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. */
82 /* These members are valid only within bridge_reconfigure(). */
83 const char *type; /* Usually same as cfg->type. */
84 const struct ovsrec_interface *cfg;
88 struct uuid uuid; /* UUID of this "mirror" record in database. */
89 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
90 struct bridge *bridge;
92 const struct ovsrec_mirror *cfg;
96 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
97 struct bridge *bridge;
100 const struct ovsrec_port *cfg;
102 /* An ordinary bridge port has 1 interface.
103 * A bridge port for bonding has at least 2 interfaces. */
104 struct list ifaces; /* List of "struct iface"s. */
108 struct hmap_node node; /* In 'all_bridges'. */
109 char *name; /* User-specified arbitrary name. */
110 char *type; /* Datapath type. */
111 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
112 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
113 const struct ovsrec_bridge *cfg;
115 /* OpenFlow switch processing. */
116 struct ofproto *ofproto; /* OpenFlow switch. */
119 struct hmap ports; /* "struct port"s indexed by name. */
120 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
121 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
123 /* Port mirroring. */
124 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
126 /* Used during reconfiguration. */
127 struct shash wanted_ports;
129 /* Synthetic local port if necessary. */
130 struct ovsrec_port synth_local_port;
131 struct ovsrec_interface synth_local_iface;
132 struct ovsrec_interface *synth_local_ifacep;
135 /* All bridges, indexed by name. */
136 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
138 /* OVSDB IDL used to obtain configuration. */
139 static struct ovsdb_idl *idl;
141 /* We want to complete daemonization, fully detaching from our parent process,
142 * only after we have completed our initial configuration, committed our state
143 * to the database, and received confirmation back from the database server
144 * that it applied the commit. This allows our parent process to know that,
145 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
146 * to have already been filled in. (It is only "very likely" rather than
147 * certain because there is always a slim possibility that the transaction will
148 * fail or that some other client has added new bridges, ports, etc. while
149 * ovs-vswitchd was configuring using an old configuration.)
151 * We only need to do this once for our initial configuration at startup, so
152 * 'initial_config_done' tracks whether we've already done it. While we are
153 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
154 * itself and is otherwise NULL. */
155 static bool initial_config_done;
156 static struct ovsdb_idl_txn *daemonize_txn;
158 /* Most recently processed IDL sequence number. */
159 static unsigned int idl_seqno;
161 /* Track changes to port connectivity. */
162 static uint64_t connectivity_seqno = LLONG_MIN;
164 /* Status update to database.
166 * Some information in the database must be kept as up-to-date as possible to
167 * allow controllers to respond rapidly to network outages. Those status are
168 * updated via the 'status_txn'.
170 * We use the global connectivity sequence number to detect the status change.
171 * Also, to prevent the status update from sending too much to the database,
172 * we check the return status of each update transaction and do not start new
173 * update if the previous transaction status is 'TXN_INCOMPLETE'.
175 * 'statux_txn' is NULL if there is no ongoing status update.
177 static struct ovsdb_idl_txn *status_txn;
179 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
180 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
181 #define STATUS_CHECK_AGAIN_MSEC 100
183 /* Each time this timer expires, the bridge fetches interface and mirror
184 * statistics and pushes them into the database. */
185 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
186 static long long int iface_stats_timer = LLONG_MIN;
188 /* Set to true to allow experimental use of OpenFlow 1.4.
189 * This is false initially because OpenFlow 1.4 is not yet safe to use: it can
190 * abort due to unimplemented features. */
191 static bool allow_of14;
193 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
194 * expensive. This can cause bridge_reconfigure() to take a long time during
195 * which no other work can be done. To deal with this problem, we limit port
196 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
197 * call to bridge_reconfigure(). If there is more work to do after the limit
198 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
199 * This allows the rest of the code to catch up on important things like
200 * forwarding packets. */
201 #define OFP_PORT_ACTION_WINDOW 10
203 static void add_del_bridges(const struct ovsrec_open_vswitch *);
204 static void bridge_run__(void);
205 static void bridge_create(const struct ovsrec_bridge *);
206 static void bridge_destroy(struct bridge *);
207 static struct bridge *bridge_lookup(const char *name);
208 static unixctl_cb_func bridge_unixctl_dump_flows;
209 static unixctl_cb_func bridge_unixctl_reconnect;
210 static size_t bridge_get_controllers(const struct bridge *br,
211 struct ovsrec_controller ***controllersp);
212 static void bridge_collect_wanted_ports(struct bridge *,
213 const unsigned long *splinter_vlans,
214 struct shash *wanted_ports);
215 static void bridge_delete_ofprotos(void);
216 static void bridge_delete_or_reconfigure_ports(struct bridge *);
217 static void bridge_del_ports(struct bridge *,
218 const struct shash *wanted_ports);
219 static void bridge_add_ports(struct bridge *,
220 const struct shash *wanted_ports);
222 static void bridge_configure_datapath_id(struct bridge *);
223 static void bridge_configure_netflow(struct bridge *);
224 static void bridge_configure_forward_bpdu(struct bridge *);
225 static void bridge_configure_mac_table(struct bridge *);
226 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
227 static void bridge_configure_ipfix(struct bridge *);
228 static void bridge_configure_stp(struct bridge *);
229 static void bridge_configure_tables(struct bridge *);
230 static void bridge_configure_dp_desc(struct bridge *);
231 static void bridge_configure_remotes(struct bridge *,
232 const struct sockaddr_in *managers,
234 static void bridge_pick_local_hw_addr(struct bridge *,
235 uint8_t ea[ETH_ADDR_LEN],
236 struct iface **hw_addr_iface);
237 static uint64_t bridge_pick_datapath_id(struct bridge *,
238 const uint8_t bridge_ea[ETH_ADDR_LEN],
239 struct iface *hw_addr_iface);
240 static uint64_t dpid_from_hash(const void *, size_t nbytes);
241 static bool bridge_has_bond_fake_iface(const struct bridge *,
243 static bool port_is_bond_fake_iface(const struct port *);
245 static unixctl_cb_func qos_unixctl_show;
247 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
248 static void port_del_ifaces(struct port *);
249 static void port_destroy(struct port *);
250 static struct port *port_lookup(const struct bridge *, const char *name);
251 static void port_configure(struct port *);
252 static struct lacp_settings *port_configure_lacp(struct port *,
253 struct lacp_settings *);
254 static void port_configure_bond(struct port *, struct bond_settings *);
255 static bool port_is_synthetic(const struct port *);
257 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
258 static void run_system_stats(void);
260 static void bridge_configure_mirrors(struct bridge *);
261 static struct mirror *mirror_create(struct bridge *,
262 const struct ovsrec_mirror *);
263 static void mirror_destroy(struct mirror *);
264 static bool mirror_configure(struct mirror *);
265 static void mirror_refresh_stats(struct mirror *);
267 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
268 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
269 const struct ovsrec_port *);
270 static bool iface_is_internal(const struct ovsrec_interface *iface,
271 const struct ovsrec_bridge *br);
272 static const char *iface_get_type(const struct ovsrec_interface *,
273 const struct ovsrec_bridge *);
274 static void iface_destroy(struct iface *);
275 static void iface_destroy__(struct iface *);
276 static struct iface *iface_lookup(const struct bridge *, const char *name);
277 static struct iface *iface_find(const char *name);
278 static struct iface *iface_from_ofp_port(const struct bridge *,
279 ofp_port_t ofp_port);
280 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
281 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
282 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
283 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
284 static void iface_configure_cfm(struct iface *);
285 static void iface_refresh_cfm_stats(struct iface *);
286 static void iface_refresh_stats(struct iface *);
287 static void iface_refresh_netdev_status(struct iface *);
288 static void iface_refresh_ofproto_status(struct iface *);
289 static bool iface_is_synthetic(const struct iface *);
290 static ofp_port_t iface_get_requested_ofp_port(
291 const struct ovsrec_interface *);
292 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
294 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
296 * This is deprecated. It is only for compatibility with broken device drivers
297 * in old versions of Linux that do not properly support VLANs when VLAN
298 * devices are not used. When broken device drivers are no longer in
299 * widespread use, we will delete these interfaces. */
301 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
302 static bool vlan_splinters_enabled_anywhere;
304 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
305 static unsigned long int *collect_splinter_vlans(
306 const struct ovsrec_open_vswitch *);
307 static void configure_splinter_port(struct port *);
308 static void add_vlan_splinter_ports(struct bridge *,
309 const unsigned long int *splinter_vlans,
310 struct shash *ports);
313 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
315 struct shash iface_hints;
316 static bool initialized = false;
323 shash_init(&iface_hints);
326 for (i = 0; i < cfg->n_bridges; i++) {
327 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
330 for (j = 0; j < br_cfg->n_ports; j++) {
331 struct ovsrec_port *port_cfg = br_cfg->ports[j];
334 for (k = 0; k < port_cfg->n_interfaces; k++) {
335 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
336 struct iface_hint *iface_hint;
338 iface_hint = xmalloc(sizeof *iface_hint);
339 iface_hint->br_name = br_cfg->name;
340 iface_hint->br_type = br_cfg->datapath_type;
341 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
343 shash_add(&iface_hints, if_cfg->name, iface_hint);
349 ofproto_init(&iface_hints);
351 shash_destroy_free_data(&iface_hints);
355 /* Public functions. */
357 /* Initializes the bridge module, configuring it to obtain its configuration
358 * from an OVSDB server accessed over 'remote', which should be a string in a
359 * form acceptable to ovsdb_idl_create(). */
361 bridge_init(const char *remote)
363 /* Create connection to database. */
364 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
365 idl_seqno = ovsdb_idl_get_seqno(idl);
366 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
367 ovsdb_idl_verify_write_only(idl);
369 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
370 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
371 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
372 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
373 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
374 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
375 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
377 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
378 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
379 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
381 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
382 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
383 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
385 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
386 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
387 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
388 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
389 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
390 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
391 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
392 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
393 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
394 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
395 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
396 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
397 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
398 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
399 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
400 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
401 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
402 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
403 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
404 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
406 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
407 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
408 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
409 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
411 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
413 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
415 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
416 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
418 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
419 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
420 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
421 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
423 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
424 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
425 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
426 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
427 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
429 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
431 /* Register unixctl commands. */
432 unixctl_command_register("qos/show", "interface", 1, 1,
433 qos_unixctl_show, NULL);
434 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
435 bridge_unixctl_dump_flows, NULL);
436 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
437 bridge_unixctl_reconnect, NULL);
447 struct bridge *br, *next_br;
449 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
452 ovsdb_idl_destroy(idl);
455 /* Enables use of OpenFlow 1.4. This is off by default because OpenFlow 1.4 is
456 * not yet safe to use: it can abort due to unimplemented features. */
458 bridge_enable_of14(void)
463 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
464 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
465 * responsible for freeing '*managersp' (with free()).
467 * You may be asking yourself "why does ovs-vswitchd care?", because
468 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
469 * should not be and in fact is not directly involved in that. But
470 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
471 * it has to tell in-band control where the managers are to enable that.
472 * (Thus, only managers connected in-band are collected.)
475 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
476 struct sockaddr_in **managersp, size_t *n_managersp)
478 struct sockaddr_in *managers = NULL;
479 size_t n_managers = 0;
483 /* Collect all of the potential targets from the "targets" columns of the
484 * rows pointed to by "manager_options", excluding any that are
487 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
488 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
490 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
491 sset_find_and_delete(&targets, m->target);
493 sset_add(&targets, m->target);
497 /* Now extract the targets' IP addresses. */
498 if (!sset_is_empty(&targets)) {
501 managers = xmalloc(sset_count(&targets) * sizeof *managers);
502 SSET_FOR_EACH (target, &targets) {
503 struct sockaddr_storage ss;
505 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
507 && ss.ss_family == AF_INET) {
508 managers[n_managers++] = *(struct sockaddr_in *) &ss;
512 sset_destroy(&targets);
514 *managersp = managers;
515 *n_managersp = n_managers;
519 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
521 unsigned long int *splinter_vlans;
522 struct sockaddr_in *managers;
523 struct bridge *br, *next;
524 int sflow_bridge_number;
527 COVERAGE_INC(bridge_reconfigure);
529 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
530 OFPROTO_FLOW_LIMIT_DEFAULT));
531 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
532 OFPROTO_MAX_IDLE_DEFAULT));
535 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
536 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
538 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
539 * to 'ovs_cfg', with only very minimal configuration otherwise.
541 * This is mostly an update to bridge data structures. Nothing is pushed
542 * down to ofproto or lower layers. */
543 add_del_bridges(ovs_cfg);
544 splinter_vlans = collect_splinter_vlans(ovs_cfg);
545 HMAP_FOR_EACH (br, node, &all_bridges) {
546 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
547 bridge_del_ports(br, &br->wanted_ports);
549 free(splinter_vlans);
551 /* Start pushing configuration changes down to the ofproto layer:
553 * - Delete ofprotos that are no longer configured.
555 * - Delete ports that are no longer configured.
557 * - Reconfigure existing ports to their desired configurations, or
558 * delete them if not possible.
560 * We have to do all the deletions before we can do any additions, because
561 * the ports to be added might require resources that will be freed up by
562 * deletions (they might especially overlap in name). */
563 bridge_delete_ofprotos();
564 HMAP_FOR_EACH (br, node, &all_bridges) {
566 bridge_delete_or_reconfigure_ports(br);
570 /* Finish pushing configuration changes to the ofproto layer:
572 * - Create ofprotos that are missing.
574 * - Add ports that are missing. */
575 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
579 error = ofproto_create(br->name, br->type, &br->ofproto);
581 VLOG_ERR("failed to create bridge %s: %s", br->name,
582 ovs_strerror(error));
583 shash_destroy(&br->wanted_ports);
588 HMAP_FOR_EACH (br, node, &all_bridges) {
589 bridge_add_ports(br, &br->wanted_ports);
590 shash_destroy(&br->wanted_ports);
593 reconfigure_system_stats(ovs_cfg);
595 /* Complete the configuration. */
596 sflow_bridge_number = 0;
597 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
598 HMAP_FOR_EACH (br, node, &all_bridges) {
601 /* We need the datapath ID early to allow LACP ports to use it as the
602 * default system ID. */
603 bridge_configure_datapath_id(br);
605 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
608 port_configure(port);
610 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
611 iface_set_ofport(iface->cfg, iface->ofp_port);
612 iface_configure_cfm(iface);
613 iface_configure_qos(iface, port->cfg->qos);
614 iface_set_mac(br, port, iface);
615 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
619 bridge_configure_mirrors(br);
620 bridge_configure_forward_bpdu(br);
621 bridge_configure_mac_table(br);
622 bridge_configure_remotes(br, managers, n_managers);
623 bridge_configure_netflow(br);
624 bridge_configure_sflow(br, &sflow_bridge_number);
625 bridge_configure_ipfix(br);
626 bridge_configure_stp(br);
627 bridge_configure_tables(br);
628 bridge_configure_dp_desc(br);
630 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
631 /* XXX: Remove this warning message eventually. */
632 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
633 " moved to the Open_vSwitch table. Ignoring its"
634 " setting in the bridge table.");
639 /* The ofproto-dpif provider does some final reconfiguration in its
640 * ->type_run() function. We have to call it before notifying the database
641 * client that reconfiguration is complete, otherwise there is a very
642 * narrow race window in which e.g. ofproto/trace will not recognize the
643 * new configuration (sometimes this causes unit test failures). */
647 /* Delete ofprotos which aren't configured or have the wrong type. Create
648 * ofprotos which don't exist but need to. */
650 bridge_delete_ofprotos(void)
657 /* Delete ofprotos with no bridge or with the wrong type. */
660 ofproto_enumerate_types(&types);
661 SSET_FOR_EACH (type, &types) {
664 ofproto_enumerate_names(type, &names);
665 SSET_FOR_EACH (name, &names) {
666 br = bridge_lookup(name);
667 if (!br || strcmp(type, br->type)) {
668 ofproto_delete(name, type);
672 sset_destroy(&names);
673 sset_destroy(&types);
677 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
679 if (*n >= *allocated) {
680 ports = x2nrealloc(ports, allocated, sizeof *ports);
682 ports[(*n)++] = port;
687 bridge_delete_or_reconfigure_ports(struct bridge *br)
689 struct ofproto_port ofproto_port;
690 struct ofproto_port_dump dump;
692 struct sset ofproto_ports;
693 struct port *port, *port_next;
695 /* List of "ofp_port"s to delete. We make a list instead of deleting them
696 * right away because ofproto implementations aren't necessarily able to
697 * iterate through a changing list of ports in an entirely robust way. */
704 sset_init(&ofproto_ports);
706 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
707 * that are not configured in the database. (This commonly happens when
708 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
710 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
711 * that have the wrong OpenFlow port number (and arrange to add them back
712 * with the correct OpenFlow port number). */
713 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
714 ofp_port_t requested_ofp_port;
717 sset_add(&ofproto_ports, ofproto_port.name);
719 iface = iface_lookup(br, ofproto_port.name);
721 /* No such iface is configured, so we should delete this
724 * As a corner case exception, keep the port if it's a bond fake
726 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
727 && !strcmp(ofproto_port.type, "internal")) {
733 if (strcmp(ofproto_port.type, iface->type)
734 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
735 /* The interface is the wrong type or can't be configured.
740 /* If the requested OpenFlow port for 'iface' changed, and it's not
741 * already the correct port, then we might want to temporarily delete
742 * this interface, so we can add it back again with the new OpenFlow
744 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
745 if (iface->ofp_port != OFPP_LOCAL &&
746 requested_ofp_port != OFPP_NONE &&
747 requested_ofp_port != iface->ofp_port) {
748 ofp_port_t victim_request;
749 struct iface *victim;
751 /* Check for an existing OpenFlow port currently occupying
752 * 'iface''s requested port number. If there isn't one, then
753 * delete this port. Otherwise we need to consider further. */
754 victim = iface_from_ofp_port(br, requested_ofp_port);
759 /* 'victim' is a port currently using 'iface''s requested port
760 * number. Unless 'victim' specifically requested that port
761 * number, too, then we can delete both 'iface' and 'victim'
762 * temporarily. (We'll add both of them back again later with new
763 * OpenFlow port numbers.)
765 * If 'victim' did request port number 'requested_ofp_port', just
766 * like 'iface', then that's a configuration inconsistency that we
767 * can't resolve. We might as well let it keep its current port
769 victim_request = iface_get_requested_ofp_port(victim->cfg);
770 if (victim_request != requested_ofp_port) {
771 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
772 iface_destroy(victim);
781 iface_destroy(iface);
782 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
784 for (i = 0; i < n; i++) {
785 ofproto_port_del(br->ofproto, del[i]);
789 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
790 * that we didn't see when we iterated through the datapath, i.e. ports
791 * that disappeared underneath use. This is an unusual situation, but it
792 * can happen in some cases:
794 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
795 * idea but could happen).
797 * - The port represented a device that disappeared, e.g. a tuntap
798 * device destroyed via "tunctl -d", a physical Ethernet device
799 * whose module was just unloaded via "rmmod", or a virtual NIC for a
800 * VM whose VM was just terminated. */
801 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
802 struct iface *iface, *iface_next;
804 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
805 if (!sset_contains(&ofproto_ports, iface->name)) {
806 iface_destroy__(iface);
810 if (list_is_empty(&port->ifaces)) {
814 sset_destroy(&ofproto_ports);
818 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
819 bool with_requested_port)
821 struct shash_node *port_node;
823 SHASH_FOR_EACH (port_node, wanted_ports) {
824 const struct ovsrec_port *port_cfg = port_node->data;
827 for (i = 0; i < port_cfg->n_interfaces; i++) {
828 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
829 ofp_port_t requested_ofp_port;
831 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
832 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
833 struct iface *iface = iface_lookup(br, iface_cfg->name);
836 iface_create(br, iface_cfg, port_cfg);
844 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
846 /* First add interfaces that request a particular port number. */
847 bridge_add_ports__(br, wanted_ports, true);
849 /* Then add interfaces that want automatic port number assignment.
850 * We add these afterward to avoid accidentally taking a specifically
851 * requested port number. */
852 bridge_add_ports__(br, wanted_ports, false);
856 port_configure(struct port *port)
858 const struct ovsrec_port *cfg = port->cfg;
859 struct bond_settings bond_settings;
860 struct lacp_settings lacp_settings;
861 struct ofproto_bundle_settings s;
864 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
865 configure_splinter_port(port);
874 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
875 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
876 s.slaves[s.n_slaves++] = iface->ofp_port;
881 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
885 /* Get VLAN trunks. */
888 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
892 if (cfg->vlan_mode) {
893 if (!strcmp(cfg->vlan_mode, "access")) {
894 s.vlan_mode = PORT_VLAN_ACCESS;
895 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
896 s.vlan_mode = PORT_VLAN_TRUNK;
897 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
898 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
899 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
900 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
902 /* This "can't happen" because ovsdb-server should prevent it. */
903 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
904 "back to trunk mode", port->name, cfg->vlan_mode);
905 s.vlan_mode = PORT_VLAN_TRUNK;
909 s.vlan_mode = PORT_VLAN_ACCESS;
911 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
915 s.vlan_mode = PORT_VLAN_TRUNK;
918 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
921 /* Get LACP settings. */
922 s.lacp = port_configure_lacp(port, &lacp_settings);
926 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
927 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
928 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
931 s.lacp_slaves = NULL;
934 /* Get bond settings. */
935 if (s.n_slaves > 1) {
936 s.bond = &bond_settings;
937 port_configure_bond(port, &bond_settings);
940 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
941 netdev_set_miimon_interval(iface->netdev, 0);
946 ofproto_bundle_register(port->bridge->ofproto, port, &s);
954 /* Pick local port hardware address and datapath ID for 'br'. */
956 bridge_configure_datapath_id(struct bridge *br)
958 uint8_t ea[ETH_ADDR_LEN];
960 struct iface *local_iface;
961 struct iface *hw_addr_iface;
964 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
965 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
967 int error = netdev_set_etheraddr(local_iface->netdev, ea);
969 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
970 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
971 "Ethernet address: %s",
972 br->name, ovs_strerror(error));
975 memcpy(br->ea, ea, ETH_ADDR_LEN);
977 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
978 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
979 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
980 ofproto_set_datapath_id(br->ofproto, dpid);
983 dpid_string = xasprintf("%016"PRIx64, dpid);
984 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
988 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
991 bridge_get_allowed_versions(struct bridge *br)
993 uint32_t allowed_versions;
995 if (!br->cfg->n_protocols)
998 allowed_versions = ofputil_versions_from_strings(br->cfg->protocols,
999 br->cfg->n_protocols);
1001 allowed_versions &= ~(1u << OFP14_VERSION);
1003 return allowed_versions;
1006 /* Set NetFlow configuration on 'br'. */
1008 bridge_configure_netflow(struct bridge *br)
1010 struct ovsrec_netflow *cfg = br->cfg->netflow;
1011 struct netflow_options opts;
1014 ofproto_set_netflow(br->ofproto, NULL);
1018 memset(&opts, 0, sizeof opts);
1020 /* Get default NetFlow configuration from datapath.
1021 * Apply overrides from 'cfg'. */
1022 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1023 if (cfg->engine_type) {
1024 opts.engine_type = *cfg->engine_type;
1026 if (cfg->engine_id) {
1027 opts.engine_id = *cfg->engine_id;
1030 /* Configure active timeout interval. */
1031 opts.active_timeout = cfg->active_timeout;
1032 if (!opts.active_timeout) {
1033 opts.active_timeout = -1;
1034 } else if (opts.active_timeout < 0) {
1035 VLOG_WARN("bridge %s: active timeout interval set to negative "
1036 "value, using default instead (%d seconds)", br->name,
1037 NF_ACTIVE_TIMEOUT_DEFAULT);
1038 opts.active_timeout = -1;
1041 /* Add engine ID to interface number to disambiguate bridgs? */
1042 opts.add_id_to_iface = cfg->add_id_to_interface;
1043 if (opts.add_id_to_iface) {
1044 if (opts.engine_id > 0x7f) {
1045 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1046 "another vswitch, choose an engine id less than 128",
1049 if (hmap_count(&br->ports) > 508) {
1050 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1051 "another port when more than 508 ports are used",
1057 sset_init(&opts.collectors);
1058 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1061 if (ofproto_set_netflow(br->ofproto, &opts)) {
1062 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1064 sset_destroy(&opts.collectors);
1067 /* Set sFlow configuration on 'br'. */
1069 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1071 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1072 struct ovsrec_controller **controllers;
1073 struct ofproto_sflow_options oso;
1074 size_t n_controllers;
1078 ofproto_set_sflow(br->ofproto, NULL);
1082 memset(&oso, 0, sizeof oso);
1084 sset_init(&oso.targets);
1085 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1087 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1088 if (cfg->sampling) {
1089 oso.sampling_rate = *cfg->sampling;
1092 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1094 oso.polling_interval = *cfg->polling;
1097 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1099 oso.header_len = *cfg->header;
1102 oso.sub_id = (*sflow_bridge_number)++;
1103 oso.agent_device = cfg->agent;
1105 oso.control_ip = NULL;
1106 n_controllers = bridge_get_controllers(br, &controllers);
1107 for (i = 0; i < n_controllers; i++) {
1108 if (controllers[i]->local_ip) {
1109 oso.control_ip = controllers[i]->local_ip;
1113 ofproto_set_sflow(br->ofproto, &oso);
1115 sset_destroy(&oso.targets);
1118 /* Returns whether a IPFIX row is valid. */
1120 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1122 return ipfix && ipfix->n_targets > 0;
1125 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1127 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1128 const struct bridge *br)
1130 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1133 /* Set IPFIX configuration on 'br'. */
1135 bridge_configure_ipfix(struct bridge *br)
1137 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1138 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1139 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1140 struct ofproto_ipfix_bridge_exporter_options be_opts;
1141 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1142 size_t n_fe_opts = 0;
1144 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1145 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1150 if (!valid_be_cfg && n_fe_opts == 0) {
1151 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1156 memset(&be_opts, 0, sizeof be_opts);
1158 sset_init(&be_opts.targets);
1159 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1161 if (be_cfg->sampling) {
1162 be_opts.sampling_rate = *be_cfg->sampling;
1164 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1166 if (be_cfg->obs_domain_id) {
1167 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1169 if (be_cfg->obs_point_id) {
1170 be_opts.obs_point_id = *be_cfg->obs_point_id;
1172 if (be_cfg->cache_active_timeout) {
1173 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1175 if (be_cfg->cache_max_flows) {
1176 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1180 if (n_fe_opts > 0) {
1181 struct ofproto_ipfix_flow_exporter_options *opts;
1182 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1184 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1185 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1186 opts->collector_set_id = fe_cfg->id;
1187 sset_init(&opts->targets);
1188 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1189 fe_cfg->ipfix->n_targets);
1190 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1191 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1192 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1193 ? *fe_cfg->ipfix->cache_max_flows : 0;
1199 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1203 sset_destroy(&be_opts.targets);
1206 if (n_fe_opts > 0) {
1207 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1209 for (i = 0; i < n_fe_opts; i++) {
1210 sset_destroy(&opts->targets);
1218 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1219 struct ofproto_port_stp_settings *port_s,
1220 int *port_num_counter, unsigned long *port_num_bitmap)
1222 const char *config_str;
1223 struct iface *iface;
1225 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1226 port_s->enable = false;
1229 port_s->enable = true;
1232 /* STP over bonds is not supported. */
1233 if (!list_is_singleton(&port->ifaces)) {
1234 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1236 port_s->enable = false;
1240 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1242 /* Internal ports shouldn't participate in spanning tree, so
1244 if (!strcmp(iface->type, "internal")) {
1245 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1246 port_s->enable = false;
1250 /* STP on mirror output ports is not supported. */
1251 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1252 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1253 port_s->enable = false;
1257 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1259 unsigned long int port_num = strtoul(config_str, NULL, 0);
1260 int port_idx = port_num - 1;
1262 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1263 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1264 port_s->enable = false;
1268 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1269 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1270 port->name, port_num);
1271 port_s->enable = false;
1274 bitmap_set1(port_num_bitmap, port_idx);
1275 port_s->port_num = port_idx;
1277 if (*port_num_counter >= STP_MAX_PORTS) {
1278 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1279 port_s->enable = false;
1283 port_s->port_num = (*port_num_counter)++;
1286 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1288 port_s->path_cost = strtoul(config_str, NULL, 10);
1290 enum netdev_features current;
1293 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1294 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1295 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1298 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1300 port_s->priority = strtoul(config_str, NULL, 0);
1302 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1306 /* Set spanning tree configuration on 'br'. */
1308 bridge_configure_stp(struct bridge *br)
1310 if (!br->cfg->stp_enable) {
1311 ofproto_set_stp(br->ofproto, NULL);
1313 struct ofproto_stp_settings br_s;
1314 const char *config_str;
1316 int port_num_counter;
1317 unsigned long *port_num_bitmap;
1319 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1321 uint8_t ea[ETH_ADDR_LEN];
1323 if (eth_addr_from_string(config_str, ea)) {
1324 br_s.system_id = eth_addr_to_uint64(ea);
1326 br_s.system_id = eth_addr_to_uint64(br->ea);
1327 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1328 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1331 br_s.system_id = eth_addr_to_uint64(br->ea);
1334 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1336 br_s.priority = strtoul(config_str, NULL, 0);
1338 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1341 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1343 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1345 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1348 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1350 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1352 br_s.max_age = STP_DEFAULT_MAX_AGE;
1355 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1357 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1359 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1362 /* Configure STP on the bridge. */
1363 if (ofproto_set_stp(br->ofproto, &br_s)) {
1364 VLOG_ERR("bridge %s: could not enable STP", br->name);
1368 /* Users must either set the port number with the "stp-port-num"
1369 * configuration on all ports or none. If manual configuration
1370 * is not done, then we allocate them sequentially. */
1371 port_num_counter = 0;
1372 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1373 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1374 struct ofproto_port_stp_settings port_s;
1375 struct iface *iface;
1377 port_configure_stp(br->ofproto, port, &port_s,
1378 &port_num_counter, port_num_bitmap);
1380 /* As bonds are not supported, just apply configuration to
1381 * all interfaces. */
1382 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1383 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1385 VLOG_ERR("port %s: could not enable STP", port->name);
1391 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1392 && port_num_counter) {
1393 VLOG_ERR("bridge %s: must manually configure all STP port "
1394 "IDs or none, disabling", br->name);
1395 ofproto_set_stp(br->ofproto, NULL);
1397 bitmap_free(port_num_bitmap);
1402 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1404 const struct port *port = port_lookup(br, name);
1405 return port && port_is_bond_fake_iface(port);
1409 port_is_bond_fake_iface(const struct port *port)
1411 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1415 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1417 struct bridge *br, *next;
1418 struct shash new_br;
1421 /* Collect new bridges' names and types. */
1422 shash_init(&new_br);
1423 for (i = 0; i < cfg->n_bridges; i++) {
1424 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1425 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1427 if (strchr(br_cfg->name, '/')) {
1428 /* Prevent remote ovsdb-server users from accessing arbitrary
1429 * directories, e.g. consider a bridge named "../../../etc/". */
1430 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1432 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1433 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1437 /* Get rid of deleted bridges or those whose types have changed.
1438 * Update 'cfg' of bridges that still exist. */
1439 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1440 br->cfg = shash_find_data(&new_br, br->name);
1441 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1442 br->cfg->datapath_type))) {
1447 /* Add new bridges. */
1448 for (i = 0; i < cfg->n_bridges; i++) {
1449 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1450 struct bridge *br = bridge_lookup(br_cfg->name);
1452 bridge_create(br_cfg);
1456 shash_destroy(&new_br);
1459 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1460 * Returns 0 if successful, otherwise a positive errno value. */
1462 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1463 struct netdev *netdev)
1465 return netdev_set_config(netdev, &iface_cfg->options);
1468 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1469 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1471 * If successful, returns 0 and stores the network device in '*netdevp'. On
1472 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1474 iface_do_create(const struct bridge *br,
1475 const struct ovsrec_interface *iface_cfg,
1476 const struct ovsrec_port *port_cfg,
1477 ofp_port_t *ofp_portp, struct netdev **netdevp)
1479 struct netdev *netdev = NULL;
1482 if (netdev_is_reserved_name(iface_cfg->name)) {
1483 VLOG_WARN("could not create interface %s, name is reserved",
1489 error = netdev_open(iface_cfg->name,
1490 iface_get_type(iface_cfg, br->cfg), &netdev);
1492 VLOG_WARN("could not open network device %s (%s)",
1493 iface_cfg->name, ovs_strerror(error));
1497 error = iface_set_netdev_config(iface_cfg, netdev);
1502 *ofp_portp = iface_pick_ofport(iface_cfg);
1503 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1508 VLOG_INFO("bridge %s: added interface %s on port %d",
1509 br->name, iface_cfg->name, *ofp_portp);
1511 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1512 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1520 netdev_close(netdev);
1524 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1525 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1526 * automatically allocated for the iface. Takes ownership of and
1527 * deallocates 'if_cfg'.
1529 * Return true if an iface is successfully created, false otherwise. */
1531 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1532 const struct ovsrec_port *port_cfg)
1534 struct netdev *netdev;
1535 struct iface *iface;
1536 ofp_port_t ofp_port;
1540 /* Do the bits that can fail up front. */
1541 ovs_assert(!iface_lookup(br, iface_cfg->name));
1542 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1544 iface_clear_db_record(iface_cfg);
1548 /* Get or create the port structure. */
1549 port = port_lookup(br, port_cfg->name);
1551 port = port_create(br, port_cfg);
1554 /* Create the iface structure. */
1555 iface = xzalloc(sizeof *iface);
1556 list_push_back(&port->ifaces, &iface->port_elem);
1557 hmap_insert(&br->iface_by_name, &iface->name_node,
1558 hash_string(iface_cfg->name, 0));
1560 iface->name = xstrdup(iface_cfg->name);
1561 iface->ofp_port = ofp_port;
1562 iface->netdev = netdev;
1563 iface->type = iface_get_type(iface_cfg, br->cfg);
1564 iface->cfg = iface_cfg;
1565 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1566 hash_ofp_port(ofp_port));
1568 /* Populate initial status in database. */
1569 iface_refresh_stats(iface);
1570 iface_refresh_netdev_status(iface);
1572 /* Add bond fake iface if necessary. */
1573 if (port_is_bond_fake_iface(port)) {
1574 struct ofproto_port ofproto_port;
1576 if (ofproto_port_query_by_name(br->ofproto, port->name,
1578 struct netdev *netdev;
1581 error = netdev_open(port->name, "internal", &netdev);
1583 ofp_port_t fake_ofp_port = OFPP_NONE;
1584 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1585 netdev_close(netdev);
1587 VLOG_WARN("could not open network device %s (%s)",
1588 port->name, ovs_strerror(error));
1591 /* Already exists, nothing to do. */
1592 ofproto_port_destroy(&ofproto_port);
1599 /* Set forward BPDU option. */
1601 bridge_configure_forward_bpdu(struct bridge *br)
1603 ofproto_set_forward_bpdu(br->ofproto,
1604 smap_get_bool(&br->cfg->other_config,
1609 /* Set MAC learning table configuration for 'br'. */
1611 bridge_configure_mac_table(struct bridge *br)
1613 const char *idle_time_str;
1616 const char *mac_table_size_str;
1619 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1620 idle_time = (idle_time_str && atoi(idle_time_str)
1621 ? atoi(idle_time_str)
1622 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1624 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1625 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1626 ? atoi(mac_table_size_str)
1629 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1633 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1634 const struct port *fake_br, struct iface **hw_addr_iface)
1636 struct hmapx mirror_output_ports;
1638 bool found_addr = false;
1642 /* Mirror output ports don't participate in picking the local hardware
1643 * address. ofproto can't help us find out whether a given port is a
1644 * mirror output because we haven't configured mirrors yet, so we need to
1645 * accumulate them ourselves. */
1646 hmapx_init(&mirror_output_ports);
1647 for (i = 0; i < br->cfg->n_mirrors; i++) {
1648 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1649 if (m->output_port) {
1650 hmapx_add(&mirror_output_ports, m->output_port);
1654 /* Otherwise choose the minimum non-local MAC address among all of the
1656 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1657 uint8_t iface_ea[ETH_ADDR_LEN];
1658 struct iface *candidate;
1659 struct iface *iface;
1661 /* Mirror output ports don't participate. */
1662 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1666 /* Choose the MAC address to represent the port. */
1668 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1669 /* Find the interface with this Ethernet address (if any) so that
1670 * we can provide the correct devname to the caller. */
1671 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1672 uint8_t candidate_ea[ETH_ADDR_LEN];
1673 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1674 && eth_addr_equals(iface_ea, candidate_ea)) {
1679 /* Choose the interface whose MAC address will represent the port.
1680 * The Linux kernel bonding code always chooses the MAC address of
1681 * the first slave added to a bond, and the Fedora networking
1682 * scripts always add slaves to a bond in alphabetical order, so
1683 * for compatibility we choose the interface with the name that is
1684 * first in alphabetical order. */
1685 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1686 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1691 /* The local port doesn't count (since we're trying to choose its
1692 * MAC address anyway). */
1693 if (iface->ofp_port == OFPP_LOCAL) {
1697 /* For fake bridges we only choose from ports with the same tag */
1698 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1699 if (!port->cfg->tag) {
1702 if (*port->cfg->tag != *fake_br->cfg->tag) {
1708 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1714 /* Compare against our current choice. */
1715 if (!eth_addr_is_multicast(iface_ea) &&
1716 !eth_addr_is_local(iface_ea) &&
1717 !eth_addr_is_reserved(iface_ea) &&
1718 !eth_addr_is_zero(iface_ea) &&
1719 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1721 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1722 *hw_addr_iface = iface;
1728 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1729 *hw_addr_iface = NULL;
1732 hmapx_destroy(&mirror_output_ports);
1736 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1737 struct iface **hw_addr_iface)
1740 *hw_addr_iface = NULL;
1742 /* Did the user request a particular MAC? */
1743 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1744 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1745 if (eth_addr_is_multicast(ea)) {
1746 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1747 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1748 } else if (eth_addr_is_zero(ea)) {
1749 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1755 /* Find a local hw address */
1756 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
1759 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1760 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1761 * an interface on 'br', then that interface must be passed in as
1762 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1763 * 'hw_addr_iface' must be passed in as a null pointer. */
1765 bridge_pick_datapath_id(struct bridge *br,
1766 const uint8_t bridge_ea[ETH_ADDR_LEN],
1767 struct iface *hw_addr_iface)
1770 * The procedure for choosing a bridge MAC address will, in the most
1771 * ordinary case, also choose a unique MAC that we can use as a datapath
1772 * ID. In some special cases, though, multiple bridges will end up with
1773 * the same MAC address. This is OK for the bridges, but it will confuse
1774 * the OpenFlow controller, because each datapath needs a unique datapath
1777 * Datapath IDs must be unique. It is also very desirable that they be
1778 * stable from one run to the next, so that policy set on a datapath
1781 const char *datapath_id;
1784 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1785 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1789 if (!hw_addr_iface) {
1791 * A purely internal bridge, that is, one that has no non-virtual
1792 * network devices on it at all, is difficult because it has no
1793 * natural unique identifier at all.
1795 * When the host is a XenServer, we handle this case by hashing the
1796 * host's UUID with the name of the bridge. Names of bridges are
1797 * persistent across XenServer reboots, although they can be reused if
1798 * an internal network is destroyed and then a new one is later
1799 * created, so this is fairly effective.
1801 * When the host is not a XenServer, we punt by using a random MAC
1802 * address on each run.
1804 const char *host_uuid = xenserver_get_host_uuid();
1806 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1807 dpid = dpid_from_hash(combined, strlen(combined));
1813 return eth_addr_to_uint64(bridge_ea);
1817 dpid_from_hash(const void *data, size_t n)
1819 uint8_t hash[SHA1_DIGEST_SIZE];
1821 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1822 sha1_bytes(data, n, hash);
1823 eth_addr_mark_random(hash);
1824 return eth_addr_to_uint64(hash);
1828 iface_refresh_netdev_status(struct iface *iface)
1832 enum netdev_features current;
1833 enum netdev_flags flags;
1834 const char *link_state;
1835 uint8_t mac[ETH_ADDR_LEN];
1836 int64_t bps, mtu_64, ifindex64, link_resets;
1839 if (iface_is_synthetic(iface)) {
1843 if (iface->change_seq == netdev_get_change_seq(iface->netdev)) {
1847 iface->change_seq = netdev_get_change_seq(iface->netdev);
1851 if (!netdev_get_status(iface->netdev, &smap)) {
1852 ovsrec_interface_set_status(iface->cfg, &smap);
1854 ovsrec_interface_set_status(iface->cfg, NULL);
1857 smap_destroy(&smap);
1859 error = netdev_get_flags(iface->netdev, &flags);
1861 const char *state = flags & NETDEV_UP ? "up" : "down";
1863 ovsrec_interface_set_admin_state(iface->cfg, state);
1865 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1868 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1869 ovsrec_interface_set_link_state(iface->cfg, link_state);
1871 link_resets = netdev_get_carrier_resets(iface->netdev);
1872 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
1874 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1875 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1877 ovsrec_interface_set_duplex(iface->cfg,
1878 netdev_features_is_full_duplex(current)
1880 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1882 ovsrec_interface_set_duplex(iface->cfg, NULL);
1883 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1886 error = netdev_get_mtu(iface->netdev, &mtu);
1889 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1891 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1894 error = netdev_get_etheraddr(iface->netdev, mac);
1896 char mac_string[32];
1898 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1899 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1901 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1904 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1905 * if there is no valid ifindex number. */
1906 ifindex64 = netdev_get_ifindex(iface->netdev);
1907 if (ifindex64 < 0) {
1910 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1914 iface_refresh_ofproto_status(struct iface *iface)
1919 if (iface_is_synthetic(iface)) {
1923 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
1927 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1929 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1932 iface_refresh_cfm_stats(iface);
1935 error = ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
1936 iface->ofp_port, &smap);
1938 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
1940 smap_destroy(&smap);
1943 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1946 iface_refresh_cfm_stats(struct iface *iface)
1948 const struct ovsrec_interface *cfg = iface->cfg;
1949 struct ofproto_cfm_status status;
1952 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1953 iface->ofp_port, &status);
1955 /* Do nothing if there is no status change since last update. */
1956 } else if (error > 0) {
1957 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1958 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1959 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1960 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1961 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1962 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1964 const char *reasons[CFM_FAULT_N_REASONS];
1965 int64_t cfm_health = status.health;
1966 int64_t cfm_flap_count = status.flap_count;
1967 bool faulted = status.faults != 0;
1970 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1973 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1974 int reason = 1 << i;
1975 if (status.faults & reason) {
1976 reasons[j++] = cfm_fault_reason_to_str(reason);
1979 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1981 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1983 if (status.remote_opstate >= 0) {
1984 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1985 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1987 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1990 ovsrec_interface_set_cfm_remote_mpids(cfg,
1991 (const int64_t *)status.rmps,
1993 if (cfm_health >= 0) {
1994 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1996 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2004 iface_refresh_stats(struct iface *iface)
2006 #define IFACE_STATS \
2007 IFACE_STAT(rx_packets, "rx_packets") \
2008 IFACE_STAT(tx_packets, "tx_packets") \
2009 IFACE_STAT(rx_bytes, "rx_bytes") \
2010 IFACE_STAT(tx_bytes, "tx_bytes") \
2011 IFACE_STAT(rx_dropped, "rx_dropped") \
2012 IFACE_STAT(tx_dropped, "tx_dropped") \
2013 IFACE_STAT(rx_errors, "rx_errors") \
2014 IFACE_STAT(tx_errors, "tx_errors") \
2015 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2016 IFACE_STAT(rx_over_errors, "rx_over_err") \
2017 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2018 IFACE_STAT(collisions, "collisions")
2020 #define IFACE_STAT(MEMBER, NAME) + 1
2021 enum { N_IFACE_STATS = IFACE_STATS };
2023 int64_t values[N_IFACE_STATS];
2024 char *keys[N_IFACE_STATS];
2027 struct netdev_stats stats;
2029 if (iface_is_synthetic(iface)) {
2033 /* Intentionally ignore return value, since errors will set 'stats' to
2034 * all-1s, and we will deal with that correctly below. */
2035 netdev_get_stats(iface->netdev, &stats);
2037 /* Copy statistics into keys[] and values[]. */
2039 #define IFACE_STAT(MEMBER, NAME) \
2040 if (stats.MEMBER != UINT64_MAX) { \
2042 values[n] = stats.MEMBER; \
2047 ovs_assert(n <= N_IFACE_STATS);
2049 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2054 br_refresh_stp_status(struct bridge *br)
2056 struct smap smap = SMAP_INITIALIZER(&smap);
2057 struct ofproto *ofproto = br->ofproto;
2058 struct ofproto_stp_status status;
2060 if (ofproto_get_stp_status(ofproto, &status)) {
2064 if (!status.enabled) {
2065 ovsrec_bridge_set_status(br->cfg, NULL);
2069 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2070 STP_ID_ARGS(status.bridge_id));
2071 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2072 STP_ID_ARGS(status.designated_root));
2073 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2075 ovsrec_bridge_set_status(br->cfg, &smap);
2076 smap_destroy(&smap);
2080 port_refresh_stp_status(struct port *port)
2082 struct ofproto *ofproto = port->bridge->ofproto;
2083 struct iface *iface;
2084 struct ofproto_port_stp_status status;
2087 if (port_is_synthetic(port)) {
2091 /* STP doesn't currently support bonds. */
2092 if (!list_is_singleton(&port->ifaces)) {
2093 ovsrec_port_set_status(port->cfg, NULL);
2097 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2098 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2102 if (!status.enabled) {
2103 ovsrec_port_set_status(port->cfg, NULL);
2107 /* Set Status column. */
2109 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2110 smap_add(&smap, "stp_state", stp_state_name(status.state));
2111 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2112 smap_add(&smap, "stp_role", stp_role_name(status.role));
2113 ovsrec_port_set_status(port->cfg, &smap);
2114 smap_destroy(&smap);
2118 port_refresh_stp_stats(struct port *port)
2120 struct ofproto *ofproto = port->bridge->ofproto;
2121 struct iface *iface;
2122 struct ofproto_port_stp_stats stats;
2124 int64_t int_values[3];
2126 if (port_is_synthetic(port)) {
2130 /* STP doesn't currently support bonds. */
2131 if (!list_is_singleton(&port->ifaces)) {
2135 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2136 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2140 if (!stats.enabled) {
2141 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2145 /* Set Statistics column. */
2146 keys[0] = "stp_tx_count";
2147 int_values[0] = stats.tx_count;
2148 keys[1] = "stp_rx_count";
2149 int_values[1] = stats.rx_count;
2150 keys[2] = "stp_error_count";
2151 int_values[2] = stats.error_count;
2153 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2154 ARRAY_SIZE(int_values));
2158 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2160 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2164 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2166 bool enable = enable_system_stats(cfg);
2168 system_stats_enable(enable);
2170 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2175 run_system_stats(void)
2177 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2180 stats = system_stats_run();
2182 struct ovsdb_idl_txn *txn;
2183 struct ovsdb_datum datum;
2185 txn = ovsdb_idl_txn_create(idl);
2186 ovsdb_datum_from_smap(&datum, stats);
2187 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2189 ovsdb_idl_txn_commit(txn);
2190 ovsdb_idl_txn_destroy(txn);
2197 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2200 case OFPCR12_ROLE_EQUAL:
2202 case OFPCR12_ROLE_MASTER:
2204 case OFPCR12_ROLE_SLAVE:
2206 case OFPCR12_ROLE_NOCHANGE:
2208 return "*** INVALID ROLE ***";
2213 refresh_controller_status(void)
2217 const struct ovsrec_controller *cfg;
2221 /* Accumulate status for controllers on all bridges. */
2222 HMAP_FOR_EACH (br, node, &all_bridges) {
2223 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2226 /* Update each controller in the database with current status. */
2227 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2228 struct ofproto_controller_info *cinfo =
2229 shash_find_data(&info, cfg->target);
2232 struct smap smap = SMAP_INITIALIZER(&smap);
2233 const char **values = cinfo->pairs.values;
2234 const char **keys = cinfo->pairs.keys;
2237 for (i = 0; i < cinfo->pairs.n; i++) {
2238 smap_add(&smap, keys[i], values[i]);
2241 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2242 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2244 ovsrec_controller_set_status(cfg, &smap);
2245 smap_destroy(&smap);
2247 ovsrec_controller_set_is_connected(cfg, false);
2248 ovsrec_controller_set_role(cfg, NULL);
2249 ovsrec_controller_set_status(cfg, NULL);
2253 ofproto_free_ofproto_controller_info(&info);
2263 /* Let each datapath type do the work that it needs to do. */
2265 ofproto_enumerate_types(&types);
2266 SSET_FOR_EACH (type, &types) {
2267 ofproto_type_run(type);
2269 sset_destroy(&types);
2271 /* Let each bridge do the work that it needs to do. */
2272 HMAP_FOR_EACH (br, node, &all_bridges) {
2273 ofproto_run(br->ofproto);
2280 static struct ovsrec_open_vswitch null_cfg;
2281 const struct ovsrec_open_vswitch *cfg;
2283 bool vlan_splinters_changed;
2286 ovsrec_open_vswitch_init(&null_cfg);
2290 if (ovsdb_idl_is_lock_contended(idl)) {
2291 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2292 struct bridge *br, *next_br;
2294 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2295 "disabling this process (pid %ld) until it goes away",
2296 (long int) getpid());
2298 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2301 /* Since we will not be running system_stats_run() in this process
2302 * with the current situation of multiple ovs-vswitchd daemons,
2303 * disable system stats collection. */
2304 system_stats_enable(false);
2306 } else if (!ovsdb_idl_has_lock(idl)) {
2309 cfg = ovsrec_open_vswitch_first(idl);
2311 /* Initialize the ofproto library. This only needs to run once, but
2312 * it must be done after the configuration is set. If the
2313 * initialization has already occurred, bridge_init_ofproto()
2314 * returns immediately. */
2315 bridge_init_ofproto(cfg);
2317 /* Once the value of flow-restore-wait is false, we no longer should
2318 * check its value from the database. */
2319 if (cfg && ofproto_get_flow_restore_wait()) {
2320 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2321 "flow-restore-wait", false));
2326 /* Re-configure SSL. We do this on every trip through the main loop,
2327 * instead of just when the database changes, because the contents of the
2328 * key and certificate files can change without the database changing.
2330 * We do this before bridge_reconfigure() because that function might
2331 * initiate SSL connections and thus requires SSL to be configured. */
2332 if (cfg && cfg->ssl) {
2333 const struct ovsrec_ssl *ssl = cfg->ssl;
2335 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2336 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2339 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2340 * usage has changed. */
2341 vlan_splinters_changed = false;
2342 if (vlan_splinters_enabled_anywhere) {
2343 HMAP_FOR_EACH (br, node, &all_bridges) {
2344 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2345 vlan_splinters_changed = true;
2351 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2352 struct ovsdb_idl_txn *txn;
2354 idl_seqno = ovsdb_idl_get_seqno(idl);
2355 txn = ovsdb_idl_txn_create(idl);
2356 bridge_reconfigure(cfg ? cfg : &null_cfg);
2359 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2362 /* If we are completing our initial configuration for this run
2363 * of ovs-vswitchd, then keep the transaction around to monitor
2364 * it for completion. */
2365 if (initial_config_done) {
2366 ovsdb_idl_txn_commit(txn);
2367 ovsdb_idl_txn_destroy(txn);
2369 initial_config_done = true;
2370 daemonize_txn = txn;
2374 if (daemonize_txn) {
2375 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2376 if (status != TXN_INCOMPLETE) {
2377 ovsdb_idl_txn_destroy(daemonize_txn);
2378 daemonize_txn = NULL;
2380 /* ovs-vswitchd has completed initialization, so allow the
2381 * process that forked us to exit successfully. */
2382 daemonize_complete();
2384 vlog_enable_async();
2386 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2390 /* Refresh interface and mirror stats if necessary. */
2391 if (time_msec() >= iface_stats_timer) {
2393 struct ovsdb_idl_txn *txn;
2395 txn = ovsdb_idl_txn_create(idl);
2396 HMAP_FOR_EACH (br, node, &all_bridges) {
2400 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2401 struct iface *iface;
2403 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2404 iface_refresh_stats(iface);
2407 port_refresh_stp_stats(port);
2410 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2411 mirror_refresh_stats(m);
2415 refresh_controller_status();
2416 ovsdb_idl_txn_commit(txn);
2417 ovsdb_idl_txn_destroy(txn); /* XXX */
2420 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2426 /* Check the need to update status. */
2427 seq = seq_read(connectivity_seq_get());
2428 if (seq != connectivity_seqno) {
2429 connectivity_seqno = seq;
2430 status_txn = ovsdb_idl_txn_create(idl);
2431 HMAP_FOR_EACH (br, node, &all_bridges) {
2434 br_refresh_stp_status(br);
2435 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2436 struct iface *iface;
2438 port_refresh_stp_status(port);
2439 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2440 iface_refresh_netdev_status(iface);
2441 iface_refresh_ofproto_status(iface);
2449 enum ovsdb_idl_txn_status status;
2451 status = ovsdb_idl_txn_commit(status_txn);
2452 /* Do not destroy "status_txn" if the transaction is
2453 * "TXN_INCOMPLETE". */
2454 if (status != TXN_INCOMPLETE) {
2455 ovsdb_idl_txn_destroy(status_txn);
2469 ovsdb_idl_wait(idl);
2470 if (daemonize_txn) {
2471 ovsdb_idl_txn_wait(daemonize_txn);
2475 ofproto_enumerate_types(&types);
2476 SSET_FOR_EACH (type, &types) {
2477 ofproto_type_wait(type);
2479 sset_destroy(&types);
2481 if (!hmap_is_empty(&all_bridges)) {
2484 HMAP_FOR_EACH (br, node, &all_bridges) {
2485 ofproto_wait(br->ofproto);
2487 poll_timer_wait_until(iface_stats_timer);
2490 /* If the status database transaction is 'TXN_INCOMPLETE' in this run,
2491 * register a timeout in 'STATUS_CHECK_AGAIN_MSEC'. Else, wait on the
2492 * global connectivity sequence number. Note, this also helps batch
2493 * multiple status changes into one transaction. */
2495 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2497 seq_wait(connectivity_seq_get(), connectivity_seqno);
2500 system_stats_wait();
2503 /* Adds some memory usage statistics for bridges into 'usage', for use with
2504 * memory_report(). */
2506 bridge_get_memory_usage(struct simap *usage)
2513 ofproto_enumerate_types(&types);
2514 SSET_FOR_EACH (type, &types) {
2515 ofproto_type_get_memory_usage(type, usage);
2517 sset_destroy(&types);
2519 HMAP_FOR_EACH (br, node, &all_bridges) {
2520 ofproto_get_memory_usage(br->ofproto, usage);
2524 /* QoS unixctl user interface functions. */
2526 struct qos_unixctl_show_cbdata {
2528 struct iface *iface;
2532 qos_unixctl_show_queue(unsigned int queue_id,
2533 const struct smap *details,
2534 struct iface *iface,
2537 struct netdev_queue_stats stats;
2538 struct smap_node *node;
2541 ds_put_cstr(ds, "\n");
2543 ds_put_format(ds, "Queue %u:\n", queue_id);
2545 ds_put_cstr(ds, "Default:\n");
2548 SMAP_FOR_EACH (node, details) {
2549 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2552 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2554 if (stats.tx_packets != UINT64_MAX) {
2555 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2558 if (stats.tx_bytes != UINT64_MAX) {
2559 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2562 if (stats.tx_errors != UINT64_MAX) {
2563 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2566 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2567 queue_id, ovs_strerror(error));
2572 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2573 const char *argv[], void *aux OVS_UNUSED)
2575 struct ds ds = DS_EMPTY_INITIALIZER;
2576 struct smap smap = SMAP_INITIALIZER(&smap);
2577 struct iface *iface;
2579 struct smap_node *node;
2581 iface = iface_find(argv[1]);
2583 unixctl_command_reply_error(conn, "no such interface");
2587 netdev_get_qos(iface->netdev, &type, &smap);
2589 if (*type != '\0') {
2590 struct netdev_queue_dump dump;
2591 struct smap details;
2592 unsigned int queue_id;
2594 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2596 SMAP_FOR_EACH (node, &smap) {
2597 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2600 smap_init(&details);
2601 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2602 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2604 smap_destroy(&details);
2606 unixctl_command_reply(conn, ds_cstr(&ds));
2608 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2609 unixctl_command_reply_error(conn, ds_cstr(&ds));
2612 smap_destroy(&smap);
2616 /* Bridge reconfiguration functions. */
2618 bridge_create(const struct ovsrec_bridge *br_cfg)
2622 ovs_assert(!bridge_lookup(br_cfg->name));
2623 br = xzalloc(sizeof *br);
2625 br->name = xstrdup(br_cfg->name);
2626 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2629 /* Derive the default Ethernet address from the bridge's UUID. This should
2630 * be unique and it will be stable between ovs-vswitchd runs. */
2631 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2632 eth_addr_mark_random(br->default_ea);
2634 hmap_init(&br->ports);
2635 hmap_init(&br->ifaces);
2636 hmap_init(&br->iface_by_name);
2637 hmap_init(&br->mirrors);
2639 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2643 bridge_destroy(struct bridge *br)
2646 struct mirror *mirror, *next_mirror;
2647 struct port *port, *next_port;
2649 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2652 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2653 mirror_destroy(mirror);
2656 hmap_remove(&all_bridges, &br->node);
2657 ofproto_destroy(br->ofproto);
2658 hmap_destroy(&br->ifaces);
2659 hmap_destroy(&br->ports);
2660 hmap_destroy(&br->iface_by_name);
2661 hmap_destroy(&br->mirrors);
2668 static struct bridge *
2669 bridge_lookup(const char *name)
2673 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2674 if (!strcmp(br->name, name)) {
2681 /* Handle requests for a listing of all flows known by the OpenFlow
2682 * stack, including those normally hidden. */
2684 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2685 const char *argv[], void *aux OVS_UNUSED)
2690 br = bridge_lookup(argv[1]);
2692 unixctl_command_reply_error(conn, "Unknown bridge");
2697 ofproto_get_all_flows(br->ofproto, &results);
2699 unixctl_command_reply(conn, ds_cstr(&results));
2700 ds_destroy(&results);
2703 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2704 * connections and reconnect. If BRIDGE is not specified, then all bridges
2705 * drop their controller connections and reconnect. */
2707 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2708 const char *argv[], void *aux OVS_UNUSED)
2712 br = bridge_lookup(argv[1]);
2714 unixctl_command_reply_error(conn, "Unknown bridge");
2717 ofproto_reconnect_controllers(br->ofproto);
2719 HMAP_FOR_EACH (br, node, &all_bridges) {
2720 ofproto_reconnect_controllers(br->ofproto);
2723 unixctl_command_reply(conn, NULL);
2727 bridge_get_controllers(const struct bridge *br,
2728 struct ovsrec_controller ***controllersp)
2730 struct ovsrec_controller **controllers;
2731 size_t n_controllers;
2733 controllers = br->cfg->controller;
2734 n_controllers = br->cfg->n_controller;
2736 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2742 *controllersp = controllers;
2744 return n_controllers;
2748 bridge_collect_wanted_ports(struct bridge *br,
2749 const unsigned long int *splinter_vlans,
2750 struct shash *wanted_ports)
2754 shash_init(wanted_ports);
2756 for (i = 0; i < br->cfg->n_ports; i++) {
2757 const char *name = br->cfg->ports[i]->name;
2758 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2759 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2763 if (bridge_get_controllers(br, NULL)
2764 && !shash_find(wanted_ports, br->name)) {
2765 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2766 br->name, br->name);
2768 ovsrec_interface_init(&br->synth_local_iface);
2769 ovsrec_port_init(&br->synth_local_port);
2771 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2772 br->synth_local_port.n_interfaces = 1;
2773 br->synth_local_port.name = br->name;
2775 br->synth_local_iface.name = br->name;
2776 br->synth_local_iface.type = "internal";
2778 br->synth_local_ifacep = &br->synth_local_iface;
2780 shash_add(wanted_ports, br->name, &br->synth_local_port);
2783 if (splinter_vlans) {
2784 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2788 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2789 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2790 * 'br' needs to complete its configuration. */
2792 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2794 struct shash_node *port_node;
2795 struct port *port, *next;
2797 /* Get rid of deleted ports.
2798 * Get rid of deleted interfaces on ports that still exist. */
2799 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2800 port->cfg = shash_find_data(wanted_ports, port->name);
2804 port_del_ifaces(port);
2808 /* Update iface->cfg and iface->type in interfaces that still exist. */
2809 SHASH_FOR_EACH (port_node, wanted_ports) {
2810 const struct ovsrec_port *port = port_node->data;
2813 for (i = 0; i < port->n_interfaces; i++) {
2814 const struct ovsrec_interface *cfg = port->interfaces[i];
2815 struct iface *iface = iface_lookup(br, cfg->name);
2816 const char *type = iface_get_type(cfg, br->cfg);
2821 } else if (!strcmp(type, "null")) {
2822 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2823 " may be removed in February 2013. Please email"
2824 " dev@openvswitch.org with concerns.",
2827 /* We will add new interfaces later. */
2833 /* Initializes 'oc' appropriately as a management service controller for
2836 * The caller must free oc->target when it is no longer needed. */
2838 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2839 struct ofproto_controller *oc)
2841 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2842 oc->max_backoff = 0;
2843 oc->probe_interval = 60;
2844 oc->band = OFPROTO_OUT_OF_BAND;
2846 oc->burst_limit = 0;
2847 oc->enable_async_msgs = true;
2850 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2852 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2853 struct ofproto_controller *oc)
2857 oc->target = c->target;
2858 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2859 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2860 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2861 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2862 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2863 oc->burst_limit = (c->controller_burst_limit
2864 ? *c->controller_burst_limit : 0);
2865 oc->enable_async_msgs = (!c->enable_async_messages
2866 || *c->enable_async_messages);
2867 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2868 if (dscp < 0 || dscp > 63) {
2869 dscp = DSCP_DEFAULT;
2874 /* Configures the IP stack for 'br''s local interface properly according to the
2875 * configuration in 'c'. */
2877 bridge_configure_local_iface_netdev(struct bridge *br,
2878 struct ovsrec_controller *c)
2880 struct netdev *netdev;
2881 struct in_addr mask, gateway;
2883 struct iface *local_iface;
2886 /* If there's no local interface or no IP address, give up. */
2887 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2888 if (!local_iface || !c->local_ip
2889 || !inet_pton(AF_INET, c->local_ip, &ip)) {
2893 /* Bring up the local interface. */
2894 netdev = local_iface->netdev;
2895 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2897 /* Configure the IP address and netmask. */
2898 if (!c->local_netmask
2899 || !inet_pton(AF_INET, c->local_netmask, &mask)
2901 mask.s_addr = guess_netmask(ip.s_addr);
2903 if (!netdev_set_in4(netdev, ip, mask)) {
2904 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2905 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2908 /* Configure the default gateway. */
2909 if (c->local_gateway
2910 && inet_pton(AF_INET, c->local_gateway, &gateway)
2911 && gateway.s_addr) {
2912 if (!netdev_add_router(netdev, gateway)) {
2913 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2914 br->name, IP_ARGS(gateway.s_addr));
2919 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2920 * in either string are treated as equal to any number of slashes in the other,
2921 * e.g. "x///y" is equal to "x/y".
2923 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2924 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2925 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2926 * 'b' against a prefix of 'a'.
2929 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2931 const char *b_start = b;
2933 if (b - b_start >= b_stoplen) {
2935 } else if (*a != *b) {
2937 } else if (*a == '/') {
2938 a += strspn(a, "/");
2939 b += strspn(b, "/");
2940 } else if (*a == '\0') {
2950 bridge_configure_remotes(struct bridge *br,
2951 const struct sockaddr_in *managers, size_t n_managers)
2953 bool disable_in_band;
2955 struct ovsrec_controller **controllers;
2956 size_t n_controllers;
2958 enum ofproto_fail_mode fail_mode;
2960 struct ofproto_controller *ocs;
2964 /* Check if we should disable in-band control on this bridge. */
2965 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2968 /* Set OpenFlow queue ID for in-band control. */
2969 ofproto_set_in_band_queue(br->ofproto,
2970 smap_get_int(&br->cfg->other_config,
2971 "in-band-queue", -1));
2973 if (disable_in_band) {
2974 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2976 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2979 n_controllers = bridge_get_controllers(br, &controllers);
2981 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2984 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2985 for (i = 0; i < n_controllers; i++) {
2986 struct ovsrec_controller *c = controllers[i];
2988 if (!strncmp(c->target, "punix:", 6)
2989 || !strncmp(c->target, "unix:", 5)) {
2990 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2993 if (!strncmp(c->target, "unix:", 5)) {
2994 /* Connect to a listening socket */
2995 whitelist = xasprintf("unix:%s/", ovs_rundir());
2996 if (strchr(c->target, '/') &&
2997 !equal_pathnames(c->target, whitelist,
2998 strlen(whitelist))) {
2999 /* Absolute path specified, but not in ovs_rundir */
3000 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3001 "controller \"%s\" due to possibility for "
3002 "remote exploit. Instead, specify socket "
3003 "in whitelisted \"%s\" or connect to "
3004 "\"unix:%s/%s.mgmt\" (which is always "
3005 "available without special configuration).",
3006 br->name, c->target, whitelist,
3007 ovs_rundir(), br->name);
3012 whitelist = xasprintf("punix:%s/%s.controller",
3013 ovs_rundir(), br->name);
3014 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3015 /* Prevent remote ovsdb-server users from accessing
3016 * arbitrary Unix domain sockets and overwriting arbitrary
3018 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3019 "controller \"%s\" due to possibility of "
3020 "overwriting local files. Instead, specify "
3021 "whitelisted \"%s\" or connect to "
3022 "\"unix:%s/%s.mgmt\" (which is always "
3023 "available without special configuration).",
3024 br->name, c->target, whitelist,
3025 ovs_rundir(), br->name);
3034 bridge_configure_local_iface_netdev(br, c);
3035 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3036 if (disable_in_band) {
3037 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3042 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3043 bridge_get_allowed_versions(br));
3044 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3047 /* Set the fail-mode. */
3048 fail_mode = !br->cfg->fail_mode
3049 || !strcmp(br->cfg->fail_mode, "standalone")
3050 ? OFPROTO_FAIL_STANDALONE
3051 : OFPROTO_FAIL_SECURE;
3052 ofproto_set_fail_mode(br->ofproto, fail_mode);
3054 /* Configure OpenFlow controller connection snooping. */
3055 if (!ofproto_has_snoops(br->ofproto)) {
3059 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3060 ovs_rundir(), br->name));
3061 ofproto_set_snoops(br->ofproto, &snoops);
3062 sset_destroy(&snoops);
3067 bridge_configure_tables(struct bridge *br)
3069 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3073 n_tables = ofproto_get_n_tables(br->ofproto);
3075 for (i = 0; i < n_tables; i++) {
3076 struct ofproto_table_settings s;
3079 s.max_flows = UINT_MAX;
3082 s.n_prefix_fields = 0;
3083 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3085 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3086 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3089 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3090 s.max_flows = *cfg->flow_limit;
3092 if (cfg->overflow_policy
3093 && !strcmp(cfg->overflow_policy, "evict")) {
3095 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3096 for (k = 0; k < cfg->n_groups; k++) {
3097 const char *string = cfg->groups[k];
3100 msg = mf_parse_subfield__(&s.groups[k], &string);
3102 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3103 "'groups' (%s)", br->name, i, msg);
3105 } else if (*string) {
3106 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3107 "element '%s' contains trailing garbage",
3108 br->name, i, cfg->groups[k]);
3114 /* Prefix lookup fields. */
3115 s.n_prefix_fields = 0;
3116 for (k = 0; k < cfg->n_prefixes; k++) {
3117 const char *name = cfg->prefixes[k];
3118 const struct mf_field *mf = mf_from_name(name);
3120 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3124 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3125 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3126 "%s", br->name, name);
3129 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3130 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3131 "field not used: %s", br->name, name);
3134 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3136 if (s.n_prefix_fields > 0) {
3138 struct ds ds = DS_EMPTY_INITIALIZER;
3139 for (k = 0; k < s.n_prefix_fields; k++) {
3141 ds_put_char(&ds, ',');
3143 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3145 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3146 br->name, i, ds_cstr(&ds));
3151 ofproto_configure_table(br->ofproto, i, &s);
3155 for (; j < br->cfg->n_flow_tables; j++) {
3156 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3157 "%"PRId64" not supported by this datapath", br->name,
3158 br->cfg->key_flow_tables[j]);
3163 bridge_configure_dp_desc(struct bridge *br)
3165 ofproto_set_dp_desc(br->ofproto,
3166 smap_get(&br->cfg->other_config, "dp-desc"));
3169 /* Port functions. */
3171 static struct port *
3172 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3176 port = xzalloc(sizeof *port);
3178 port->name = xstrdup(cfg->name);
3180 list_init(&port->ifaces);
3182 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3186 /* Deletes interfaces from 'port' that are no longer configured for it. */
3188 port_del_ifaces(struct port *port)
3190 struct iface *iface, *next;
3191 struct sset new_ifaces;
3194 /* Collect list of new interfaces. */
3195 sset_init(&new_ifaces);
3196 for (i = 0; i < port->cfg->n_interfaces; i++) {
3197 const char *name = port->cfg->interfaces[i]->name;
3198 const char *type = port->cfg->interfaces[i]->type;
3199 if (strcmp(type, "null")) {
3200 sset_add(&new_ifaces, name);
3204 /* Get rid of deleted interfaces. */
3205 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3206 if (!sset_contains(&new_ifaces, iface->name)) {
3207 iface_destroy(iface);
3211 sset_destroy(&new_ifaces);
3215 port_destroy(struct port *port)
3218 struct bridge *br = port->bridge;
3219 struct iface *iface, *next;
3222 ofproto_bundle_unregister(br->ofproto, port);
3225 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3226 iface_destroy__(iface);
3229 hmap_remove(&br->ports, &port->hmap_node);
3235 static struct port *
3236 port_lookup(const struct bridge *br, const char *name)
3240 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3242 if (!strcmp(port->name, name)) {
3250 enable_lacp(struct port *port, bool *activep)
3252 if (!port->cfg->lacp) {
3253 /* XXX when LACP implementation has been sufficiently tested, enable by
3254 * default and make active on bonded ports. */
3256 } else if (!strcmp(port->cfg->lacp, "off")) {
3258 } else if (!strcmp(port->cfg->lacp, "active")) {
3261 } else if (!strcmp(port->cfg->lacp, "passive")) {
3265 VLOG_WARN("port %s: unknown LACP mode %s",
3266 port->name, port->cfg->lacp);
3271 static struct lacp_settings *
3272 port_configure_lacp(struct port *port, struct lacp_settings *s)
3274 const char *lacp_time, *system_id;
3277 if (!enable_lacp(port, &s->active)) {
3281 s->name = port->name;
3283 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3285 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3286 ETH_ADDR_SCAN_ARGS(s->id))) {
3287 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3288 " address.", port->name, system_id);
3292 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3295 if (eth_addr_is_zero(s->id)) {
3296 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3300 /* Prefer bondable links if unspecified. */
3301 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3303 s->priority = (priority > 0 && priority <= UINT16_MAX
3305 : UINT16_MAX - !list_is_short(&port->ifaces));
3307 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3308 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3310 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3311 "lacp-fallback-ab", false);
3317 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3319 int priority, portid, key;
3321 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3322 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3324 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3326 if (portid <= 0 || portid > UINT16_MAX) {
3327 portid = ofp_to_u16(iface->ofp_port);
3330 if (priority <= 0 || priority > UINT16_MAX) {
3331 priority = UINT16_MAX;
3334 if (key < 0 || key > UINT16_MAX) {
3338 s->name = iface->name;
3340 s->priority = priority;
3345 port_configure_bond(struct port *port, struct bond_settings *s)
3347 const char *detect_s;
3348 struct iface *iface;
3349 int miimon_interval;
3351 s->name = port->name;
3353 if (port->cfg->bond_mode) {
3354 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3355 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3356 port->name, port->cfg->bond_mode,
3357 bond_mode_to_string(s->balance));
3360 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3362 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3363 * active-backup. At some point we should remove this warning. */
3364 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3365 " in previous versions, the default bond_mode was"
3366 " balance-slb", port->name,
3367 bond_mode_to_string(s->balance));
3369 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3370 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3371 "please use another bond type or disable flood_vlans",
3375 miimon_interval = smap_get_int(&port->cfg->other_config,
3376 "bond-miimon-interval", 0);
3377 if (miimon_interval <= 0) {
3378 miimon_interval = 200;
3381 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3382 if (!detect_s || !strcmp(detect_s, "carrier")) {
3383 miimon_interval = 0;
3384 } else if (strcmp(detect_s, "miimon")) {
3385 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3386 "defaulting to carrier", port->name, detect_s);
3387 miimon_interval = 0;
3390 s->up_delay = MAX(0, port->cfg->bond_updelay);
3391 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3392 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3393 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3394 "bond-rebalance-interval", 10000);
3395 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3396 s->rebalance_interval = 1000;
3399 s->fake_iface = port->cfg->bond_fake_iface;
3401 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3402 "lacp-fallback-ab", false);
3404 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3405 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3409 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3410 * instead of obtaining it from the database. */
3412 port_is_synthetic(const struct port *port)
3414 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3417 /* Interface functions. */
3420 iface_is_internal(const struct ovsrec_interface *iface,
3421 const struct ovsrec_bridge *br)
3423 /* The local port and "internal" ports are always "internal". */
3424 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3427 /* Returns the correct network device type for interface 'iface' in bridge
3430 iface_get_type(const struct ovsrec_interface *iface,
3431 const struct ovsrec_bridge *br)
3435 /* The local port always has type "internal". Other ports take
3436 * their type from the database and default to "system" if none is
3438 if (iface_is_internal(iface, br)) {
3441 type = iface->type[0] ? iface->type : "system";
3444 return ofproto_port_open_type(br->datapath_type, type);
3448 iface_destroy__(struct iface *iface)
3451 struct port *port = iface->port;
3452 struct bridge *br = port->bridge;
3454 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3455 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3458 if (iface->ofp_port != OFPP_NONE) {
3459 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3462 list_remove(&iface->port_elem);
3463 hmap_remove(&br->iface_by_name, &iface->name_node);
3465 netdev_close(iface->netdev);
3473 iface_destroy(struct iface *iface)
3476 struct port *port = iface->port;
3478 iface_destroy__(iface);
3479 if (list_is_empty(&port->ifaces)) {
3485 static struct iface *
3486 iface_lookup(const struct bridge *br, const char *name)
3488 struct iface *iface;
3490 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3491 &br->iface_by_name) {
3492 if (!strcmp(iface->name, name)) {
3500 static struct iface *
3501 iface_find(const char *name)
3503 const struct bridge *br;
3505 HMAP_FOR_EACH (br, node, &all_bridges) {
3506 struct iface *iface = iface_lookup(br, name);
3515 static struct iface *
3516 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3518 struct iface *iface;
3520 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3522 if (iface->ofp_port == ofp_port) {
3529 /* Set Ethernet address of 'iface', if one is specified in the configuration
3532 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3534 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3535 struct iface *hw_addr_iface;
3537 if (strcmp(iface->type, "internal")) {
3541 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3543 } else if (port->cfg->fake_bridge) {
3544 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3545 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3550 if (iface->ofp_port == OFPP_LOCAL) {
3551 VLOG_ERR("interface %s: ignoring mac in Interface record "
3552 "(use Bridge record to set local port's mac)",
3554 } else if (eth_addr_is_multicast(mac)) {
3555 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3558 int error = netdev_set_etheraddr(iface->netdev, mac);
3560 VLOG_ERR("interface %s: setting MAC failed (%s)",
3561 iface->name, ovs_strerror(error));
3567 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3569 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3571 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3572 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3573 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3577 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3578 * sets the "ofport" field to -1.
3580 * This is appropriate when 'if_cfg''s interface cannot be created or is
3581 * otherwise invalid. */
3583 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3585 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3586 iface_set_ofport(if_cfg, OFPP_NONE);
3587 ovsrec_interface_set_status(if_cfg, NULL);
3588 ovsrec_interface_set_admin_state(if_cfg, NULL);
3589 ovsrec_interface_set_duplex(if_cfg, NULL);
3590 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3591 ovsrec_interface_set_link_state(if_cfg, NULL);
3592 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3593 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3594 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3595 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3596 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3597 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3598 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3599 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3604 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3606 union ovsdb_atom atom;
3608 atom.integer = target;
3609 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3613 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3615 struct ofpbuf queues_buf;
3617 ofpbuf_init(&queues_buf, 0);
3619 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3620 netdev_set_qos(iface->netdev, NULL, NULL);
3622 const struct ovsdb_datum *queues;
3623 struct netdev_queue_dump dump;
3624 unsigned int queue_id;
3625 struct smap details;
3629 /* Configure top-level Qos for 'iface'. */
3630 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3632 /* Deconfigure queues that were deleted. */
3633 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3635 smap_init(&details);
3636 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3637 if (!queue_ids_include(queues, queue_id)) {
3638 netdev_delete_queue(iface->netdev, queue_id);
3641 smap_destroy(&details);
3643 /* Configure queues for 'iface'. */
3645 for (i = 0; i < qos->n_queues; i++) {
3646 const struct ovsrec_queue *queue = qos->value_queues[i];
3647 unsigned int queue_id = qos->key_queues[i];
3649 if (queue_id == 0) {
3653 if (queue->n_dscp == 1) {
3654 struct ofproto_port_queue *port_queue;
3656 port_queue = ofpbuf_put_uninit(&queues_buf,
3657 sizeof *port_queue);
3658 port_queue->queue = queue_id;
3659 port_queue->dscp = queue->dscp[0];
3662 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3665 struct smap details;
3667 smap_init(&details);
3668 netdev_set_queue(iface->netdev, 0, &details);
3669 smap_destroy(&details);
3673 if (iface->ofp_port != OFPP_NONE) {
3674 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
3675 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
3677 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3678 port_queues, n_queues);
3681 netdev_set_policing(iface->netdev,
3682 iface->cfg->ingress_policing_rate,
3683 iface->cfg->ingress_policing_burst);
3685 ofpbuf_uninit(&queues_buf);
3689 iface_configure_cfm(struct iface *iface)
3691 const struct ovsrec_interface *cfg = iface->cfg;
3692 const char *opstate_str;
3693 const char *cfm_ccm_vlan;
3694 struct cfm_settings s;
3695 struct smap netdev_args;
3697 if (!cfg->n_cfm_mpid) {
3698 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3702 s.check_tnl_key = false;
3703 smap_init(&netdev_args);
3704 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3705 const char *key = smap_get(&netdev_args, "key");
3706 const char *in_key = smap_get(&netdev_args, "in_key");
3708 s.check_tnl_key = (key && !strcmp(key, "flow"))
3709 || (in_key && !strcmp(in_key, "flow"));
3711 smap_destroy(&netdev_args);
3713 s.mpid = *cfg->cfm_mpid;
3714 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3715 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3716 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3718 if (s.interval <= 0) {
3722 if (!cfm_ccm_vlan) {
3724 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3725 s.ccm_vlan = CFM_RANDOM_VLAN;
3727 s.ccm_vlan = atoi(cfm_ccm_vlan);
3728 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3733 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3735 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3737 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3738 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3740 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3743 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3744 * instead of obtaining it from the database. */
3746 iface_is_synthetic(const struct iface *iface)
3748 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3752 iface_validate_ofport__(size_t n, int64_t *ofport)
3754 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3755 ? u16_to_ofp(*ofport)
3760 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3762 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3766 iface_pick_ofport(const struct ovsrec_interface *cfg)
3768 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3769 return (requested_ofport != OFPP_NONE
3771 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3774 /* Port mirroring. */
3776 static struct mirror *
3777 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3781 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3782 if (uuid_equals(uuid, &m->uuid)) {
3790 bridge_configure_mirrors(struct bridge *br)
3792 const struct ovsdb_datum *mc;
3793 unsigned long *flood_vlans;
3794 struct mirror *m, *next;
3797 /* Get rid of deleted mirrors. */
3798 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3799 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3800 union ovsdb_atom atom;
3802 atom.uuid = m->uuid;
3803 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3808 /* Add new mirrors and reconfigure existing ones. */
3809 for (i = 0; i < br->cfg->n_mirrors; i++) {
3810 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3811 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3813 m = mirror_create(br, cfg);
3816 if (!mirror_configure(m)) {
3821 /* Update flooded vlans (for RSPAN). */
3822 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3823 br->cfg->n_flood_vlans);
3824 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3825 bitmap_free(flood_vlans);
3828 static struct mirror *
3829 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3833 m = xzalloc(sizeof *m);
3834 m->uuid = cfg->header_.uuid;
3835 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3837 m->name = xstrdup(cfg->name);
3843 mirror_destroy(struct mirror *m)
3846 struct bridge *br = m->bridge;
3849 ofproto_mirror_unregister(br->ofproto, m);
3852 hmap_remove(&br->mirrors, &m->hmap_node);
3859 mirror_collect_ports(struct mirror *m,
3860 struct ovsrec_port **in_ports, int n_in_ports,
3861 void ***out_portsp, size_t *n_out_portsp)
3863 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3864 size_t n_out_ports = 0;
3867 for (i = 0; i < n_in_ports; i++) {
3868 const char *name = in_ports[i]->name;
3869 struct port *port = port_lookup(m->bridge, name);
3871 out_ports[n_out_ports++] = port;
3873 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3874 "port %s", m->bridge->name, m->name, name);
3877 *out_portsp = out_ports;
3878 *n_out_portsp = n_out_ports;
3882 mirror_configure(struct mirror *m)
3884 const struct ovsrec_mirror *cfg = m->cfg;
3885 struct ofproto_mirror_settings s;
3888 if (strcmp(cfg->name, m->name)) {
3890 m->name = xstrdup(cfg->name);
3894 /* Get output port or VLAN. */
3895 if (cfg->output_port) {
3896 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3897 if (!s.out_bundle) {
3898 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3899 m->bridge->name, m->name);
3902 s.out_vlan = UINT16_MAX;
3904 if (cfg->output_vlan) {
3905 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3906 "output vlan; ignoring output vlan",
3907 m->bridge->name, m->name);
3909 } else if (cfg->output_vlan) {
3910 /* The database should prevent invalid VLAN values. */
3911 s.out_bundle = NULL;
3912 s.out_vlan = *cfg->output_vlan;
3914 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3915 m->bridge->name, m->name);
3919 /* Get port selection. */
3920 if (cfg->select_all) {
3921 size_t n_ports = hmap_count(&m->bridge->ports);
3922 void **ports = xmalloc(n_ports * sizeof *ports);
3927 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3937 /* Get ports, dropping ports that don't exist.
3938 * The IDL ensures that there are no duplicates. */
3939 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3940 &s.srcs, &s.n_srcs);
3941 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3942 &s.dsts, &s.n_dsts);
3945 /* Get VLAN selection. */
3946 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3949 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3952 if (s.srcs != s.dsts) {
3961 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3963 * This is deprecated. It is only for compatibility with broken device drivers
3964 * in old versions of Linux that do not properly support VLANs when VLAN
3965 * devices are not used. When broken device drivers are no longer in
3966 * widespread use, we will delete these interfaces. */
3968 static struct ovsrec_port **recs;
3969 static size_t n_recs, allocated_recs;
3971 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3972 * splinters are reconfigured. */
3974 register_rec(struct ovsrec_port *rec)
3976 if (n_recs >= allocated_recs) {
3977 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3979 recs[n_recs++] = rec;
3982 /* Frees all of the ports registered with register_reg(). */
3984 free_registered_recs(void)
3988 for (i = 0; i < n_recs; i++) {
3989 struct ovsrec_port *port = recs[i];
3992 for (j = 0; j < port->n_interfaces; j++) {
3993 struct ovsrec_interface *iface = port->interfaces[j];
3998 smap_destroy(&port->other_config);
3999 free(port->interfaces);
4007 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4010 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4012 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4016 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4019 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4020 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4021 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4024 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4025 * use, returns NULL.
4027 * Updates 'vlan_splinters_enabled_anywhere'. */
4028 static unsigned long int *
4029 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4031 unsigned long int *splinter_vlans;
4032 struct sset splinter_ifaces;
4033 const char *real_dev_name;
4034 struct shash *real_devs;
4035 struct shash_node *node;
4039 /* Free space allocated for synthesized ports and interfaces, since we're
4040 * in the process of reconstructing all of them. */
4041 free_registered_recs();
4043 splinter_vlans = bitmap_allocate(4096);
4044 sset_init(&splinter_ifaces);
4045 vlan_splinters_enabled_anywhere = false;
4046 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4047 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4050 for (j = 0; j < br_cfg->n_ports; j++) {
4051 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4054 for (k = 0; k < port_cfg->n_interfaces; k++) {
4055 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4057 if (vlan_splinters_is_enabled(iface_cfg)) {
4058 vlan_splinters_enabled_anywhere = true;
4059 sset_add(&splinter_ifaces, iface_cfg->name);
4060 vlan_bitmap_from_array__(port_cfg->trunks,
4066 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4067 bitmap_set1(splinter_vlans, *port_cfg->tag);
4072 if (!vlan_splinters_enabled_anywhere) {
4073 free(splinter_vlans);
4074 sset_destroy(&splinter_ifaces);
4078 HMAP_FOR_EACH (br, node, &all_bridges) {
4080 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4084 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4085 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4086 * device to be created for it. */
4087 bitmap_set0(splinter_vlans, 0);
4088 bitmap_set0(splinter_vlans, 4095);
4090 /* Delete all VLAN devices that we don't need. */
4092 real_devs = vlandev_get_real_devs();
4093 SHASH_FOR_EACH (node, real_devs) {
4094 const struct vlan_real_dev *real_dev = node->data;
4095 const struct vlan_dev *vlan_dev;
4096 bool real_dev_has_splinters;
4098 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4100 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4101 if (!real_dev_has_splinters
4102 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4103 struct netdev *netdev;
4105 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4106 if (!netdev_get_in4(netdev, NULL, NULL) ||
4107 !netdev_get_in6(netdev, NULL)) {
4108 /* It has an IP address configured, so we don't own
4109 * it. Don't delete it. */
4111 vlandev_del(vlan_dev->name);
4113 netdev_close(netdev);
4120 /* Add all VLAN devices that we need. */
4121 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4124 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4125 if (!vlandev_get_name(real_dev_name, vid)) {
4126 vlandev_add(real_dev_name, vid);
4133 sset_destroy(&splinter_ifaces);
4135 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4136 free(splinter_vlans);
4139 return splinter_vlans;
4142 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4145 configure_splinter_port(struct port *port)
4147 struct ofproto *ofproto = port->bridge->ofproto;
4148 ofp_port_t realdev_ofp_port;
4149 const char *realdev_name;
4150 struct iface *vlandev, *realdev;
4152 ofproto_bundle_unregister(port->bridge->ofproto, port);
4154 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4157 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4158 realdev = iface_lookup(port->bridge, realdev_name);
4159 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4161 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4165 static struct ovsrec_port *
4166 synthesize_splinter_port(const char *real_dev_name,
4167 const char *vlan_dev_name, int vid)
4169 struct ovsrec_interface *iface;
4170 struct ovsrec_port *port;
4172 iface = xmalloc(sizeof *iface);
4173 ovsrec_interface_init(iface);
4174 iface->name = xstrdup(vlan_dev_name);
4175 iface->type = "system";
4177 port = xmalloc(sizeof *port);
4178 ovsrec_port_init(port);
4179 port->interfaces = xmemdup(&iface, sizeof iface);
4180 port->n_interfaces = 1;
4181 port->name = xstrdup(vlan_dev_name);
4182 port->vlan_mode = "splinter";
4183 port->tag = xmalloc(sizeof *port->tag);
4186 smap_add(&port->other_config, "realdev", real_dev_name);
4192 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4193 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4194 * 1-bit in the 'splinter_vlans' bitmap. */
4196 add_vlan_splinter_ports(struct bridge *br,
4197 const unsigned long int *splinter_vlans,
4198 struct shash *ports)
4202 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4203 * we're modifying 'ports'. */
4204 for (i = 0; i < br->cfg->n_ports; i++) {
4205 const char *name = br->cfg->ports[i]->name;
4206 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4209 for (j = 0; j < port_cfg->n_interfaces; j++) {
4210 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4212 if (vlan_splinters_is_enabled(iface_cfg)) {
4213 const char *real_dev_name;
4216 real_dev_name = iface_cfg->name;
4217 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4218 const char *vlan_dev_name;
4220 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4222 && !shash_find(ports, vlan_dev_name)) {
4223 shash_add(ports, vlan_dev_name,
4224 synthesize_splinter_port(
4225 real_dev_name, vlan_dev_name, vid));
4234 mirror_refresh_stats(struct mirror *m)
4236 struct ofproto *ofproto = m->bridge->ofproto;
4237 uint64_t tx_packets, tx_bytes;
4240 size_t stat_cnt = 0;
4242 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4243 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4247 if (tx_packets != UINT64_MAX) {
4248 keys[stat_cnt] = "tx_packets";
4249 values[stat_cnt] = tx_packets;
4252 if (tx_bytes != UINT64_MAX) {
4253 keys[stat_cnt] = "tx_bytes";
4254 values[stat_cnt] = tx_bytes;
4258 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);