC address among all of the
1594 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1595 uint8_t iface_ea[ETH_ADDR_LEN];
1596 struct iface *candidate;
1597 struct iface *iface;
1599 /* Mirror output ports don't participate. */
1600 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1604 /* Choose the MAC address to represent the port. */
1606 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1607 /* Find the interface with this Ethernet address (if any) so that
1608 * we can provide the correct devname to the caller. */
1609 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1610 uint8_t candidate_ea[ETH_ADDR_LEN];
1611 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1612 && eth_addr_equals(iface_ea, candidate_ea)) {
1617 /* Choose the interface whose MAC address will represent the port.
1618 * The Linux kernel bonding code always chooses the MAC address of
1619 * the first slave added to a bond, and the Fedora networking
1620 * scripts always add slaves to a bond in alphabetical order, so
1621 * for compatibility we choose the interface with the name that is
1622 * first in alphabetical order. */
1623 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1624 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1629 /* The local port doesn't count (since we're trying to choose its
1630 * MAC address anyway). */
1631 if (iface->ofp_port == OFPP_LOCAL) {
1635 /* For fake bridges we only choose from ports with the same tag */
1636 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1637 if (!port->cfg->tag) {
1640 if (*port->cfg->tag != *fake_br->cfg->tag) {
1646 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1652 /* Compare against our current choice. */
1653 if (!eth_addr_is_multicast(iface_ea) &&
1654 !eth_addr_is_local(iface_ea) &&
1655 !eth_addr_is_reserved(iface_ea) &&
1656 !eth_addr_is_zero(iface_ea) &&
1657 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1659 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1660 *hw_addr_iface = iface;
1666 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1667 *hw_addr_iface = NULL;
1670 hmapx_destroy(&mirror_output_ports);
1674 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1675 struct iface **hw_addr_iface)
1678 *hw_addr_iface = NULL;
1680 /* Did the user request a particular MAC? */
1681 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1682 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1683 if (eth_addr_is_multicast(ea)) {
1684 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1685 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1686 } else if (eth_addr_is_zero(ea)) {
1687 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1693 /* Find a local hw address */
1694 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
1697 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1698 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1699 * an interface on 'br', then that interface must be passed in as
1700 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1701 * 'hw_addr_iface' must be passed in as a null pointer. */
1703 bridge_pick_datapath_id(struct bridge *br,
1704 const uint8_t bridge_ea[ETH_ADDR_LEN],
1705 struct iface *hw_addr_iface)
1708 * The procedure for choosing a bridge MAC address will, in the most
1709 * ordinary case, also choose a unique MAC that we can use as a datapath
1710 * ID. In some special cases, though, multiple bridges will end up with
1711 * the same MAC address. This is OK for the bridges, but it will confuse
1712 * the OpenFlow controller, because each datapath needs a unique datapath
1715 * Datapath IDs must be unique. It is also very desirable that they be
1716 * stable from one run to the next, so that policy set on a datapath
1719 const char *datapath_id;
1722 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1723 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1727 if (!hw_addr_iface) {
1729 * A purely internal bridge, that is, one that has no non-virtual
1730 * network devices on it at all, is difficult because it has no
1731 * natural unique identifier at all.
1733 * When the host is a XenServer, we handle this case by hashing the
1734 * host's UUID with the name of the bridge. Names of bridges are
1735 * persistent across XenServer reboots, although they can be reused if
1736 * an internal network is destroyed and then a new one is later
1737 * created, so this is fairly effective.
1739 * When the host is not a XenServer, we punt by using a random MAC
1740 * address on each run.
1742 const char *host_uuid = xenserver_get_host_uuid();
1744 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1745 dpid = dpid_from_hash(combined, strlen(combined));
1751 return eth_addr_to_uint64(bridge_ea);
1755 dpid_from_hash(const void *data, size_t n)
1757 uint8_t hash[SHA1_DIGEST_SIZE];
1759 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1760 sha1_bytes(data, n, hash);
1761 eth_addr_mark_random(hash);
1762 return eth_addr_to_uint64(hash);
1766 iface_refresh_status(struct iface *iface)
1770 enum netdev_features current;
1774 uint8_t mac[ETH_ADDR_LEN];
1778 if (iface_is_synthetic(iface)) {
1784 if (!netdev_get_status(iface->netdev, &smap)) {
1785 ovsrec_interface_set_status(iface->cfg, &smap);
1787 ovsrec_interface_set_status(iface->cfg, NULL);
1790 smap_destroy(&smap);
1792 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1793 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1795 ovsrec_interface_set_duplex(iface->cfg,
1796 netdev_features_is_full_duplex(current)
1798 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1800 ovsrec_interface_set_duplex(iface->cfg, NULL);
1801 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1804 error = netdev_get_mtu(iface->netdev, &mtu);
1807 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1809 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1812 error = netdev_get_etheraddr(iface->netdev, mac);
1814 char mac_string[32];
1816 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1817 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1819 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1822 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1823 * if there is no valid ifindex number. */
1824 ifindex64 = netdev_get_ifindex(iface->netdev);
1825 if (ifindex64 < 0) {
1828 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1831 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1834 iface_refresh_cfm_stats(struct iface *iface)
1836 const struct ovsrec_interface *cfg = iface->cfg;
1837 struct ofproto_cfm_status status;
1839 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1840 iface->ofp_port, &status)) {
1841 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1842 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1843 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1844 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1845 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1846 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1848 const char *reasons[CFM_FAULT_N_REASONS];
1849 int64_t cfm_health = status.health;
1850 int64_t cfm_flap_count = status.flap_count;
1851 bool faulted = status.faults != 0;
1854 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1857 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1858 int reason = 1 << i;
1859 if (status.faults & reason) {
1860 reasons[j++] = cfm_fault_reason_to_str(reason);
1863 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1865 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1867 if (status.remote_opstate >= 0) {
1868 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1869 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1871 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1874 ovsrec_interface_set_cfm_remote_mpids(cfg,
1875 (const int64_t *)status.rmps,
1877 if (cfm_health >= 0) {
1878 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1880 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1888 iface_refresh_stats(struct iface *iface)
1890 #define IFACE_STATS \
1891 IFACE_STAT(rx_packets, "rx_packets") \
1892 IFACE_STAT(tx_packets, "tx_packets") \
1893 IFACE_STAT(rx_bytes, "rx_bytes") \
1894 IFACE_STAT(tx_bytes, "tx_bytes") \
1895 IFACE_STAT(rx_dropped, "rx_dropped") \
1896 IFACE_STAT(tx_dropped, "tx_dropped") \
1897 IFACE_STAT(rx_errors, "rx_errors") \
1898 IFACE_STAT(tx_errors, "tx_errors") \
1899 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1900 IFACE_STAT(rx_over_errors, "rx_over_err") \
1901 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1902 IFACE_STAT(collisions, "collisions")
1904 #define IFACE_STAT(MEMBER, NAME) + 1
1905 enum { N_IFACE_STATS = IFACE_STATS };
1907 int64_t values[N_IFACE_STATS];
1908 char *keys[N_IFACE_STATS];
1911 struct netdev_stats stats;
1913 if (iface_is_synthetic(iface)) {
1917 /* Intentionally ignore return value, since errors will set 'stats' to
1918 * all-1s, and we will deal with that correctly below. */
1919 netdev_get_stats(iface->netdev, &stats);
1921 /* Copy statistics into keys[] and values[]. */
1923 #define IFACE_STAT(MEMBER, NAME) \
1924 if (stats.MEMBER != UINT64_MAX) { \
1926 values[n] = stats.MEMBER; \
1931 ovs_assert(n <= N_IFACE_STATS);
1933 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1938 br_refresh_stp_status(struct bridge *br)
1940 struct smap smap = SMAP_INITIALIZER(&smap);
1941 struct ofproto *ofproto = br->ofproto;
1942 struct ofproto_stp_status status;
1944 if (ofproto_get_stp_status(ofproto, &status)) {
1948 if (!status.enabled) {
1949 ovsrec_bridge_set_status(br->cfg, NULL);
1953 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1954 STP_ID_ARGS(status.bridge_id));
1955 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1956 STP_ID_ARGS(status.designated_root));
1957 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1959 ovsrec_bridge_set_status(br->cfg, &smap);
1960 smap_destroy(&smap);
1964 port_refresh_stp_status(struct port *port)
1966 struct ofproto *ofproto = port->bridge->ofproto;
1967 struct iface *iface;
1968 struct ofproto_port_stp_status status;
1971 if (port_is_synthetic(port)) {
1975 /* STP doesn't currently support bonds. */
1976 if (!list_is_singleton(&port->ifaces)) {
1977 ovsrec_port_set_status(port->cfg, NULL);
1981 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1982 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1986 if (!status.enabled) {
1987 ovsrec_port_set_status(port->cfg, NULL);
1991 /* Set Status column. */
1993 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1994 smap_add(&smap, "stp_state", stp_state_name(status.state));
1995 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1996 smap_add(&smap, "stp_role", stp_role_name(status.role));
1997 ovsrec_port_set_status(port->cfg, &smap);
1998 smap_destroy(&smap);
2002 port_refresh_stp_stats(struct port *port)
2004 struct ofproto *ofproto = port->bridge->ofproto;
2005 struct iface *iface;
2006 struct ofproto_port_stp_stats stats;
2008 int64_t int_values[3];
2010 if (port_is_synthetic(port)) {
2014 /* STP doesn't currently support bonds. */
2015 if (!list_is_singleton(&port->ifaces)) {
2019 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2020 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2024 if (!stats.enabled) {
2025 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2029 /* Set Statistics column. */
2030 keys[0] = "stp_tx_count";
2031 int_values[0] = stats.tx_count;
2032 keys[1] = "stp_rx_count";
2033 int_values[1] = stats.rx_count;
2034 keys[2] = "stp_error_count";
2035 int_values[2] = stats.error_count;
2037 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2038 ARRAY_SIZE(int_values));
2042 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2044 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2048 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2050 bool enable = enable_system_stats(cfg);
2052 system_stats_enable(enable);
2054 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2059 run_system_stats(void)
2061 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2064 stats = system_stats_run();
2066 struct ovsdb_idl_txn *txn;
2067 struct ovsdb_datum datum;
2069 txn = ovsdb_idl_txn_create(idl);
2070 ovsdb_datum_from_smap(&datum, stats);
2071 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2073 ovsdb_idl_txn_commit(txn);
2074 ovsdb_idl_txn_destroy(txn);
2081 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2084 case OFPCR12_ROLE_EQUAL:
2086 case OFPCR12_ROLE_MASTER:
2088 case OFPCR12_ROLE_SLAVE:
2090 case OFPCR12_ROLE_NOCHANGE:
2092 return "*** INVALID ROLE ***";
2097 refresh_controller_status(void)
2101 const struct ovsrec_controller *cfg;
2105 /* Accumulate status for controllers on all bridges. */
2106 HMAP_FOR_EACH (br, node, &all_bridges) {
2107 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2110 /* Update each controller in the database with current status. */
2111 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2112 struct ofproto_controller_info *cinfo =
2113 shash_find_data(&info, cfg->target);
2116 struct smap smap = SMAP_INITIALIZER(&smap);
2117 const char **values = cinfo->pairs.values;
2118 const char **keys = cinfo->pairs.keys;
2121 for (i = 0; i < cinfo->pairs.n; i++) {
2122 smap_add(&smap, keys[i], values[i]);
2125 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2126 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2128 ovsrec_controller_set_status(cfg, &smap);
2129 smap_destroy(&smap);
2131 ovsrec_controller_set_is_connected(cfg, false);
2132 ovsrec_controller_set_role(cfg, NULL);
2133 ovsrec_controller_set_status(cfg, NULL);
2137 ofproto_free_ofproto_controller_info(&info);
2142 * Some information in the database must be kept as up-to-date as possible to
2143 * allow controllers to respond rapidly to network outages. We call these
2144 * statistics "instant" stats.
2146 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2147 * assuming that they've changed. The only means we have to determine whether
2148 * they have changed are:
2150 * - Try to commit changes to the database. If nothing changed, then
2151 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2154 * - instant_stats_run() is called late in the run loop, after anything that
2155 * might change any of the instant stats.
2157 * We use these two facts together to avoid waking the process up every
2158 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2161 /* Minimum interval between writing updates to the instant stats to the
2163 #define INSTANT_INTERVAL_MSEC 100
2165 /* Current instant stats database transaction, NULL if there is no ongoing
2167 static struct ovsdb_idl_txn *instant_txn;
2169 /* Next time (in msec on monotonic clock) at which we will update the instant
2171 static long long int instant_next_txn = LLONG_MIN;
2173 /* True if the run loop has run since we last saw that the instant stats were
2174 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2175 * to refresh the instant stats. */
2176 static bool instant_stats_could_have_changed;
2179 instant_stats_run(void)
2181 enum ovsdb_idl_txn_status status;
2183 instant_stats_could_have_changed = true;
2189 if (time_msec() < instant_next_txn) {
2192 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2194 seq = seq_read(connectivity_seq_get());
2195 if (seq == connectivity_seqno) {
2198 connectivity_seqno = seq;
2200 instant_txn = ovsdb_idl_txn_create(idl);
2201 HMAP_FOR_EACH (br, node, &all_bridges) {
2202 struct iface *iface;
2205 br_refresh_stp_status(br);
2207 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2208 port_refresh_stp_status(port);
2211 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2212 enum netdev_flags flags;
2214 const char *link_state;
2215 int64_t link_resets;
2218 if (iface_is_synthetic(iface)) {
2222 current = ofproto_port_is_lacp_current(br->ofproto,
2226 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2228 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2231 error = netdev_get_flags(iface->netdev, &flags);
2233 const char *state = flags & NETDEV_UP ? "up" : "down";
2234 ovsrec_interface_set_admin_state(iface->cfg, state);
2236 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2239 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2240 ovsrec_interface_set_link_state(iface->cfg, link_state);
2242 link_resets = netdev_get_carrier_resets(iface->netdev);
2243 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2245 iface_refresh_cfm_stats(iface);
2248 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2250 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2251 smap_destroy(&smap);
2256 status = ovsdb_idl_txn_commit(instant_txn);
2257 if (status != TXN_INCOMPLETE) {
2258 ovsdb_idl_txn_destroy(instant_txn);
2261 if (status == TXN_UNCHANGED) {
2262 instant_stats_could_have_changed = false;
2267 instant_stats_wait(void)
2270 ovsdb_idl_txn_wait(instant_txn);
2271 } else if (instant_stats_could_have_changed) {
2272 poll_timer_wait_until(instant_next_txn);
2283 /* Let each datapath type do the work that it needs to do. */
2285 ofproto_enumerate_types(&types);
2286 SSET_FOR_EACH (type, &types) {
2287 ofproto_type_run(type);
2289 sset_destroy(&types);
2291 /* Let each bridge do the work that it needs to do. */
2292 HMAP_FOR_EACH (br, node, &all_bridges) {
2293 ofproto_run(br->ofproto);
2300 static struct ovsrec_open_vswitch null_cfg;
2301 const struct ovsrec_open_vswitch *cfg;
2303 bool vlan_splinters_changed;
2306 ovsrec_open_vswitch_init(&null_cfg);
2310 if (ovsdb_idl_is_lock_contended(idl)) {
2311 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2312 struct bridge *br, *next_br;
2314 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2315 "disabling this process (pid %ld) until it goes away",
2316 (long int) getpid());
2318 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2321 /* Since we will not be running system_stats_run() in this process
2322 * with the current situation of multiple ovs-vswitchd daemons,
2323 * disable system stats collection. */
2324 system_stats_enable(false);
2326 } else if (!ovsdb_idl_has_lock(idl)) {
2329 cfg = ovsrec_open_vswitch_first(idl);
2331 /* Initialize the ofproto library. This only needs to run once, but
2332 * it must be done after the configuration is set. If the
2333 * initialization has already occurred, bridge_init_ofproto()
2334 * returns immediately. */
2335 bridge_init_ofproto(cfg);
2337 /* Once the value of flow-restore-wait is false, we no longer should
2338 * check its value from the database. */
2339 if (cfg && ofproto_get_flow_restore_wait()) {
2340 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2341 "flow-restore-wait", false));
2346 /* Re-configure SSL. We do this on every trip through the main loop,
2347 * instead of just when the database changes, because the contents of the
2348 * key and certificate files can change without the database changing.
2350 * We do this before bridge_reconfigure() because that function might
2351 * initiate SSL connections and thus requires SSL to be configured. */
2352 if (cfg && cfg->ssl) {
2353 const struct ovsrec_ssl *ssl = cfg->ssl;
2355 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2356 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2359 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2360 * usage has changed. */
2361 vlan_splinters_changed = false;
2362 if (vlan_splinters_enabled_anywhere) {
2363 HMAP_FOR_EACH (br, node, &all_bridges) {
2364 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2365 vlan_splinters_changed = true;
2371 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2372 struct ovsdb_idl_txn *txn;
2374 idl_seqno = ovsdb_idl_get_seqno(idl);
2375 txn = ovsdb_idl_txn_create(idl);
2376 bridge_reconfigure(cfg ? cfg : &null_cfg);
2379 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2382 /* If we are completing our initial configuration for this run
2383 * of ovs-vswitchd, then keep the transaction around to monitor
2384 * it for completion. */
2385 if (initial_config_done) {
2386 ovsdb_idl_txn_commit(txn);
2387 ovsdb_idl_txn_destroy(txn);
2389 initial_config_done = true;
2390 daemonize_txn = txn;
2394 if (daemonize_txn) {
2395 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2396 if (status != TXN_INCOMPLETE) {
2397 ovsdb_idl_txn_destroy(daemonize_txn);
2398 daemonize_txn = NULL;
2400 /* ovs-vswitchd has completed initialization, so allow the
2401 * process that forked us to exit successfully. */
2402 daemonize_complete();
2404 vlog_enable_async();
2406 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2410 /* Refresh interface and mirror stats if necessary. */
2411 if (time_msec() >= iface_stats_timer) {
2413 struct ovsdb_idl_txn *txn;
2415 txn = ovsdb_idl_txn_create(idl);
2416 HMAP_FOR_EACH (br, node, &all_bridges) {
2420 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2421 struct iface *iface;
2423 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2424 iface_refresh_stats(iface);
2425 iface_refresh_status(iface);
2428 port_refresh_stp_stats(port);
2431 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2432 mirror_refresh_stats(m);
2436 refresh_controller_status();
2437 ovsdb_idl_txn_commit(txn);
2438 ovsdb_idl_txn_destroy(txn); /* XXX */
2441 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2445 instant_stats_run();
2454 ovsdb_idl_wait(idl);
2455 if (daemonize_txn) {
2456 ovsdb_idl_txn_wait(daemonize_txn);
2460 ofproto_enumerate_types(&types);
2461 SSET_FOR_EACH (type, &types) {
2462 ofproto_type_wait(type);
2464 sset_destroy(&types);
2466 if (!hmap_is_empty(&all_bridges)) {
2469 HMAP_FOR_EACH (br, node, &all_bridges) {
2470 ofproto_wait(br->ofproto);
2472 poll_timer_wait_until(iface_stats_timer);
2475 system_stats_wait();
2476 instant_stats_wait();
2479 /* Adds some memory usage statistics for bridges into 'usage', for use with
2480 * memory_report(). */
2482 bridge_get_memory_usage(struct simap *usage)
2489 ofproto_enumerate_types(&types);
2490 SSET_FOR_EACH (type, &types) {
2491 ofproto_type_get_memory_usage(type, usage);
2493 sset_destroy(&types);
2495 HMAP_FOR_EACH (br, node, &all_bridges) {
2496 ofproto_get_memory_usage(br->ofproto, usage);
2500 /* QoS unixctl user interface functions. */
2502 struct qos_unixctl_show_cbdata {
2504 struct iface *iface;
2508 qos_unixctl_show_queue(unsigned int queue_id,
2509 const struct smap *details,
2510 struct iface *iface,
2513 struct netdev_queue_stats stats;
2514 struct smap_node *node;
2517 ds_put_cstr(ds, "\n");
2519 ds_put_format(ds, "Queue %u:\n", queue_id);
2521 ds_put_cstr(ds, "Default:\n");
2524 SMAP_FOR_EACH (node, details) {
2525 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2528 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2530 if (stats.tx_packets != UINT64_MAX) {
2531 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2534 if (stats.tx_bytes != UINT64_MAX) {
2535 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2538 if (stats.tx_errors != UINT64_MAX) {
2539 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2542 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2543 queue_id, ovs_strerror(error));
2548 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2549 const char *argv[], void *aux OVS_UNUSED)
2551 struct ds ds = DS_EMPTY_INITIALIZER;
2552 struct smap smap = SMAP_INITIALIZER(&smap);
2553 struct iface *iface;
2555 struct smap_node *node;
2557 iface = iface_find(argv[1]);
2559 unixctl_command_reply_error(conn, "no such interface");
2563 netdev_get_qos(iface->netdev, &type, &smap);
2565 if (*type != '\0') {
2566 struct netdev_queue_dump dump;
2567 struct smap details;
2568 unsigned int queue_id;
2570 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2572 SMAP_FOR_EACH (node, &smap) {
2573 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2576 smap_init(&details);
2577 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2578 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2580 smap_destroy(&details);
2582 unixctl_command_reply(conn, ds_cstr(&ds));
2584 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2585 unixctl_command_reply_error(conn, ds_cstr(&ds));
2588 smap_destroy(&smap);
2592 /* Bridge reconfiguration functions. */
2594 bridge_create(const struct ovsrec_bridge *br_cfg)
2598 ovs_assert(!bridge_lookup(br_cfg->name));
2599 br = xzalloc(sizeof *br);
2601 br->name = xstrdup(br_cfg->name);
2602 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2605 /* Derive the default Ethernet address from the bridge's UUID. This should
2606 * be unique and it will be stable between ovs-vswitchd runs. */
2607 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2608 eth_addr_mark_random(br->default_ea);
2610 hmap_init(&br->ports);
2611 hmap_init(&br->ifaces);
2612 hmap_init(&br->iface_by_name);
2613 hmap_init(&br->mirrors);
2615 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2619 bridge_destroy(struct bridge *br)
2622 struct mirror *mirror, *next_mirror;
2623 struct port *port, *next_port;
2625 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2628 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2629 mirror_destroy(mirror);
2632 hmap_remove(&all_bridges, &br->node);
2633 ofproto_destroy(br->ofproto);
2634 hmap_destroy(&br->ifaces);
2635 hmap_destroy(&br->ports);
2636 hmap_destroy(&br->iface_by_name);
2637 hmap_destroy(&br->mirrors);
2644 static struct bridge *
2645 bridge_lookup(const char *name)
2649 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2650 if (!strcmp(br->name, name)) {
2657 /* Handle requests for a listing of all flows known by the OpenFlow
2658 * stack, including those normally hidden. */
2660 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2661 const char *argv[], void *aux OVS_UNUSED)
2666 br = bridge_lookup(argv[1]);
2668 unixctl_command_reply_error(conn, "Unknown bridge");
2673 ofproto_get_all_flows(br->ofproto, &results);
2675 unixctl_command_reply(conn, ds_cstr(&results));
2676 ds_destroy(&results);
2679 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2680 * connections and reconnect. If BRIDGE is not specified, then all bridges
2681 * drop their controller connections and reconnect. */
2683 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2684 const char *argv[], void *aux OVS_UNUSED)
2688 br = bridge_lookup(argv[1]);
2690 unixctl_command_reply_error(conn, "Unknown bridge");
2693 ofproto_reconnect_controllers(br->ofproto);
2695 HMAP_FOR_EACH (br, node, &all_bridges) {
2696 ofproto_reconnect_controllers(br->ofproto);
2699 unixctl_command_reply(conn, NULL);
2703 bridge_get_controllers(const struct bridge *br,
2704 struct ovsrec_controller ***controllersp)
2706 struct ovsrec_controller **controllers;
2707 size_t n_controllers;
2709 controllers = br->cfg->controller;
2710 n_controllers = br->cfg->n_controller;
2712 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2718 *controllersp = controllers;
2720 return n_controllers;
2724 bridge_collect_wanted_ports(struct bridge *br,
2725 const unsigned long int *splinter_vlans,
2726 struct shash *wanted_ports)
2730 shash_init(wanted_ports);
2732 for (i = 0; i < br->cfg->n_ports; i++) {
2733 const char *name = br->cfg->ports[i]->name;
2734 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2735 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2739 if (bridge_get_controllers(br, NULL)
2740 && !shash_find(wanted_ports, br->name)) {
2741 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2742 br->name, br->name);
2744 ovsrec_interface_init(&br->synth_local_iface);
2745 ovsrec_port_init(&br->synth_local_port);
2747 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2748 br->synth_local_port.n_interfaces = 1;
2749 br->synth_local_port.name = br->name;
2751 br->synth_local_iface.name = br->name;
2752 br->synth_local_iface.type = "internal";
2754 br->synth_local_ifacep = &br->synth_local_iface;
2756 shash_add(wanted_ports, br->name, &br->synth_local_port);
2759 if (splinter_vlans) {
2760 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2764 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2765 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2766 * 'br' needs to complete its configuration. */
2768 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2770 struct shash_node *port_node;
2771 struct port *port, *next;
2773 /* Get rid of deleted ports.
2774 * Get rid of deleted interfaces on ports that still exist. */
2775 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2776 port->cfg = shash_find_data(wanted_ports, port->name);
2780 port_del_ifaces(port);
2784 /* Update iface->cfg and iface->type in interfaces that still exist. */
2785 SHASH_FOR_EACH (port_node, wanted_ports) {
2786 const struct ovsrec_port *port = port_node->data;
2789 for (i = 0; i < port->n_interfaces; i++) {
2790 const struct ovsrec_interface *cfg = port->interfaces[i];
2791 struct iface *iface = iface_lookup(br, cfg->name);
2792 const char *type = iface_get_type(cfg, br->cfg);
2797 } else if (!strcmp(type, "null")) {
2798 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2799 " may be removed in February 2013. Please email"
2800 " dev@openvswitch.org with concerns.",
2803 /* We will add new interfaces later. */
2809 /* Initializes 'oc' appropriately as a management service controller for
2812 * The caller must free oc->target when it is no longer needed. */
2814 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2815 struct ofproto_controller *oc)
2817 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2818 oc->max_backoff = 0;
2819 oc->probe_interval = 60;
2820 oc->band = OFPROTO_OUT_OF_BAND;
2822 oc->burst_limit = 0;
2823 oc->enable_async_msgs = true;
2826 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2828 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2829 struct ofproto_controller *oc)
2833 oc->target = c->target;
2834 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2835 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2836 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2837 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2838 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2839 oc->burst_limit = (c->controller_burst_limit
2840 ? *c->controller_burst_limit : 0);
2841 oc->enable_async_msgs = (!c->enable_async_messages
2842 || *c->enable_async_messages);
2843 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2844 if (dscp < 0 || dscp > 63) {
2845 dscp = DSCP_DEFAULT;
2850 /* Configures the IP stack for 'br''s local interface properly according to the
2851 * configuration in 'c'. */
2853 bridge_configure_local_iface_netdev(struct bridge *br,
2854 struct ovsrec_controller *c)
2856 struct netdev *netdev;
2857 struct in_addr mask, gateway;
2859 struct iface *local_iface;
2862 /* If there's no local interface or no IP address, give up. */
2863 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2864 if (!local_iface || !c->local_ip
2865 || !inet_pton(AF_INET, c->local_ip, &ip)) {
2869 /* Bring up the local interface. */
2870 netdev = local_iface->netdev;
2871 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2873 /* Configure the IP address and netmask. */
2874 if (!c->local_netmask
2875 || !inet_pton(AF_INET, c->local_netmask, &mask)
2877 mask.s_addr = guess_netmask(ip.s_addr);
2879 if (!netdev_set_in4(netdev, ip, mask)) {
2880 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2881 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2884 /* Configure the default gateway. */
2885 if (c->local_gateway
2886 && inet_pton(AF_INET, c->local_gateway, &gateway)
2887 && gateway.s_addr) {
2888 if (!netdev_add_router(netdev, gateway)) {
2889 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2890 br->name, IP_ARGS(gateway.s_addr));
2895 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2896 * in either string are treated as equal to any number of slashes in the other,
2897 * e.g. "x///y" is equal to "x/y".
2899 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2900 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2901 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2902 * 'b' against a prefix of 'a'.
2905 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2907 const char *b_start = b;
2909 if (b - b_start >= b_stoplen) {
2911 } else if (*a != *b) {
2913 } else if (*a == '/') {
2914 a += strspn(a, "/");
2915 b += strspn(b, "/");
2916 } else if (*a == '\0') {
2926 bridge_configure_remotes(struct bridge *br,
2927 const struct sockaddr_in *managers, size_t n_managers)
2929 bool disable_in_band;
2931 struct ovsrec_controller **controllers;
2932 size_t n_controllers;
2934 enum ofproto_fail_mode fail_mode;
2936 struct ofproto_controller *ocs;
2940 /* Check if we should disable in-band control on this bridge. */
2941 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2944 /* Set OpenFlow queue ID for in-band control. */
2945 ofproto_set_in_band_queue(br->ofproto,
2946 smap_get_int(&br->cfg->other_config,
2947 "in-band-queue", -1));
2949 if (disable_in_band) {
2950 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2952 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2955 n_controllers = bridge_get_controllers(br, &controllers);
2957 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2960 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2961 for (i = 0; i < n_controllers; i++) {
2962 struct ovsrec_controller *c = controllers[i];
2964 if (!strncmp(c->target, "punix:", 6)
2965 || !strncmp(c->target, "unix:", 5)) {
2966 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2969 if (!strncmp(c->target, "unix:", 5)) {
2970 /* Connect to a listening socket */
2971 whitelist = xasprintf("unix:%s/", ovs_rundir());
2972 if (strchr(c->target, '/') &&
2973 !equal_pathnames(c->target, whitelist,
2974 strlen(whitelist))) {
2975 /* Absolute path specified, but not in ovs_rundir */
2976 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2977 "controller \"%s\" due to possibility for "
2978 "remote exploit. Instead, specify socket "
2979 "in whitelisted \"%s\" or connect to "
2980 "\"unix:%s/%s.mgmt\" (which is always "
2981 "available without special configuration).",
2982 br->name, c->target, whitelist,
2983 ovs_rundir(), br->name);
2988 whitelist = xasprintf("punix:%s/%s.controller",
2989 ovs_rundir(), br->name);
2990 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2991 /* Prevent remote ovsdb-server users from accessing
2992 * arbitrary Unix domain sockets and overwriting arbitrary
2994 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2995 "controller \"%s\" due to possibility of "
2996 "overwriting local files. Instead, specify "
2997 "whitelisted \"%s\" or connect to "
2998 "\"unix:%s/%s.mgmt\" (which is always "
2999 "available without special configuration).",
3000 br->name, c->target, whitelist,
3001 ovs_rundir(), br->name);
3010 bridge_configure_local_iface_netdev(br, c);
3011 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3012 if (disable_in_band) {
3013 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3018 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3019 bridge_get_allowed_versions(br));
3020 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3023 /* Set the fail-mode. */
3024 fail_mode = !br->cfg->fail_mode
3025 || !strcmp(br->cfg->fail_mode, "standalone")
3026 ? OFPROTO_FAIL_STANDALONE
3027 : OFPROTO_FAIL_SECURE;
3028 ofproto_set_fail_mode(br->ofproto, fail_mode);
3030 /* Configure OpenFlow controller connection snooping. */
3031 if (!ofproto_has_snoops(br->ofproto)) {
3035 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3036 ovs_rundir(), br->name));
3037 ofproto_set_snoops(br->ofproto, &snoops);
3038 sset_destroy(&snoops);
3043 bridge_configure_tables(struct bridge *br)
3045 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3049 n_tables = ofproto_get_n_tables(br->ofproto);
3051 for (i = 0; i < n_tables; i++) {
3052 struct ofproto_table_settings s;
3055 s.max_flows = UINT_MAX;
3058 s.n_prefix_fields = 0;
3059 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3061 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3062 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3065 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3066 s.max_flows = *cfg->flow_limit;
3068 if (cfg->overflow_policy
3069 && !strcmp(cfg->overflow_policy, "evict")) {
3071 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3072 for (k = 0; k < cfg->n_groups; k++) {
3073 const char *string = cfg->groups[k];
3076 msg = mf_parse_subfield__(&s.groups[k], &string);
3078 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3079 "'groups' (%s)", br->name, i, msg);
3081 } else if (*string) {
3082 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3083 "element '%s' contains trailing garbage",
3084 br->name, i, cfg->groups[k]);
3090 /* Prefix lookup fields. */
3091 s.n_prefix_fields = 0;
3092 for (k = 0; k < cfg->n_prefixes; k++) {
3093 const char *name = cfg->prefixes[k];
3094 const struct mf_field *mf = mf_from_name(name);
3096 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3100 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3101 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3102 "%s", br->name, name);
3105 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3106 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3107 "field not used: %s", br->name, name);
3110 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3112 if (s.n_prefix_fields > 0) {
3114 struct ds ds = DS_EMPTY_INITIALIZER;
3115 for (k = 0; k < s.n_prefix_fields; k++) {
3117 ds_put_char(&ds, ',');
3119 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3121 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3122 br->name, i, ds_cstr(&ds));
3127 ofproto_configure_table(br->ofproto, i, &s);
3131 for (; j < br->cfg->n_flow_tables; j++) {
3132 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3133 "%"PRId64" not supported by this datapath", br->name,
3134 br->cfg->key_flow_tables[j]);
3139 bridge_configure_dp_desc(struct bridge *br)
3141 ofproto_set_dp_desc(br->ofproto,
3142 smap_get(&br->cfg->other_config, "dp-desc"));
3145 /* Port functions. */
3147 static void iface_destroy__(struct iface *);
3149 static struct port *
3150 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3154 port = xzalloc(sizeof *port);
3156 port->name = xstrdup(cfg->name);
3158 list_init(&port->ifaces);
3160 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3164 /* Deletes interfaces from 'port' that are no longer configured for it. */
3166 port_del_ifaces(struct port *port)
3168 struct iface *iface, *next;
3169 struct sset new_ifaces;
3172 /* Collect list of new interfaces. */
3173 sset_init(&new_ifaces);
3174 for (i = 0; i < port->cfg->n_interfaces; i++) {
3175 const char *name = port->cfg->interfaces[i]->name;
3176 const char *type = port->cfg->interfaces[i]->type;
3177 if (strcmp(type, "null")) {
3178 sset_add(&new_ifaces, name);
3182 /* Get rid of deleted interfaces. */
3183 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3184 if (!sset_contains(&new_ifaces, iface->name)) {
3185 iface_destroy(iface);
3189 sset_destroy(&new_ifaces);
3193 port_destroy(struct port *port)
3196 struct bridge *br = port->bridge;
3197 struct iface *iface, *next;
3200 ofproto_bundle_unregister(br->ofproto, port);
3203 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3204 iface_destroy__(iface);
3207 hmap_remove(&br->ports, &port->hmap_node);
3213 static struct port *
3214 port_lookup(const struct bridge *br, const char *name)
3218 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3220 if (!strcmp(port->name, name)) {
3228 enable_lacp(struct port *port, bool *activep)
3230 if (!port->cfg->lacp) {
3231 /* XXX when LACP implementation has been sufficiently tested, enable by
3232 * default and make active on bonded ports. */
3234 } else if (!strcmp(port->cfg->lacp, "off")) {
3236 } else if (!strcmp(port->cfg->lacp, "active")) {
3239 } else if (!strcmp(port->cfg->lacp, "passive")) {
3243 VLOG_WARN("port %s: unknown LACP mode %s",
3244 port->name, port->cfg->lacp);
3249 static struct lacp_settings *
3250 port_configure_lacp(struct port *port, struct lacp_settings *s)
3252 const char *lacp_time, *system_id;
3255 if (!enable_lacp(port, &s->active)) {
3259 s->name = port->name;
3261 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3263 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3264 ETH_ADDR_SCAN_ARGS(s->id))) {
3265 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3266 " address.", port->name, system_id);
3270 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3273 if (eth_addr_is_zero(s->id)) {
3274 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3278 /* Prefer bondable links if unspecified. */
3279 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3281 s->priority = (priority > 0 && priority <= UINT16_MAX
3283 : UINT16_MAX - !list_is_short(&port->ifaces));
3285 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3286 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3288 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3289 "lacp-fallback-ab", false);
3295 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3297 int priority, portid, key;
3299 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3300 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3302 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3304 if (portid <= 0 || portid > UINT16_MAX) {
3305 portid = ofp_to_u16(iface->ofp_port);
3308 if (priority <= 0 || priority > UINT16_MAX) {
3309 priority = UINT16_MAX;
3312 if (key < 0 || key > UINT16_MAX) {
3316 s->name = iface->name;
3318 s->priority = priority;
3323 port_configure_bond(struct port *port, struct bond_settings *s)
3325 const char *detect_s;
3326 struct iface *iface;
3327 int miimon_interval;
3329 s->name = port->name;
3331 if (port->cfg->bond_mode) {
3332 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3333 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3334 port->name, port->cfg->bond_mode,
3335 bond_mode_to_string(s->balance));
3338 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3340 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3341 * active-backup. At some point we should remove this warning. */
3342 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3343 " in previous versions, the default bond_mode was"
3344 " balance-slb", port->name,
3345 bond_mode_to_string(s->balance));
3347 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3348 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3349 "please use another bond type or disable flood_vlans",
3353 miimon_interval = smap_get_int(&port->cfg->other_config,
3354 "bond-miimon-interval", 0);
3355 if (miimon_interval <= 0) {
3356 miimon_interval = 200;
3359 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3360 if (!detect_s || !strcmp(detect_s, "carrier")) {
3361 miimon_interval = 0;
3362 } else if (strcmp(detect_s, "miimon")) {
3363 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3364 "defaulting to carrier", port->name, detect_s);
3365 miimon_interval = 0;
3368 s->up_delay = MAX(0, port->cfg->bond_updelay);
3369 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3370 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3371 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3372 "bond-rebalance-interval", 10000);
3373 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3374 s->rebalance_interval = 1000;
3377 s->fake_iface = port->cfg->bond_fake_iface;
3379 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3380 "lacp-fallback-ab", false);
3382 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3383 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3387 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3388 * instead of obtaining it from the database. */
3390 port_is_synthetic(const struct port *port)
3392 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3395 /* Interface functions. */
3398 iface_is_internal(const struct ovsrec_interface *iface,
3399 const struct ovsrec_bridge *br)
3401 /* The local port and "internal" ports are always "internal". */
3402 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3405 /* Returns the correct network device type for interface 'iface' in bridge
3408 iface_get_type(const struct ovsrec_interface *iface,
3409 const struct ovsrec_bridge *br)
3413 /* The local port always has type "internal". Other ports take
3414 * their type from the database and default to "system" if none is
3416 if (iface_is_internal(iface, br)) {
3419 type = iface->type[0] ? iface->type : "system";
3422 return ofproto_port_open_type(br->datapath_type, type);
3426 iface_destroy__(struct iface *iface)
3429 struct port *port = iface->port;
3430 struct bridge *br = port->bridge;
3432 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3433 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3436 if (iface->ofp_port != OFPP_NONE) {
3437 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3440 list_remove(&iface->port_elem);
3441 hmap_remove(&br->iface_by_name, &iface->name_node);
3443 netdev_close(iface->netdev);
3451 iface_destroy(struct iface *iface)
3454 struct port *port = iface->port;
3456 iface_destroy__(iface);
3457 if (list_is_empty(&port->ifaces)) {
3463 static struct iface *
3464 iface_lookup(const struct bridge *br, const char *name)
3466 struct iface *iface;
3468 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3469 &br->iface_by_name) {
3470 if (!strcmp(iface->name, name)) {
3478 static struct iface *
3479 iface_find(const char *name)
3481 const struct bridge *br;
3483 HMAP_FOR_EACH (br, node, &all_bridges) {
3484 struct iface *iface = iface_lookup(br, name);
3493 static struct iface *
3494 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3496 struct iface *iface;
3498 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3500 if (iface->ofp_port == ofp_port) {
3507 /* Set Ethernet address of 'iface', if one is specified in the configuration
3510 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3512 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3513 struct iface *hw_addr_iface;
3515 if (strcmp(iface->type, "internal")) {
3519 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3521 } else if (port->cfg->fake_bridge) {
3522 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3523 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3528 if (iface->ofp_port == OFPP_LOCAL) {
3529 VLOG_ERR("interface %s: ignoring mac in Interface record "
3530 "(use Bridge record to set local port's mac)",
3532 } else if (eth_addr_is_multicast(mac)) {
3533 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3536 int error = netdev_set_etheraddr(iface->netdev, mac);
3538 VLOG_ERR("interface %s: setting MAC failed (%s)",
3539 iface->name, ovs_strerror(error));
3545 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3547 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3549 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3550 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3551 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3555 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3556 * sets the "ofport" field to -1.
3558 * This is appropriate when 'if_cfg''s interface cannot be created or is
3559 * otherwise invalid. */
3561 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3563 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3564 iface_set_ofport(if_cfg, OFPP_NONE);
3565 ovsrec_interface_set_status(if_cfg, NULL);
3566 ovsrec_interface_set_admin_state(if_cfg, NULL);
3567 ovsrec_interface_set_duplex(if_cfg, NULL);
3568 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3569 ovsrec_interface_set_link_state(if_cfg, NULL);
3570 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3571 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3572 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3573 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3574 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3575 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3576 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3577 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3582 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3584 union ovsdb_atom atom;
3586 atom.integer = target;
3587 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3591 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3593 struct ofpbuf queues_buf;
3595 ofpbuf_init(&queues_buf, 0);
3597 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3598 netdev_set_qos(iface->netdev, NULL, NULL);
3600 const struct ovsdb_datum *queues;
3601 struct netdev_queue_dump dump;
3602 unsigned int queue_id;
3603 struct smap details;
3607 /* Configure top-level Qos for 'iface'. */
3608 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3610 /* Deconfigure queues that were deleted. */
3611 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3613 smap_init(&details);
3614 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3615 if (!queue_ids_include(queues, queue_id)) {
3616 netdev_delete_queue(iface->netdev, queue_id);
3619 smap_destroy(&details);
3621 /* Configure queues for 'iface'. */
3623 for (i = 0; i < qos->n_queues; i++) {
3624 const struct ovsrec_queue *queue = qos->value_queues[i];
3625 unsigned int queue_id = qos->key_queues[i];
3627 if (queue_id == 0) {
3631 if (queue->n_dscp == 1) {
3632 struct ofproto_port_queue *port_queue;
3634 port_queue = ofpbuf_put_uninit(&queues_buf,
3635 sizeof *port_queue);
3636 port_queue->queue = queue_id;
3637 port_queue->dscp = queue->dscp[0];
3640 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3643 struct smap details;
3645 smap_init(&details);
3646 netdev_set_queue(iface->netdev, 0, &details);
3647 smap_destroy(&details);
3651 if (iface->ofp_port != OFPP_NONE) {
3652 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
3653 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
3655 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3656 port_queues, n_queues);
3659 netdev_set_policing(iface->netdev,
3660 iface->cfg->ingress_policing_rate,
3661 iface->cfg->ingress_policing_burst);
3663 ofpbuf_uninit(&queues_buf);
3667 iface_configure_cfm(struct iface *iface)
3669 const struct ovsrec_interface *cfg = iface->cfg;
3670 const char *opstate_str;
3671 const char *cfm_ccm_vlan;
3672 struct cfm_settings s;
3673 struct smap netdev_args;
3675 if (!cfg->n_cfm_mpid) {
3676 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3680 s.check_tnl_key = false;
3681 smap_init(&netdev_args);
3682 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3683 const char *key = smap_get(&netdev_args, "key");
3684 const char *in_key = smap_get(&netdev_args, "in_key");
3686 s.check_tnl_key = (key && !strcmp(key, "flow"))
3687 || (in_key && !strcmp(in_key, "flow"));
3689 smap_destroy(&netdev_args);
3691 s.mpid = *cfg->cfm_mpid;
3692 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3693 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3694 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3696 if (s.interval <= 0) {
3700 if (!cfm_ccm_vlan) {
3702 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3703 s.ccm_vlan = CFM_RANDOM_VLAN;
3705 s.ccm_vlan = atoi(cfm_ccm_vlan);
3706 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3711 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3713 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3715 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3716 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3718 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3721 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3722 * instead of obtaining it from the database. */
3724 iface_is_synthetic(const struct iface *iface)
3726 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3730 iface_validate_ofport__(size_t n, int64_t *ofport)
3732 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3733 ? u16_to_ofp(*ofport)
3738 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3740 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3744 iface_pick_ofport(const struct ovsrec_interface *cfg)
3746 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3747 return (requested_ofport != OFPP_NONE
3749 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3752 /* Port mirroring. */
3754 static struct mirror *
3755 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3759 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3760 if (uuid_equals(uuid, &m->uuid)) {
3768 bridge_configure_mirrors(struct bridge *br)
3770 const struct ovsdb_datum *mc;
3771 unsigned long *flood_vlans;
3772 struct mirror *m, *next;
3775 /* Get rid of deleted mirrors. */
3776 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3777 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3778 union ovsdb_atom atom;
3780 atom.uuid = m->uuid;
3781 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3786 /* Add new mirrors and reconfigure existing ones. */
3787 for (i = 0; i < br->cfg->n_mirrors; i++) {
3788 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3789 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3791 m = mirror_create(br, cfg);
3794 if (!mirror_configure(m)) {
3799 /* Update flooded vlans (for RSPAN). */
3800 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3801 br->cfg->n_flood_vlans);
3802 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3803 bitmap_free(flood_vlans);
3806 static struct mirror *
3807 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3811 m = xzalloc(sizeof *m);
3812 m->uuid = cfg->header_.uuid;
3813 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3815 m->name = xstrdup(cfg->name);
3821 mirror_destroy(struct mirror *m)
3824 struct bridge *br = m->bridge;
3827 ofproto_mirror_unregister(br->ofproto, m);
3830 hmap_remove(&br->mirrors, &m->hmap_node);
3837 mirror_collect_ports(struct mirror *m,
3838 struct ovsrec_port **in_ports, int n_in_ports,
3839 void ***out_portsp, size_t *n_out_portsp)
3841 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3842 size_t n_out_ports = 0;
3845 for (i = 0; i < n_in_ports; i++) {
3846 const char *name = in_ports[i]->name;
3847 struct port *port = port_lookup(m->bridge, name);
3849 out_ports[n_out_ports++] = port;
3851 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3852 "port %s", m->bridge->name, m->name, name);
3855 *out_portsp = out_ports;
3856 *n_out_portsp = n_out_ports;
3860 mirror_configure(struct mirror *m)
3862 const struct ovsrec_mirror *cfg = m->cfg;
3863 struct ofproto_mirror_settings s;
3866 if (strcmp(cfg->name, m->name)) {
3868 m->name = xstrdup(cfg->name);
3872 /* Get output port or VLAN. */
3873 if (cfg->output_port) {
3874 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3875 if (!s.out_bundle) {
3876 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3877 m->bridge->name, m->name);
3880 s.out_vlan = UINT16_MAX;
3882 if (cfg->output_vlan) {
3883 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3884 "output vlan; ignoring output vlan",
3885 m->bridge->name, m->name);
3887 } else if (cfg->output_vlan) {
3888 /* The database should prevent invalid VLAN values. */
3889 s.out_bundle = NULL;
3890 s.out_vlan = *cfg->output_vlan;
3892 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3893 m->bridge->name, m->name);
3897 /* Get port selection. */
3898 if (cfg->select_all) {
3899 size_t n_ports = hmap_count(&m->bridge->ports);
3900 void **ports = xmalloc(n_ports * sizeof *ports);
3905 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3915 /* Get ports, dropping ports that don't exist.
3916 * The IDL ensures that there are no duplicates. */
3917 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3918 &s.srcs, &s.n_srcs);
3919 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3920 &s.dsts, &s.n_dsts);
3923 /* Get VLAN selection. */
3924 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3927 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3930 if (s.srcs != s.dsts) {
3939 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3941 * This is deprecated. It is only for compatibility with broken device drivers
3942 * in old versions of Linux that do not properly support VLANs when VLAN
3943 * devices are not used. When broken device drivers are no longer in
3944 * widespread use, we will delete these interfaces. */
3946 static struct ovsrec_port **recs;
3947 static size_t n_recs, allocated_recs;
3949 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3950 * splinters are reconfigured. */
3952 register_rec(struct ovsrec_port *rec)
3954 if (n_recs >= allocated_recs) {
3955 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3957 recs[n_recs++] = rec;
3960 /* Frees all of the ports registered with register_reg(). */
3962 free_registered_recs(void)
3966 for (i = 0; i < n_recs; i++) {
3967 struct ovsrec_port *port = recs[i];
3970 for (j = 0; j < port->n_interfaces; j++) {
3971 struct ovsrec_interface *iface = port->interfaces[j];
3976 smap_destroy(&port->other_config);
3977 free(port->interfaces);
3985 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3988 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3990 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3994 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3997 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3998 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3999 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4002 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4003 * use, returns NULL.
4005 * Updates 'vlan_splinters_enabled_anywhere'. */
4006 static unsigned long int *
4007 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4009 unsigned long int *splinter_vlans;
4010 struct sset splinter_ifaces;
4011 const char *real_dev_name;
4012 struct shash *real_devs;
4013 struct shash_node *node;
4017 /* Free space allocated for synthesized ports and interfaces, since we're
4018 * in the process of reconstructing all of them. */
4019 free_registered_recs();
4021 splinter_vlans = bitmap_allocate(4096);
4022 sset_init(&splinter_ifaces);
4023 vlan_splinters_enabled_anywhere = false;
4024 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4025 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4028 for (j = 0; j < br_cfg->n_ports; j++) {
4029 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4032 for (k = 0; k < port_cfg->n_interfaces; k++) {
4033 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4035 if (vlan_splinters_is_enabled(iface_cfg)) {
4036 vlan_splinters_enabled_anywhere = true;
4037 sset_add(&splinter_ifaces, iface_cfg->name);
4038 vlan_bitmap_from_array__(port_cfg->trunks,
4044 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4045 bitmap_set1(splinter_vlans, *port_cfg->tag);
4050 if (!vlan_splinters_enabled_anywhere) {
4051 free(splinter_vlans);
4052 sset_destroy(&splinter_ifaces);
4056 HMAP_FOR_EACH (br, node, &all_bridges) {
4058 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4062 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4063 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4064 * device to be created for it. */
4065 bitmap_set0(splinter_vlans, 0);
4066 bitmap_set0(splinter_vlans, 4095);
4068 /* Delete all VLAN devices that we don't need. */
4070 real_devs = vlandev_get_real_devs();
4071 SHASH_FOR_EACH (node, real_devs) {
4072 const struct vlan_real_dev *real_dev = node->data;
4073 const struct vlan_dev *vlan_dev;
4074 bool real_dev_has_splinters;
4076 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4078 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4079 if (!real_dev_has_splinters
4080 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4081 struct netdev *netdev;
4083 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4084 if (!netdev_get_in4(netdev, NULL, NULL) ||
4085 !netdev_get_in6(netdev, NULL)) {
4086 /* It has an IP address configured, so we don't own
4087 * it. Don't delete it. */
4089 vlandev_del(vlan_dev->name);
4091 netdev_close(netdev);
4098 /* Add all VLAN devices that we need. */
4099 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4102 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4103 if (!vlandev_get_name(real_dev_name, vid)) {
4104 vlandev_add(real_dev_name, vid);
4111 sset_destroy(&splinter_ifaces);
4113 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4114 free(splinter_vlans);
4117 return splinter_vlans;
4120 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4123 configure_splinter_port(struct port *port)
4125 struct ofproto *ofproto = port->bridge->ofproto;
4126 ofp_port_t realdev_ofp_port;
4127 const char *realdev_name;
4128 struct iface *vlandev, *realdev;
4130 ofproto_bundle_unregister(port->bridge->ofproto, port);
4132 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4135 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4136 realdev = iface_lookup(port->bridge, realdev_name);
4137 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4139 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4143 static struct ovsrec_port *
4144 synthesize_splinter_port(const char *real_dev_name,
4145 const char *vlan_dev_name, int vid)
4147 struct ovsrec_interface *iface;
4148 struct ovsrec_port *port;
4150 iface = xmalloc(sizeof *iface);
4151 ovsrec_interface_init(iface);
4152 iface->name = xstrdup(vlan_dev_name);
4153 iface->type = "system";
4155 port = xmalloc(sizeof *port);
4156 ovsrec_port_init(port);
4157 port->interfaces = xmemdup(&iface, sizeof iface);
4158 port->n_interfaces = 1;
4159 port->name = xstrdup(vlan_dev_name);
4160 port->vlan_mode = "splinter";
4161 port->tag = xmalloc(sizeof *port->tag);
4164 smap_add(&port->other_config, "realdev", real_dev_name);
4170 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4171 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4172 * 1-bit in the 'splinter_vlans' bitmap. */
4174 add_vlan_splinter_ports(struct bridge *br,
4175 const unsigned long int *splinter_vlans,
4176 struct shash *ports)
4180 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4181 * we're modifying 'ports'. */
4182 for (i = 0; i < br->cfg->n_ports; i++) {
4183 const char *name = br->cfg->ports[i]->name;
4184 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4187 for (j = 0; j < port_cfg->n_interfaces; j++) {
4188 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4190 if (vlan_splinters_is_enabled(iface_cfg)) {
4191 const char *real_dev_name;
4194 real_dev_name = iface_cfg->name;
4195 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4196 const char *vlan_dev_name;
4198 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4200 && !shash_find(ports, vlan_dev_name)) {
4201 shash_add(ports, vlan_dev_name,
4202 synthesize_splinter_port(
4203 real_dev_name, vlan_dev_name, vid));
4212 mirror_refresh_stats(struct mirror *m)
4214 struct ofproto *ofproto = m->bridge->ofproto;
4215 uint64_t tx_packets, tx_bytes;
4218 size_t stat_cnt = 0;
4220 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4221 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4225 if (tx_packets != UINT64_MAX) {
4226 keys[stat_cnt] = "tx_packets";
4227 values[stat_cnt] = tx_packets;
4230 if (tx_bytes != UINT64_MAX) {
4231 keys[stat_cnt] = "tx_bytes";
4232 values[stat_cnt] = tx_bytes;
4236 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);